SMJ

‌E-ISSN 2228-8082

Volume 78, Number 9, September 2026

MONTHLY

Siriraj Medical Journal

The world-leading biomedical science of Thailand

Indexed by

THAILAND SECTION 1954

https://he02.tci-thaijo.org/index.php/sirirajmedj/index

E-mail: sijournal92@gmail.com

ORIGINAL ARTICLE CORRIGENDUM

‌ORIGINAL ARTICLE‌

649 Thyroid Disorders in Preterm Infants: Prevalence, Etiologies, and Thyroid

Function Test Characteristics

Pichaya Vathanasakpuban, Buranee Yangthara, Nithiwat Vatanavicharn, Somporn Liammongkolkul, Ornsuda Lertbannaphong

SMJ

Siriraj Medical Journal

The world-leading biomedical science of Thailand

Volume 78 Number 9

September 2026

660 Cost-Effectiveness Analysis of Rimegepant Versus Eletriptan for Patients with Acute Episodic Migraine in

Thailand: Establishing Dominance from a Societal Perspective

Jadesada Lertsirimunkong, Wiwat Thavornwattanayong, Jidapha Pradab, Jidapa Khongbun, Kitiphong Khunphong, Nattanun Panyasrilert

671 Impact of Air Pollution on Acne Vulgaris: A Systematic Review

Pawat Udompataikul, Natthasith Chatpimolkul, Nattapat Chandraruangnabha, Kan Tangkijngamwong,

Phongsaphak Santimagorn, Nanticha Faromkao

683 Exploratory Fetal Doppler Thresholds for Predicting Composite Fetal Compromise in Preeclampsia: A Single-Centre Prospective Diagnostic Accuracy Study

Linh My Duong, Nghia Quang Bui, Dan Thanh Nguyen Hong, Hien Thi Nguyen

695 Design and Validation of a Low-Cost IoT-Based Medical Gas Monitoring System for Hospital Storage Rooms

Navapadol Kittiamornkul, Tawiwan Sareebot, Sumaman Buntoung, Pranomkorn Choosri

709 The Effect of Sit-To-Stand and Heel-Raise Exercises on Dynamic Balance Function in Sedentary Male Adolescents

723

Sleep Habits and Behavioral Problems in Preschool-aged Children with ADHD

Prakasit Wannapaschaiyong, Amornrat Penphattarakul, Pat Rojmahamongkol, Sureelak Sutchritpongsa

724

The Relationship Between Primary Caregivers’ Psychosocial Factors and Self-esteem in Children and Adolescents with ADHD: An Exploratory Cross-sectional Study

Prakasit Wannapaschaiyong, Amornrat Penphattarakul, Pat Rojmahamongkol, Sureelak Sutchritpongsa

CORRIGENDUM

Fifi Florensia, I Putu Alit Pawana, Yudith Dian Prawitri, Atika

‌Executive Editor:

SIRIRAJ MEDICAL JOURNAL

SMJ

https://he02.tci-thaijo.org/index.php/sirirajmedj/index

Professor Apichat Asavamongkolkul, Mahidol University, Thailand

Editorial Director:

Professor Aasis Unnanuntana, Mahidol University, Thailand

Associate Editors

Assistant Professor Adisorn Ratanayotha, Mahidol University, Thailand Pieter Dijkstra, University of Groningen, Netherlands

Professor Phunchai Charatcharoenwitthaya, Mahidol University, Thailand Professor Varut Lohsiriwat, Mahidol University, Thailand

International Editorial Board Members

Editor-in-Chief:

Professor Thawatchai Akaraviputh, Mahidol University, Thailand

Allen Finley, Delhousie University, Canada

Christopher Khor, Singapore General Hospital, Singapore Ciro Isidoro, University of Novara, Italy

David S. Sheps, University of Florida, USA

David Wayne Ussery, University of Arkansas for Medical Sciences, USA Dennis J. Janisse, Medical College of Wisconsin, USA

Dong-Wan Seo, University of Ulsan College of Medicine, Republic of Korea Folker Meyer, Argonne National Laboratory, USA

Frans Laurens Moll, University Medical Center Ultrecht, Netherlands George S. Baillie, University of Glasgow, United Kingdom

Gustavo Saposnik, Unity Health Toronto, St. Micheal Hospital, Canada Harland Winter, Harvard Medical School, USA

Hidemi Goto, Nagoya University Graduate School of Medicine, Japan Ichizo Nishino, National Institute of Neuroscience NCNP, Japan Intawat Nookaew, University of Arkansas for Medical Sciences, USA James P. Doland, Oregon Health & Science University, USA

John Hunter, Oregon Health & Science University, USA

Karl Thomas Moritz, Swedish University of Agricultural Sciences, Sweden Kazuo Hara, Aichi Cancer Center Hospital, Japan

Keiichi Akita, Institute of Science Toko, Japan Kyoichi Takaori, Kyoto University Hospital, Japan Marcela Hermoso Ramello, University of Chile, Chile Marianne Hokland, University of Aarhus, Denmark

Matthew S. Dunne, Institute of Food, Nutrition, and Health, Switzerland Mazakayu Yamamoto, Tokyo Women’s Medical University, Japan Mitsuhiro Kida, Kitasato University & Hospital, Japan

Moses Rodriguez, Mayo Clinic, USA

Nam H. CHO, Ajou University School of Medicine and Hospital, Republic of Korea Nima Rezaei, Tehran University of Medical Sciences, Iran

Noritaka Isogai, Kinki University, Japan

Philip A. Brunell, State University of New York At Buffalo, USA Philip Board, Australian National University, Australia Ramanuj Dasgupta, Genome Institution of Singapore

Richard J. Deckelbaum, Columbia University, USA Robert W. Mann, University of Hawaii, USA

Robin CN Williamson, Royal Postgraduate Medical School, United Kingdom Sara Schwanke Khilji, Oregon Health & Science University, USA

Seigo Kitano, Oita University, Japan Seiji Okada, Kumamoto University

Shomei Ryozawa, Saitama Medical University, Japan Shuji Shimizu, Kyushu University Hospital, Japan

Stanlay James Rogers, University of California, San Francisco, USA Stephen Dalton, Chinese University of HK & Kyoto University

Tai-Soon Yong, Yonsei University, Republic of Korea Tomohisa Uchida, Oita University, Japan

Victor Manuel Charoenrook de la Fuente, Centro de Oftalmologia Barraquer, Spain Wikrom Karnsakul, Johns Hopkins Children’s Center, USA

Yasushi Sano, Director of Gastrointestinal Center, Japan Yik Ying Teo, National University of Singapore, Singapore Yoshiki Hirooka, Nagoya University Hospital, Japan

Yozo Miyake, Aichi Medical University, Japan Yuji Murata, Aizenbashi Hospital, Japan

Editorial Board Members

Vitoon Chinswangwatanakul, Mahidol University, Thailand Jarupim Soongswang, Mahidol University, Thailand

Jaturat Kanpittaya, Khon Kaen University, Thailand Nopphol Pausawasdi, Mahidol University, Thailand Nopporn Sittisombut, Chiang Mai University, Thailand Pa-thai Yenchitsomanus, Mahidol University, Thailand Pornprom Muangman, Mahidol University, Thailand Prasit Wattanapa, Mahidol University, Thailand

Prasert Auewarakul, Mahidol University, Thailand Somboon Kunathikom, Mahidol University, Thailand Supakorn Rojananin, Mahidol University, Thailand

Suttipong Wacharasindhu, Chulalongkorn University, Thailand Vasant Sumethkul, Mahidol University, Thailand

Watchara Kasinrerk, Chiang Mai University, Thailand Wiroon Laupattrakasem, Khon Kaen University, Thailand Yuen Tanniradorn, Chulalongkorn University, Thailand

Editorial Assistant: Nuchpraweepawn Saleeon, Mahidol University, Thailand

Proofreader: Amornrat Sangkaew, Mahidol University, Thailand, Nuchpraweepawn Saleeon, Mahidol University, Thailand

‌Thyroid Disorders in Preterm Infants: Prevalence, Etiologies, and Thyroid Function Test Characteristics

Pichaya Vathanasakpuban, M.D.1, Buranee Yangthara, M.D.2, Nithiwat Vatanavicharn, M.D.3, Somporn Liammongkolkul, M.Sc.3, Ornsuda Lertbannaphong, M.D.1,*

1Division of Endocrinology and Metabolism, Department of Pediatrics, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok 10700,

Thailand, 2Division of Neonatology, Department of Pediatrics, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok 10700, Thailand,

3Division of Medical Genetics, Department of Pediatrics, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok 10700, Thailand.

*Corresponding author: Ornsuda Lertbannaphong E-mail: ornsuda.ler@mahidol.ac.th

Received 20 April 2026 Revised 18 July 2026 Accepted 18 July 2026 ORCID ID: http://orcid.org/0000-0003-2087-5149 https://doi.org/10.33192/smj.v78i9.281716

All material is licensed under terms of the Creative Commons Attribution 4.0 International (CC-BY-NC-ND 4.0) license unless otherwise stated.

ABSTRACT

Objective: To determine the prevalence, etiologies, risk factors, and characteristics of thyroid disorders in preterm infants.

Materials and Methods: This retrospective review included eligible infants with gestational age (GA) < 37 weeks born at Siriraj Hospital in 2022. Infants with genetic syndromes, chromosomal abnormalities, or maternal thyroid disease were excluded. Thyroid-stimulating hormone (TSH) from neonatal dried blood spots was measured using time-resolved fluoroimmunoassay at 48–72 hours and again at 2 weeks of age (cutoff ≥ 12 mIU/L), followed by serum thyroid function tests (electrochemiluminescence immunoassay) obtained at 4 weeks until term-corrected age (cutoff ≥ 12.53 mIU/L). Risk factors were identified using multiple logistic regression.

Results: Among 603 preterm infants, the mean gestational age was 35 weeks and the mean birth weight was 2,230 g.

Thyroid disorder prevalence was 2.99% (18/603), with hyperthyrotropinemia being the most common diagnosis. Multiple variables were analyzed but only the 5-minute Apgar score (P = 0.013; adjusted odds ratio 0.635; 95% CI 0.443–0.910) and maternal preeclampsia (P = 0.039; adjusted odds ratio 5.064; 95% CI 1.084–23.661) were associated with hyperthyrotropinemia. TSH levels increased from the first to second week of life in infants with GA 28–33

weeks, but decreased in those with GA 34–36 weeks. Free thyroxine increased with GA in infants < 34 weeks but decreased in GA 34–36 weeks.

Conclusions: The prevalence of thyroid disorders was approximately 3%, with hyperthyrotropinemia being the most

common. Lower 5-minute Apgar scores and maternal preeclampsia were associated with hyperthyrotropinemia. TSH and free thyroxine levels tended to increase in infants with GA < 34 weeks but decrease in those with GA

> 34 weeks. Abnormal TFTs were detected in subsequent weeks of life, supporting the practice of repeated TSH screening in preterm infants until term. More large-scale, prospective studies are needed to follow preterm infants with thyroid disorders to assess cognitive, motor, and behavioral outcomes.

Keywords: Congenital hypothyroidism; hyperthyrotropinemia; hypothyroxinemia; maternal preeclampsia; preterm infants; thyroid disorders (Siriraj Med J 2026;78(9):649-659)

INTRODUCTION

The thyroid hormone is essential for the development and function of multiple organ systems, particularly the brain,1 prompting the establishment of newborn screening programs for congenital thyroid disorders worldwide. In Thailand, the newborn screening program uses thyroid-stimulating hormone (TSH) obtained from umbilical cord blood or neonatal dried blood spots (DBSs), collected only once at 48–72 hours of age, whereas some countries use both TSH and free thyroxine (FT4). Healthy term singleton infants are screened once. Previously, preterm infants required two screenings: the first at 48–72 hours and the second at 2 weeks of age. Recently, the Thai Society for Pediatric Endocrinology (TSPE) recommended that screening should be performed every 2 weeks until term-corrected age because some preterm infants may exhibit low TSH levels, delayed response to thyrotropin-releasing hormone, or delayed TSH elevation. https://thaipedendo. org/guideline-for-congenital-hypothyroidism/ In addition, they may initially have low total thyroxine, free thyroxine, and triiodothyronine, which gradually increase with age.1 A previous study reported a 46.7% prevalence of

thyroid disorders in preterm infants, including congenital hypothyroidism (4.79%), hyperthyrotropinemia (5.99%), and transient hypothyroxinemia of prematurity (23.95%).2 Beyond prematurity, several risk factors may adversely affect thyroid function tests (TFTs), such as multiple pregnancies, neonatal comorbidities, and medication exposure. Given these considerations, including the recent TSPE recommendations, this study aimed to determine the prevalence, etiologies, and risk factors of thyroid disorders in preterm infants. Increased frequency of TFT screenings may lead to more detection of abnormal results in preterm infants, such as congenital hypothyroidism, potentially improving neurodevelopmental outcomes. However, the benefit of prompt treatment in transient conditions, such as transient hypothyroxinemia, is uncertain. Additionally, this study characterized TFT patterns in this population, which may facilitate appropriate interpretation and identification of associated risk factors.

MATERIALS AND METHODS

Study population and data collection

A retrospective chart review was conducted among

inborn preterm infants with gestational age (GA) less than 37 weeks who were born at Siriraj Hospital between January 1, 2022, and December 31, 2022. Data were extracted from electronic medical records. Investigators were not blinded to outcomes during data collection. Preterm infants with genetic syndromes, multiple congenital anomalies, chromosomal abnormalities, or maternal history of thyroid disease were excluded. GA was determined using first- or second-trimester ultrasonography. If unavailable, the last menstrual period was used; if both were unavailable, gestational age was assessed via physical examination using the Ballard score.

Collected data included demographic data of preterm infants, including GA, birth weight, length, head circumference, and Apgar score, as well as maternal age, history of preeclampsia, gestational diabetes, hypertension, and dexamethasone use. Comorbidities were recorded as present or absent and included respiratory conditions (respiratory distress syndrome, apnea of prematurity, transient tachypnea of the newborn, and bronchopulmonary dysplasia), cardiovascular conditions (patent ductus arteriosus and congenital heart disease), and other comorbidities (neonatal sepsis, intraventricular hemorrhage, necrotizing enterocolitis, hemangioma). Medication exposure was documented as used or not used and included inotropic agents (dopamine and dobutamine), respiratory stimulants (aminophylline and caffeine), surfactant, antibiotics (cefotaxime, metronidazole), and other medications (prostaglandin, furosemide, heparin, iodine, total parenteral nutrition, hepatitis B immunoglobulin, steroids, alpha-adrenergic agonists, opioids, paracetamol). To illustrate gestational age-specific patterns of postnatal TSH and FT4, median hormone concentrations at 48–72 hours, 2 weeks, and 4 weeks of life were calculated for each gestational age group (24–27, 28–30, 31–33, and 34–36 weeks). Only thyroid function test results within the established reference ranges for preterm infants were included in this descriptive analysis. Repeated specimens from the same infant were not treated as independent observations; instead, the plots represent descriptive, selected-specimen summaries.

Newborn screening protocol

The institutional protocol for newborn screening in preterm infants consisted of measuring TSH from DBSs at 48–72 hours of age for the first screening and at 2 weeks for the second screening. For subsequent screenings, TFTs, including serum TSH and FT4, were performed. TFTs were obtained in all infants with abnormal TSH results from DBS screening. In infants who were not of Thai nationality, newborn screening was either not

performed or TFTs were used instead. Because DBS screening measured only TSH, central hypothyroidism and transiently low TSH levels could not be detected.

Thyroxine treatment evaluation

Preterm infants who received thyroxine treatment were evaluated for discontinuation at 3 years of age. After thyroxine withdrawal for 2–4 weeks, TFTs were reassessed. If results were normal, TFTs were subsequently monitored every 3–6 months during the first year and every 1–2 years thereafter. However, if TSH was ≥ 5 mIU/L or free thyroxine was low, lifelong thyroxine therapy was considered.

Laboratory methods

The DBS collection procedure followed the guideline of the Department of Medical Sciences, Ministry of Public Health, Thailand (https://www.neoscreen.go.th/images/ pdf/guildeline.pdf). Time-resolved fluoroimmunoassay (Delfia Neonatal hTSH) was used to measure TSH levels from DBS samples, with abnormal results defined as TSH

≥ 12 mIU/L. Electrochemiluminescence immunoassay (Cobas e411 and Cobas e601 analyzers; Roche) was used to measure TFTs. An abnormally elevated serum TSH level was defined as ≥ 12.53 mIU/L.

Definitions

Reference ranges for thyroid hormones in preterm infants were derived from a literature review.3 Abnormal TFT patterns were defined as follows:

• Congenital hypothyroidism: elevated TSH with

  decreased free thyroxine

• Hyperthyrotropinemia: elevated TSH with normal free thyroxine

• Hypothyroxinemia: normal TSH with decreased free thyroxine

• Central hypothyroidism: decreased or normal TSH with decreased free thyroxine

• Transiently high free thyroxine: normal TSH with elevated free thyroxine

• Transiently low TSH: decreased TSH with normal free thyroxine

• Non-thyroidal illness: decreased or normal TSH with decreased or normal free thyroxine, decreased or normal thyroxine, or decreased free or total triiodothyronine

Statistical analysis

Statistical analysis was performed using IBM SPSS Statistics version 25 (IBM Corp, Armonk, NY, USA). Categorical variables were reported as frequencies and percentages. Continuous data were reported as medians

with IQRs for non-normally distributed data and as means with SDs for normally distributed data. Fisher’s exact test was used for comparisons of categorical data, whereas the Mann–Whitney U test was used for comparisons of continuous data. Univariable analyses were conducted for all thyroid function categories and no significant associations were identified apart from hyperthyrotropinemia. Variables with a p value < 0.2 were subsequently included in multivariable logistic regression analysis. A p value < 0.05 was considered statistically significant.

Ethical considerations

This study was approved by the Ethics Committee of the Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand (COA no. Si 361/2023).

RESULTS

Screening results

Among 664 inborn preterm infants, 603 were included in the study. During the first week of life, DBS screening was performed in 588 of 603 infants, whereas TFTs were performed in 26 of 603 infants. Among these,

3 DBS samples were positive, and 4 abnormal TFT results were identified, including one case of congenital hypothyroidism, two cases of hyperthyrotropinemia, and one case of transiently high free thyroxine.

During the second week of life, an additional 206 DBS screenings and 69 TFTs were performed, yielding four additional abnormal TFT results: two cases of hyperthyrotropinemia, one case of hypothyroxinemia, and one case of transiently elevated free thyroxine.

After the fourth week of life, 17 DBS screenings and 431 TFTs were performed, resulting in 10 additional abnormal TFT results. These included one case of congenital hypothyroidism, four cases of hyperthyrotropinemia, three cases of hypothyroxinemia, one case of transiently low TSH, and one case of central hypothyroidism, as detailed in Fig 1.

Characteristics of preterm infants

The characteristics of preterm infants with abnormal and normal TFTs are presented as median (IQR) values in Table 1. Compared with infants with normal TFTs, those with abnormal TFTs or thyroid disorders had statistically significantly lower GA (P = 0.009), birth

Fig 1. Flowchart of newborn thyroid screening using dried blood spots and thyroid function tests in preterm infants.

Of 664 preterm infants screened, 61 were excluded (maternal thyroid disease, n = 33; congenital anomaly, n = 6; transferred, n = 4; died, n = 11; no TFT performed, n = 7), leaving 603 infants enrolled. Screening was performed at three time points: first week, second week, and after the fourth week of life. Abnormal TFT results were classified by diagnosis.

Abbreviations: CH, congenital hypothyroidism; DBS, dried blood spot; FT4, free thyroxine; HT, hyperthyrotropinemia; NBS, newborn screening; TFT, thyroid function test; TSH, thyroid-stimulating hormone

TABLE 1. Characteristics of preterm infants with abnormal and normal thyroid function test results.

Data are presented as median (IQR) and percentages.

aThe Mann–Whitney U test was used for non-normally distributed continuous variables.

bFisher’s exact test was used for categorical variables. A p-value < 0.05 indicates statistical significance.

Abbreviations: GA, gestational age; IQR, interquartile range; TFT, thyroid function test; SGA, small for gestational age; RDS, respiratory

distress syndrome; NEC, necrotizing enterocolitis

weight, birth length, and head circumference (all P

< 0.001). Apgar scores at 1 and 5 minutes were also statistically significantly lower in infants with abnormal TFTs (P = 0.034 and P = 0.007, respectively).

Prevalence of thyroid disorders

The overall prevalence of thyroid disorders in this cohort of preterm infants was 2.99% (18/603) (95% CI 1.77-4.67). The disorders were classified as follows: hyperthyrotropinemia 1.33% (8/603) (95% CI 0.57-2.59),

hypothyroxinemia 0.66% (4/603) (95% CI 0.18-1.68)

congenital hypothyroidism 0.33% (2/603) (95% CI 0.04-

1.19), transiently high free thyroxine 0.33% (2/603) (95%

CI 0.04-1.19), central hypothyroidism 0.17% (1/603) (95% CI 0.00-0.92), and transiently low TSH 0.17% (1/603) (95% CI 0.00-0.92).

Risk factor analysis

The characteristics of preterm infants and their mothers, including comorbidities and medication exposure, were initially evaluated using univariable logistic regression. The results of multivariable logistic regression analysis

of risk factors associated with thyroid disorders are presented in Table 2. Iodine exposure was observed in 3/8 (37.5%) patients in the hyperthyrotropinemia group, 4/4 (100%) in the hypothyroxinemia group and 118/585 (20%) in the normal TFT group.

The 5-minute Apgar score in infants with hyperthyrotropinemia was statistically significantly lower (median 8.5) than that in infants with normal TFTs (median 9; P = 0.013; adjusted odds ratio 0.635; 95% CI 0.443–0.910). Maternal preeclampsia was associated with hyperthyrotropinemia group (37.5%) than in the normal TFT group (12.48%; P = 0.039; adjusted odds ratio 5.064; 95% CI 1.084–23.661). No statistically significant difference in comorbidities was observed between infants with thyroid disorders and those with normal TFTs in the multivariable logistic regression analysis.

Gestational Age-Specific Patterns of Normal Postnatal Thyroid Function in Preterm Infants

The characteristics of normal TFT results in preterm infants are illustrated in Figs 2 and 3.

Median TSH levels increased from the first to the

TABLE 2. Multiple logistic regression analysis of risk factors for hyperthyrotropinemia in preterm infants

aaOR were calculated using a multivariate logistic regression model. Apgar scores are presented as median (IQR). Maternal preeclampsia is presented as n/N (%).

A p-value < 0.05 indicates statistical significance.

Abbreviations: aOR, adjusted odds ratio; CI, confidence interval; Hyper-TSH, hyperthyrotropinemia; IQR, interquartile range; TFT, thyroid function test

Fig 2. Median thyroid-stimulating hormone levels in preterm infants by gestational age and week of life.

Median TSH levels (mIU/L) are plotted for four gestational age groups (24–27, 28–30, 31–33, and 34–36 weeks) at three time points (first week, second week, and fourth week of life). Sample sizes (n) for each gestational age group and time point are shown below the x-axis. TSH levels generally increased from the first to second week in infants with gestational age 28–33 weeks but decreased in those with gestational age 34–36 weeks.

Abbreviations: n, number of thyroid function test specimens; TFT, thyroid function test; TSH, thyroid-stimulating hormone

Fig 3. Median free thyroxine levels in preterm infants by gestational age and week of life.

Median free thyroxine levels (ng/dL) are plotted for four gestational age groups (24–27, 28–30, 31–33, and 34–36 weeks) at three time points (first week, second week, and fourth week of life). Sample sizes (n) for each gestational age group and time point are shown below the x-axis. Free thyroxine levels trended upward with increasing gestational age in infants under 34 weeks but decreased in the 34–36 weeks subgroup. Abbreviations: FT4, free thyroxine; n, number of thyroid function test specimens; TFT, thyroid function test

second week of life in preterm infants with GA 28–33 weeks, particularly in the 31–33 weeks subgroup. In contrast, TSH decreased in infants with GA 34–36 weeks. Median free thyroxine levels increased with increasing

GA in infants born before 34 weeks of gestation but decreased in the 34–36 weeks subgroup.

Clinical characteristics and management

Demographic characteristics, TFT results, management, and clinical course of preterm infants with abnormal TFTs are summarized in Table 3. Eight of the 18 patients received thyroxine treatment: four for hyperthyrotropinemia, one for hypothyroxinemia, two for congenital hypothyroidism, and one for central hypothyroidism. The decision to initiate treatment was made at the discretion of the attending physician.

Outcomes by diagnosis and time to start treatment

• Hyperthyrotropinemia: Among the eight patients with hyperthyrotropinemia, six had normal results on both the first and second DBS screenings, while two had an abnormal first DBS followed by a normal second DBS. During subsequent screenings, elevated serum TSH levels were detected in five patients on the third screening, two patients on the fourth screening and one patient on the fifth screening. Treatment was initiated because TSH levels progressively increased during follow-up. Among the four treated patients, one had normal TFTs at 3 years of age and subsequently discontinued thyroxine. The remaining three patients were either lost to follow-up, transferred to another hospital, or had not yet reached three years of age for re-evaluation.

• Congenital hypothyroidism: Two patients were

  diagnosed with congenital hypothyroidism. The first patient had elevated TSH levels on both the first and second DBS screenings, and the diagnosis was subsequently confirmed by TFTs. The second patient had normal TSH levels on both initial DBS screenings; however, elevated TSH and low FT4 levels were detected approximately 6 weeks after birth.

• Hypothyroxinemia, transiently high FT4, central

hypothyroidism and transiently low TSH: There were 4 patients with hypothyroxinemia. They were diagnosed around 1 month after birth. During the follow-up period, TFTs returned to normal in 3 patients. However, one patient demonstrated decreasing FT4 levels so the thyroxine treatment was started. This patient continued thyroxine treatment at 3 years of age. There were 2 patients with transiently high free T4. This condition was spontaneously resolved on 35 and 47 days after birth. There was one patient with central hypothyroidism. The diagnosis

was made due to persistently low FT4 and TSH levels. The patient with central hypothyroidism also remained on thyroxine treatment. There was one patient with transiently low TSH. The diagnosis was made on day of life 31. The TSH level was spontaneously returned to normal level on day of life 52.

Time to resolution in untreated patients

Among the 10 of 18 preterm infants with abnormal TFTs who did not receive treatment, the mean time to resolution varied according to diagnosis. There solution times were 29-50 days for hyperthyrotropinemia, 40-78 days for hypothyroxinemia, 35-47 days for transiently elevated free thyroxine, and 52 days for transiently low TSH.

DISCUSSION

In this study, the prevalence of thyroid disorders among preterm infants was 2.99% (18/603). The most common diagnosis was hyperthyrotropinemia, accounting for 44.4% of cases, followed by hypothyroxinemia (22.2%), congenital hypothyroidism (11.1%), transiently high free thyroxine (11.1%), central hypothyroidism (5.6%), and transiently low TSH (5.6%).

A lower 5-minute Apgar score and maternal preeclampsia were associated with hyperthyrotropinemia. TSH levels increased from the first to the second week of life in preterm infants with GA 28–33 weeks but decreased in those with GA 34–36 weeks. Free thyroxine levels increased with advancing GA in preterm infants born before 34 weeks of gestation but decreased in the 34–36 weeks subgroup.

The prevalence of thyroid disorders in preterm infants observed in our study differed from that reported in previous studies. Despite using lower TSH cutoff values, which theoretically increase screening sensitivity, our study demonstrated a lower prevalence of thyroid disorders than Lee et al.4 (12%). This discrepancy is likely attributable to differences in baseline characteristics, as Lee et al’s cohort had lower gestational ages, lower birth weights, and lower Apgar scores; all of which are established risk factors for transient thyroid dysfunction.5-8 Similarly, Ali et al.9 reported a thyroid disorder prevalence of 10%. Such variations between studies may be explained by differences in exclusion criteria, particularly regarding maternal thyroid disease history.

