Volume 74, No.4: 2022 Siriraj Medical Journal
https://he02.tci-thaijo.org/index.php/sirirajmedj/index
256
Nisachon Khorwanichakij, M.D.*, Smith Kungwankiattichai, M.D.**, Weerapat Owattanapanich, M.D.**
*Division of Hematology, Department of Medicine, Chaophraya Yommarat Hospital, Suphanburi 72000, ailand, **Division of Hematology,
Department of Medicine, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok 10700, ailand.
Validation of Several Formulas to Differentiate
Thalassemia from Iron Deciency Anemia and
Proposal of a Thalassemia–Iron Deciency
Discrimination (TID) Predictive Score
ABSTRACT
Objective: is study aimed to validate the sensitivity analysis of all the available formulas for their ability to
dierentiate between IDA and thalassemia and propose a novel formula to improve the sensitivity of all thalassemia
subtypes screening.
Materials and Methods: We conducted a 5-year, single-center, Cohort study on 227 microcytic anemia patients
diagnosed between June 2015 and September 2020 at Chaophraya Yommarat Hospital, Suphanburi, ailand to
validate the sensitivity of all the available formulas and invent the novel predictive score.
Results: Approximately three-quarters of our cases were all subtypes of thalassemia diseases while 26.9% were IDA.
e sensitivity of almost all the previous formulas for thalassemia prediction ranged between 13.9%-44.0%, while
the specicity varied between 0%–98.4%. Nevertheless, the sensitivity of the formulas that had favorable sensitivity