The prevalence of hyperthyrotropinemia in the present study was 1.33% (8/603). However, previous studies have reported a prevalence ranging from 2.0% to 13.1% in preterm infants.4-6,9 Consistent with evidence that lower GA and small-for-GA status are key risk

TABLE 3. Demographics data, thyroid function test results, management, and clinical course of preterm infants with abnormal thyroid function tests.

4

Hyperthyrotropinemia

33+6

1940 33

38+4

11.7

1.13

—

50

41 —

5 Hyperthyrotropinemia 36+5 2450 48 43+4 11.3 1.77 5 — — Re-evaluation

pending at 3 y of age

8

Hyperthyrotropinemia

33+5

1600 50

40+6

11.3

1.18

6.5

—

— Followed at another

hospital

discontinued

10 Hypothyroxinemia

34+6

1870

36

40

3.14 0.79

—

43

41 —

9 Hypothyroxinemia 35 1650 26 38+5 3.12 0.78 — 40 40+5 —

12 Hypothyroxinemia

29

610

27

3

11 Hypothyroxinemia 31+1 850 24 34+4 5.07 0.67 — 78 42+2 —

3

37+5

3.01

2.26

—

35

40+5

—

3

18 Transiently low TSH

27+6

1120

31

32+2

0.48

1.26

—

52

35+2

—

17 Central hypothyroidism 36+4 2410 38 42 4.44 0.66 4.9 — — T4 continued

“—” indicates not done or not applicable.

Abbreviations: CRA, corrected age; DOL, day of life; GA, gestational age; T3, triiodothyronine; T4, thyroxine; TFT, thyroid function test; TSH, thyroid-stimulating hormone

factors for hyperthyrotropinemia5,6, Grob et al.5 reported lower mean GA (30.6 vs. 35 weeks) and lower mean birth weight (1525 vs. 2230 g) than those observed in the current study. In our cohort, hyperthyrotropinemia was the most common abnormal TFT pattern, in contrast to the predominance of hypothyroxinemia reported by Mohamed et al.2 and Rehman et al.10 The lower incidence of hypothyroxinemia in our study may be attributed to our primary use of TSH-based screening. Implementation of combined TSH and FT4 screening may improve detection of hypothyroxinemia in preterm infants.

The high incidence of hyperthyrotropinemia observed in our study may be associated with iodine exposure. Specifically, 37.5%(3/8) of infantswith hyperthyrotropinemia and 100% (4/4) of those with hypothyroxinemia had been exposed to iodine-based disinfectants during clinical procedures, such as umbilical venous catheterization, umbilical arterial catheterization, or lumbar puncture. Iodine exposure from surgical disinfectants is statistically linked to transient hyperthyrotropinemia in infants, consistent with retrospective data from Chung et al.11 In addition, previous evidence suggests that preterm neonates are more susceptible to the thyroid-suppressive effects of iodine than full-term infants.12,13 This vulnerability likely occurs because the thyroid’s ability to reduce iodine trapping in response to an iodine bolus — the protective Wolff–Chaikoff effect — does not fully mature until 36–40 weeks GA.14 When feasible, alternative non-iodine antiseptics should be considered for neonatal procedures to prevent iatrogenic thyroid suppression. The prevalence of hypothyroxinemia and congenital hypothyroidism in the current study (0.66% and 0.33%, respectively) was lower than that reported by Maneenil

et al.16 (4.8% and 3.0%, respectively). This discrepancy

may be explained by the lower mean GA and birth weight observed in their cohort. Therefore, clinicians should expect a higher rate of thyroid disorders and monitor infants more closely when managing preterm or low birth weight infants.

In the present study, multivariable logistic regression identified a low 5-minute Apgar score as a significant risk factor for hyperthyrotropinemia in preterm infants. This finding is consistent with results reported by Chung et al,11 who demonstrated a similar association. This relationship can be explained by the fact that low Apgar

scores reflect perinatal distress, such as hypoxia, which can trigger thyrotropin-releasing hormone secretion and subsequently increase TSH levels.

Maternal preeclampsia was also identified as an association factor for hyperthyrotropinemia in preterm infants. This association may be explained by a compensatory

increase in neonatal TSH secretion in response to relatively low maternal free thyroxine levels. Belet et al.17 found that infants born to preeclamptic mothers had lower thyroid hormones and higher TSH levels compared to healthy preterm infants, suggesting intrauterine hypothyroidism. Furthermore, a meta-analysis by Toloza et al.18 reported that maternal subclinical hypothyroidism is associated with an increased risk of preeclampsia. This relationship may result from relative thyroid hormone deficiency leading to reduced production of vasoactive substances such as nitric oxide, resulting in impaired vasorelaxation, increased vascular resistance, and ultimately hypertension.

In the present study, TSH levels increased from the first to second week of life in preterm infants with GA 28–33 weeks but decreased in those with GA 34–36 weeks. This pattern is similar to that demonstrated by Williams et al,21 who showed that TSH values increased between the first and second weeks of life in preterm infants with GA 23–34 weeks. This transient TSH elevation may represent a mechanism to stimulate the immature thyroid gland and compensate for the loss of maternal thyroxine.1 Free thyroxine levels in this study increased slightly from the first to second week in preterm infants with GA 28–33 weeks. This finding may be explained by a combination of immature thyroid hormone production, reduced production of thyroid-binding proteins by the liver, and the influence of non-thyroidal illness.1

Among the eight patients who received thyroxine treatment, only one continued treatment beyond 3 years of age, whereas one patient had not yet reached 3 years of age. These findings emphasize that thyroid disorders in preterm infants are frequently transient and tend to improve with time as the hypothalamic-pituitary-thyroid axis matures.1

Limitations and strength

This study had several limitations. First, the use of TSH alone as the primary screening test in most cases may have underestimated cases of central hypothyroidism and hypothyroxinemia. In addition, not all preterm infants underwent screening at all three recommended time points, which may have introduced selection bias. Second, the retrospective study design limited control over potential confounding factors, although we adjusted for these using multivariable logistic regression analysis. Third, this was a single-center study, which may not be representative of the global population. Fourth, the decisions regarding treatment and follow-up were made by the attending physician, which may have introduced classification bias. The strengths of this study was that the use of electronic medical records minimized missing

data and allowed for a comprehensive analysis of the variables.

CONCLUSION

The prevalence of thyroid disorders among preterm infants in this study was approximately 3%, with hyperthyrotropinemia being the most common. Lower 5-minute Apgar scores and maternal preeclampsia were identified as factors associated with hyperthyrotropinemia. TSH and free thyroxine levels trended upward in preterm infants with GA < 34 weeks but decreased in those with GA ≥ 34 weeks.

Clinical implications

Thyroid disorders in preterm infants were relatively common, especially hyperthyrotropinemia. In this study, abnormal TFT results were detected in subsequent weeks of life, supporting the practice of repeated TSH screening in preterm infants until term. Although TSH-based screening alone can detect congenital hypothyroidism, comprehensive TFTs provide additional information on other abnormalities beyond what TSH screening can identify. Future prospective studies are needed to standardize screening strategies and minimize diagnostic variability. In addition, long-term neurodevelopmental follow-up of preterm infants with thyroid dysfunction is warranted to ensure that subtle or delayed sequelae are properly identified and managed.

Data Availability Statement

All data generated or analyzed during the present study are included in this published article. Further details are available for non-commercial purposes from the corresponding author on reasonable request.

DECLARATIONS

Grants and Funding Information

The authors did not receive grants or funding for this research.

Conflict of Interest

The authors declare no potential conflicts of interest.

Registration Number of Clinical Trial

Not applicable.

Author Contributions

Conceptualization and methodology, P.V., B.Y.,

N.V. and O.L.; Investigation, B.Y., N.V., S.L. and O.L.; Formal analysis, P.V., B.Y. and O.L. ; Visualization and writing – original draft, P.V. and O.L. ; Writing –

review and editing, P.V. and O.L. ; Supervision, O.L. All authors have read and agreed to the final version of the manuscript.

Use of Artificial Intelligence

The submitted work was produced without the use of artificial intelligence-assisted technologies.

Institutional Review Board Statement

This research adhered to established ethical standards and received formal authorization from the Siriraj Institutional Review Board (COA no. Si 361/2023). Written informed consent was not obtained because this is a retrospective study.

REFERENCES

1. LaFranchi SH. Thyroid function in preterm/low birth weight infants: impact on diagnosis and management of thyroid dysfunction. Front Endocrinol (Lausanne). 2021;12:666207.

2. Mohamed AIAE, Ahmed El-Masry HM, Hassan AA, Sayed Younis MM, Aladawy MA, Hashem AM. Screening for thyroid dysfunction among preterm infants at Manfalout General Hospital, Assiut, Egypt. Annals of Neonatology. 2022;4(1): 71-90.

3. Zhu L, Zhang X, He X, Yang X, Wang Y, Wang C et al. Reference intervals for serum thyroid hormones in preterm hospitalized infants. J Pediatr Endocrinol Metab. 2013;26(5-6):463-7.

4. Lee JH, Kim SW, Jeon GW, Sin JB. Thyroid dysfunction in very low birth weight preterm infants. Korean J Pediatr. 2015;58(6): 224-9.

5. Grob F, Gutiérrez M, Leguizamón L, Fabres J. Hyperthyrotropinemia is common in preterm infants who are born small for gestational age. J Pediatr Endocrinol Metab. 2020;33(3):375-82.

6. Jeong MH, Lee N, Bae MH, Han YM, Park KH, Byun SY. Risk factors for delayed hyperthyrotropinemia in late preterm infants. Neonatal Med. 2019;26(4):204-12.

7. Rameshbabum M, Subramanian S, Sethuraman G. Thyroid function profiles in small versus appropriate for gestational age stable neonates with implications for screening cutoffs: a retrospective cross-sectional study. J Matern Fetal Neonatal Med. 2025;38(1):2517762.

8. Bosch-Giménez VM, Palazón-Bru A, Blasco-Barbero Á, Juste-Ruiz M, Rizo-Baeza MM, Cortés-Castell E. Multivariate analysis of thyrotropin in preterm newborns based on adequacy of weight for gestational age. Thyroid 2017;27(1):120-4.

9. Ali H, Mohamad HA, Aboelela MGM. Thyroid Function in Healthy Preterm Infants at the NICU of Assiut University Children’s Hospital: An Observational Study. J Curr Med Res Pract. 2025;10(4):123-38.

10. Rehman A, Khan WI, Quddusi AI, Nadeem A, Iqbal I. Thyroid function disorders in low birth weight/preterm neonates admitted in NICU, Children Hospital Complex Multan. Professional Med J. 2020;27(08):1649-55.

11. Chung HR, Shin CH, Yang SW, Choi CW, Kim BI, Kim EK, et al. High incidence of thyroid dysfunction in preterm infants. J Korean Med Sci. 2009;24:627-31.

12. Linder N, Davidovitch N, Reichman B, Kuint J, Lubin D,

    
    

    Meyerovitch J, et al. Topical iodine-containing antiseptic and subclinical hypothyroidims in preterm infants. J Pediatr. 1997;131(3):434-9.

13. Smerdely P, Boyages S, Wu D, Leslie G, John E, Lim A, et al. Topical iodine-containing antiseptics and neonatal hypothyroidism in very-low-birthweight infants. Lancet. 1989;2(8664):661-4.

14. Brown RS. The thyroid. In: Brook CGD, Clayton P, Brown R, editors. Brook’s clinical pediatric endocrinology. 6th ed. Oxford: Wiley-Blackwell; 2009. p. 250-82.

15. Prasad HK, Pulluru P, Venugopalan L, Murugesan G, Ramanathan

    S. Congenital hypothyroidism in Indian preterm babies–screening, prevalence, and aetiology. Pediatr Endocrinol Diabetes Metab. 2021;27(2):82-6.

16. Maneenil G, Juthong M, Thatrimontrichai A, Janjindamai W, Dissaneevate S, Jaruratanasirikul S. Incidence and risk factors of transient hypothyroxinemia of prematurity: a prospective cohort study. Turk J Pediatr. 2023;65:562-71.

17. Belet N, Imdat H, Yanık F, Küçüködük Ş. Thyroid function tests in preterm infants born to preeclamptic mothers with

    placental insufficiency. J Pediatr Endocrinol Metab. 2003;16(8): 1131-5.

18. Toloza FJ, Derakhshan A, Männistö T, Bliddal S, Popova PV, Carty DM, et al. Association between maternal thyroid function and risk of gestational hypertension and pre-eclampsia: a systematic review and individual-participant data meta-analysis. Lancet Diabetes Endocrinol. 2022;10(4):243-52.

19. Deshpande S, Yelikar K, Patil S, Andurkar S. Maternal thyroid hormone status in pre-eclampsia: a tertiary care hospital based study. Int J Reprod Contracept Obstet Gynecol. 2015;4(6): 1853-7.

20. Qublan H, Al-Kaisi I, Hindawi I, Hiasat M, Awamleh I, Hamaideh A, et al. Severe pre-eclampsia and maternal thyroid function. J Obstet Gynaecol. 2003;23(3):244-6.

21. Williams FL, Simpson J, Delahunty C, Ogston SA, Bongers-Schokking JJ, Murphy N et al. Developmental trends in cord and postpartum serum thyroid hormones in preterm infants. J Clin Endocrinol Metab. 2004;89(11):5314-20.

‌Cost-Effectiveness Analysis of Rimegepant Versus Eletriptan for Patients with Acute Episodic Migraine in Thailand: Establishing Dominance from a Societal Perspective

Jadesada Lertsirimunkong, PharmD1, Wiwat Thavornwattanayong, FCCP., FACP.2,*, Jidapha Pradab, PharmD3, Jidapa Khongbun, PharmD3, Kitiphong Khunphong, PharmD3, Nattanun Panyasrilert, PharmD3

1Department of Social and Administrative Pharmacy, College of Pharmacy, Rangsit University, Pathum Thani, Thailand, 2Health Innovation and

Research Outcomes Group, Faculty of Pharmacy, Silpakorn University, Nakorn-pathom, Thailand, 3Faculty of Pharmacy, Silpakorn University, Nakorn-pathom, Thailand.

*Corresponding author: Wiwat Thavornwattanayong E-mail: ajwiwat@gmail.com

Received 10 June 2026 Revised 1 July 2026 Accepted 7 July 2026 ORCID ID: http://orcid.org/0000-0002-6617-8587 https://doi.org/10.33192/smj.v78i9.282940

All material is licensed under terms of the Creative Commons Attribution 4.0 International (CC-BY-NC-ND 4.0) license unless otherwise stated.

INTRODUCTION

According to the Global Burden of Disease (GBD) 2019 report, the age-standardized incidence rate (ASIR) of migraine across countries ranges from 962 to 1,336 cases per 100,000 population. In Thailand, the ASIR is the highest among Non-High-Income Southeast Asian countries.1,2 The highest prevalence is observed in individuals aged 30–39 years, and it is significantly more common in females than in males, with prevalence rates of 18% in women, 6% in men, and 4% in children. Moreover, approximately 85% of patients experience moderate to severe pain. The World Health Organization (WHO) therefore recognizes migraine as a major cause of functional disability. If migraine is untreated or inadequately managed, attacks may last from 4 to 72 hours, severely affecting patients’ quality of life. In Thailand, the economic burden of migraine has been estimated to amount to several tens of billions of Thai baht annually.

Current acute migraine management guidelines in

Thailand (2022) and in other countries3,4 recommend triptans as first-line therapy for patients with moderate to severe attacks. In Thailand, the triptans commonly prescribed are sumatriptan and eletriptan.5 However, the effectiveness of sumatriptan has important limitations, with reports indicating that approximately 30–45% of patients6,7 do not respond to treatment even at the maximum recommended dose. Evidence from a systematic review and network meta-analysis by Karlsson et al. (2024)8

demonstrates that eletriptan provides several advantages over sumatriptan, including higher therapeutic gain, greater consistency of response6,8,9, improved sustained response, and higher rates of pain freedom at 2 hours. These findings make eletriptan an attractive option for patients who do not respond to sumatriptan. Nevertheless, all triptans share a key limitation related to their mechanism of action–agonism at the 5-HT1B receptor–which induces vasoconstriction.8 This effect may increase the risk of cardiovascular or cerebrovascular events and has been associated with chest tightness in 42–78% of patients8, potentially leading to myocardial ischemia, myocardial infarction, or stroke.10 Furthermore, the use of triptans on more than 10 days per month for over 3 consecutive months has been associated with medication-overuse headache (MOH).11

Recently, newer medications specifically targeting the calcitonin gene–related peptide receptor (CGRP receptor antagonists), known as gepants, have been developed for the acute treatment of migraine. Although several gepants exist, only two–rimegepant and ubrogepant–are currently available in Thailand.4 Rimegepant is formulated as an oral disintegrating tablet (ODT)4,5, which helps reduce issues related to nausea or vomiting during attacks. Clinical studies have shown that rimegepant reduces the recurrence of migraine attacks more effectively than other gepants and offers significantly higher rates of sustained 24-hour pain freedom versus placebo. Compared with

eletriptan, rimegepant can reduce the risk of headache recurrence by approximately 71%8 and demonstrates lower rates of adverse events, such as nausea and vomiting (less than 3%).12 Importantly, no evidence of hepatotoxicity in long-term use of rimegepant13, and it has not been associated with MOH, even with repeated use.12,14,15

Economic evaluations of rimegepant from other countries have produced inconsistent results. In China, rimegepant was found to be not cost-effective due to its high drug cost relative to its clinical benefits.16 In contrast, an evaluation conducted in the United Kingdom comparing rimegepant with best supportive care demonstrated that rimegepant significantly improved health outcomes and was considered cost-effective within the UK healthcare context.15 Conversely, an assessment by the Institute for Clinical and Economic Review 2020 (ICER) in the United States12 concluded that rimegepant was not cost-effective when compared with triptans. These discrepancies reflect variations in drug costs, healthcare system structures, and time horizons applied in each economic model, highlighting why results from other countries cannot be directly applied to the Thai context.

To date, no cost-effectiveness analysis (CEA) comparing rimegepant with eletriptan has been conducted in Thailand. The availability of such economic evidence would support policymakers in evaluating both the clinical benefits and the economic value of these treatments, which is particularly crucial in the resource-limited Thai healthcare system. A clear understanding of the cost-effectiveness of rimegepant and eletriptan is essential for designing policies that promote equitable access to effective migraine treatments. Such evidence would not only benefit patients but also contribute to reducing the national burden of migraine and improving the overall efficiency of the healthcare system in the long term.

MATERIALS AND METHODS

Study design

This study was a model-based health economic

evaluation using a decision tree and Markov model to compare the costs and humanistic outcomes, presented in quality-adjusted life years (QALYs), of different treatment strategies for acute episodic migraine in Thailand. A lifetime horizon with 30-day cycles was implemented with a societal perspective. The incremental cost-effectiveness ratio (ICER) was calculated by taking the difference in the total costs of treatment between rimegepant and eletriptan and dividing it by the difference in QALYs for the same treatments. Future costs and QALYs were discounted at an annual rate of 3%.17

This study has been reviewed and approved by the Human Research Ethics Committee of Silpakorn University (COE Number: COE 68.1010-173-9080).

Intervention

This study compared 75 mg of rimegepant (taken as needed) with 40 mg of eletriptan in patients with acute episodic migraine.18 In the eletriptan group, patients with an inadequate response to the initial 40 mg dose could take a second 40 mg tablet at least 2 hours later.4

Decision model

This study employed a hybrid decision tree and Markov model to evaluate the cost-effectiveness through Microsoft Excel 2019. The model was developed based on a previous international cost-effectiveness model for rimegepant15,16 and was adapted to the Thai context.4

The analysis was conducted in a population not using preventive migraine medication to avoid any potential confounding effects from such therapies on the outcomes of the two acute treatments under investigation. The base-case analysis focused on a hypothetical cohort of patients aged 18 and older with moderate to severe acute episodic migraine, evaluated over a lifetime time horizon. The model simulates a single migraine episode, defined as one continuous headache attack with a duration not exceeding 72 hours from onset to resolution. This model focused on a single headache attack.

Fig 1. Decision tree and Markov model for health state transitions in moderate to severe acute episodic migraine attack.

This model was validated by three specialists: a neurologist, a clinical pharmacist specializing in neurology, and a pharmacoeconomist. The target population consists of patients who previously had an inadequate response to the maximum daily dose of sumatriptan (200 mg/day).18 Patients diagnosed with moderate to severe acute episodic migraine entered the model, based on the criteria set by the International Classification of Headache Disorders (ICHD-3).11 It is assumed that patients entering the model have no comorbidities, are not using other acute or prophylactic medications, and do not receive any other treatments outside of the model’s pathways.

This cohort enters the model at the initial state, first sustained pain, and is subsequently divided into two treatment arms. The medication is to be taken as early as possible at the onset of a migraine headache, or within one hour of symptom onset. Following medication administration, the treatment outcome is assessed using the pain score. Patients were divided into two groups based on their pain scores. Those with a pain score of 0 were classified as the response group, while those with a pain score greater than 0 were placed in the no response group. The response group is defined as individuals who became completely pain-free (pain score = 0/10) within 2 hours of taking the medication. In contrast, the no response group consists of individuals who did not achieve a pain-free state within the same timeframe.4,19 This classification aligns with the standard primary endpoint outlined in ICHD-3, and the guidelines from the International Headache Society (IHS).3

Rimegepant treatment pathway: patients in the

first sustained pain state received 75 mg of rimegepant. If they did not respond, they transitioned to the second sustained pain state and were advised to proceed to the emergency room (ER), which consists of a single intravenous (IV) dose of ketorolac 30 mg.4 Rimegepant was administered as a single maximum recommended daily dose.18 Eletriptan treatment pathway: patients in the first sustained pain state are initiated on 40 mg of eletriptan. If they did not respond, they transitioned to the second sustained pain state and received an additional 40 mg of eletriptan (the maximum recommended daily dose is 80 mg).4 If there was still no response, it was considered a treatment failure, and they transitioned to the third sustained pain state, which leads to seeking ER service.4

Patients who responded to treatment at any stage entered remission, defined as complete headache relief after responding to medication or treatment, without any adverse events. They remained in this state until experiencing a recurrence of migraine, at which point they returned to the initial first sustained pain state in

the model. The model also included a death state that accounts for all-cause mortality, which was not directly linked to migraine or its treatments.19 This assumption was consistent with previous findings indicating no significant association between migraine and all-cause mortality.4,15 To inform this model, we utilized the mortality rate4,15 from the Thai population, which was 8 deaths per 1,000 people per year.20 In this model, the death state can be reached from any health state.

Model assumptions

1. The baseline number of monthly migraine days (MMD) is assumed to be 6 days per month based on studies in the Thai context.21 In the second year, it was assumed that patients taking eletriptan would experience 6 migraine episodes per month, while those taking rimegepant would have a reduced frequency of 3 migraine episodes per month. This is consistent with clinical studies showing an approximate 50% reduction in MMD within one year, provided that patients use rimegepant for every migraine attack.22 It is also assumed that eletriptan does not affect the frequency of migraines.23

2. Eletriptan adverse events include mild asthenia, somnolence, dizziness, nausea, chest symptoms, and vomiting.8

3. Rimegepant adverse events include nausea, dizziness, and urinary tract infection (UTI).8

4. Chest symptom-related adverse events include sensations of tightness, heaviness, or discomfort in the chest and arms, shortness of breath (dyspnea), palpitations, and occasionally chest pain. These are assumed to be generally mild, transient, and self-resolving without requiring treatment.24

5. Management of nausea and vomiting was one tablet of 10 mg domperidone taken up to three times daily as needed when symptoms occurred.

6. Management of UTI was one tablet of 250 mg ciprofloxacin taken twice daily for 3 days.25

7. ER service was assumed to be effective for all patients with sustained pain.

8. Patients were assumed to have full adherence to their prescribed medication.

Time horizon

As most clinical trials evaluating acute migraine treatments included patients aged 18 years and older,12 the decision tree and Markov model in this study were developed to follow the treatment of acute migraine over a lifetime, from a starting age of 18 until death. The average life expectancy in Thailand was 75.6 years.26 A cycle length of 30 days was considered appropriate to align with the standard assessment of MMD.

TABLE 1. Parameters used in the model.

TABLE 1. Parameters used in the model. (Continued)

Abbreviation: *ADR; Adverse drug reaction

Clinical effectiveness and transition probabilities

This systematic review was conducted to derive clinical outcome probabilities and transition probabilities for the cost-effectiveness model in acute episodic migraine, following a structured and predefined protocol consistent with standard methodological guidance for systematic reviews. A comprehensive literature search was performed across electronic databases, including PubMed, Scopus, and the Cochrane Library. The search aimed to identify relevant randomized controlled trials (RCTs) evaluating rimegepant and eletriptan in adults with acute migraine. Furthermore, meta-analysis and network meta-analysis based on systematic reviews with protocols similar to the aforementioned protocol was also be included if they report clinical outcome probabilities and transition probabilities relevant to the cost-effectiveness model of acute episodic migraine.

Eligible studies included RCTs with parallel, cross-over, or factorial designs, as well as systematic reviews with meta-analysis published in Thai or English that reported quantitative outcomes suitable for estimating transitions between response and non-response states. Non-RCTs, non-original articles, animal studies, studies involving ineligible populations, studies with insufficient data, and duplicate publications were excluded. The model population comprised adults (≥18 years) diagnosed with acute episodic migraine of moderate-to-severe intensity according to ICHD-3 criteria, with

a diagnosis duration of at least one year. Patients with chronic migraine, mild attacks, other primary headache disorders, contraindication or intolerance to triptans, use of migraine prophylaxis, pregnancy or breastfeeding, significant drug–drug interactions, or allergy to triptans or gepants were excluded.

Finally, the Risk of Bias 2 (RoB 2) tool was used to assess the risk of bias in the included study before its data were used in the analysis.

Costs

All costs were expressed in Thai baht as presented in Table 1. Drug costs for rimegepant and eletriptan were obtained from official procurement prices and median drug prices in Thailand. The unit cost of rimegepant 75 mg was derived from pricing data from tertiary care hospitals28,29 while the cost of eletriptan 40 mg and other drugs were obtained from national reference prices.30,31 Costs of outpatient visits and ER services were derived from 2025 standard service charge schedules of public hospitals affiliated with the Ministry of Public Health, Thailand.29 Direct non-medical costs, including food and transportation expenses incurred during healthcare utilization, were obtained from the standard cost lists for health technology assessment in Thailand.38 All costs were adjusted to 2025 values using the consumer price index (CPI) from the Bureau of Trade and Economic Indices, Ministry of Commerce, Thailand.

Utility values

QALYs were used as a measure of humanistic outcomes, which were calculated by utility weight multiplied by life years. Utility weight reflected health-related quality of life associated with different health states and adverse drug reactions in patients with acute episodic migraine. Utility values ranged from 0 (death) to 1 (perfect health). This approach allowed the model to capture changes in quality of life associated with both migraine symptoms and treatment-related outcomes during an acute migraine episode. Utility values for migraine-related health states and adverse drug reactions were obtained from previously published studies. All utility values used in the analysis are summarized in Table 1.

Sensitivity analysis

The one-way sensitivity analysis involves varying individual parameters with their highest and lowest values to observe their effect on the ICER, which was presented in a tornado diagram. The probabilistic analysis examines the combined uncertainty of multiple parameters through 1,000 random iterations using Monte Carlo simulations. The result was presented by a cost-effectiveness plane between incremental cost and incremental QALYs.

RESULTS

Cost-effectiveness analysis

The cost-effectiveness analysis of rimegepant compared with eletriptan for patients with acute episodic migraine in Thailand showed that the total costs of rimegepant and eletriptan treatments were 2,442,622.42 baht and 2,450,782.33 baht, respectively. Regarding health outcomes, the QALYs gained from rimegepant and eletriptan were

25.21 and 17.59, respectively. The rimegepant strategy is considered a dominant strategy because it yields more QALY and has lower overall treatment costs. Therefore, rimegepant can be considered the more cost-effective option for treating acute episodic migraine in the Thai context.

Sensitivity analysis

The one-way sensitivity analysis presented as a tornado diagram in Fig 2 demonstrated that the parameter with the greatest influence on the ICER was the cost of rimegepant. When the cost of rimegepant 75 mg increased, the ICER also rose, as the drug cost is the primary direct cost component. The next most influential parameters were the probabilities of non-response to eletriptan 80 mg and eletriptan 40 mg, followed by the probability of non-response to rimegepant 75 mg.

TABLE 2. Cost-effectiveness results.

Total cost

(Baht)

QALYs

(Years)

ICER of rimegepant compared to eletriptan

THB per QALY gained*

Rimegepant 2,442,622.42 25.21 -1,071.79* Dominant*

*Negative ICER due to higher effectiveness and lower costs of rimegepant compared to eletriptan

Fig 2. Tornado diagram depicts the results of a one-way sensitivity analysis

The probabilistic sensitivity analysis results are shown in Fig 3 as a cost-effectiveness plane illustrating the relationship between incremental costs and incremental QALYs of rimegepant compared with eletriptan. All parameters were randomized 1,000 times using Monte Carlo simulation. The blue dot indicated the base-case ICER. The simulation results primarily fell within the quadrants of the cost-effectiveness plane indicating incremental gains in quality-adjusted life years (QALYs) with rimegepant. Regarding incremental costs, 52.7% of the simulation results fell in the fourth quadrant, indicating lower costs and greater QALYs with rimegepant, while 47.3% fell in the first quadrant, indicating higher costs and greater QALYs. These results imply a 52.7% probability that the total cost of treatment with rimegepant would be lower than that with eletriptan and a 47.3% probability that it would be higher.

DISCUSSION

This study represents the first economic evaluation to compare the cost-effectiveness of rimegepant and eletriptan for the treatment of moderate to severe migraine. The findings demonstrate that rimegepant is a more cost-effective option than eletriptan within the context and cost structure of the Thai healthcare system.

In comparison with international studies, the cost-effectiveness outcomes of rimegepant are dependent on several factors, such as the model type, comparator, time horizon, and differing costs. Our results are not consistent with studies from China16 and the United States.12 The Chinese study, which compared rimegepant to a placebo using a decision tree model with a 2-hour time horizon,

was too short to capture long-term factors such as the impact of adverse events and the reduction in monthly migraine days. In terms of cost, it only included drug costs, omitting other relevant expenses like physician consultation fees and emergency room visits. Meanwhile, the study from the United States (ICER) utilized a semi-Markov model with a 48-hour cycle over a 2-year time horizon, comparing rimegepant against triptans. While it considered the quality-of-life impact from adverse events, it did not incorporate the benefit of rimegepant in reducing MMD.23 Its cost analysis was comprehensive, including drug costs, physician visits, and ER service visits, but it lacked the costs associated with managing adverse events.

Conversely, our findings align with a study from the United Kingdom15, which found rimegepant to be cost-effective compared to best supportive care. That study used a long-term Markov model and included comprehensive costs beyond just the drug itself, although it similarly lacked costs for managing adverse events.

This study’s distinction lies in being the first economic evaluation model for the gepants class developed to specifically reflect the costs and structure of the Thai healthcare system. Critically, it employs a hybrid model, combining a short-term (2-hour) decision tree to simulate the acute attack with a long-term Markov model to capture the impacts of recurrence and ADRs. The cost analysis is more comprehensive, including drug costs, physician visits, emergency room visits, ADR management costs, travel, and food expenses, making it more thorough than the aforementioned studies. Most importantly, it integrates the benefit of rimegepant in reducing monthly

Fig 3. Cost-effectiveness plane of treatment with rimegepant compared eletriptan for patients with acute episodic migraine in Thailand

migraine days, allowing it to reflect the true value of a drug with both acute and preventive properties. This gives the study a significant advantage for direct application in policymaking within the Thai context.

The key factor driving the superior cost-effectiveness of rimegepant over eletriptan is the gain in the QALYs, resulting from better overall quality of life outcomes. This is primarily due to a lower incidence of adverse events, such as nausea, vomiting, and somnolence8, which leads to higher utility values during treatment. Additionally, rimegepant’s ability to reduce MMD23 shortens the time patients spend in a migraine state. Although rimegepant has a significantly higher per-treatment unit cost than eletriptan28-31, when considering the entire treatment pathway, it reduces overall costs. The lower rate of adverse events decreases expenditures on managing them, while the reduction in MMDs leads to less frequent medication use and a lower risk of treatment failure, which in turn reduces subsequent costly ER service visits.4,29 Considering the reduction in MMDs as a treatment benefit is crucial, given that the frequency of migraine days directly impacts treatment costs and is a primary determinant of patients’ quality of life. However, if the reduction in MMDs is less than 34% from baseline—resulting in more than four migraine days per month—the cost-effectiveness outcomes may differ from those reported in this study.

The uncertainty analysis in this study was conducted

to validate the model’s economic findings. The PSA, which involved 1,000 Monte Carlo simulations simultaneously varying all parameters, demonstrated that rimegepant consistently remained the more cost-effective option. This indicates that the cost-effectiveness of rimegepant is not solely dependent on its price but is a result of the model’s overall cost and effectiveness structure. Nevertheless, the drug response rates for both rimegepant and eletriptan remain critical variables. If future large-scale studies were to show lower efficacy for rimegepant or a higher non-response rate than assumed in this model, it could increase the number of patients entering more costly treatment pathways and potentially alter the cost-effectiveness conclusion.

However, the application of these findings must consider certain limitations. First, the MMD reduction benefit of rimegepant was established in studies lasting approximately one year. It is possible that long-term use could lead to tolerance or diminished efficacy, which would mean our study might overestimate its benefits. Conversely, if the drug’s preventive effect remains stable over the long term, its true cost-effectiveness could be even higher than estimated here. Another key limitation is that our model directs patients with treatment failure

to the emergency room. In practice4, some patients may manage failure by using a triptan combined with an NSAID, a lower-cost alternative. The exclusion of this pathway might have overestimated the costs in the eletriptan arm. Although guidelines recommend combining a triptan with an NSAID (specifically sumatriptan with naproxen)4, a literature review found no direct efficacy data for the combination of eletriptan and naproxen. Therefore, this option was not included to avoid introducing further uncertainty into the model. Finally, the calculation of utility was simplified. In reality, a patient can experience multiple ADRs simultaneously, which would have a cumulative impact on quality of life and QALYs, potentially causing our model to over- or underestimate the true cost-effectiveness.

In conclusion, the evidence presented demonstrates that rimegepant is a more cost-effective option than eletriptan for the treatment of moderate-to-severe migraine within the Thai context. The uncertainty analyses, both one-way and probabilistic, confirm this result, with outcomes consistently clustering in the quadrant representing a net health benefit. These findings are therefore reliable for informing healthcare policy decisions in Thailand, particularly for migraine patients for whom triptans are not a suitable option. Nonetheless, the collection of long-term, real-world effectiveness data is recommended to confirm these findings.

CONCLUSIONS

Rimegepant has demonstrated cost-effectiveness in treating acute episodic migraine in Thailand, as it is a dominant strategy, generatings higher QALYs while incurring lower overall costs. Therefore, rimegepant should be a treatment option for acute episodic migraine in patients who have previously experienced an inadequate response to the maximum daily dose of sumatriptan.

Data Availability Statement

The data supporting the findings of this study are available from the corresponding author upon reasonable request.

ACKNOWLEDGEMENTS

The authors would like to express appreciation to Assistant Professor Dr. Surath Tanprawet from the Faculty of Medicine, Chiang Mai University; Assistant Professor Dr. Chutatip Suphanklang and Associate Professor Dr. Nattiya Kapol from the Faculty of Pharmacy, Silpakorn University, for their generous support and valuable contributions in reviewing and validating the accuracy of the research model.

DECLARATIONS

Grants and Funding Information

This research did not receive any specific grants from funding agencies in the public, commercial, or not-for-profit sectors.

Conflict of Interest

The authors declare no conflict of interest.

Registration Number of Clinical Trial

Not applicable.

Author Contributions

Conceptualization and Design study protocol, J.L., W.T.; Methodology, J.L., W.T.; Collecting information and Data interpretation. W.T., J.P., J.K., K.K., N.P.; Analysis and results interpretation, J.L.; Visualization and writing, J.L., W.T., J.P., J.K., K.K., N.P.; Original draft, J.L., W.T.; Writing – review and editing, J.L., W.T. All authors have read and agreed to the final version of the manuscript.

Use of Artificial Intelligence

Artificial intelligence tools were used for language editing and grammar refinement. All scientific content, interpretation, and conclusions were developed and verified by the authors.

Institutional Review Board Statement

This study has been reviewed and approved by the Human Research Ethics Committee of Silpakorn University (COE Number: COE 68.1010-173-9080).

REFERENCES

1. Ge R, Chang J. Disease burden of migraine and tension-type headache in non-high-income East and Southeast Asia from 1990 to 2019. J Headache Pain. 2023;24(1):32.

2. Zhao Y, Yi Y, Zhou H, Pang Q, Wang J. The burden of migraine and tension-type headache in Asia from 1990 to 2021. J Headache Pain. 2025;26(1):49.

3. Acute Migraine Evidence Report, January 10, 2020 [Internet]. Boston: Institute for Clinical and Economic Review; 2020 [cited 2025 Feb 18]. Available from: https://icer.org/wp-content/ uploads/2020/10/ICER_Acute-Migraine_Evidence_Report_011020_ updated_011320_-2.pdf.

4. Clinical Practice Guideline for the Management of Migraine [Internet]. Bangkok: Neurological Institute of Thailand; 2022 [cited 2024 Jan 25]. Available from: https://thaigpfm.org/news/ cpg-migraine/.

5. Lohavisavapanich R. New Drugs for Migraine Treatment. Thai J Hosp Pharm. 2024;34(1):52–68.

6. Ferrari MD, Roon KI, Lipton RB, Goadsby PJ. Oral triptans (serotonin 5-HT1B/1D agonists) in acute migraine treatment: A meta-analysis of 53 trials. Lancet. 2001;358(9294):1668–75.

7. Goadsby PJ, Lipton RB, Ferrari MD. Migraine — Current Understanding and Treatment. N Engl J Med 2002;346(4):257–70.

8. Karlsson WK, Ostinelli EG, Zhuang ZA, Kokoti L, Christensen RH, Al-Khazali HM, et al. Comparative effects of drug interventions for the acute management of migraine episodes in adults: Systematic review and network meta-analysis. BMJ. 2024; 368:e080107.

9. Goadsby PJ, Ferrari, Olesen J, Stovner LJ, Senard JM, Jackson NC, et al. Eletriptan in acute migraine: A double-blind, placebo-controlled comparison to sumatriptan. Neurology. 2000;54(1): 156–63.

10. Tepper SJ, Rapoport AM, Sheftell FD. Mechanisms of Action of the 5-HT1B/1D Receptor Agonists. Arch Neurol. 2002;59(7):1084–8.

11. Olesen J. Headache Classification Committee of the International Headache Society (IHS) The International Classification of Headache Disorders, 3rd edition. Cephalalgia. 2018;38:1–211.

12. Steven J. Atlas DRT, Foluso Agboola, Todd A. Lee, Rick Chapman, Steven D. Pearson, David M. Rind. Acute Migraine for Treatment (Evidence Report), Febuary 25, 2020 [Internet]. Boston: Institute for Clinical and Economic Review; 2020 [cited 2025 Jan 1]. Available from: http://icer-review.org/ material/acute-migraine-evidence-report/.

13. Croop R, Goadsby PJ, Stock DA, Conway CM, Forshaw M, Stock EG, et al. Efficacy, safety, and tolerability of rimegepant orally disintegrating tablet for the acute treatment of migraine: a randomised, phase 3, double-blind, placebo-controlled trial. Lancet. 2019;394(10200):737–45.

14. Henson B, Hollingsworth H, Nevois E, Herndon C. Calcitonin Gene-Related Peptide (CGRP) Antagonists and Their Use in Migraines. J Pain Palliat Care Pharmacother. 2020;34(1):22–31.

15. Johnston K, Powell LC, Popoff E, L’Italien GJ, Pawinski R, Ahern A, et al. Cost-effectiveness of rimegepant oral lyophilisate compared to best supportive care for the acute treatment of migraine in the UK. J Med Econ. 2024;27(1):627–43.

16. Tian S, Yang Y, Tan S, Luo J, Yang C, Liu Q, et al. Cost-effectiveness analysis of rimegepant for on-demand acute treatment of migraine in China. Front Neurol. 2024;15:1411576.

17. Health technology assessment manual for Thailand (2021 updated edition) [Internet]. Nonthaburi: Health Intervention and Technology Assessment Program 2021 [cited 2025 Aug 11]. Available from: https://www.hitap.net/th/document/health-technology-assessment-manual-thailand-2021/.

18. Ailani J, Burch RC, Robbins MS. The American Headache Society Consensus Statement: Update on integrating new migraine treatments into clinical practice. Headache. 2021;61(7):1021–39.

19. Outcome Measures [Internet]. The British Pain Society & Faculty of Pain Medicine; 2019 [cited 2025 May 22]. Available from: https://www.britishpainsociety.org/static/uploads/ resources/files/Outcome_Measures_January_2019.pdf.

20. Demographic: General demographic information [Internet]. Bangkok: Thailand Board of Investment; 2024 [cited 2026 Feb 4]. Available from: https://www.boi.go.th/index.php?page=demographic

21. Asawavichienjinda T. Treatment gaps in migraine preventive treatment-eligible patients: A hospital-based study in Thailand. Chula Med J. 2025;69(1):51–8.

22. Zhang M, Guo A, Wu J, Wang H, Zhang Y, Dong H, et al. Rimegepant for the acute treatment of migraine: A phase 3, multicenter, open-label, long-term safety and effectiveness study in adults from China. Cephalalgia. 2025;45(10):3331024251371686.

23. Johnston K, Harris L, Powell L, Popoff E, Coric V, L’Italien

    
    

    G, et al. Monthly migraine days, tablet utilization, and quality of life associated with Rimegepant – post hoc results from an open label safety study (BHV3000–201). J Headache Pain. 2022;23(1):10.

24. Dodick DW. Triptans and chest symptoms: The role of pulmonary vasoconstriction. Cephalalgia. 2004;24:298–304.

25. Clinical Practice Guidelines for the Treatment of Urinary Tract Infections 2025 Bangkok: The Infectious Disease Association of Thailand; 2025 [cited 2025 May 22]. Available from: https:// www.idthai.org/Contents/Views/?d=%2117%219%21%21%211045.

26. Life expectancy and health-adjusted life expectancy of the Thai population [Internet]. International Health Policy Program, Burden of Disease Thailand unit; 2019 [cited 2024 May 22]. Available from: https://le-hale.bodthai.net/.

27. Ornello R, Caponnetto V, Ahmed F, Al-Khazali HM, Ambrosini A, Ashina S, et al. Evidence-based guidelines for the pharmacological treatment of migraine. Cephalalgia. 2025;45(4): 1-326.

28. Summary report of drug utilization estimation: Rimegepant 75 mg orally disintegrating tablet [Internet]. Pathum Thani: Thammasat University Hospital; 2022 [cited 2026 Feb 4]. Available from: https://www.hospital.tu.ac.th/tuhpharmacy/ upload/news/files/2025082719292968.pdf.

29. Manual of healthcare service charge rates for Thai nationals 2025 [Internet]. Nonthaburi: Bureau of Health Administration, Ministry of Public Health; 2025 [cited 2026 Feb 4]. Available from: https://phdb.moph.go.th/main/upload/web_news_files/ jqsd49irwcg0ogswgk.pdf.

30. Median price of drugs [Internet]. Nonthaburi: Drug and Medical Supply Information Center, Ministry of Public Health; 2019 [cited 2026 Feb 4]. Available from: https://dmsic.moph. go.th/index/drugsearch/3.

31. Notification of the Ministry of Public Health Re: Healthcare service rates of service units under the Ministry of Public Health 2025, Special Section 45 Ng (2025).

32. Oliveira Gonçalves AS, Panteli D, Neeb L, Kurth T, Aigner A. HIT-6 and EQ-5D-5L in patients with migraine: assessment of common latent constructs and development of a mapping algorithm. Eur J Health Econ. 2022;23(1):47–57.

33. Xu R, Insinga RP, Golden W, Hu XH. EuroQol (EQ-5D) health utility scores for patients with migraine. Qual Life Res. 2011; 20(4):601–8.

34. Ito K, Mitobe Y, Inoue R, Momoeda M. Impact of nausea/ vomiting on EQ-5D-5L utility scores in patients taking iron preparations for heavy menstrual bleeding or anemia. BMC Womens Health. 2023;23(1):505.

35. Matza LS, Deger KA, Vo P, Maniyar F, Goadsby PJ. Health state utilities associated with attributes of migraine preventive treatments based on patient and general population preferences. Qual Life Res. 2019;28(9):2359–72.

36. Betts MB, Rane P, Bergrath E, Chitnis M, Bhutani MK, Gulea C, et al. Utility value estimates in cardiovascular disease and the effect of changing elicitation methods: A systematic literature review. Health Qual Life Outcomes. 2020;18(1):251.

37. Millier A, Amri I, Boyer L, Auquier P, Toumi M. Utility decrements associated with side effects in schizophrenia. J Med Econ. 2014;17(12):853–61.

38. Standard cost list development for health technology assessment [Internet]. Nonthaburi: Health Intervention and Technology Assessment Program, Ministry of Public Health 2010 [cited 2026 Feb 4]. Available from: https://costingmenu.hitap.net/.

‌Impact of Air Pollution on Acne Vulgaris: A Systematic Review

Pawat Udompataikul, M.D., Natthasith Chatpimolkul, M.D., Nattapat Chandraruangnabha, M.D., Kan Tangkijngamwong, M.D., Phongsaphak Santimagorn, M.D., Nanticha Faromkao, M.D.*

Department of Dermatology, Faculty of Medicine, Srinakharinwirot University, Bangkok 10110, Thailand.

*Corresponding author: Nanticha Faromkao E-mail: pangrum109@gmail.com

Received 15 May 2026 Revised 14 July 2026 Accepted 15 July 2026 ORCID ID: http://orcid.org/0000-0003-1343-5612 https://doi.org/10.33192/smj.v78i9.282324

All material is licensed under terms of the Creative Commons Attribution 4.0 International (CC-BY-NC-ND 4.0) license unless otherwise stated.

ABSTRACT

Objective: This study aimed to systematically review the existing evidence of the impact of air pollutants on acne vulgaris.

Materials and Methods: The studies assessed the relationship between air pollution and acne, in terms of the number of acne lesions, severity of acne, number of patients who visited the hospital with acne complaints, and impact on quality of life, which were the eligibility criteria for the outcome of this study. The studies that were not in the English language, case reports, conference abstracts, and studies with duplicate publication data were excluded. Risk of Bias in Non-randomized Studies (ROBINS-E) was used to assess the risk of bias. Descriptive analysis was conducted because of different outcomes reported among the studies.

Results: 226 articles were identified through the literature search. Four studies were included. El Haddad, et al,

reported that NO2 exposure aggravated acne vulgaris. Xiang Li, et al, presented that exposure to Particulate matter10 (PM10), sulfur dioxide (SO2), and nitrogen dioxide (NO2) in Xi’an, China, as well as PM2.5 in Chongqing, China showed inconclusive results regarding increases in outpatient visits of acne vulgaris. Wei Liu, et al demonstrated that exposure to SO2 decreases the outpatient visits of acne vulgaris.

Conclusion: Studies suggested a possible association between NO2 and acne vulgaris. PM2.5, PM10 and SO2 presented

controversial outcomes. Since current studies had a risk of bias and methodological limitations, a prospective cohort or panel studies using personal or high-resolution exposure estimates, standardized measures of acne incidence and severity, prespecified lag periods, and adjustment of confounders were suggested for a better understanding of the correlation.

Keywords: Acne; air pollution; NO2, PM2.5, PM10, SO2 (Siriraj Med J 2026;78(9):671-682)

INTRODUCTION

Air pollution is considered the leading environmental risk factor causing several diseases and premature deaths worldwide. In 2019, approximately 99% of the global population was living in air conditions worse than the World Health Organization (WHO) safety standard. Moreover, the trend of air pollution has been increasing every year.1,2 Air pollutants are produced by various sources. Particulate matter (PM) originates from various sources, such as forest fires, cooking, biomass burning, construction, diesel engines, road dust, and industrial processes. The sources of nitrogen dioxide (NO2) are wood burning, vehicle emissions, and waste burning. Finally, sulfur dioxide (SO2) comes from vehicle emissions, fuel incineration, crude oil, and coal.3

Air pollutants can cause and worsen diseases, especially in the respiratory and cardiovascular systems. Other than systemic diseases, recent studies have suggested that various air pollutants such as PM, NO2, SO2, carbon monoxide (CO), and ozone (O3) can aggravate inflammatory dermatoses.4,5 For instance, exposure to PM2.5 and PM10 increased the number of outpatients with eczema in Guangzhou, China.6 The pathogenic mechanism of air pollution triggering inflammatory dermatosis has been proposed. Permeation of different types of air pollution

through the skin causes an induction of inflammatory cascade via Aryl hydrocarbon receptors.5,7 Reactive oxygen species are also generated, causing depletion of enzymatic and nonenzymatic antioxidant capacity and an increase in proinflammatory cytokines, which consequently lead to cutaneous inflammation.5,8

Acne vulgaris is a common chronic inflammatory cutaneous disease of the pilosebaceous unit that primarily presents as papules, pustules, comedones, or nodules on the face, upper arms, chest, trunk, and back.9 The global prevalence (for all ages) is approximately 9.38%, with adolescents and young adults being the most affected.10 In addition, adverse psychological conditions such as depression and anxiety are commonly found in people with acne.11 Jirawattanadon, et al demonstrated that adult female acne can have a moderate effect on the quality of life, particularly in patients with moderate to severe acne.12

There are several contributing factors for acne, for instance cosmetic products, food, and air pollution. Certain cosmetic products, so-called acnegenic and comedogenic agents, can cause acne vulgaris, particularly oil-based products.9 Shaving was also associated with the severity of adult male acne.13 Regarding food, dairy products and high glycemic-load diet were demonstrated

to exacerbate acne.14,15 Air pollution is also a significant issue for the Acne Association. Several observational studies have explored the effects of air pollution on acne development.16-19 Nevertheless, the results have remained controversial. For this reason, there was a need to conduct a systematic review of the relationship between exposure to air pollutants and the occurrence of acne vulgaris.

MATERIALS AND METHODS

This systematic review was conducted using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) as the reporting standard.20

1. Information Sources and Search Strategy

   The databases searched were Medline (Pubmed), Cochrane Library, Scopus, and Embase, which was conducted on March 31, 2024. Articles published after the search date were not included. The details of keywords and search strategies are outlined in Table S1

2. Eligibility Criteria

   This study included studies which explored the relationship between air pollution and acne. To elaborate, the population included were individuals with acne vulgaris and the exposure were air pollutants. Moreover, the outcomes had to include at least one of the following: number of acne lesions, severity of acne, number of patients who visited the hospital with acne complaints, or impact on quality of life. The setting of the studies was not restricted by geographic location or healthcare setting and the study designs included were observational studies. This study excluded articles that were not in English since the review team did not have sufficient resources for reliable translation and assessment of non-English publications. Case reports and conference abstracts were excluded because of the insufficient methodological and outcome information. Lastly, studies with duplicate publication data were excluded to avoid overrepresenting findings from the same underlying study. For retrospective time series studies, the results were reported using the term “lag days” which means the number of days between the exposure of air pollution and the acne outcome.

3. Data Collection and Selection process

   Two researchers (NSC., NPC.) independently screened all records identified from the databases in accordance with the keywords. All retrieved records were imported into EndNote 21, and duplicate records were removed before screening. Titles and abstracts of the remaining records were independently screened by the same two reviewers. Inter-rater agreement during title and abstract screening was assessed using Cohen’s kappa statistic. Any disagreements during the screening

   process were resolved through discussion between the two reviewers. If consensus could not be reached, two additional researchers (K.T. and P.S.) were consulted for adjudication. The full texts of eligible reports were subsequently retrieved and independently assessed for eligibility by the same two reviewers. The extracted variables included study characteristics (author, publication year, country, study design, and study period), population characteristics (sample size and participant characteristics), exposure characteristics (air pollutants assessed, exposure assessment method, and pollutant concentrations), outcome definitions, statistical methods, and main findings. No authors of the included studies were contacted because the relevant information was available in the published reports. Missing information was recorded as not reported, and no assumptions were made to replace missing data. Numerical data extracted from the reports were checked against the original publications to minimize transcription errors.

4. Quality Assessment

Risk Of Bias In Non-randomized Studies (ROBINS-E) tool by Higgins et al, published in 2024, was used to assess the quality of the included observational studies. Four researchers (NSC., NPC., K.T., and P.S.) collaboratively assessed the risk of bias of the included studies. Biases included those arising from confounding, measurement of exposure, selection of participants into the study, post-exposure interventions, missing data, measurement of outcome, and selection of the reported result. The final judgement was determined by consensus. These were evaluated and reported as low risk of bias, some concerns, high risk of bias, or very high risk of bias.21

Statistical analyses

This study used descriptive methods in reporting study data, since there were variances in study characteristics, participant’s characteristics, assessment methods, and outcomes.

RESULTS

1. Study selection and characteristics

   226 research records were selected by two researchers from four databases; 27 studies from Medline (Pubmed), one study from Cochrane Library, 63 studies from Scopus, and 135 studies from Embase. Subsequently, 47 duplicated articles were excluded, leaving a total of 179 records. Then, 179 records underwent title and abstract screening independently by two researchers (NSC., NPC.). The inter-rater agreement was substantial (κ=0.78). Discrepancies between the two reviewers were discussed. When agreement could not be achieved, two

   
   

   additional researchers (K.T. and P.S.) adjudicated the disagreements. As a result, 175 studies were excluded. The reasons were as follows: 165 studies were unrelated to air pollution, eight studies were review articles, and two studies were conference abstract. Eventually, four articles were sought for retrieval and assessed for eligibility though full text screening, and four studies were included, as summarized in Fig 1. All of these were observational studies, with one being a cross-sectional study, and the others being retrospective time series studies. The cross-sectional study reported the presence or absence of acne according to the dermatologist’s diagnosis as an outcome. On the other hand, the three retrospective time series studies presented the number of patients who visited the outpatient clinic with acne complaints as an outcome. Finally, the air pollutants that were taken into

   account in these studies included PM10, PM2.5, NO2, and SO2.

   
   

2. Summary of quality assessments

   The assessment for risk of bias of these four studies was assessed independently by four researchers according to the ROBINS-E tool. If there were any disagreements, discussions took place to reach a consensus. The Claudia El Haddad, et al, study conducted in Lebanon, was reported as some concerns. In contrast, the remaining studies were assessed as overall high risk of bias, due to the high risk of bias in their first domain (bias of confounding) without any very high risk of bias in other domains. The ROBINS-E traffic light plot was presented in Fig 2. The detailed evaluations of the included studies were provided in Table 1.

   
   

   
   

   Fig 1. Flowchart of identification, screening, eligibility, and exclusion of retrieved studies

   
   

   
   

   Fig 2. Risk of bias assessment using ROBINS-E tool

   
   

3. Summary of the findings (Table 2)

There were four studies included in this article. First, the study in Lebanon analyzed various variables, including air pollution in association with absence or presence of acne16. Regarding the bivariate analysis, exposure to SO2,

O3, PM10, and PM2.5 presented insignificant results.

Whereas in logistic regression using the presence or absence of acne as a dependent variable, severe NO2 exposure compared to absence was significantly associated with higher odds of having acne (adjusted odds ratio = 8.24, p<0.05).

In Xiang Li, et al.’s study, 27,190 outpatient visits for post-adolescent acne were evaluated in Xi’an, China.17 It was reported as a percentage change of daily visits with an increase in pollutant concentrations per 10 μg/m3 at different lag days. One- and two-pollutant models of PM10, SO2, and NO2 were evaluated. The one-pollutant model included both genders and a separate analysis for males and females. Regarding PM10, it showed significant results of the increment in visits in all lag days except lag 4, 5, and 7 days, while for the female models, it showed a significant increase in visits in all lag days, except for lag 7 days. In contrast, the male models demonstrated insignificant results across all lag days. Concerning SO2, it presented significant incremental results in visits in lag 3, 0-4, 0-5, 0-6, and 0-7 days, while in both male and female models, the results were significant in lag 1 day and lag 0-7 days, respectively. NO2 demonstrated significant results in all lag days, except lag days 4, 5, 6, and 7 in both genders and female models. Finally, all lag

days, except lag 2, 4, 5, 6, and 7 days, showed significant results in male models. Regarding the two-pollutant models, the percentage change in visits was demonstrated based on an increase in pollutant concentrations per 10 μg/m³ in lag 0-7 days. PM10, when evaluated with SO2 and NO2 showed a significant increase in visits by 0.86% (95% CI, 0.51–1.21%) and 0.76% (95% CI, 0.38–1.14%),

respectively. On the other hand, insignificant results were presented when SO2 was evaluated with PM10 and NO2. Finally, NO₂ remained significantly associated after adjustment for SO₂ (3.75%, 95% CI, 1.25–6.25%) but

not after adjustment for PM₁₀ (0.90%, 95% CI, −1.38

to 3.19%).

Xiang Li, et al conducted a study in Chongqing, China, evaluating 120,842 acne vulgaris outpatient visits in Chongqing in relation with PM2.5. Percentage increases in the number of visits in association with increases in PM2.5 concentration per 10 μg/m3 at different lag days were presented.18 One- and two-pollutant models were evaluated between different genders (male and female) and age groups (<25 years old and ≥25 years old). Concerning the one-pollutant model, for all gender models and female models, PM2.5 was positively correlated with an increase in visits in all lag periods, except lag day 0. In male models, it showed significant results of the negative association of PM2.5 on outpatient visits in lag 4, 5, 6, 7, 0-4, 0-5, 0-6, and 0-7 days. Regarding the age-specific analysis, in patients aged less than 25 years old, positive correlations were noted in lag 0, 4, 5, 6, 0-1,

0-3, 0-4, 0-5, 0-6, and 0-7 days. Meanwhile, in patients

TABLE 1. Characteristics of the included studies.

Original Article SMJ

TABLE 2. Summary of the findings.

TABLE 2. Summary of the findings. (Continued)

* - p<0.05, **-p<0.01, ***-p<0.001

aged 25 years or older, significant results were present in all lag periods, except lag 0, 1, 0-1, and 0-2 days. The two-pollutant models were evaluated in lag 0-7 days. To elaborate, when analyzing PM2.5 with SO2, NO2, CO, and O3, it was found that an increase in PM2.5 significantly correlated with increases in visits in models that analyzed both sexes, females, and patients aged 25 years or older. Conversely, in male models, PM2.5, when evaluated with NO2 showed insignificant results. In patients less than 25 years old, evaluating PM2.5 with NO2 and O3 illustrated insignificant results as well.

The study of Wei Liu, et al, in Beijing, 59,530 outpatient visits were analyzed. PM2.5, PM10, SO2, and NO2 were included in one- and two-pollutant models19.

Regarding the one-pollutant models of PM2.5, lag 0-3

days and 0-4 days showed significant associations of PM2.5 (RR 0.98 95% CI 0.963–0.997) and RR 0.981 (95%

CI 0.963–0.999), respectively) in decreasing outpatient visits of acne vulgaris, while the results in other lag days demonstrated insignificant results. For PM10, it showed a significant association with an RR of 0.975 (95% CI 0.954–0.996) in lag 0-7 days, as well. Meanwhile, for SO2, all lag days, which were shown in the study, showed significant results of SO2 reducing outpatient visits for acne vulgaris. In contrast, it was found that NO2 demonstrated to be associated with increase in outpatient visits in Lag 0 days and 0-2 days with RRs of

1.016 (95% CI 1.001-1.031) and 1.020 (95% CI 1.001-

1.038), respectively. Concerning the two-pollutant model of PM2.5 and PM10, when evaluated with SO2, it showed a significant result in increasing outpatient visits with RRs of 1.016 (95% CI 1.00-1.031) and 1.023 (95% CI

1.009-1.039), respectively. On the other hand, both gases, when analyzed with NO2 showed insignificant results. For the two-pollutant model of NO2, significant results were shown in terms of increasing outpatient visits when analyzed with PM2.5, PM10, and SO2. Finally, SO2 showed significant associations in reducing outpatient visits when examined with other pollutants, with statistical significance (p<0.05).

1. PM2.5

   There were three studies that illustrated the correlation between PM2.5 and acne vulgaris. The results were inconsistent. To elaborate, for the study conducted in Chongqing, the one and two-pollutant models reported both significant and insignificant results in increasing the outpaint visits of acne vulgaris. In addition, the study evaluated in Lebanon demonstrated insignificant results. Conversely, there were two lag periods that reported significant associations of the reduction in outpatient

   visits in the one-pollutant model of the Beijing study. However, in the two-pollutant model, it demonstrated both significant and insignificant results of the negative associations.

2. PM10

   Concerning PM10 and acne vulgaris, there were three studies that reported the correlation. The results were controversial. The Lebanese study reported insignificant results. For the Xi’an study, the one-pollutant model showed both significant and insignificant positive correlations. Moreover, in the two-pollutant model, all of the results were significant in indicating an adverse impact on acne. In contrast, in the study conducted in Beijing, the one-pollutant model reported a significant association of the reduction in outpatient visits impact in lag 0-7 days. Lastly, it demonstrated both significant and insignificant results of the negative associations in the two-pollutant model.

3. NO2

   Three studies reported the relation between NO2 and acne vulgaris. To elucidate, the study conducted in Lebanon showed that severe NO2 exposure was significantly related with the presence of acne (adjusted odds ratio (aOR) 8.24 (95% CI 1.36–49.90)). Moreover, for the Xi’an study, the one-pollutant model demonstrated significant results in increase in outpatient visits of acne vulgaris in two lag days. Furthermore, the two-pollutant model reported significant results when analyzed with other pollutants. Finally, for the study conducted in Beijing, the one-pollutant model reported both significant and insignificant results in increasing acne vulgaris outpatient visits. However, in the two-pollutant model, significant results were found when analyzed with SO2 (RR 1.059, 95% CI 1.038–1.080), PM₁₀ (RR 1.024, 95% CI 1.002–1.046),

   and PM₂.₅ (RR 1.025, 95% CI 1.006–1.044).

   
   

4. SO2

Regarding SO2, there were three studies which mentioned the correlation between SO2 and acne vulgaris. Insignificant results were found in the Lebanese study. Furthermore, for the Xi’an study, the one-pollutant model showed both significant and insignificant results in the negative association of SO2 on acne vulgaris. In addition, insignificant results were shown in the two-pollutant model. In contrast, for the study conducted in Beijing, both one and two-pollutant models demonstrated significant results, which indicated that exposure to SO2 reduced outpatient visits of acne vulgaris.

DISCUSSION

There are four mechanisms in the pathogenesis of acne vulgaris, inflammation, follicular hyperkeratinization, bacterial dysbiosis with Cutibacterium acnes predominant, and increased sebum production.9 There have been several studies that have proposed the mechanisms of the PMs’ effects on acne vulgaris. First, they directly disrupt skin barrier function. Moreover, oxidative stress caused by PMs disrupts lipids, proteins, DNA, and cellular functions. They also activate inflammatory responses through pathways such as AhR, NF-κB, and MAPK. Not only do PMs stimulate follicular inflammation, but they also increase sebum production.3,5,22,23 NO2 affects pilosebaceous units with two mechanisms. It has a direct deleterious effect by generating free radicals, causing oxidative stress, thus epithelial barrier functions are disrupted.3,5,24 In addition, it was found that NO2 causes bacterial dysbiosis, which alters skin homeostasis.25 For SO2, the proposed mechanisms were the induction of oxidative stress and inflammation via the activation of the NF-κB pathway to inhibit the ERK1/2 and p38 pathways.3,26 Therefore, all four air pollutants could aggravate acne vulgaris.

Across the four included studies, the associations between air pollutants and acne-related outcomes were heterogeneous. Regarding PM2.5, studies have reported inconsistent results. The studies assessed in Chongqing showed some significant results in the positive correlation in both one and two-pollutant models. However, for the Beijing study, significant associations of the reduction in outpatient visits were found in two lag days. For PM10, the results were controversial. The Xi’an study demonstrated some significant results in both one and two-pollutant models that the outpatient visits of acne vulgaris was increased by its exposure to PM10, while the study conducted in Beijing showed that one of the results in the one-pollutant model indicated associations of the reduction in outpatient visits. These heterogeneous findings may reflect differences in study design, exposure assessment, confounding factors and lag structures.

For NO2, the significant possible associations with

acne vulgaris have been shown in three studies.16,18,17 The average NO2 concentrations exceeded the concentration limit according to 2024 WHO guidelines2, which was more than 40 μg/m3 (annual mean) in all three studies. The two-pollutant model of the Chongqing study showed insignificant results for patients aged less than 25 years old when NO2 was assessed with PM2.5. Furthermore, the Beijing study reported insignificant results in the two-pollutant model when NO2 was analyzed with PM2.5 and PM10 The plausible explanation is correlations

between air pollutants, changes in standard errors after adjustment for other pollutants, exposure measurement error, or model instability.

Concerning SO2, the Xi’an study demonstrated both significant and insignificant results of SO2 in association with acne vulgaris. The Xi’an study reported heterogeneous outcome between SO₂ exposure and acne outpatient visits across different lag periods, whereas the Beijing study reported significant associations of SO₂ exposure in reducing outpatient visits. However, this correlation should not be interpreted as evidence of a protective effect of SO₂ against acne. The observed heterogeneity between the Xi’an and Beijing studies may be confounded by differences in study design, exposure patterns, population characteristics, and model instability. Therefore, the reason for the differing associations remains uncertain. Finally, the associations between acne vulgaris and exposure to PM2.5, PM10, and SO2 in the Lebanese study showed insignificant results. One possible explanation was that the Lebanese study had a lower sample size (372 participants) than Beijing, Chongqing, and Xi’an studies (120,842, 59,325, and 27,190 participants, respectively). However, differences in study design, exposure classification, and outcome ascertainment also limit direct comparison with the other studies.

The strengths of this study are as follows. With the

current increase in air pollution, the conduct and results of this study are useful. To explain, this study provides valuable information for researchers, physicians, and individuals about the impact of pollution on acne vulgaris. Also, this article raises public awareness of its significance. There were some limitations to this systematic review. To begin with, there were four studies included from the initial 226 studies found from searching. Therefore, the existing evidence is insufficient to establish definite causal relationships. Second, the four included studies have differences in exposure assessment. To elaborate, the three retrospective time-series studies used fixed monitoring stations as aggregate exposure measures. As a result, it could not represent the individual exposure levels. In the Lebanese cross-sectional study, although acne was assessed at the individual level, exposure was classified using area-level pollution data, which resulted in individual-level exposure misclassification. Moreover, this study reported the association of acne vulgaris with four air pollutants: PM2.5, PM10, NO2, and SO2. The evidence for other air pollutants, for example, O3 and CO was limited. To elaborate, only the Lebanese study reported the association between O3 and acne vulgaris. In addition, the number of studies that met the eligibility criteria was limited, with differences in outcome reports.

As a result, it was not feasible to conduct meta-analyses. Lastly, bias due to confounding factors was found in the three observational studies. Several confounders that were not adjusted, for instance, cosmetic use, diet, and stress, could alter the results. Moreover, the outcome of those studies, suspected of confounding bias, was an increase in outpatient visits, which might reflect health-seeking behavior rather than true disease burden. Hence, they did not represent direct measurement of the disease severity. Concerning the Lebanese study, the outcome of that study was obtained by answering questionnaires. Therefore, the results could be unreliable and have recall and reporting biases. This study suggests that future research on the relationships between air pollutants and acne vulgaris should be conducted as a prospective time series study with adjusted variables.

CONCLUSIONS

From the studies included, NO2 has a possible association on acne vulgaris. However, PM2.5, PM10 and SO2 showed controversial results. Several articles have proposed the effect of air pollutants on acne exacerbation. Nevertheless, the current evidence should be interpreted with caution due to the limited number of studies, heterogeneity in study designs and outcomes, and the high risk of bias. A prospective cohort or panel studies using personal or high-resolution exposure estimates, standardized measures of acne incidence and severity, prespecified lag periods, and adjustment of confounders are suggested for more reliable estimation of the relationship.

Data Availability Statement

The flowchart representing identification, screening, eligibility, and exclusion of retrieved studies were presented in Fig 1. The detailed information of the search strategies was demonstrated in Table S1. The risk of bias assessment using ROBINS-E tool was showed in Fig 2.

ACKNOWLEDGEMENT

We would like to thank all the researchers for their contributions to this systematic review.

DECLARATIONS

Grants and Funding Information

We declare that there was no funding for the development of this article.

Conflict of Interest

We declare no conflict of interest.

Registration Number of Clinical Trial

Not applicable.

Author Contributions

Conceptualization, P.U., NSC., NPC., K.T., P.S., and N.F.; Methodology, P.U. and N.F.; Data curation, P.U., NSC., NPC., K.T., and P.S.; Formal analysis, P.U.; Project administration, N.F.; Software, N.F.; Supervision, N.F.; Validation, N.F.; Visualization, N.F.; Writing – original draft, P.U.; Writing – review and editing, P.U. and N.F. All authors have read and agreed to the final version of the manuscript.

Use of Artificial Intelligence

We declare that there was no use of Generative AI and AI-assisted technologies in this article.

Institutional Review Board Statement

This systematic review analyzed data from previously published studies and did not recruit human participants. Therefore, institutional review board approval was not applicable.

REFERENCES

1. Fuller R, Landrigan PJ, Balakrishnan K, Bathan G, Bose-O’Reilly S, Brauer M, et al. Pollution and health: a progress update. Lancet Planet Health. 2022;6(6):e535-e547.

2. World Health Organization. Ambient (outdoor) air pollution [Internet]. Geneva: WHO;2024 Oct 24. Available from: https:// www.who.int/news-room/fact-sheets/detail/ambient-(outdoor)-air-quality-and-health

3. Denisow-Pietrzyk M. Human Skin Reflects Air Pollution – a Review of the Mechanisms and Clinical Manifestations of Environment-Derived Skin Pathologies. Pol J Environ Stud. 2021;30(4):3433-44.

4. Abolhasani R, Araghi F, Tabary M, Aryannejad A, Mashinchi B, Robati RM. The impact of air pollution on skin and related disorders: A comprehensive review. Dermatol Ther. 2021;34(2): e14840.

5. Han HS, Seok J, Park KY. Air Pollution and Skin Diseases. Ann Dermatol. 2025;37:53-67.

6. Yang Y, Zhang J, Fu L, Jing D, Sun B, Chen Y, et al. Short-term exposure of PM2.5 and PM10 increases the number of outpatients with eczema in Guangzhou: A time-series study. Front Public Health. 2023;10:930545.

7. Afaq F, Zaid MA, Pelle E, Khan N, Syed DN, Matsui MS, et al. Aryl hydrocarbon receptor is an ozone sensor in human skin. J Invest Dermatol. 2009;129(10):2396–403.

8. Valacchi G, Sticozzi C, Pecorelli A, Cervellati F, Cervellati C, Maioli E. Cutaneous responses to environmental stressors. Ann N Y Acad Sci. 2012; 1271:75–81.

9. Sutaria AH, Masood S, Saleh HM, Schlessinger J. Acne Vulgaris. 2023 Aug 17. StatPearls [Internet]. Treasure Island (FL): StatPearls Publishing; 2024 Jan.

10. Heng AHS, Chew FT. Systematic review of the epidemiology

    
    

    of acne vulgaris. Sci Rep. 2020;10(1):5754.

11. Yazici K, Baz K, Yazici AE, Köktürk A, Tot S, Demirseren D, et al. Disease-specific quality of life is associated with anxiety and depression in patients with acne. J Eur Acad Dermatol Venereol. 2004;18(4):435-9.

12. Jirawattanadon P, Chularojanamontri L, Nanchaipruek Y, Chantrapanichkul P, Tuchinda P, Kulthanan K. Associated Factors of Quality of Life in Adult Female Acne Coexisting with Hyperandrogenism and Polycystic Ovarian Syndrome using the Dermatology Life Quality Index. Siriraj Med. J. 2023;75(10): 744-51.

13. Likitwattananurak C, Chularojanamontri L, Tuchinda P, Kulthanan K. Clinical features of adult male acne in a tropical country: a prospective cross-sectional study. Siriraj Med J. 2023; 75(2):85-91.

14. Adebamowo CA, Spiegelman D, Danby FW, Frazier AL, Willett WC, Holmes MD. High school dietary dairy intake and teenage acne. J Am Acad Dermatol. 2005;52(2):207-14.

15. Roengritthidet K, Kamanamool N, Udompataikul M, Rojhirunsakool S, Khunket S, Kanokrungsee S. Association Between Diet and Acne Severity: A Cross-sectional Study in Thai Adolescents and Adults. Acta Derm Venereol. 2021;101(12):569.

16. El Haddad C, Gerbaka N, Hallit S, Tabet C. Association between exposure to ambient air pollution and occurrence of inflammatory acne in the adult population. BMC Public Health. 2021;21(1):1664.

17. Li X, An SJ, Liu XL, Ji AL, Cao Y, Xiang Y, et al. The Association Between Short-Term Air Pollution Exposure and Post-Adolescent Acne: The Evidence from a Time Series Analysis in Xi’an, China. Clin Cosmet Investig Dermatol. 2021;14:723-31.

18. Li X, Zhou LX, Yang LL, Huang XL, Wang N, Hu YG, et al. The relationship between short-term PM2.5 exposure and outpatient visits for acne vulgaris in Chongqing, China: a

    time-series study. Environ Sci Pollut Res Int. 2022;29(40):61502-11.

19. Liu W, Pan X, Vierkötter A, Guo Q, Wang X, Wang Q, et al. A Time-Series Study of the Effect of Air Pollution on Outpatient Visits for Acne Vulgaris in Beijing. Skin Pharmacol Physiol. 2018;31(2): 107-13.

20. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71.

21. Higgins JPT, Morgan RL, Rooney AA, Taylor KW, Thayer KA, Silva RA, et al. A tool to assess risk of bias in non-randomized follow-up studies of exposure effects (ROBINS-E). Environment International. 2024;186:108602.

22. Paik K, Na JI, Huh CH, Shin JW. Particulate Matter and Its Molecular Effects on Skin: Implications for Various Skin Diseases. Int J Mol Sci. 2024;25(18):9888.

23. Piao MJ, Ahn MJ, Kang KH, Ryu YS, Hyun YJ, Shilnikova K, et al. Particulate matter 2.5 damages skin cells by inducing oxidative stress, subcellular organelle dysfunction, and apoptosis. Arch Toxicol. 2018;92(6):2077-91

24. Eberlein-König B, Przybilla B, Kühnl P, Gebefügi I, Kleinschmidt J, Ring J, et al. Influence of airborne nitrogen dioxide or formaldehyde on parameters of skin function and cellular activation in patients with atopic eczema and control subjects. J Allergy Clin Immunol. 1998;101(1):141-3.

25. Janvier X, Alexandre S, Boukerb AM, Souak D, Maillot O, Barreau M, et al. Deleterious Effects of an Air Pollutant (NO2) on a Selection of Commensal Skin Bacterial Strains, Potential Contributor to Dysbiosis? Front Microbiol. 2020;11:591839.

26. Liang J, Liu L, Kang X, Hu F, Mao L. Mechanism underlying the effect of SO2-induced oxidation on human skin keratinocytes. Medicine (Baltimore). 2020;99(48):e23152.

‌Exploratory Fetal Doppler Thresholds for Predicting Composite Fetal Compromise in Preeclampsia: A Single-Centre Prospective Diagnostic Accuracy Study

Linh My Duong, M.D., Ph.D.1, Nghia Quang Bui, M.D., Ph.D.2, Dan Thanh Nguyen Hong, M.D., M.Sc.3, Hien Thi Nguyen, M.D., Ph.D.4,*

1Department of Obstetrics and Gynecology, Faculty of Medicine, Can Tho University of Medicine and Pharmacy, Can Tho 94000, Vietnam, 2Department

of Pediatrics, Faculty of Medicine, Can Tho University of Medicine and Pharmacy, Can Tho 94000, Vietnam, 3Division of Obstetrics and Gynecology, Vinh Long General Hospital, Vinh Long 85000, Vietnam, 4Department of Nutrition and Food Safety, Faculty of Public Health, Can Tho University of Medicine and Pharmacy, Can Tho 94000, Vietnam.

*Corresponding author: Hien Thi Nguyen E-mail: nthien@ctump.edu.vn

Received 9 April 2026 Revised 25 July 2026 Accepted 25 July 2026 ORCID ID: http://orcid.org/0000-0002-3997-5943 https://doi.org/10.33192/smj.v78i9.281484

All material is licensed under terms of the Creative Commons Attribution 4.0 International (CC-BY-NC-ND 4.0) license unless otherwise stated.

ABSTRACT

Objective: To assess the diagnostic accuracy of umbilical artery (UA), middle cerebral artery (MCA), and cerebroplacental ratio (CPR) Doppler indices for composite fetal compromise in preeclampsia and derive exploratory cut-offs.

Materials and Methods: This prospective study included 77 women with preeclampsia at one Vietnamese centre (June 2021–June 2023). UA and MCA Doppler were performed within 48 hours before delivery; CPR was MCA pulsatility index divided by UA pulsatility index. The prespecified reference standard was at least one of category III cardiotocography (CTG), thick meconium-stained amniotic fluid, or umbilical arterial pH <7.10. CTG classification was blinded, ascertainment was complete, and delivery decisions did not use Doppler thresholds. Analyses were prespecified for <37 and ≥37 weeks. Cut-offs maximised the Youden index; optimism was assessed using 2,000 bootstrap replicates, and gestational-age interaction was tested.

Results: Fetal compromise occurred in 35/77 women (45.5%): 18/40 before 37 weeks and 17/37 thereafter. All

components were strictly nested; thus, the composite identified the same 35 women as category III CTG. Cord acidaemia occurred in five women, all before 37 weeks. CPR provided the best discrimination (AUC 0.78; 95% CI 0.65–0.87). CPR <1.10 yielded 74.3% sensitivity and 78.6% specificity, decreasing to 71.3% and 76.5% after optimism correction. Gestational-age interactions were non-significant (p=0.22 and p=0.46), but stratum-specific thresholds differed (<1.08 versus <1.22); gestational-age independence was not established.

Conclusion: CPR outperformed single-vessel indices, but estimates were imprecise and internally derived. These

thresholds are hypothesis-generating and require external validation before clinical use.

Keywords: Preeclampsia; cerebroplacental ratio; doppler ultrasonography; fetal distress; adverse perinatal outcome; umbilical artery; middle cerebral artery; diagnostic accuracy (Siriraj Med J 2026;78(9):683-694)

INTRODUCTION

Preeclampsia remains a leading cause of maternal morbidity and mortality worldwide, affecting approximately 2–8% of pregnancies.1 Impaired uteroplacental perfusion may restrict fetal growth and ultimately result in tissue hypoxia.2-5,6

In response to hypoxia, the fetus redistributes cardiac output to preserve perfusion of vital organs. These haemodynamic adaptations can be assessed non-invasively using fetal Doppler velocimetry.7,8 artery (UA) and middle cerebral artery (MCA) indices have been widely studied in hypertensive pregnancies, while the cerebroplacental ratio (CPR) integrates placental resistance and fetal cerebral adaptation. These indices have been associated with non-reassuring fetal status and adverse perinatal outcomes, although their sensitivity is modest and they should not be interpreted as stand-alone tests of fetal wellbeing.9-11

Reported Doppler reference values and diagnostic thresholds vary between cohorts, reflecting differences in measurement techniques, outcome definitions, case-mix and disease severity.12-14 Evidence from Southeast Asian clinical settings remains limited; however, a cohort comprising only women with preeclampsia and no healthy comparison group cannot establish population-specific standards or attribute observed thresholds to ethnicity, maternal characteristics or placental morphology.

Accordingly, this single-centre study aimed to explore the diagnostic performance of UA, MCA and CPR Doppler indices for a prespecified composite of in utero fetal compromise among women with preeclampsia in Vietnam and to generate candidate cut-off values for subsequent external validation. The study was not intended to establish population reference thresholds.

MATERIALS AND METHODS

Study design and setting

This was a prospective diagnostic accuracy study conducted at Can Tho City Obstetrics and Gynecology Hospital, a tertiary referral center in Vietnam, from June 2021 to June 2023. The study objective was to evaluate the predictive utility of Doppler ultrasound indices for anticipating in-utero fetal compromise among pregnant women complicated by preeclampsia. Reporting followed the STARD 2015 statement.

Study population

Women were enrolled consecutively if they had: (1) a singleton pregnancy with a viable fetus; (2) gestational age established by first-trimester crown–rump length;

(3) preeclampsia meeting the operational criteria below; and (4) provided written informed consent. Pregnancies complicated by eclampsia, HELLP syndrome, clinical chorioamnionitis or major fetal structural anomalies were

excluded, as were women with pre-gestational diabetes mellitus, chronic kidney disease or severe maternal cardiovascular disease.

Preeclampsia was defined as new-onset hypertension after 20 weeks in a previously normotensive woman, with systolic blood pressure ≥140 mmHg or diastolic blood pressure ≥90 mmHg on two occasions at least 4 hours apart. Blood pressure was measured after at least 10 minutes of rest, with the woman seated, the arm supported at heart level and an appropriately sized cuff. Severe hypertension (≥160/110 mmHg) could be confirmed within a shorter interval to permit timely treatment.

Diagnosis additionally required proteinuria of ≥300 mg/24 hours, a urine protein-to-creatinine ratio ≥0.3, or dipstick ≥2+ when quantitative measurement was unavailable. In the absence of proteinuria, preeclampsia was diagnosed when hypertension was accompanied by thrombocytopenia (<100 × 10⁹/L), renal insufficiency (serum creatinine >1.1 mg/dL or doubling of baseline), liver transaminases ≥2 times the upper limit of normal, pulmonary oedema, persistent unexplained headache unresponsive to medication, or visual symptoms. Severe features comprised severe-range blood pressure or any of these end-organ manifestations.

The approved protocol specified a minimum gestational age of 34 weeks. However, 14 women examined at 28.5–33.9 weeks were enrolled because severe features necessitated early delivery. This protocol deviation is reported explicitly. These women were retained in the primary analysis, and a sensitivity analysis excluding them was performed.

Sample size and sampling strategy

The sample size was calculated for a diagnostic accuracy study based on the anticipated specificity of the Doppler indices. The required number of participants without the target condition was estimated using the formula for a single proportion:

The observed prevalence was higher than anticipated: 35 of 77 women had the composite outcome, leaving 42 without it. For the observed specificity of 78.6%, the exact 95% confidence interval was 63.2–89.7%, corresponding to an achieved precision of approximately 13 percentage points rather than the planned 8 percentage points. The study was therefore considered exploratory.

Clinical and doppler ultrasound procedures

Doppler velocimetry was performed using a Samsung Medison WS80A ultrasound system with a 1.0–7.0 MHz curvilinear transabdominal transducer (CA1-7AD). All examinations were performed by one obstetrician with 15 years of Doppler experience, with the woman semi-recumbent, during fetal quiescence and in the absence of fetal breathing movements.

The umbilical artery (UA) was sampled at a free loop of cord. The middle cerebral artery (MCA) was examined in an axial section of the fetal head at the level of the sphenoid bones, with the sample volume placed in the proximal third of the vessel near its origin from the internal carotid artery. The insonation angle was maintained below 30° with angle correction, and the wall filter was set at 50 Hz. Measurements were obtained from three consecutive uniform, artefact-free waveforms. Each vessel was measured three times, and the mean was recorded.

Resistance index, pulsatility index (PI) and systolic-to-diastolic ratio were generated by automated waveform tracing, with manual correction when required. No waveform demonstrated absent or reversed end-diastolic flow. Formal intraobserver repeatability was not assessed and is acknowledged as a limitation. The cerebroplacental ratio (CPR) was calculated as MCA-PI divided by UA-PI and was recomputed directly from the recorded component indices.

Only Doppler examinations performed within 48 hours before delivery were analysed. Examinations preceded active labour and uterotonic administration.

Under institutional policy, antenatal corticosteroids were

Z

2

n = 1-α/2

× psp × (1-psp)

d2

administered only before 34 weeks; therefore, only the 14 women examined before 34 weeks could have been

The sample size was planned for estimation of specificity. Assuming a specificity of 0.90, a precision of

0.08 and a 95% confidence level, 54 participants without fetal compromise were required. Adjustment for an anticipated fetal-compromise prevalence of 23.1% from an earlier study at the same institution yielded a minimum total sample of 71; therefore, 77 women were enrolled consecutively.

exposed. A sensitivity analysis excluding all 14 potentially exposed women was performed to assess possible residual corticosteroid effects. No participant received magnesium sulphate before the qualifying cardiotocogram.

Outcome measures and definitions

The participant-level reference standard was a prespecified composite of in utero fetal compromise, defined as at least one of: (1) category III cardiotocography

(CTG) according to the ACOG three-tier system; (2) thick meconium-stained amniotic fluid observed at delivery; or (3) umbilical arterial cord-blood pH <7.10 measured immediately after delivery.

Doppler assessment preceded completion of the reference standard. CTG was performed during the same admission, meconium status was recorded at delivery and cord arterial blood was collected immediately after birth. All three components were ascertained in all 77 participants. CTG tracings were classified by a reader blinded to the Doppler findings.

Clinicians making delivery decisions were not formally blinded to Doppler results. However, the institutional protocol contained no Doppler-based delivery criterion, and isolated Doppler abnormalities did not prompt delivery. Delivery decisions were based on severe maternal features or persistent category III CTG.

Category III CTG was both a component of the composite and a potential trigger for delivery; therefore, the reference standard was not fully independent of the clinical pathway. Individual component frequencies and overlaps were reported separately because CTG, meconium and cord acidaemia differ in timing, objectivity and clinical significance. The composite was not interpreted as a neonatal outcome.

Statistical analysis

Statistical analyses were performed using SPSS software, version 22.0. Continuous variables were presented as means ± standard deviations, while categorical data were summarized as frequencies and percentages.

Gestational-age strata (<37 and ≥37 weeks) were derived directly from gestational age at the qualifying Doppler examination rather than from pre-coded grouping variables.

Discrimination was evaluated using receiver operating characteristic curves. Areas under the curve (AUCs) were reported with DeLong 95% confidence intervals and tested against an AUC of 0.5. Candidate cut-offs were selected using the maximum Youden index. Sensitivity, specificity, positive predictive value and negative predictive value were calculated from complete 2 × 2 tables and reported with exact Clopper–Pearson 95% confidence intervals. Because cut-offs were selected and evaluated in the same dataset, all thresholds were considered exploratory. Optimism in the primary CPR analysis was assessed using 2,000 bootstrap replicates, with cut-point re-selection in each replicate. Apparent and optimism-corrected estimates were reported; no external validation was

performed.

Effect modification by gestational age was assessed

using likelihood-ratio tests comparing logistic models with and without CPR-by-gestational-age interaction terms, with gestational age modelled both as the prespecified binary stratum and as a continuous variable. Stratum-specific AUCs were compared using the DeLong test. Two-sided p-values <0.05 were considered statistically significant. No adjustment was made for multiple comparisons; therefore, all analyses were interpreted descriptively.

Ethical considerations

The study protocol was reviewed and approved by the Institutional Review Board of Can Tho University of Medicine and Pharmacy (Reference No. 86/PCT-HĐĐĐ, dated March 30, 2021). All procedures were conducted in accordance with the Declaration of Helsinki. Before participation, all women were fully briefed on the study objectives and provided written informed consent.

RESULTS

Characteristics of research subjects

Seventy-seven women with preeclampsia were included: 40 (51.9%) underwent the qualifying Doppler examination before 37 weeks and 37 (48.1%) at or beyond 37 weeks. Baseline characteristics are shown in Table 1. The hypertensive disorder in the index pregnancy was classified into mutually exclusive categories: 57 women (74.0%) had de novo preeclampsia and 20 (26.0%) had preeclampsia superimposed on chronic hypertension. Previous preeclampsia was recorded separately as an obstetric-history variable (8/77, 10.4%).

Composite fetal compromise occurred in 35 of 77 women (45.5%): 18 of 40 (45.0%) before 37 weeks and

17 of 37 (45.9%) at or beyond 37 weeks. Category III cardiotocography (CTG) occurred in 35 women, thick meconium-stained amniotic fluid in 32, and cord arterial pH < 7.10 in 5.

Doppler indices and exploratory cut-offs for composite fetal compromise

UA-PI and UA-S/D were higher before 37 weeks than at or beyond 37 weeks, whereas MCA-RI and MCA-PI did not differ between strata (Table 2). CPR was lower before 37 weeks (1.15 ± 0.31 vs 1.34 ± 0.45; p = 0.039).

Diagnostic performance of individual Doppler indices

The gestational-age–stratified ROC curves are shown in Fig 1.

Table 3 presents the complete 2 × 2 classification for each exploratory cut-off. Before 37 weeks, 18 women had the composite outcome and 22 did not; at or beyond 37 weeks, the corresponding numbers were 17 and 20. No

TABLE 1. Characteristics of research subjects.

Note: Data are n (%) unless otherwise stated. De novo and superimposed preeclampsia are mutually exclusive; previous preeclampsia is reported separately. CTG counts denote women; 37 Category III episodes occurred among 35 women.

Abbreviations: CTG, cardiotocography; SD, standard deviation

TABLE 2. Doppler indices in the two gestational-age strata.

Note: Data are mean ± SD; aequal variances assumed, bnot assumed; *p < 0.05.

Abbreviations: CPR, cerebroplacental ratio; MCA, middle cerebral artery; PI, pulsatility index; RI, resistance index; S/D, systolic/diastolic ratio; UA, umbilical artery.

Fig 1. Gestational-age–stratified ROC curves for UA-RI, UA-PI, MCA-RI and MCA-PI in predicting composite fetal compromise.

TABLE 3. Diagnostic performance of Doppler indices for composite fetal compromise, by gestational-age stratum.

Note: Cut-offs were selected by the maximum Youden index within each stratum and are exploratory. AUCs were calculated from the continuous markers; AUC 95% CIs are DeLong intervals. Sensitivity, specificity, PPV and NPV 95% CIs are exact (Clopper–Pearson). Because cut-offs were selected and evaluated in the same data, the estimates are optimistically biased.

Abbreviations: AUC, area under the curve; CI, confidence interval; FN, false negative; FP, false positive; GA, gestational age; NPV, negative

predictive value; PPV, positive predictive value; TN, true negative; TP, true positive.

Doppler measurement or reference-standard component was missing. CPR had the highest AUC in both strata (0.77 and 0.79).

Cerebroplacental ratio

In the overall cohort, CPR showed an AUC of 0.78 (95% CI, 0.65–0.87; p < 0.001). At the exploratory cut-off of CPR < 1.10, sensitivity was 74.3%, specificity 78.6%, PPV 74.3%, and NPV 78.6%. Sensitivity and specificity were 77.8% and 72.7% before 37 weeks and 70.6% and

85.0% at or beyond 37 weeks, respectively (Table 4).

Bootstrap correction reduced overall sensitivity to 71.3% and specificity to 76.5%. The CPR-by-gestational-age interaction was not significant for gestational age as a binary stratum (p = 0.22) or a continuous variable (p = 0.46), and the stratum-specific AUCs did not differ (DeLong p = 0.86). These tests had low power; therefore, gestational-age independence was not established. The overall ROC curve is shown in Fig 2.

Sensitivity analysis

Under institutional policy, only the 14 women examined before 34 weeks could have received antenatal corticosteroids. Excluding these women left 63 participants, including 30 with the composite outcome. Results were essentially unchanged: AUC 0.79 (95% CI, 0.65–0.88;

p < 0.001), with 76.7% sensitivity (95% CI, 57.7–90.1)

and 78.8% specificity (61.1–91.0) at CPR < 1.10.

Composition of the composite outcome

All 5 women with cord arterial pH < 7.10 also had thick meconium and Category III CTG, and all 32 women with thick meconium also had Category III CTG. Three

women had Category III CTG alone, 27 had Category III CTG with meconium but without acidaemia, and 5 had all three components. No woman had meconium or acidaemia without Category III CTG, and no cord-gas sample was missing. Cord acidaemia occurred only before 37 weeks.

DISCUSSION

Demographic and clinical characteristics of the study population

The clinical profile of our cohort reflects a high-risk preeclamptic population, with a notable proportion of women of advanced maternal age, multiparity, and pre-

Fig 2. Receiver operating characteristic curve of CPR for predicting composite fetal compromise in the overall cohort.

TABLE 4. Continuous CPR AUC and diagnostic performance of the fixed exploratory threshold CPR < 1.10, by gestational-age stratum.

Note: The whole-cohort threshold was applied unchanged to both strata.

Abbreviations: AUCs are for continuous CPR. Bootstrap optimism correction was performed only for the overall cohort. AUC, area under the curve; CI, confidence interval; CPR, cerebroplacental ratio; FN, false negative; FP, false positive; NPV, negative predictive value; PPV, positive predictive value; TN, true negative; TP, true positive.

TABLE 5. Components of the composite outcome and their overlap.

Data are n (%). The mutually exclusive positive combinations sum to the composite total. No participant had meconium or acidaemia without Category III CTG. CTG, cardiotocography.

existing vascular disease. Nearly half of the participants were aged ≥ 35 years, and chronic hypertension or superimposed preeclampsia was present in a clinically relevant subset. These findings are consistent with recent evidence suggesting that increasing maternal age and accumulated cardiovascular risk may contribute substantially to the burden of preeclampsia in contemporary obstetric populations.15 Instead, contemporary clinical settings increasingly support a “cumulative cardiovascular risk” paradigm.16,17

Although nulliparity has traditionally been considered a major immunological risk factor for preeclampsia18-20, the predominance of multiparous women in our cohort suggests that underlying vascular and metabolic vulnerability may partly offset the protective effect of prior pregnancies. Pre-existing endothelial dysfunction and impaired vascular adaptation may predispose these women to abnormal placentation, placental insufficiency, and subsequent fetal compromise.21

This high-risk maternal background supports the clinical relevance of population-specific Doppler thresholds. Standard cut-offs derived from other populations may not fully capture fetal hemodynamic adaptation in Vietnamese women, particularly in settings with different maternal anthropometric and cardiovascular profiles. Locally derived thresholds may therefore improve risk stratification and support timely obstetric decision-making in preeclamptic pregnancies.22

The gestational-age distribution was sparse across

28.5–40.3 weeks. Although formal binary and continuous interaction analyses did not detect effect modification, the subgroup analyses were underpowered and the exploratory candidate thresholds differed between strata. Gestational-age centiles were not estimated because observations per gestational week were insufficient. Gestational-age independence was therefore not established, and the pooled CPR threshold of < 1.10 is retained only as an exploratory candidate; the title has been revised accordingly.23,24

Predictive Value of Individual Doppler Indices: Umbilical Artery and Middle Cerebral Artery

The hemodynamic cascade of fetal compromise in preeclampsia follows a predictable pathophysiological continuum, initiating with placental vascular resistance and progressing to compensatory cerebral vasodilation. Early in the disease process, defective trophoblastic invasion and failure of spiral artery remodeling result in increased impedance in the fetoplacental circulation, clinically manifested as elevated UA indices. As placental insufficiency worsens and intrauterine hypoxia becomes chronic, fetal chemoreceptors trigger a vital adaptive mechanism, the "brain-sparing" effect, resulting in profound cerebral vasodilation and a subsequent drop in MCA impedance.25

Placental dysfunction exhibits distinct severity by gestational age, with early-onset preeclampsia (< 37 weeks) demonstrating a more profound degree of defective placental development than the late-onset phenotype.

These findings align with recommendations by Maulik et al.26,27 and Pei Zhou et al.18 However, the moderate specificity of UA indices in our study reiterates that elevated placental resistance alone does not invariably translate into immediate fetal compromise, necessitating further evaluation of fetal adaptation via the MCA.

The negative predictive values of individual Doppler indices were modest and are dependent on outcome prevalence. Moreover, this study did not evaluate expectant management, compare delivery strategies, or assess neonatal benefit. A reassuring UA or MCA result therefore cannot establish that delaying delivery is safe; these indices should be interpreted together with maternal condition, cardiotocography, fetal growth, amniotic fluid, and gestational age.2,22,28,29

Diagnostic performance of the cerebroplacental ratio The cerebroplacental ratio (CPR) is the most robust predictor of fetal compromise because it captures both the pathological trigger and the consequent fetal adaptive response. Unlike single-vessel Doppler indices that provide only a partial view of fetal hemodynamics, CPR seamlessly integrates the "cause" (increased placental vascular resistance reflected by the UA) and the "effect" (fetal compensatory cerebral vasodilation reflected by the MCA).23 This integration provides a superior, multifaceted measure of the maternal-fetal hemodynamic interplay, making it exquisitely sensitive to the progressive deterioration of placental function before overt clinical symptoms

appear.

The rates of small-for-gestational-age (SGA) fetuses (9.1%) and low-dose aspirin (LDA) prophylaxis (9.1%) observed in this study are consistent with the characteristics of a high-risk obstetric cohort. Although not all SGA fetuses are growth restricted, SGA associated with placental dysfunction may exhibit increased fetoplacental vascular resistance, reflected by an elevated UA-PI. Chronic fetal hypoxemia may also trigger cerebral vasodilatation, the “brain-sparing” response, leading to a lower MCA-PI and, consequently, a reduced CPR.22,30,31 Previous studies have shown that prophylactic LDA may influence uterine artery pulsatility index trajectories and uteroplacental perfusion in pregnancies at increased risk of pre-eclampsia.32,33 Therefore, LDA use may have influenced maternal uterine artery Doppler profiles in the present cohort; however, current evidence does not establish that aspirin masks fetal UA, MCA, or CPR abnormalities specifically among SGA pregnancies.

The exploratory CPR threshold of < 1.10 lies within

the range reported across previous cohorts, in which proposed operating points have varied from values near

1.0 to substantially higher cut-offs.24,34,35 This variation may reflect differences in outcome definitions, disease severity, gestational-age distribution, acquisition methods, and sampling. Because ethnicity-related physiology, maternal BMI, placental volume, and a normotensive reference distribution were not evaluated, a causal population-specific explanation would be speculative. The present threshold should therefore be interpreted only as a single-centre candidate for external testing.

The observed specificity does not establish CPR as a definitive rule-in test, and its positive predictive value is cohort- and prevalence-dependent. The threshold was derived and assessed in the same small dataset, and no delivery strategy was evaluated. A low CPR should therefore not be interpreted as evidence of fetal compensatory exhaustion or used alone to abandon expectant management or expedite delivery.

Strengths and limitations

This prospective study provides preliminary single-centre evidence on the diagnostic performance of fetal Doppler indices in preeclamptic pregnancies in Southeast Asia. Its strengths include consecutive enrolment, standardised Doppler examinations and complete ascertainment of the reference-standard components. The resulting candidate cut-offs may inform the design of future validation studies.

Several limitations constrain interpretation. The small sample and gestational-age subgrouping reduced statistical precision, while the limited number of events precluded multivariable adjustment. The achieved number of women without the outcome was below that required by the original precision calculation. Moreover, the cut-offs were derived and evaluated in the same dataset; bootstrap correction quantified but could not eliminate optimism or replace external validation. The absence of a normotensive control group precluded derivation of physiological or population-specific reference values, and intraobserver repeatability was not assessed. Although outcome ascertainment was complete, the composite was coextensive with Category III CTG; consequently, the estimates primarily reflect prediction of this partly subjective surveillance finding rather than independently confirmed fetal hypoxaemia, acidaemia or neonatal morbidity. Category III CTG also triggered delivery under the institutional protocol, linking the reference standard to clinical management, although CTG classification was blinded to Doppler findings and Doppler results did not determine delivery. The protocol discrepancy involving examinations before 34 weeks was addressed through sensitivity analysis. Finally, the single-centre

tertiary-care setting limits generalisability. All proposed thresholds should therefore be considered exploratory.

Clinical implications and future directions

These findings do not support a rule-in/rule-out pathway or the use of CPR alone to determine delivery timing. The candidate threshold may be useful for hypothesis generation, but clinical management should continue to rely on integrated maternal and fetal assessment until diagnostic performance and clinical benefit have been demonstrated independently.

Prospective multicentre studies with larger samples are needed to externally validate a pre-specified threshold and assess gestational-age-adjusted centiles using objective fetal and neonatal outcomes. Only after independent validation should comparative studies determine whether adding CPR to existing surveillance improves outcomes without increasing iatrogenic prematurity. Future work may also examine whether combining fetal Doppler indices with maternal angiogenic markers, including the sFlt-1/PlGF ratio, improves prediction in validated multivariable models

CONCLUSIONS

In this single-centre cohort, CPR showed greater discrimination than individual UA or MCA indices, but the proposed cut-offs remain exploratory and unvalidated. Because the composite outcome was coextensive with Category III CTG, the estimates primarily concern prediction of that surveillance finding rather than objectively confirmed fetal hypoxaemia, acidaemia or neonatal morbidity. The study does not establish population-specific reference thresholds, gestational-age independence, fetal compensatory exhaustion or a rule-in/rule-out clinical pathway. External validation in adequately powered multicentre cohorts is required before these values can inform clinical management.

Data Availability Statement

The anonymized dataset supporting this study has been deposited in the Zenodo repository (https://doi. org/10.5281/zenodo.21928086 ). The data are distributed under the Creative Commons Attribution 4.0 International License, allowing unrestricted use for scientific and academic purposes.

ACKNOWLEDGEMENTS

The authors would like to express their sincere gratitude to the pregnant women for their participation and consent in this study. Special thanks are also extended to Can Tho University of Medicine and Pharmacy and

Can Tho Obstetrics and Gynecology Hospital for their institutional assistance and approval throughout the research process.

DECLARATIONS

Grants and Funding Information

This research received no external funding.

Conflict of Interest

All authors declare that they have no financial or commercial relationships, such as consultancies, stock ownership, equity holdings, or patent/licensing agreements, that could be perceived as potential conflicts of interest related to the content of this manuscript.

Registration number of clinical trial

This study was not registered as a clinical trial.

Author Contributions

Conceptualization and methodology, L.M.D., N.Q.B., and H.T.N.; Investigation, L.M.D. and D.T.N.H.; Formal analysis, N.Q.B. and H.T.N.; Visualization and writing – original draft, L.M.D. and N.Q.B.; Writing – review and editing, L.M.D., N.Q.B., D.T.N.H., and H.T.N.;

Supervision, H.T.N. All authors have read and agreed to the final version of the manuscript.

Use of Artificial Intelligence

During the preparation of this work, the authors used Grammarly and ChatGPT (OpenAI) strictly to improve the readability, clarity and language of the manuscript. The authors reviewed and edited all outputs with full human oversight and control, assume full responsibility for the final content of the publication, and confirm that no artificial intelligence tool is listed as an author or co-author.

Institutional Review Board Statement

The study protocol was reviewed and approved by the Institutional Review Board of Can Tho University of Medicine and Pharmacy (Reference No. 86/PCT-HĐĐĐ, dated March 30, 2021). The study was conducted in accordance with the Declaration of Helsinki. All participants were fully informed of the study objectives and procedures, and written informed consent was obtained from each participant before enrolment.

REFERENCES

1. Practice ACoO. ACOG Practice bulletin No. 202: diagnosis and management of preeclampsia and eclampsia. Obstetrics & Gynecology. 2019;133(1):1-25.

   
   

2. Adekanmi AJR, Adebola1,; Akinmoladun, Janet Adetinuke; Adeyinka, Abiodun Oludotun. Uterine and Umbilical Artery Doppler in Women with Pre-eclampsia and their Pregnancy Outcomes. Nigerian Postgraduate Medical Journal. 2019;26(2).

3. Chappell LC, Cluver CA, Tong S. Pre-eclampsia. The Lancet. 2021;398(10297):341-54.

4. Dimitriadis E. Pre-eclampsia. Nature Reviews Disease Primers. 2023;9(1):8.

5. Peraçoli JC. Pre-eclampsia/eclampsia. Revista Brasileira de Ginecologia e Obstetrícia. 2019;41:318-32.

6. Medjedovic E. Uric acid values along with Doppler sonography findings as a tool for preeclampsia screening. Medical Archives. 2019;73(6):408.

7. Stevenson GN. Automated visualization and quantification of spiral artery blood flow entering the first-trimester placenta, using 3-D power Doppler ultrasound. Ultrasound in Medicine & Biology. 2018;44(3):522-31.

8. Yang T, Luo H. Practical Ultrasonography in Obstetrics and Gynecology: Springer Nature; 2022.

9. Sotiriadis A. ISUOG Practice Guidelines: role of ultrasound in screening for and follow-up of pre-eclampsia. Ultrasound in Obstetrics & Gynecology. 2019;53(1):7-22.

10. Chuchotirot M, Sutanthavibul A. Prediction of Adverse Perinatal Outcome by Umbilical Artery to Middle Cerebral Artery Resistance Index Ratio in Pregnancies Complicated by Hypertensive Disorder. Siriraj Medical Journal. 2002;54(10):611-6.

11. Ounphamornlah N, Luangdansakul W, Orungrote N, Pleankong M, Petcharat K, metaneedol S, et al. Prevalence of Abnormal Cerebroplacental Ratio in Uncomplicated Full-term Pregnancy and Correlation with Adverse Perinatal Outcomes. Siriraj Medical Journal. 2024;76(8):514-21.

12. Oros D, Ruiz-Martinez S, Staines-Urias E, Conde-Agudelo A, Villar J, Fabre E, et al. Reference ranges for Doppler indices of umbilical and fetal middle cerebral arteries and cerebroplacental ratio: systematic review. Ultrasound Obstet Gynecol. 2019;53(4): 454-64.

13. Teulings N, Wood AM, Sovio U, Ozanne SE, Smith GCS, Aiken CE. Independent influences of maternal obesity and fetal sex on maternal cardiovascular adaptation to pregnancy: a prospective cohort study. Int J Obes (Lond). 2020;44(11):2246-55.

14. Drukker L, Staines-Urias E, Villar J, Barros FC, Carvalho M, Munim S, et al. International gestational age-specific centiles for umbilical artery Doppler indices: a longitudinal prospective cohort study of the INTERGROWTH-21(st) Project. Am J Obstet Gynecol. 2020;222(6):602.e1-.e15.

15. Duong LM. Early pregnancy pre-eclampsia screening using fetal medicine foundation formula: a longitudinal study. Archives of the Balkan Medical Union. 2025;60(4):555-62.

16. Gestational Hypertension and Preeclampsia: ACOG Practice Bulletin Summary, Number 222. Obstetrics & Gynecology. 2020;135(6).

17. Staff AC, Redman CW, Williams D, Leeson P, Moe K, Thilaganathan B, et al. Pregnancy and Long-Term Maternal Cardiovascular Health: Progress Through Harmonization of Research Cohorts and Biobanks. Hypertension. 2016;67(2):251-60.

18. Zhou P, Sun Y, Tan Y, An Y, Wang X, Wang L. Fetal and Neonatal Middle Cerebral Artery Hemodynamic Changes and Significance under Ultrasound Detection in Hypertensive

    Disorder Complicating Pregnancy Patients with Different Severities. Comput Math Methods Med. 2022;2022:6110228.

19. Hernández-Díaz S, Toh S, Cnattingius S. Risk of pre-eclampsia in first and subsequent pregnancies: prospective cohort study. Bmj. 2009;338:b2255.

20. Burton GJ, Redman CW, Roberts JM, Moffett A. Pre-eclampsia: pathophysiology and clinical implications. BMJ. 2019;366:l2381.

21. Kooffreh ME, Ekott M, Ekpoudom DO. The prevalence of pre-eclampsia among pregnant women in the University of Calabar Teaching Hospital, Calabar. Saudi Journal for Health Sciences. 2014;3(3).

22. Lees CC, Stampalija T, Baschat A, da Silva Costa F, Ferrazzi E, Figueras F, et al. ISUOG Practice Guidelines: diagnosis and management of small-for-gestational-age fetus and fetal growth restriction. Ultrasound Obstet Gynecol. 2020;56(2):298-312.

23. Morales-Roselló J, Khalil A, Morlando M, Papageorghiou A, Bhide A, Thilaganathan B. Changes in fetal Doppler indices as a marker of failure to reach growth potential at term. Ultrasound Obstet Gynecol. 2014;43(3):303-10.

24. Shahinaj R, Manoku N, Kroi E, Tasha I. The value of the middle cerebral to umbilical artery Doppler ratio in the prediction of neonatal outcome in patient with preeclampsia and gestational hypertension. J Prenat Med. 2010;4(2):17-21.

25. Mose JC. The role of maternal & fetal doppler in pre-eclampsia. Pregnancy Hypertens. 2014;4(3):242.

26. Maulik D, & Lees, C. C. (Eds.). Umbilical Doppler velocimetry: Normative data and diagnostic efficacy; The cerebroplacental ratio. In Doppler ultrasound in obstetrics and gynecology. ed. r, editor: Springer; 2023.

27. Maulik DE. Doppler ultrasound in obstetrics and gynecology2005.

28. Lees CC, Marlow N, van Wassenaer-Leemhuis A, Arabin B, Bilardo CM, Brezinka C, et al. 2 year neurodevelopmental and intermediate perinatal outcomes in infants with very preterm fetal growth restriction (TRUFFLE): a randomised trial. Lancet. 2015;385(9983): 2162-72.

29. Chaipakdi P. Diagnostic accuracy of the foetal umbilical artery, middle cerebral artery Doppler indices and cerebroplacental ratios for predicting adverse perinatal outcomes in pregnancy-induced hypertension. Thai Journal of Obstetrics and Gynaecology. 2024;32(2):99-106.

30. Melamed N, Baschat A, Yinon Y, Athanasiadis A, Mecacci F, Figueras F, et al. FIGO (International Federation of Gynecology and Obstetrics) initiative on fetal growth: best practice advice for screening, diagnosis, and management of fetal growth restriction. Int J Gynaecol Obstet. 2021;152(S1):3-57.

31. Miller SL, Huppi PS, Mallard C. The consequences of fetal growth restriction on brain structure and neurodevelopmental outcome. J Physiol. 2016;594(4):807-23.

32. Rolnik DL, Syngelaki A, O'Gorman N, Wright D, Poon LC, Nicolaides KH. ASPRE trial: effects of aspirin on mean arterial blood pressure and uterine artery pulsatility index trajectories in pregnancy. Ultrasound in Obstetrics & Gynecology. 2023;61(6): 691-7.

33. Trilla C, Platero J, Mora Brugués J, Nan MN, Medina C, Alejos Abad O, et al. Role of aspirin therapy in modulating uterine artery resistance and placental growth between first and second trimesters of pregnancy. Ultrasound in Obstetrics & Gynecology. 2025;66(5):641-8.

34. Alanwar A, El Nour AA, El Mandooh M, Abdelazim IA, Abbas

    
    

    L, Abbas AM, et al. Prognostic accuracy of cerebroplacental ratio for adverse perinatal outcomes in pregnancies complicated with severe pre-eclampsia; a prospective cohort study. Pregnancy Hypertension. 2018;14:86-9.

35. Agaoglu Z, Tanacan A, Ipek G, Peker A, Ozturk Agaoglu M,

Ozkavak OO, et al. The role of the cerebro-placental-uterine ratio in predicting composite adverse perinatal outcomes in patients with pregnancy-induced hypertension. Pregnancy Hypertens. 2024;37:101148.

‌Design and Validation of a Low-Cost IoT-Based Medical Gas Monitoring System for Hospital Storage Rooms

Navapadol Kittiamornkul, Ph.D.1, Tawiwan Sareebot, M.Sc.1, Sumaman Buntoung, Ph.D.2, Pranomkorn Choosri, Ph.D.1,*

1Department of Biomedical Engineering, College of Health Sciences, Christian University of Thailand, Nakhon Pathom 73000, Thailand, 2Department

of Physics, Faculty of Science, Silpakorn University, Nakhon Pathom 73000, Thailand.

*Corresponding author: Pranomkorn Choosri E-mail: pranomkorn.aum@gmail.com

Received 13 May 2026 Revised 5 August 2026 Accepted 5 August 2026 ORCID ID: http://orcid.org/0009-0007-4301-1032 https://doi.org/10.33192/smj.v78i9.282246

All material is licensed under terms of the Creative Commons Attribution 4.0 International (CC-BY-NC-ND 4.0) license unless otherwise stated.

ABSTRACT

Objective: This study aimed to develop and evaluate a low-cost medical gas monitoring system to enhance safety in hospital gas storage rooms. Continuous monitoring of oxygen and carbon dioxide levels may help identify oxygen-deficient or oxygen-enriched conditions and elevated carbon dioxide concentrations that may pose occupational safety risks to healthcare personnel.

Materials and Methods: An Internet of Things (IoT)-based device was developed to monitor medical gases and

environmental conditions, including oxygen, carbon dioxide, temperature, and relative humidity. The system enables real-time monitoring and online data visualization. Sensor performance was validated against standard reference equipment.

Results: The developed system demonstrated normalized root mean square differences (nRMSD) of 15.47%, 0.085%,

4.80%, and 5.39% for carbon dioxide, oxygen, temperature, and relative humidity, respectively, under ambient environmental conditions. The corresponding normalized mean bias differences (nMBD) were -9.70%, 0.283%, 2.54%, and -3.00%, respectively. However, validation was conducted over a narrow ambient concentration range and did not include controlled gas concentrations near the predefined alarm thresholds. Therefore, these results represent preliminary measurement agreement under ambient conditions and do not establish alarm performance or safety-monitoring performance across the intended decision range.

Conclusion: The proposed low-cost monitoring system (approximately USD 255) enables continuous real-time

monitoring of oxygen, carbon dioxide, temperature, and relative humidity under ambient environmental conditions. The system may support environmental monitoring and improve occupational safety in hospital medical gas storage areas.

Keywords: Medical gas systems; carbon dioxide; internet of things; low-cost sensor (Siriraj Med J 2026;78(9):695-708)

INTRODUCTION

In hospital environments, improper levels of medical gases suchas oxygenandcarbondioxide may pose significant health risks to healthcare personnel. Elevated carbon dioxide concentrations may cause adverse physiological effects, including respiratory acidosis, dizziness, and loss of consciousness, depending on exposure concentration and duration. In contrast, oxygen-deficient atmospheres may pose a direct risk of hypoxia, whereas oxygen-enriched atmospheres above 23.5%Vol increase the risk of accelerated combustion and fire.1 Continuous monitoring is therefore essential to ensure occupational safety, particularly in confined medical gas storage areas. According to the Occupational Safety and Health Administration (OSHA), the permissible exposure limit for carbon dioxide is 5,000 ppm as an 8-hour time-weighted average, while short-term exposure to higher concentrations may cause adverse physiological effects.2

Medical gas systems are critical components of

hospital infrastructure and are regulated by standards such as NFPA 99: Health Care Facilities Code, which provides requirements for the storage, distribution, alarm systems, and safe operation of medical gas systems according to the type of gas system and facility risk category.3 These systems encompass the production, storage, distribution, and control of medical gases to ensure safe delivery

for patient care. In addition to installed medical gas safety systems, continuous environmental monitoring of gas concentrations within storage areas may provide supplementary information on local environmental conditions and potential gas leakage.

Occupational safety frameworks from the World Health Organization emphasize the importance of safe working environments and risk prevention strategies for healthcare workers, including engineering controls and environmental monitoring where hazardous exposure may occur. Previous research conducted at Siriraj Hospital has also highlighted the importance of monitoring hazardous medical gases in hospital environments by evaluating ethylene oxide exposure and determining safe contact periods for healthcare personnel, reinforcing the need for effective environmental gas monitoring systems.4 Medical gas storage rooms are typically enclosed spaces with limited ventilation and contain large numbers of gas cylinders, including oxygen, nitrous oxide, carbon dioxide, and medical air. Biomedical engineers and maintenance personnel are required to enter these areas regularly to inspect equipment and monitor environmental conditions, often spending extended periods inside. In the presence of undetected gas leaks or abnormal gas concentrations, personnel may be exposed to hazardous conditions with potentially serious health consequences.

Medical gases, particularly oxygen and carbon dioxide, have well-documented physiological effects depending on concentration and duration of exposure. Oxygen deficiency may result in hypoxia and impaired physiological function, whereas oxygen-enriched atmospheres increase the risk of accelerated combustion and fire. Elevated carbon dioxide concentrations may cause respiratory acidosis and adverse neurological effects, including dizziness, impaired consciousness, and, at sufficiently high concentrations, loss of consciousness.5,6

Commercial gas monitoring instruments are available for various applications; however, their configurations, monitoring capabilities, certification requirements, and costs vary substantially. Recent studies have explored low-cost gas sensing technologies based on semiconductor, electrochemical, and non-dispersive infrared (NDIR) principles. However, most existing systems are limited to single-parameter measurements or lack real-time monitoring capability in confined hospital environments. Furthermore, the high cost of commercial devices limits their widespread implementation in routine hospital safety monitoring.7-10

Previous studies have primarily focused on the development and evaluation of individual gas sensing technologies or single-gas monitoring systems. In contrast, the proposed system was specifically designed for hospital medical gas storage rooms by integrating simultaneous monitoring of carbon dioxide, oxygen, temperature, and relative humidity with IoT-based real-time remote visualization and a visual warning function into a low-cost prototype. This integrated approach provides a practical solution for continuous environmental safety monitoring in confined hospital gas storage areas, thereby addressing an important gap in previous monitoring systems.

Therefore, this study aimed to develop and evaluate a low-cost prototype device for real-time monitoring of medical gas levels in hospital storage rooms using Internet of Things (IoT) technology. Recent studies have demonstrated the growing application of IoT technologies for hospital environmental and air quality monitoring, highlighting their potential to provide continuous surveillance, real-time alerts, and improved occupational safety in healthcare settings.11 The proposed system enables continuous monitoring of oxygen, carbon dioxide, temperature, and relative humidity, with real-time data transmission and alert capability. This system is designed to improve safety and reduce the risk of hazardous gas exposure among healthcare personnel.

MATERIALS AND METHODS

Materials

The prototype system was developed to provide low-cost monitoring of medical gases and environmental conditions, including carbon dioxide (CO2), oxygen (O2), room temperature (T), and relative humidity (RH). Commercially available sensors and open-source microcontroller platforms were selected to reduce development cost while maintaining acceptable measurement performance.

An Arduino Mega 2560 microcontroller was used as the main controller for sensor data acquisition and system operation. The controller was selected because of its compatibility with multiple sensors, stable performance, and ease of programming.

For wireless data communication, an AIS NB-IoT module based on the Quectel BC95-B8 chipset with SubMiniature version A (SMA) connector was integrated into the system. This module enabled real-time transmission of measured data to the Magellan IoT cloud platform for remote monitoring and data visualization. NB-IoT communication was selected because of its low power consumption, wide network coverage, and suitability for long-term monitoring applications in hospital environments.

The controller unit and communication module were electrically connected to enable automatic data collection, processing, and online transmission, as illustrated in Fig 1.

Fig 1. NB-IoT and Arduino Mega 2560 controller were used in this work.

The sensors were selected based on their commercial availability, low cost, suitable measurement range, and compatibility with the microcontroller platform. The selected components for monitoring carbon dioxide, oxygen, temperature, and relative humidity are summarized in Table 1.

The estimated total hardware cost of the developed prototype is presented in Table 2. The system was assembled using commercially available components to minimize cost while maintaining practical functionality for hospital monitoring applications.

Carbon dioxide concentration was measured using the MG-811 sensor, a metal oxide semiconductor sensor designed for CO2 detection. The sensor was selected because of its wide measurement range and rapid response time, making it suitable for detecting changes in carbon dioxide concentration in enclosed environments.

Oxygen concentration was measured using the SEN0322 electrochemical oxygen sensor (DFRobot). This sensor provides stable oxygen measurements over

a range of 0–25 %Vol with adequate resolution for environmental monitoring applications. Electrochemical sensing technology was selected because of its sensitivity and suitability for continuous oxygen monitoring.

Temperature and relative humidity were measured using the DHT22 sensor. This low-cost digital sensor integrates a thermistor and capacitive humidity sensing element, allowingsimultaneousmonitoringofenvironmental conditions relevant to gas storage room safety.

An OLED display (SSD1309) was incorporated into the prototype to provide real-time on-site visualization of measured parameters.

Prototype design and system integration

All sensors were integrated with the microcontroller unit to enable simultaneous monitoring of carbon dioxide, oxygen, temperature, and relative humidity, as shown in Fig 2. Sensor outputs were acquired and processed by the Arduino Mega 2560 controller. The processed data were transmitted wirelessly through the AIS NB-IoT

TABLE 1. Selected low-cost sensors used in the prototype.

MG-811 Generic Carbon dioxide (CO2) 350-10,000 ppm

DHT22 Aosong Temperature (T) -40oC to 80oC

TABLE 2. Estimated hardware cost of the developed prototype.

Arduino Mega 2560 51.49

CO2 sensor (MG-811) 51.49

Temperature/RH sensor (DHT22) 5.46

Housing and accessories 15.60

DEVIO NB-SHIELD I

and Arduino Mega 2560

LED: O2

O2 Sensor

MG-811 CO2 sensor

DHT22

5V

5V

A15

A14

A13

LED: CO

2

A12

SCL (I2C)

SDA (I2C) GND

GND

5V

5 V GND

A1

OLED LED 128x64

P1

AIS

NB-IoT

Fig 2. Three sensors were connected to the microcontroller and display system.

communication module based on the Quectel BC95-B8 chipset to the Magellan IoT cloud platform for remote monitoring and data visualization.

The sensing enclosure was designed using Tinkercad (Autodesk) and had external dimensions of 10 cm × 6 cm × 4 cm. The 3D model was subsequently prepared for 3D printing using FlashPrint 5 (Flashforge, USA). The enclosure was designed to accommodate the sensors and associated components. No dedicated ventilation holes were incorporated into the sensor enclosure. Ambient air could enter and exit through the cable-routing opening, allowing air exchange around the sensors. The control enclosure was a commercially available housing modified to accommodate the electronic components, display, cable connections, and antenna. The enclosure design and dimensions are shown in Fig 3.

The prototype was powered by a 12-V DC power supply. The Arduino Mega 2560 was programmed using Arduino IDE version 2.0. The firmware used the corresponding Arduino libraries for the CO2, oxygen, temperature, and relative humidity sensors. For carbon dioxide measurement, the MG-811 sensor was interfaced using the CO2Sensor library with configuration parameters of 0.99 and 100, and the library calibration function was executed during system initialization. For oxygen measurement, the DFRobot SEN0322 sensor was operated using the manufacturer’s I²C library with a collection number of 10 measurements. Oxygen calibration was performed using ambient atmospheric oxygen concentration of 20.9%Vol as the reference value before the validation experiments.

The sensors acquired measurements at approximately 5-s intervals. The measurements were processed and

A developed device (c)

2

2

Fig 3. A low-cost device design; (a) inside of control box (b) inside of sensor box and (c) the successfully developed prototype.

averaged by the firmware over the reporting period and transmitted to the Magellan IoT cloud platform at approximately 1-min intervals. During normal operation, the system automatically attempted to reconnect when the NB-IoT connection was interrupted. Data that could not be transmitted during a network interruption were not locally buffered and were therefore lost. For comparative validation, the prototype data were subsequently averaged over 2-min intervals to match the recording interval of

the corresponding reference instrument. A total of 68 paired observations were obtained for the comparative analysis of carbon dioxide, temperature, and relative humidity. Thus, the 2-min interval represented a post-processing aggregation interval for temporal alignment with the reference instrument rather than the original sensor measurement interval.

The visual warning logic was implemented using predefined concentration thresholds. Under normal conditions, the indicator light remained green. The indicator changed to red when the measured carbon dioxide concentration exceeded 5,000 ppm or when the measured oxygen concentration was below 19.5%Vol or above 23.5%Vol. These thresholds were implemented as warning criteria in the prototype; however, their alarm sensitivity and performance were not validated across the relevant concentration ranges.

Sensor data processing and calibration

The raw outputs from the sensors were processed using the manufacturers’ recommended Arduino libraries. For carbon dioxide measurement, the MG-811 sensor was interfaced using the CO2Sensor Arduino library. The library converts the analog voltage signal into carbon dioxide concentration (ppm) using an embedded conversion algorithm based on the sensor response characteristics. Prior to data acquisition, the sensor was initialized using the library calibration function to establish the baseline response.

Oxygen concentration was measured using the DFRobot SEN0322 electrochemical oxygen sensor through the manufacturer’s I²C library. Oxygen concentration (%Vol) was obtained directly using the getOxygenData() function after sensor initialization. The sensor was calibrated according to the manufacturer’s recommended procedure using ambient atmospheric oxygen concentration (20.9%Vol) as the calibration reference before the validation experiments.

Warning function verification

The warning function was functionally checked during prototype testing. For oxygen monitoring, oxygen gas was introduced into a closed chamber containing the prototype to observe the response of the visual indicator when the measured oxygen concentration increased. For carbon dioxide monitoring, the warning logic was checked by temporarily setting the CO₂ threshold to 1,000 ppm. Carbon dioxide concentration was increased using exhaled breath and combustion smoke, and the indicator changed from green to red when the measured concentration exceeded the temporary threshold.

These tests were performed only to confirm the basic operation of the visual warning logic and were not considered quantitative alarm validation. The use of exhaled breath and combustion smoke did not provide traceable or controlled CO2 concentrations and may have introduced other environmental factors or interfering substances. Therefore, the response, sensitivity, and accuracy of the deployed 5,000-ppm CO2 alarm threshold were not established in this study. Validation of alarm performance using certified CO2 concentrations across the relevant operating range and at the 5,000-ppm threshold is required in future work.

Prototype validation

Prior to field deployment, the developed prototype was validated against standard reference instruments to assess measurement performance.

Oxygen concentration was evaluated separately using a MaxO₂+ AE oxygen analyzer (Maxtec, USA; model R217P72) as the reference instrument. The analyzer has a measurement range of 0–100% O₂ and a resolution of 0.1% O₂. According to the manufacturer’s specifications, its total accuracy is ±3% of the actual oxygen level over the full operating temperature range, while accuracy and linearity are specified as ±1% of full scale under constant temperature, relative humidity, and pressure when calibrated at full scale. The analyzer uses a Max-250 series galvanic fuel-cell oxygen sensor and has a response time of approximately 15 s to reach 90% of the final value at 23 °C.

Before the validation measurements, the MAXO₂+ AE was calibrated to ambient air (20.9% O₂) using the instrument’s one-touch calibration function, following the manufacturer’s recommended procedure. No warm-up time was required according to the manufacturer’s specifications. The prototype and reference analyzer were operated simultaneously, and their measurements were synchronized and compared at 1-min intervals.

Carbon dioxide concentration, temperature, and relative humidity were evaluated using an OIZOM Polludrone V6 air quality monitoring system (serial number N002P0001) available at Silpakorn University. The reference instrument was configured with the OZCO2_1, OZTEMP_1, and OZHUM_1 sensor modules for CO₂, temperature, and relative humidity measurements, respectively. According to the manufacturer’s specifications, the CO₂ sensor uses non-dispersive infrared (NDIR) technology and has a measurement range of 0–5,000 ppm, a resolution of 1 ppm, a minimum detection level of 400 ppm, and a reported drift of ±5 ppm/year. The temperature sensor has a range of −40 to 125 °C with a resolution of 0.01 °C, while the

relative humidity sensor has a range of 0–100% RH with a resolution of 0.10% RH. The reference instrument was operated side-by-side with the prototype under ambient environmental conditions. For the instrument used in this study, raw measurements were recorded at 2-min intervals and transmitted to the manufacturer’s cloud platform. The recorded data could also be downloaded as averaged datasets over longer intervals; however, the 2-min raw measurements were used as the reference observations in the present comparative analysis. The prototype measurements, originally acquired at shorter intervals, were subsequently aggregated to 2-min intervals to match the reference recording interval. Following temporal alignment, a total of 68 paired observations were included in the comparative analysis of CO₂, temperature, and relative humidity. These paired observations were obtained from a continuous validation sequence and were therefore treated as time-series paired observations rather than independent samples. The reference measurements covered ranges of 325.06–468.53 ppm for CO₂, 35.41–37.08

°C for temperature, and 57.34–61.65% RH for relative

humidity. For oxygen validation, 11 paired observations were obtained from simultaneous measurements with the reference oxygen analyzer.

The validation was conducted under ambient environmental conditions to assess the agreement between the prototype and the reference instruments prior to practical deployment. Controlled validation using certified gas concentrations near the alarm thresholds was beyond the scope of the present study. Future studies will include controlled chamber experiments using certified gas concentrations to further evaluate the system under abnormal gas conditions.

The experimental setup for environmental parameter validation is shown in Fig 4.

Statistical analysis

The measurement performance of the developed prototype was evaluated by comparing its measurements with those obtained from the reference instruments. Agreement between the prototype and reference measurements was assessed using the normalized root mean square difference (nRMSD) and normalized mean bias difference (nMBD). The nRMSD was used to describe the magnitude of overall measurement deviation, whereas the nMBD was used to describe systematic overestimation or underestimation relative to the reference measurements.

For each paired observation, xi represents the prototype measurement, yi represents the corresponding reference measurement, and N represents the number of paired observations. The root mean square difference

Fig 4. Experimental setup for side-by-side validation of carbon dioxide, temperature, and relative humidity measurements using the prototype and OIZOM reference instrument.

(RMSD) and mean bias were calculated from the paired measurements. For normalization, the denominator was the arithmetic mean of the reference measurements (ȳ) for the corresponding validation dataset. Thus, nRMSD was calculated as RMSD divided by ȳ and expressed as a percentage, while nMBD was calculated as mean bias divided by ȳ and expressed as a percentage (as shown in Equations 1 and 2).12 This normalization expresses the measurement deviation relative to the average reference measurement and allows comparison of relative measurement deviations across parameters with different units and scales.

Because the validation data were collected as time-series measurements under ambient conditions, the paired observations were not considered statistically independent. Therefore, nRMSD and nMBD were interpreted as descriptive measures of agreement rather than evidence of alarm performance. Uncertainty was estimated using a moving-block bootstrap procedure to account for temporal dependence in the paired observations, and 95% bootstrap confidence intervals were calculated for nRMSD and nMBD. The 95% limits of agreement

(LoA) were calculated from the paired measurement differences as the mean bias ± 1.96 standard deviations of the differences.

(1)

(2)

RESULTS

Validation of environmental parameters

Carbon dioxide, temperature, and relative humidity data were recorded at 2-min intervals and compared with measurements obtained from the OIZOM air quality monitoring system. Fig 5 illustrates the temporal variations and agreement between the developed prototype and the reference instrument.

The results presented in Fig 5 demonstrate that the developed prototype was capable of tracking temporal variations in environmental parameters with trends comparable to those of the reference instrument. Although some deviations were observed, particularly in carbon

800.00

700.00

600.00

Carbon dioxide (ppm)

500.00

400.00

300.00

200.00

100.00

14:46

14:50

14:54

14:58

15:02

15:06

15:10

15:14

15:18

15:22

15:26

15:30

15:34

15:38

15:42

15:46

15:50

15:54

15:58

16:02

16:06

16:10

16:14

16:18

16:22

16:26

16:30

16:34

16:38

16:42

16:46

16:50

16:54

16:58

0.00

Time

100.00

90.00

80.00

Relative Humidity (%)

70.00

60.00

50.00

40.00

30.00

20.00

10.00

14:46

14:50

14:54

14:58

15:02

15:06

15:10

15:14

15:18

15:22

15:26

15:30

15:34

15:38

15:42

15:46

15:50

15:54

15:58

16:02

16:06

16:10

16:14

16:18

16:22

16:26

16:30

16:34

16:38

16:42

16:46

16:50

16:54

16:58

0.00

Time

80.00

70.00

60.00

Temperature (oC)

50.00

40.00

30.00

Fig 5. Comparison between prototype and OIZOM reference instrument for (A) carbon dioxide,

(B) relative humidity, and (C) temperature.

20.00

10.00

14:46

14:50

14:54

14:58

15:02

15:06

15:10

15:14

15:18

15:22

15:26

15:30

15:34

15:38

15:42

15:46

15:50

15:54

15:58

16:02

16:06

16:10

16:14

16:18

16:22

16:26

16:30

16:34

16:38

16:42

16:46

16:50

16:54

16:58

0.00

Time

dioxide measurements, the prototype showed consistent response behavior throughout the monitoring period. For carbon dioxide measurement (Fig 5A), the reference instrument recorded concentrations ranging from 325.06 to 468.53 ppm, with a mean concentration of

410.71 ppm, while the prototype measurements ranged from 333.02 to 525.29 ppm. The prototype exhibited greater measurement variability and larger deviations compared with the reference system. The nRMSD and nMBD were 15.47% and -9.70%, respectively, corresponding to an absolute RMSE of 63.52 ppm and a mean bias of -39.85 ppm. Using a moving-block bootstrap to account for temporal dependence, the 95% confidence intervals were approximately 12.96-17.66% for nRMSD and -14.17 to

-5.30% for nMBD. The negative mean bias indicates that the prototype tended to underestimate carbon dioxide concentration relative to the reference instrument. These results indicate preliminary agreement under the observed ambient concentration range and should not be interpreted as validation of CO2 alarm performance at the predefined 5,000-ppm threshold.

In contrast, relative humidity and temperature measurements (Fig 5B and Fig 5C) demonstrated relatively close agreement with the reference instrument. The lower normalized RMSD (nRMSD) and normalized MBD (nMBD) values suggest that the DHT22 sensor provided stable environmental monitoring performance under ambient operating conditions. Minor differences between the datasets may be attributed to sensor positioning, airflow variation, and differences in sensor response time between the prototype and the reference device.

For relative humidity, the normalized RMSD (nRMSD) and normalized MBD (nMBD) were 5.39% and -3.00%, respectively, corresponding to an absolute RMSE of

3.18 %RH and a mean bias of -1.77 %RH. The negative

mean bias indicates a slight underestimation of relative humidity compared with the reference instrument.

For temperature, the normalized RMSD (nRMSD) and normalized MBD (nMBD) were 4.80% and 2.54%, respectively, corresponding to an absolute RMSE of 1.75

°C and a mean bias of 0.92 °C. The positive mean bias indicates a slight overestimation of temperature relative to the reference instrument.

Validation of oxygen measurement

Oxygen concentration data were recorded separately at 1-min intervals and compared with a standard oxygen analyzer. Fig 6 shows strong agreement between the developed prototype and the reference analyzer. The normalized RMSD (nRMSD) and normalized MBD (nMBD) were 0.085% and 0.283%, respectively, corresponding to an absolute RMSE of 0.017 %Vol and a mean bias of

0.056 %Vol.

The oxygen monitoring results shown in Fig 6 demonstrated very strong agreement between the developed prototype and the reference oxygen analyzer. The low normalized RMSD (0.085%) and normalized MBD (0.283%) values indicate good agreement between the prototype and the reference analyzer under ambient environmental conditions.

The measured oxygen concentrations remained close to ambient atmospheric levels with minimal fluctuation, and the prototype successfully reproduced these variations in a manner consistent with the reference instrument. These results suggest that the developed system has the potential for continuous oxygen monitoring under normal ambient conditions.

To further verify the response of the oxygen sensor under oxygen-enriched conditions, an additional functional verification was performed by introducing oxygen gas

24.00

22.00

20.00

Oxygen (%)

18.00

16.00

14.00

12.00

10.00

13.30 13.31 13.32 13.33 13.34 13.35 13.36 13.37 13.38 13.39 13.40

Time

Fig 6. Comparison of oxygen concentration measured by the prototype and reference oxygen analyzer.

into a closed chamber. A total of 15 paired observations were compared with the reference oxygen analyzer. The prototype showed good agreement with the reference instrument, with a normalized RMSD (nRMSD) of 2.35%, a normalized MBD (nMBD) of -2.33%, an absolute RMSE of 0.512 %Vol, and a mean bias of -0.508 %Vol. However, the oxygen concentration did not exceed the upper alarm threshold of 23.5 %Vol, and no oxygen-deficient condition below 19.5 %Vol was tested. Therefore, these results demonstrate the prototype’s ability to track increasing oxygen concentrations under the tested conditions but do not validate the upper or lower oxygen alarm thresholds or establish alarm sensitivity, response time, or false-negative performance.

Although the oxygen concentration did not exceed the upper alarm threshold (23.5 %Vol), the results confirmed that the prototype responded consistently to increasing oxygen concentrations under oxygen-enriched conditions. The overall measurement performance of the developed prototype compared with the reference instruments is

summarized in Table 3.

Table 4 summarizes the characteristics of the validation datasets and the agreement between the prototype and reference instruments. The table provides the observed reference measurement ranges, number of paired observations, reference instrument specifications, and 95% limits of agreement for each measured parameter. The results indicate that the validation data covered only ambient environmental conditions, with relatively narrow concentration ranges for carbon dioxide, temperature, and relative humidity, while the oxygen measurements were also obtained primarily under ambient conditions. These characteristics should be considered when interpreting the measurement agreement and the applicability of the results to conditions outside the tested ranges.

Field deployment in hospital environment

Following laboratory validation, the prototype was installed in a medical gas storage room at a practical hospital, as shown in Fig 7. The prototype was operated for approximately one week to demonstrate its performance under routine operating conditions. The sensing unit

TABLE 3. Normalized and absolute measurement errors of the developed prototype compared with the reference instruments.

TABLE 4. Summary of validation data characteristics and agreement between the prototype and reference instruments.

Fig 7. Installation of the developed monitoring system in a medical gas storage room at practical hospital.

was mounted on the wall at a height of approximately

1.8 m above the floor, close to the medical gas cylinder storage area. The location was selected as a practical monitoring point near the potential gas-release source while maintaining a fixed position that did not interfere with routine cylinder handling and storage operations. The storage room was a relatively enclosed space equipped with a single ventilation fan. The selected location was intended to provide representative monitoring of the local room environment; however, it was not intended to characterize spatial gas concentration gradients throughout the room. Gas distribution following a release may vary depending on the release location, cylinder arrangement, room geometry, and ventilation conditions. The placement was considered in the context of the need for ventilation and gas monitoring in enclosed medical gas storage areas. However, the present study did not evaluate spatial gas distribution or determine an optimal sensor location for the entire storage room. The deployment therefore demonstrated monitoring operation at the selected location rather than spatially resolved gas monitoring throughout the storage room.

The system successfully transmitted measurement data in real time via the NB-IoT network. Collected data were displayed on the Magellan IoT platform, allowing remote monitoring of carbon dioxide, oxygen, temperature, and relative humidity values (Fig 8). The platform also enabled historical data review and continuous observation of environmental conditions.

No reference instrument was operated alongside the prototype during the hospital deployment. Therefore, this phase of the study was considered an operational demonstration rather than field validation of sensor measurement performance. The purpose of the deployment was to demonstrate the practical operation of the proposed monitoring system, including real-time environmental monitoring and NB-IoT communication under routine hospital conditions. The prototype was operated for approximately one week and demonstrated measurement acquisition and transmission of environmental data to the cloud platform during the deployment period. As this deployment was intended primarily to demonstrate practical system operation rather than quantitative sensor validation, the measurements obtained during

Fig 8. Real-time monitoring dashboard on the Magellan IoT platform.

this phase were not used for statistical comparison with a reference instrument. The hospital deployment therefore provides a preliminary demonstration of the feasibility of implementing the proposed monitoring system under routine hospital conditions.

Alert function performance

The developed system demonstrated the basic operation of the real-time visual warning function during prototype testing. Under normal test conditions, the indicator light remained green when the measured gas concentrations were below the predefined warning thresholds. When the measured values exceeded the programmed thresholds during functional testing, the indicator changed from green to red.

DISCUSSION

The present study demonstrated that the developed low-cost IoT-based monitoring system was capable of measuring oxygen, carbon dioxide, temperature, and relative humidity with good agreement with the reference instruments under ambient environmental conditions and was capable of real-time monitoring of oxygen, carbon dioxide, temperature, and relative humidity in hospital environments. Among all tested parameters, oxygen concentration showed the highest agreement with the reference instrument, whereas carbon dioxide measurements exhibited greater deviation but maintained similar temporal trends.

In practical hospital settings, monitoring temporal changes in gas concentration may provide useful supplementary information even when measurement accuracy is limited. Although the carbon dioxide sensor showed higher measurement error, its measurements followed temporal variations in concentration under the tested ambient conditions. However, its performance for quantitative monitoring near the 5,000-ppm alarm threshold was not established. This limitation is particularly relevant for enclosed medical gas storage rooms, where ventilation conditions and potential gas leakage may result in changes in indoor gas concentrations.

The high agreement observed for oxygen measurement under ambient environmental conditions suggests that the proposed system has the potential for continuous oxygen monitoring. However, additional validation under oxygen concentrations near and beyond the alarm thresholds is required before confirming alarm performance under abnormal oxygen conditions.

Conventional studies have reported the use of low-cost gas sensing technologies based on semiconductor, electrochemical, and non-dispersive infrared (NDIR)

principles. Recent reviews have also highlighted the growing adoption of Internet of Things (IoT)-enabled environmental monitoring systems in hospitals, demonstrating their ability to provide continuous surveillance, remote data access, automated alerts, and improved environmental safety.13,14 However, most reported systems have primarily focused on general indoor environmental monitoring (e.g., temperature, humidity, and indoor air quality) or single-parameter sensing, with limited attention to continuous monitoring of medical gas storage rooms where multiple gas and environmental parameters must be monitored simultaneously. In contrast, the present system integrates oxygen, carbon dioxide, temperature, and relative humidity sensing with real-time wireless data transmission through an IoT platform, providing a practical and low-cost solution for hospital safety monitoring.

The successful installation and operation of the

prototype in a medical gas storage room at a practical hospital further demonstrate its practical feasibility. Real-time remote monitoring may facilitate continuous observation of environmental conditions and provide timely information to support safety management, improve response time, and strengthen occupational safety within hospital facilities.

The visual warning system may improve decision-making by providing immediate visual alerts when predefined alarm thresholds are exceeded before personnel enter enclosed storage areas.

In addition, the use of commercially available low-cost components may make this approach potentially attractive for hospitals in resource-limited settings, where access to conventional imported monitoring systems may be constrained by cost or procurement requirements. The estimated prototype component cost was lower than the listed prices of the commercial monitoring systems included in Table 5; however, these systems differ in sensor configurations, monitoring capabilities, certification, and procurement conditions. Therefore, the cost comparison should be interpreted as an indicative comparison of listed equipment prices rather than a like-for-like assessment of total system cost.

This study has several limitations. First, validation was conducted under ambient environmental conditions using side-by-side comparison with reference instruments rather than controlled gas concentrations. Consequently, the performance of the proposed system at carbon dioxide concentrations near the alarm threshold (5,000 ppm) and oxygen concentrations outside the normal environmental range (19.5–23.5%Vol) was not evaluated. Therefore, the present findings should be interpreted as evidence

TABLE 5. Approximate cost comparison between the developed prototype and commercially available systems.

Note: Prices shown are indicative listed prices obtained from the respective vendors or publicly available sources at the time of comparison and may vary according to sensor configuration, accessories, calibration, certification, shipping, installation, maintenance, and procurement conditions. The compared systems are not necessarily equivalent in terms of gas coverage, data logging, remote monitoring, certification, or service support.

of preliminary validation under ambient environmental conditions rather than comprehensive performance validation across the full operating range of the sensors. In addition, repeatability, response time, and long-term sensor stability were not investigated. Future studies should include controlled chamber experiments using certified gas concentrations to further evaluate alarm performance under abnormal environmental conditions. Future studies should focus on improving carbon dioxide sensor calibration, evaluating long-term field performance, integrating automated alert systems such as mobile notifications, and expanding the platform to

monitor additional medical gases.

CONCLUSIONS

The developed low-cost IoT-based medical gas monitoring system demonstrated preliminary measurement agreement with reference instruments for oxygen, carbon dioxide, temperature, and relative humidity under ambient environmental conditions. The prototype also demonstrated the feasibility of real-time data transmission and environmental monitoring in a hospital medical gas storage room. However, validation was limited to ambient conditions and did not cover the predefined alarm thresholds or the full intended decision range. Therefore, the present findings support the feasibility of the prototype as a low-cost environmental monitoring platform but do not establish its suitability as a safety alarm device. Further validation using controlled and certified gas concentrations, including concentrations around the relevant alarm thresholds, is required before safety-related deployment can be established.

Data Availability Statement

The data that support the findings of this study are not publicly available due to institutional and safety considerations. However, the data are available from the corresponding author upon reasonable request.

ACKNOWLEDGEMENTS

The authors would like to sincerely thank RFS Company Limited, Ramathibodi Hospital, Bangkok, Thailand, for kindly providing the facilities and technical guidance throughout this research work.

DECLARATIONS

Grants and Funding Information

This research is financially supported by the Thailand Science Research and Innovation (TSRI) National Science, Research and Innovation Fund (NSRF) through the Fundamental Fund (Fiscal Year 2026) at Christian University of Thailand (Grant Number 484613; Funding Approval No. FRB690050/0528).

Conflict of Interest

The research team declares no conflict of interest in this study.

Registration Number of Clinical Trial

Not applicable.

Author Contributions

Conceptualization and methodology, P.C and N.K.; Investigation, T.S and S.B.; Formal analysis, Visualization and writing – original draft, N.K. and P.C; Writing – review

and editing, S.B.; Supervision, P.C. All authors have read and agreed to the final version of the manuscript.

Use of Artificial Intelligence

During the preparation of this manuscript, the authors used ChatGPT to improve the language and readability of the manuscript. The authors subsequently reviewed and edited the generated content as necessary and take full responsibility for the final version submitted for publication.

Institutional Review Board Statement

This study was approved by the Human Research Ethics Committee of Christian University of Thailand (Approval No. บ. 08/2568).

REFERENCES

1. Azuma K, Kagi N, Yanagi U, Osawa H. Effects of low-level inhalation exposure to carbon dioxide in indoor environments: a short review on human health and psychomotor performance. Environ Int. 2018;121 (Pt1):51-56.

2. Satish U, Mendell MJ, Shekhar K, Hotchi T, Sullivan D, Streufert S, Fisk WJ. Is CO2 an indoor pollutant? Direct effects of low-to-moderate CO2 concentrations on human decision-making performance. Environ Health Perspect. 2012;120(12): 1671-7.

3. Divine T. Basics of NFPA 99 changes for hospital design: how engineers should navigate the changes to the 2018 edition of NFPA 99. Consult Specif Eng. 2020;57(5):10+.

4. Cherdchoo-ngarm P, Kolladaruengkri T, Chomchai S, Pisalsarakij D, Amornrat A, Deechoo T. Measurement of ethylene oxide gas and calculation of a safe period for contact at the medical sterilization unit in Siriraj Hospital. Siriraj Med J. 2004;56(3): 129-34.

5. Clark JM, Lambertsen CJ. Pulmonary oxygen toxicity: a review. Pharmacol Rev. 1971;23(2):37-133.

6. McArdle WD, Katch FI, Katch VL. Essentials of exercise physiology. 5th ed. Philadelphia: Wolters Kluwer; 2019.

7. Kwon J, Ahn G, Kim G, Kim JC, Kim H. A study on NDIR-

   based CO2 sensor to apply remote air quality monitoring system. In: Proceedings of the 2009 ICCAS-SICE; 2009 Aug; Fukuoka, Japan. p. 1683-7.

8. Santonico M, Pennazza G, Parente FR, Grasso S, Zompanti A, Stornelli V, et al. A gas sensor device for oxygen and carbon dioxide detection. Proceedings. 2017;1:447.

9. Brown SL, Goulsbra CS, Evans MG, Heath T, Shuttleworth E. Low cost CO2 sensing: a simple microcontroller approach with calibration and field use. HardwareX. 2020;8:e00136.

10. Yasuda T, Yonemura S, Tani A. Comparison of the characteristics of small commercial NDIR CO2 sensor models and development of a portable CO2 measurement device. Sensors. 2012;12:3641-55.

11. Sribudda S, Harnpicharnchai K, Nankongnab N, Thongkum W. Development of a hospital air quality monitoring application utilizing IoT technology and Google Workspace: A case study of a hospital in Mahasarakham, Thailand. Stud Health Technol Inform. 2026;338:409-13.

12. Reindl DT, Beckman WA, Duffie JA. Evaluation of hourly tilted surface radiation models. Sol Energy. 1990;45(1):9-17. doi: 10.1016/0038-092X(90)90061-G

13. Araújo T, Silva L, Aguiar A, Moreira A. Calibration assessment of low-cost carbon dioxide sensors using the extremely randomized trees algorithm. Sensors. 2023;23(13):6153.

14. Rajagopal S, Lyngdoh DC, Anathapadmanabhan UK, Soangra R, Vasan N. IoT-enabled solutions for environmental monitoring in hospitals. Open Biomed Eng J. 2023;17.

15. Comcube Co., Ltd. Extech Instruments CO260 CO2/CO data logger [Internet]. Bangkok: Comcube Co., Ltd.; [cited 2026 Apr 10]. Available from: https://www.comcube.co.th/product/co260/.

16. Henan Yuante Technology Co., Ltd. Sky2000 multi-gas detector [Internet]. Zhengzhou: Henan Yuante Technology Co., Ltd.; [cited 2026 Apr 10]. Available from: https://yuante.en.made-in-china.com/product/PSXxCKFVMwpa/China-CO2-Co-O2-Nh3-Multi-Gas-Detector-for-Chicken-House.html.

17. Advanced Energy. Innova 1314i highly accurate, stable, quantitative, and remotely controllable gas monitoring system [Internet]. Denver (CO): Advanced Energy Industries, Inc.; [cited 2026 Apr 10]. Available from: https://emin.com.mm/ advancedinnova1314i-advanced-energy-innova-1314i-gas-monitor-60cm3-60954/pr.html.

‌The Effect of Sit-To-Stand and Heel-Raise Exercises on Dynamic Balance Function in Sedentary Male Adolescents

Fifi Florensia, M.D.1,3,*, I Putu Alit Pawana, M.D.1,3, Yudith Dian Prawitri, M.D.2,3, Atika, B.Sc.4

1Department of Physical Medicine and Rehabilitation, Dr. Soetomo General Academic Hospital, Surabaya, Indonesia, 2Department of Physical Medicine and Rehabilitation, Airlangga University Hospital, Surabaya, Indonesia, 3Faculty of Medicine, Universitas Airlangga, Indonesia, 4Department of Public Health and Preventive Medicine, Faculty of Medicine, Universitas Airlangga, Surabaya, Indonesia.

*Corresponding author: Fifi Florensia E-mail: fififlorensiamd@gmail.com

Received 4 May 2026 Revised 5 July 2026 Accepted 8 July 2026 ORCID ID: http://orcid.org/0009-0001-1431-8066 https://doi.org/10.33192/smj.v78i9.282055

All material is licensed under terms of the Creative Commons Attribution 4.0 International (CC-BY-NC-ND 4.0) license unless otherwise stated.

INTRODUCTION

Adolescence represents a critical developmental window that shapes long-term health trajectories. Globally, individuals aged 10–24 years account for nearly a quarter of the population, making this group a strategic target for preventive health interventions. Investments in adolescent health are known to yield a “triple dividend”, benefiting current well-being, future adult health, and even the next generation.1,2 Despite being generally perceived as a healthy phase of life, many non-communicable diseases (NCDs) in adulthood originate from modifiable risk behaviors established during adolescence, particularly physical inactivity and sedentary lifestyles.3

In recent decades, sedentary behavior has emerged as a dominant lifestyle pattern among adolescents. The rapid expansion of digital technology has shifted daily activities toward prolonged sitting, including screen-based entertainment and academic tasks. Alarmingly, more than 80% of adolescents worldwide fail to meet recommended physical activity levels.4 This prevalence is even higher in Southeast Asia, reaching 73.4%, and in Indonesia, where it has increased from 83.6% in 2007 to 87.1% in 2015.5 Such behavioral shifts have profound implications for physical fitness, neuromuscular function, and overall motor development. Among these, dynamic balance, a key component of functional movement and injury prevention, appears particularly vulnerable to reduced physical activity.

Dynamic balance plays a central role in maintaining postural stability during movement and is an essential component of fundamental motor skills (FMS). Adequate

development of FMS during adolescence is strongly associated with higher levels of physical activity, better metabolic health, and improved motor competence.6,7 Conversely, insufficient motor skill development contributes to a cycle of inactivity and declining physical function. Empirical evidence indicates that adolescents with sedentary lifestyles exhibit poorer dynamic balance, increasing their risk of falls, postural abnormalities, and reduced functional capacity in daily activities.8,9

From a physiological perspective, prolonged sedentary behavior is associated with reduced muscle strength, impaired neuromuscular coordination, and altered biomechanical alignment. These changes are often accompanied by increased adiposity and decreased lean muscle mass, both of which negatively affect postural control and stability.10–12 Furthermore, the concept of Exercise Deficit Disorder (EDD) highlights a growing concern that many adolescents fail to achieve the minimum level of physical activity required for optimal musculoskeletal and neuromotor development, potentially leading to long-term health consequences.13

Given this context, effective and practical interventions are needed to improve neuromuscular function and dynamic balance in sedentary adolescents. Functional exercises have gained increasing attention due to their ability to mimic daily movements and simultaneously enhance strength, coordination, and motor control. Among these, the sit-to-stand (STS) exercise is widely recognized for its effectiveness in strengthening lower limb muscles and improving functional mobility across various populations.6,14–16 However, evidence regarding

its impact on balance outcomes remains inconsistent, suggesting that STS alone may not be sufficient to optimize postural control.17

To address this limitation, combining STS with heel-raise (HR) exercises may offer a more comprehensive approach. While STS primarily targets major muscle groups such as the quadriceps and gluteal muscles, heel-raise exercises specifically activate the plantar-flexor muscles, including the gastrocnemius and soleus, which are crucial for postural stability and balance control. Previous studies have shown that heel-raise training can increase soleus muscle thickness and plantar-flexor strength, contributing significantly to improved postural control.18 Additionally, such combined functional movements may enhance proprioceptive input and neuromuscular coordination, both of which are essential for maintaining dynamic stability.19–21

Despite these promising theoretical advantages, empirical evidence examining the combined effects of sit-to-stand and heel-raise exercises on dynamic balance in sedentary adolescents remains limited. Most existing studies focus on isolated interventions or adult populations, with little attention to integrated functional training approaches in adolescents. This gap highlights the need for targeted research to evaluate whether such combined exercises can effectively improve balance outcomes in this population.

Therefore, this study aims to investigate the effect of sit-to-stand and heel-raise exercises on dynamic balance function in sedentary male adolescents. By addressing this gap, the study contributes to the growing literature on adolescent motor development and offers a simple, low-cost, and scalable intervention that can be implemented in school or community settings to promote physical health and prevent future functional decline.

MATERIALS AND METHODS

This study employed a randomized single-blind controlled trial with a two-group pretest–posttest design. Participants were randomly allocated into an intervention group and a control group. Dynamic balance was assessed at baseline and after the intervention period using the Y-Balance Test. Outcome assessment was conducted by five independents, trained healthcare assessor who was blinded to group allocation.

The study was conducted between February and October 2025 at a senior high school in Surabaya, Indonesia. All exercise interventions and assessments were performed within the school environment under supervised conditions. The study population consisted of male adolescents aged 15–17 years. Permission to

conduct the study was obtained from the school principal before participant recruitment. Eligible participants were identified through an initial screening process. Students who met the criteria for sedentary behavior based on the International Physical Activity Questionnaire (IPAQ) and demonstrated adequate baseline static balance (one-leg stance >10 seconds) were invited to participate. Recruitment was voluntary, and no financial incentives were provided. Because all participants were minors, written informed consent was obtained from their parents or legal guardians, while written assent was obtained from each participant before enrollment. Participants were informed that they could withdraw from the study at any time without any academic or personal consequences. Physical activity was assessed using the International Physical Activity Questionnaire–Long Form (IPAQ-LF), which records the frequency (days/week) and duration (minutes/day) of physical activity performed during the previous seven days across four activity domains: work-related activity, transportation, domestic and household activity, and recreation, sport, and leisure-time activity. Sitting time was recorded separately. Following the IPAQ scoring protocol, walking, moderate-intensity activity, and vigorous-intensity activity were assigned metabolic equivalent values and expressed as MET-min/week. Domain-specific scores were calculated as the assigned MET value multiplied by the duration in minutes and frequency in days, and the total physical activity score was calculated by summing the scores across all domains. Participants were categorized as having low, moderate, or high physical activity levels according to the established IPAQ criteria. For this study, only participants classified in the low physical activity category. Moderate physical activity was defined as meeting at least one of the following criteria: (1) vigorous-intensity activity on at least 3 days for at least 20 minutes/day; (2) moderate-intensity activity and/or walking on at least 5 days for at least 30 minutes/day; or (3) any combination of walking, moderate-intensity activity, and vigorous-intensity activity on at least 5 days achieving a minimum total physical activity of 600 MET-min/week. High physical activity was defined as either: (1) vigorous-intensity activity on at least 3 days achieving at least 1,500 MET-min/week; or (2) any combination of walking, moderate-intensity activity, and vigorous-intensity activity on 7 days achieving at least 3,000 MET-min/week. Participants who did not meet the criteria for either category were classified as having low physical activity. In accordance with the standard IPAQ scoring protocol, walking, moderate-intensity activity, and vigorous-intensity activity were assigned values of

3.3, 4.0, and 8.0 METs, respectively. Exclusion criteria included a history of musculoskeletal, neurological, cardiovascular, or other medical conditions that could affect participation or outcomes. Participants who missed more than 20% of training sessions or withdrew consent were classified as dropouts.

The sample-size calculation used the formula

demonstrated the correct technique for both exercises. Blood pressure, heart rate, respiratory rate, body temperature, and peripheral oxygen saturation were measured 15 minutes before exercise. Participants then completed a standardized dynamic warm-up consisting of shoulder, trunk, hip, and knee flexion–extension; ankle dorsiflexion–plantar flexion; and ankle inversion–eversion. Each

n1 = n2 =

2SD2 (Z1/2α+Zβ)2 (X1-X2)

movement was performed for 10 seconds and repeated three times, followed by 1 minute of marching in place.

The previous study reported mean scores of 89.46

group 1 and 96.26 group 2, corresponding to an expected mean change of 6.80 points. Using an estimated variance of 53.86 (standard deviation, 7.34). With a two-sided α of 0.05 (1.96) and 80% power (0.84). The estimated minimum sample size was 9.13, which was rounded up to 10 participants per group. After adding a 20% allowance for attrition, the recruitment target was set at 12 participants per group. Thirty eligible participants were ultimately enrolled and allocated equally, resulting in 15 participants per group. The decision to enroll all 30 eligible participants was made before randomization and before the investigators examined any outcome data. Participants were allocated in a 1:1 ratio to the STS-HR intervention or control group using simple randomization. After all participants had been enrolled and assigned unique identification numbers, a researcher generated random numbers using the RAND() function in Microsoft Excel and sorted the identification numbers according to the generated values. The first 15 participants in the randomized sequence were allocated to the intervention group and the remaining 15 to the control group. No

block randomization or stratification was used.

The intervention group participated in a supervised combined sit-to-stand and heel-raise (STS-HR) exercise program three times per week for 8 consecutive weeks. The intervention was developed with reference to the American College of Sports Medicine guidelines. Each session consisted of a warm-up, three exercise sequences, and a cool-down. One exercise sequence comprised one set of sit-to-stand repetitions followed by one set of double-leg heel raises. Thus, sit-to-stand and heel-raise repetitions were counted separately rather than as a single combined repetition. Participants were instructed to obtain adequate rest and to refrain from strenuous exercise or heavy physical activity before each training and assessment session.

Preparation and warm-up

Participants were instructed to avoid strenuous exercise before each session and to wear sports clothing and barefoot. Before the first training session, the researcher

Sit-to-stand exercise

The sit-to-stand exercise was performed using a stable, armless chair with a seat height of approximately 45 cm. Participants began in an upright seated position with the hips fully supported on the chair, the knees flexed, and both feet placed flat and parallel on the floor approximately 10 cm behind the knees. The arms were crossed over the chest, and the gaze was directed forward. During the ascent phase, participants leaned the trunk slightly forward, pressed through the heels, and extended the knees and hips until reaching a fully upright standing position. During the descent phase, participants flexed the hips and knees and lowered the body under control until fully seated on the chair.

Heel-raise exercise

Immediately after completing each sit-to-stand set, participants performed the same number of double-leg heel raises. The starting position was upright standing with the feet parallel and the arms crossed over the chest. Participants raised both heels as high as possible to achieve their maximum comfortable plantar-flexion range while maintaining support through the metatarsophalangeal joints. The maximal heel-raised position was held for 2 seconds, after which the heels were lowered under control until they contacted the floor. External support was not used unless required for safety.

Cadence

Exercise cadence was controlled using a metronome set at 30 beats/min, corresponding to one beat every 2 seconds. Participants changed movement phase at each metronome beat. For the sit-to-stand exercise, the ascent and descent phases were each performed over approximately 2 seconds. For the heel raise, participants raised their heels over approximately 2 seconds, maintained the highest position for 2 seconds, and lowered their heels over approximately 2 seconds.

Volume and progression

Each training session included three sets of sit-to-

stand exercise and three sets of double-leg heel raises. The number of repetitions was increased according to a predetermined time-based progression: 8 repetitions per set during weeks 1–2, 10 repetitions during weeks 3–4, 13 repetitions during weeks 5–6, and 15 repetitions during weeks 7–8. Progression was determined by the intervention week and was not individualized according to participant performance. A participant progressed to the next training volume provided that the current volume could be completed without meeting any prespecified termination criterion.

Cool-down and monitoring

After completing the exercise, participants performed a standardized cool-down consisting of hamstring, quadriceps, calf, arm and shoulder, and spinal-rotation stretches, followed by deep breathing and relaxation. Each activity was performed for 30 seconds. Blood pressure, heart rate, respiratory rate, body temperature, and peripheral oxygen saturation were reassessed 15 minutes after the exercise. Heart rate and oxygen saturation were also monitored during training using a fingertip pulse oximeter.

Supervision and missed sessions

The intervention group received repeated supervision, physiological monitoring, exercise demonstrations, and direct contact with the research team, whereas the control group did not receive an attention-matched activity. All sessions were supervised to ensure proper technique and safety. Participants in the control group did not receive the STS-HR exercise program and were instructed to maintain their usual daily activities throughout the 8-week study period. Both groups received the same general education regarding physical activity recommendations based on WHO guidelines. However, the educational intervention was not designed as an attention-matched control for the supervised exercise sessions.

Outcome Measures

The original protocol identified Y-Balance Test performance as the study outcome but did not prespecify a single primary endpoint. For the revised principal exploratory analysis, the bilateral mean modified composite reach score at week 8 was selected. This outcome selection and the baseline-adjusted ANCOVA were introduced during manuscript revision and were therefore not prospectively prespecified. Direction-specific, limb-specific, change-score, and within-group analyses were treated as secondary exploratory analyses.

Modified Y-Balance Test

A field-based modified Y-Balance Test was conducted barefoot in a designated 3×3-m area using three floor-mounted measuring tapes arranged in the anterior, posteromedial, and posterolateral directions and separated by 120°. Before testing, the assessor explained and demonstrated the procedure. Participants completed six familiarization trials for each limb in each direction, followed by three recorded trials after an adequate rest period.

Participants stood on the test limb with the most distal aspect of the great toe positioned at the center of the grid. While maintaining single-leg stance, they reached as far as possible with the contralateral limb, lightly touched the measuring surface, and returned to the starting position under control. Testing was conducted in the following order: right anterior, left anterior, right posteromedial, left posteromedial, right posterolateral, and left posterolateral. The same order was used at baseline and follow-up.

A trial was considered invalid and repeated if the participant moved the stance foot or placed the reaching foot on the floor for support. Limb length was measured separately for each side from the anterior superior iliac spine to the most distal aspect of the medial malleolus while the participant was supine.

The bilateral mean modified composite score was calculated as the mean of the right- and left-limb modified composite scores. Higher values indicated better test performance. Because the mean rather than the maximum reach distance was used, this outcome differs from the standard YBT composite score and is therefore described as a modified composite reach score. All assessments were conducted by the same trained assessor using a standardized procedure. No study-specific intra-rater or inter-rater reliability assessment was conducted.

Other measurements

Age, height, body weight, BMI, limb length, leg dominance, baseline physical activity, and baseline dietary intake were recorded. Participants were also asked to document physical activity outside the intervention using activity diaries and dietary intake using food records. However, follow-up activity and dietary records were not quantitatively analyzed using a standardized scoring procedure. Therefore, the study could not determine reliably whether physical activity and dietary intake outside the intervention remained comparable between groups during follow-up. Baseline caloric intake, which was available from the baseline dietary assessment, was

included as a covariate in the revised exploratory ANCOVA. The unquantified follow-up activity and dietary records were not included in the model.

Statistical analysis

Baseline continuous characteristics were compared using independent-samples t-tests for approximately normally distributed variables and Mann–Whitney U tests otherwise. Baseline p-values were presented for descriptive and exploratory purposes only and were not used to determine whether randomization was successful, establish the absence of confounding, or select covariates. The revised principal exploratory analysis used ANCOVA to estimate the between-group difference in the week-8 bilateral mean modified composite reach score. Treatment group was entered as the fixed factor, with baseline bilateral mean modified composite score and baseline caloric intake entered as covariates. The treatment contrast was defined as STS-HR minus control, such that a positive estimate indicated a higher week-8 score in the STS-HR group. The adjusted between-group mean difference was reported in percentage points with its 95% CI, F statistic, p-value, and partial η².

Within-group comparisons were performed using

paired-samples t-tests when individual change scores were approximately normally distributed and Wilcoxon signed-rank tests otherwise. For paired-samples t-tests, standardized within-group changes were reported as Cohen’s, calculated as the mean individual change divided by the SD of the individual change scores. For Wilcoxon signed-rank tests, matched-pairs rank-biserial correlations were reported.

Unadjusted between-group comparisons of change scores were reported as mean differences with 95% CIs. Standardized between-group effects were expressed as Hedges’, calculated using the pooled SD of the individual change scores with a small-sample bias correction. The contrast was consistently defined as STS-HR minus control, so positive values favored the STS-HR group. Direction-specific, limb-specific, change-score, and within-group analyses were considered secondary and exploratory. No adjustment for multiple testing was applied; therefore, the corresponding p-values were interpreted descriptively. All tests were two-sided, with p<0.05 used as the nominal significance threshold. Exact p-values were reported to three decimal places, except values below 0.001, which were reported as p<0.001.

Ethical considerations and trial registration

The study protocol was approved by the Institutional Ethics Committee of the Faculty of Medicine, Universitas

Airlangga (Approval No. 206/EC/KEPK/FKUA/2025). The protocol was not published before participant enrollment and is available from the corresponding author upon reasonable request. A separate publicly available statistical analysis plan was not prepared. Permission to conduct the study was also obtained from the participating school prior to recruitment. Because all participants were minors, written informed consent was obtained from their parents or legal guardians, while written assent was obtained from each participant before enrollment. Confidentiality was maintained by assigning anonymous identification codes, and participants were free to withdraw from the study at any stage without penalty. All exercise sessions were supervised by trained physiotherapists to ensure participant safety throughout the intervention.

RESULTS

The results of this study are presented in three sections: baseline characteristics, within-group changes, and between-group comparisons of dynamic balance performance. All 15 intervention participants completed all 24 prescribed sessions, an adherence rate of 100%. No sessions were missed, no protocol deviations or withdrawals occurred, and no adverse events were observed or reported.

At baseline, total physical activity was 328.27 ±

179.67 MET-min/week in the control group and 346.87

± 154.42 MET-min/week in the intervention group. Baseline energy intake was 1,947.64 ± 242.92 and 1,965.83

± 239.39 kcal/day, respectively. Fat intake was 71.69 ±

16.68 versus 71.91 ± 20.77 g/day, carbohydrate intake

was 280.34 ± 57.96 versus 276.38 ± 48.95 g/day, and

protein intake was 52.76 ± 13.94 versus 52.50 ± 11.89 g/day. These baseline values are presented descriptively in Table 1.

Baseline demographic, anthropometric, physical activity, and dietary characteristics are summarized descriptively in Table 1. The study included 15 participants in each group, and all participants were right-leg dominant. Participants had low baseline physical activity levels, with mean values of 328.27 ± 179.67 MET-min/week in the control group and 346.87 ± 154.42 MET-min/week in the STS-HR group. Because this was a randomized trial, no hypothesis tests were used to assess baseline balance between groups. Potential baseline differences in the primary outcome were addressed through baseline-adjusted ANCOVA.

Within-group analysis (Table 2) showed no significant changes in the control group across all Y-Balance Test directions and limbs (p > 0.05), with small effect sizes indicating minimal change. In contrast, the intervention

TABLE 1. Baseline Characteristics, Physical Activity and Dietary Intake During the 8-Week Study Period.

Notes: Values are presented as mean ± SD, median (IQR), or n (%), as appropriate. BMI, body mass index; MET-min/week, metabolic equivalent minutes per week; SD, standard deviation.

a Mann Whitney U test; b Independent-samples t-test

TABLE 2. Within-Group comparison of dynamic balance function using the y-balance test after 8 weeks.

Note: Within-group effect sizes were calculated as Cohen’s (dz), using the standard deviation of the individual paired change scores as the denominator. For the non-normally distributed outcome analyzed using the Wilcoxon signed-rank test, the matched-pairs rank-biserial correlation (rrb)was reported instead.

group demonstrated significant improvements in all directions and both limbs (p = 0.001), accompanied by large effect sizes (d > 0.8). Notable gains in composite reach length reflect enhanced postural control and lower limb neuromuscular function, supporting the effectiveness of the sit-to-stand and heel-raise intervention.

In the unadjusted analysis of change scores (Table 3), the STS-HR group demonstrated greater improvements than the control group in all direction-specific Y-Balance Test outcomes. The between-group mean difference in change was 6.17 points for the right anterior direction (95% CI 2.46 to 9.88; p = 0.006; Hedges’ g = 1.108) and

9.73 points for the left anterior direction (95% CI 4.83 to 14.63; p < 0.001; Hedges’ g = 1.456). Corresponding differences were 10.02 points for the right posterolateral direction (95% CI 3.71 to 16.34; p = 0.006; Hedges’ g = 1.067), 10.34 points for the left posterolateral direction (95% CI 3.27 to 17.41; p = 0.006; Hedges’ g = 1.067),

11.91 points for the right posteromedial direction (95% CI 4.30 to 19.52; p = 0.003; Hedges’ g = 1.144), and

10.87 points for the left posteromedial direction (95% CI 4.91 to 16.83; p = 0.001; Hedges’ g = 1.326). These estimates were unadjusted for baseline scores and should be interpreted as exploratory. For the dominant/right-limb composite reach length, the mean change was 1.40

± 9.16 percentage points in the control group and 11.72

± 5.48 percentage points in the STS-HR group. The unadjusted between-group mean difference in change

was 10.32 percentage points (95% CI 4.62 to 16.02; p = 0.001), corresponding to a large standardized effect (Hedges’ g = 1.33). For the left-limb composite reach length, the mean change was 0.07 ± 7.90 percentage points in the control group and 12.51 ± 6.17 percentage points in the STS-HR group, yielding an unadjusted mean difference of 12.44 percentage points (95% CI 7.12 to 17.76; p < 0.001; Hedges’ g = 1.71). The left-limb analysis was considered exploratory.

After adjustment for baseline bilateral composite score and caloric intake, the STS-HR group had a higher bilateral mean composite Y-Balance Test score at week 8 than the control group. The adjusted between-group mean difference was 12.4 percentage points (95% CI

8.6 to 16.1; F(1,26) = 46.675; p < 0.001; partial η² =

0.642). Baseline composite score was also independently associated with the week-8 outcome (F(1,26) = 28.068; p < 0.001; partial η² = 0.519), whereas caloric intake was not (F(1,26) = 0.106; p = 0.747; partial η² = 0.004).

Safety and exercise perception

Exercise intensity was monitored using the Borg 6–20 Rating of Perceived Exertion scale. The recorded session-level RPE values ranged from 9 to 11, corresponding to very light to light perceived exertion. The RPE value remained at 9 during sessions 1–6, increased to 10 during sessions 7–18, and reached 11 from session 19 onward. These values were summarized descriptively because no

TABLE 3. Unadjusted between-group comparison of change scores in dynamic balance performance after 8 Weeks.

Note: Values are presented as mean change ± SD, where change was calculated as the week-8 value minus the baseline value. Unadjusted mean differences were calculated as the change in the STS-HR group minus the change in the control group; therefore, positive values favor the STS-HR group. Standardized between-group effect sizes were calculated as Hedges’ g using the pooled standard deviation of the individual change scores and a small-sample correction.

TABLE 4. Baseline- and calorie-adjusted ANCOVA for bilateral mean composite Y-Balance Test score at week 8.

Note: The group estimate represents the intervention group minus the control group. Positive values favor the STS-HR group. The outcome was expressed as a percentage; regression coefficients originally generated on the proportion scale were multiplied by 100 for reporting. Partial η² was calculated from the Type III F statistic and residual degrees of freedom.

Fig 1. Mean Change in Composite Reach Length (ΔCRL%) with 95% Confidence Intervals by Group and Limb

Mean changes in composite reach length (ΔCRL%) for the control and intervention groups are presented for both left and right limbs. Error bars represent 95% confidence intervals. The dashed horizontal line indicates no change (Δ = 0).

inferential analysis of RPE was prespecified. Therefore, variations in RPE should not be interpreted as evidence of physiological adaptation or improved exercise tolerance. In addition, RPE alone cannot establish the safety or broader feasibility of the intervention.

BORG Rating of Perceived Exertion (RPE) scores across 24 training sessions over 8 weeks. The training

program was progressively increased in repetitions: weeks 1–2 (8 repetitions), weeks 3–4 (10 repetitions),

weeks 5–6 (13 repetitions), and weeks 7–8 (15 repetitions). Fig 2 illustrates the recorded perceived exercise intensity throughout the 8-week intervention. Perceived exertion was assessed using the Borg 6–20 Rating of Perceived Exertion scale. The recorded RPE values ranged from

Fig 2. Session-specific Rating of Perceived Exertion during the intervention. RPE was assessed using the Borg 6–20 scale. Values of 9–11 correspond to very light to light perceived exertion. Data are presented as mean±SD for each session.

9 to 11, corresponding to very light to light perceived exertion. The RPE value remained at 9 during sessions 1–6, increased to 10 during sessions 7–18, and reached 11 during sessions 19–20. These values are presented descriptively and should not be interpreted as evidence of physiological adaptation or improved exercise tolerance. Small variations in the recorded RPE values across sessions were interpreted descriptively and were not used to infer physiological adaptation. Pain levels were assessed using the Wong–Baker Scale across 24 training sessions over 8 weeks. All participants consistently reported a pain score of 0 throughout the intervention period.

Fig 3 shows that no participants reported pain during the 8-week intervention, with Wong–Baker scores consistently at 0 across all sessions. This indicates that the sit-to-stand and heel-raise program was well tolerated and did not induce musculoskeletal discomfort despite progressive loading. No pain was reported during the monitored sessions. However, the small number of exposed participants and short follow-up preclude definitive conclusions regarding safety.

Participants’ perception of the exercise program

At the end of the intervention, all participants in the intervention group (n = 15) completed a perception questionnaire consisting of four Likert-scale statements. Overall, participants reported highly positive perceptions of the exercise program. A total of 53% strongly agreed and 40% agreed that the intervention was beneficial, while only 7% expressed uncertainty. Regarding ease of implementation and independence, 67% of participants strongly agreed and 33% agreed that the exercises were simple and could be performed independently. In terms of safety, the majority of participants perceived the intervention as safe, with 73% strongly agreeing and 27%

agreeing that the exercises did not cause serious adverse effects. Notably, 40% of participants strongly agreed and 27% agreed that they were willing to continue the exercise after the study, although 33% remained uncertain. Importantly, no participants reported disagreement or strong disagreement across any of the statements. Participants reported high acceptability, perceived safety, and ease of performing the intervention. However, willingness to continue the exercise was less consistent, with 67% agreeing or strongly agreeing and 33% remaining uncertain.

DISCUSSION

The absence of significant changes in the control group indicates that dynamic balance in sedentary male adolescents remains relatively stable in the absence of specific training stimuli. This finding suggests that time alone (8 weeks) is insufficient to induce neuromuscular adaptations that meaningfully affect balance performance. This result is consistent with previous studies, such as Hoch et al.22, which reported no significant changes in Y-Balance Test (YBT) scores following a period of routine activity without targeted intervention, and Linek et al.25, who demonstrated high reliability of YBT measurements (ICC > 0.85) in the absence of structured training. Therefore, the lack of improvement in the control group reinforces the notion that the postural control system in adolescents requires specific and targeted stimuli to improve. Furthermore, this finding supports the internal validity of the study by confirming that improvements observed in the intervention group are not attributable to measurement variability or learning effects. Conceptually, this highlights that general physical activity without a structured balance component is insufficient to produce significant improvements in postural control.

Fig 3. Pain Assessment Using the Wong–Baker Scale Across Training Sessions

Mechanisms underlying the effectiveness of STS-HR training

The significant improvements observed in the intervention group can be explained by neuromuscular adaptations involving enhanced muscle strength, motor coordination, and sensorimotor integration. The sit-to-stand (STS) component primarily activates large extensor muscle groups, such as the quadriceps and gluteal muscles, which are essential for controlling vertical movement and stabilizing the center of mass. Meanwhile, the heel-raise (HR) component specifically targets the plantar-flexor muscles (gastrocnemius–soleus complex), which play a critical role in maintaining postural stability during dynamic tasks. The integration of these two functional movements creates a training stimulus that closely resembles daily activities, thereby improving the efficiency of dynamic postural control. These findings are consistent with Lizama-Pérez et al.14, who reported improvements in muscle strength (29–38%) and muscle quality (up to 45%) following STS training, and with Muehlbauer et al.24, who demonstrated significant improvements in YBT performance after structured balance training. Moreover, the improvements observed across all reach directions (anterior, posteromedial, and posterolateral) indicate that the adaptations were global rather than direction-specific. This suggests that STS-HR training enhances not only localized muscle strength but also the coordination between multiple body segments required for maintaining dynamic stability.

Role of Proprioception and Sensorimotor Efficiency

In addition to strength gains, the observed improvements may also be attributed to enhanced proprioceptive sensitivity and more efficient sensorimotor integration. Repetitive functional movements such as STS-HR provide continuous stimulation to proprioceptive receptors in muscles and joints, thereby improving the nervous system’s ability to detect body position and respond to balance perturbations. Linek et al.25 emphasized that improvements in YBT performance are closely associated with enhanced control of the center of gravity and improved neuromuscular coordination. Furthermore, Borgmann et al.26 demonstrated that reductions in YBT performance are primarily associated with lower limb muscle fatigue, suggesting that improvements observed in this study may also reflect enhanced neuromuscular endurance. In this context, STS-HR training appears to improve not only force production capacity but also the efficiency of energy utilization during postural control. This dual adaptation likely explains the large effect sizes and consistent improvements observed among participants.

Such adaptations are particularly relevant in sedentary individuals, who typically exhibit deficits in sensorimotor control.

Magnitude of between-group differences

The STS-HR group demonstratedasizable improvement in Y-Balance Test performance compared with the control group. After adjustment for baseline bilateral composite score and caloric intake, the between-group difference in the bilateral mean modified composite reach score was

12.4 percentage points (95% CI 8.6 to 16.1; p<0.001). However, because an established minimal important difference for this modified outcome in sedentary male adolescents is unavailable, the clinical importance of the observed difference remains uncertain. The low baseline physical activity level of the participants may have provided greater opportunity for improvement, although this explanation was not directly tested. Ortega et al.27 noted that MET values below 12 reflect suboptimal aerobic fitness, which is associated with impaired postural function. Under such conditions, even relatively simple interventions can produce substantial functional gains. Additionally, the progressive overload applied in the training protocol likely facilitated gradual adaptation without inducing excessive fatigue. This is supported by Schedler et al.20, who demonstrated that increased training volume significantly improves YBT performance. Therefore, the large effect sizes observed in this study reflect both the responsiveness of the sedentary population and the effectiveness of the intervention design.

Safety and adherence of the STS-HR intervention

Complete attendance and the absence of observed adverse events indicate that the intervention was well tolerated by the 15 participants who received it. However, the small number of exposed participants and the short intervention period preclude definitive conclusions regarding safety. This can be attributed to the nature of the exercises, which are classified as closed-chain, bodyweight-based movements that promote stable load distribution and reduce injury risk. Lizama-Pérez et al.14 similarly reported no adverse effects associated with STS training in sedentary populations. Moreover, the exercise intensity remained within a light-to-moderate range (Borg scale 9–11), allowing for physiological adaptation without excessive fatigue. Borgmann et al.26 further emphasized that impairments in balance performance are more commonly associated with high-intensity fatigue rather than moderate, repetitive training. The simplicity of the STS-HR protocol, requiring no specialized equipment and closely resembling daily movements, likely contributed

to both its safety and high participant acceptance. These findings suggest that STS-HR training is not only effective but also highly feasible for implementation in school-based or community settings.

Implications of the study

This study provides preliminary evidence that an 8-week sit-to-stand and heel-raise intervention can improve Y-Balance Test performance in sedentary male adolescents. Because the exercises require minimal equipment, the program may be practical for supervised implementation in school or community settings. However, the present study did not measure muscle strength, neuromuscular adaptation, functional decline, or lower-extremity injury incidence. Therefore, the findings should not be interpreted as evidence that the intervention prevents functional decline or injury. Whether improvements in Y-Balance Test performance translate into broader functional benefits or a lower risk of injury should be examined in larger and longer-term trials that directly measure these outcomes.

Limitations of the study

The findings are limited to a small sample of sedentary male adolescents from a single Catholic senior high school in Surabaya who were required to maintain a one-leg stance for more than 10 seconds. The exclusion of female adolescents, the single-school setting, specific selection criteria, and local educational and cultural context limit generalizability to broader adolescent populations.

Because the stored dataset contained the average rather than the maximum of the three reach trials, the scores could not be recalculated using the standard YBT scoring method. This deviation may limit comparability with published normative and composite YBT values.

The study was not prospectively registered, and a publicly accessible statistical analysis plan was unavailable before participant enrollment. Consequently, prospective verification of the prespecified primary outcome and analytical decisions is not possible, which increases the risk of selective outcome reporting. The results should therefore be interpreted with appropriate caution.

Several limitations should be considered when interpreting the adjusted analysis. Formal diagnostic testing of the ANCOVA assumptions, including linearity between the covariates and outcome, homogeneity of regression slopes, normality and homoscedasticity of residuals, and the absence of influential observations, was not performed. Consequently, it cannot be confirmed that all model assumptions were satisfied. Violations of these assumptions could affect the validity of the estimated

adjusted between-group difference, its confidence interval, and the corresponding significance test. The ANCOVA findings should therefore be regarded as exploratory and interpreted cautiously.

CONCLUSIONS

This study demonstrates that an 8-week sit-to-stand and heel-raise (STS-HR) program significantly improves dynamic balance in sedentary male adolescents. Consistent improvements across all Y-Balance Test directions, along with large effect sizes, indicate clinically meaningful functional gains. These effects are likely driven by enhanced neuromuscular coordination, lower limb strength, and sensorimotor integration. The lack of change in the control group confirms that improvements were attributable to the intervention. The program was safe and feasible, with full adherence and no adverse events. Overall, the 8-week STS-HR program improved Y-Balance Test performance in the sedentary male adolescents included in this study. Its effects on functional decline, daily physical functioning, and lower-extremity injury risk remain unknown and require investigation in larger, longer-term trials.

Data Availability Statement

The data supporting the findings of this study are not publicly available due to ethical and privacy considerations involving human participants. However, the data are available from the corresponding author upon reasonable request and with permission from the institutional ethics committee.

ACKNOWLEDGEMENTS

The authors would like to express their sincere gratitude to the staff of the Department of Physical Medicine and Rehabilitation, Dr. Soetomo General Academic Hospital, for their support throughout this study. We also thank the students, principal, teachers, and staff of Frateran Catholic Senior High School, Surabaya, Indonesia, for their participation, cooperation, and assistance with participant recruitment, intervention implementation, and administrative support during this research.

DECLARATIONS

Grants and Funding Information

Dr. Soetomo Regional General Hospital’s Research Fund (No.Ref: 100.3.3/166/102.6/2026), the funder had no involvement in the design, execution, or analysis of the research, and did not influence the interpretation of the results.

Conflict of Interest

All authors declare no conflict of interest in this research.

Registration Number of Clinical Trial

This trial was not prospectively registered in a publicly accessible clinical trial registry before participant enrollment. The authors acknowledge this omission as a limitation of the study.

Author Contributions

Conceptualization and methodology, F.F., I.P.A.P., and Y.D.P.; Investigation, F.F.; Formal analysis, F.F., I.P.A.P., and A.; Visualization and writing – original draft, F.F.; Writing – review and editing, F.F., I.P.A.P., Y.D.P., and A.; Supervision, I.P.A.P. and Y.D.P. All authors have read and agreed to the final version of the manuscript.

Use of Artificial Intelligence

Artificial intelligence tools (ChatGPT 5.2 and Grammarly) were used only for language editing and proofreading; all scientific content, analysis, and interpretations were developed by the authors.

Institutional Review Board Statement

The study protocol was approved by the Institutional Ethics Committee of the Faculty of Medicine, Universitas Airlangga (Approval No. 206/EC/KEPK/FKUA/2025).

REFERENCES

1. Patton GC, Sawyer SM, Santelli JS, Ross DA, Afifi R, Allen NB, et al. Our future: a Lancet commission on adolescent health and wellbeing. Lancet. 2016;387(10036):2423–78.

2. WHO. Global accelerated action for the health of adolescents (AA-HA!): guidance to support country implementation. Geneva; 2017.

3. van Sluijs EMF, Ekelund U, Crochemore-Silva I, Guthold R, Ha A, Lubans D, et al. Physical activity behaviours in adolescence: current evidence and opportunities for intervention. Lancet. 2021;398(10298):429–42.

4. WHO. Guidelines on physical activity and sedentary behaviour. Geneva; 2020.

5. WHO. Global action plan on physical activity 2018–2030: more active people for a healthier world. Geneva; 2018.

6. Barnett LM, Lai SK, Veldman SLC, Hardy LL, Cliff DP, Morgan PJ, et al. Correlates of Gross Motor Competence in Children and Adolescents: A Systematic Review and Meta-Analysis. Sports Medicine. 2016;46(11):1663–88.

7. Stodden DF, Goodway JD, Langendorfer SJ, Roberton MA, Rudisill ME, Garcia C, et al. A Developmental Perspective on the Role of Motor Skill Competence in Physical Activity: An Emergent Relationship. Quest. 2008;60(2):290–306.

8. Schwiertz G, Brueckner D, Schedler S, Kiss R, Muehlbauer T.

   Performance and reliability of the Lower Quarter Y Balance Test in healthy adolescents from grade 6 to 11. Gait Posture. 2019; 67:142–6.

9. Balzanelli C, Spataro D, Redaelli de Zinis LO. Prevalence of Pediatric and Adolescent Balance Disorders: Analysis of a Mono-Institutional Series of 472 Patients. Children. 2021;8(11):1056.

10. Vongsirinavarat M, Kraiwong R. Falls among Older Adults with Type 2 Diabetes Mellitus with Peripheral Neuropathy. Siriraj Med J. 2021;72(2):92–8.

11. Yin L, Qin J, Chen Y, Xie J, Hong C, Huang J, et al. Impact of Body Mass Index on Static Postural Control in Adults With and Without Diabetes: A Cross-Sectional Study. Front Endocrinol (Lausanne). 2021;12:768185.

12. Marijančić V, Kezele TG, Peharec S, Dragaš-Zubalj N, Pavičić Žeželj S, Starčević-Klasan G. Relationship between Physical Activity and Sedentary Behavior, Spinal Curvatures, Endurance and Balance of the Trunk Muscles-Extended Physical Health Analysis in Young Adults. Int J Environ Res Public Health. 2023;20(20): 6938.

13. Faigenbaum AD, Stracciolini A, Myer GD. Exercise deficit disorder in youth: a hidden truth. Acta Paediatr. 2011 Nov;100(11):1423–5; discussion 1425.

14. Lizama-Pérez R, Chirosa-Ríos LJ, Contreras-Díaz G, Jerez-Mayorga D, Jiménez-Lupión D, Chirosa-Ríos IJ. Effect of sit-to-stand-based training on muscle quality in sedentary adults: a randomized controlled trial. PeerJ. 2023;11:e15665.

15. Jones CJ, Rikli RE, Beam WC. A 30-s chair-stand test as a measure of lower body strength in community-residing older adults. Res Q Exerc Sport. 1999;70(2):113–9.

16. Chaiwan P, Srisawasdi G, Prachgosin P, Aekwattanaphol P, Jirapongpathai N. Effects of Exercise Through Telerehabilitation on Balance and Walking Speed in Older Adults with Diabetic Peripheral Neuropathy: A Randomized Controlled Trial. Siriraj Med J. 2025;77(6):403–10.

17. Chaovalit S, Taylor NF, Dodd KJ. Sit-to-stand exercise programs improve sit-to-stand performance in people with physical impairments due to health conditions: a systematic review and meta-analysis. Disabil Rehabil. 2020;42(9):1202–11.

18. Fujiwara K, Toyama H, Asai H, Yaguchi C, Irei M, Naka M, et al. Effects of regular heel-raise training aimed at the soleus muscle on dynamic balance associated with arm movement in elderly women. J Strength Cond Res. 2011;25(9):2605–15.

19. Pavão SL, Dos Santos AN, Cicuto Ferreira Rocha NA. Sex and age influence on postural sway during sit-to-stand movement in children and adolescents: Cross-sectional study. Int J Dev Neurosci. 2021;81(6):520–8.

20. Schedler S, Graf SM, Muehlbauer T. Effects of Different Balance Training Volumes on Children’s Dynamic Balance. Sports Med Int Open. 2022;6(01):E32–8.

21. Tepa W, Lertwanich P, Chuensiri N. Effects of Supervised Plyometric Training on Neuromuscular and Knee Functions for Late Phase Rehabilitation Program in Patients with Anterior Cruciate Ligament Reconstruction: A Randomized Controlled Trial. Siriraj Med J. 2024;76(6):353–65.

22. Hoch MC, Welsch LA, Hartley EM, Powden CJ, Hoch JM. Y-Balance Test Performance After a Competitive Field Hockey Season: A Pretest-Posttest Study. J Sport Rehabil. 2017;26(5).

23. Lowe L, Castillo F, Gokun Y, Williams DK, Israel M, Yates C. Static and Dynamic Balance Assessmentin Healthy and Concussed

    
    

    Adolescent Athletes. Clin J Sport Med. 2022;32(4):385–90.

24. Muehlbauer T, Grundmann A, Vortkamp L, Schedler S. Effect of balance training on static and dynamic balance performance in male adolescents: role of training frequency. BMC Res Notes. 2022;15(1):365.

25. Linek P, Sikora D, Wolny T, Saulicz E. Reliability and number of trials of Y Balance Test in adolescent athletes. Musculoskelet Sci Pract. 2017;31:72–5.

26. Borgmann K, Brinkmann R, Bauer J, Hill MW, Muehlbauer T. Effect of Lower Limb Muscle Fatigue on Dynamic Balance Performance in Healthy Young Adults: Role of Arm Movement. Sports Med Int Open. 2024;8:a23462759.

27. Ortega FB, Artero EG, Ruiz JR, España-Romero V, Jiménez-Pavón D, Vicente-Rodriguez G, et al. Physical fitness levels among European adolescents: the HELENA study. Br J Sports Med. 2011;45(1): 20–9.

‌ Corrigendum

SMJ

Corrigendum to: Sleep Habits and Behavioral Problems in Preschool-aged Children with ADHD

Prakasit Wannapaschaiyong, M.D., Amornrat Penphattarakul, M.D., Pat Rojmahamongkol, M.D., Sureelak Sutchritpongsa, M.D.

Department of Pediatrics, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand.

This corrigendum relates to the following article: Wannapaschaiyong P, Penphattarakul A, Rojmahamongkol P, Sutchritpongsa S. Sleep Habits and Behavioral Problems in Preschool-aged Children with ADHD. Siriraj Med J. 2024;76(3):116–124. DOI: https://doi.org/10.33192/smj.v76i3.267462

URL: https://he02.tci-thaijo.org/index.php/sirirajmedj/article/view/267462

The author would like to correct the error in the publication of the original article1 and has informed the editor to correct as follows:

Content of Correction:

Incorrect: The COA number in the article: COA no. Si 719/2023 (IRB3)

Correct:

COA no. Si 719/2023

REFERENCE

1. Wannapaschaiyong P, Penphattarakul A, Rojmahamongkol P, Sutchritpongsa S. Sleep Habits and Behavioral Problems in Preschool-aged Children with ADHD. Siriraj Med J. 2024;76(3):116–

124. https://doi.org/10.33192/smj.v76i3.267462

Corresponding author: Amornrat Penphattarakul

E-mail: penphattarakul.nass@gmail.com

All material is licensed under terms of the Creative Commons Attribution 4.0 International (CC-BY-NC-ND 4.0) license unless otherwise stated.

‌Corrigendum to: The Relationship Between Primary Caregivers’ Psychosocial Factors and Self-esteem in Children and Adolescents with ADHD:

An Exploratory Cross-sectional Study

Prakasit Wannapaschaiyong, M.D., Amornrat Penphattarakul, M.D., Pat Rojmahamongkol, M.D., Sureelak Sutchritpongsa, M.D.

Department of Pediatrics, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand.

This corrigendum relates to the following article: Wannapaschaiyong P, Penphattarakul A, Rojmahamongkol P, Sutchritpongsa S. The Relationship Between Primary Caregivers’ Psychosocial Factors and Self-Esteem in Children and Adolescents with ADHD: An Exploratory Cross-sectional Study. Siriraj Med J. 2023;75(8):584–591.DOI: https://doi.org/10.33192/smj.v75i8.263447

URL: https://he02.tci-thaijo.org/index.php/sirirajmedj/article/view/263447

The author would like to correct the error in the publication of the original article1 and has informed the editor to correct as follows:

Content of Correction:

Incorrect: The COA number in the article: COA no. Si 661/2022 (IRB1)

Correct:

COA no. Si 661/2022

REFERENCE

1. Wannapaschaiyong P, Penphattarakul A, Rojmahamongkol P, Sutchritpongsa S. The Relationship Between Primary Caregivers’ Psychosocial Factors and Self-Esteem in Children and Adolescents with ADHD: An Exploratory Cross-sectional Study. Siriraj Med

J. 2023;75(8):584–591. https://doi.org/10.33192/smj.v75i8.263447

Corresponding author: Sureelak Sutchritpongsa

E-mail: ssureelaknok@gmail.com

All material is licensed under terms of the Creative Commons Attribution 4.0 International (CC-BY-NC-ND 4.0) license unless otherwise stated.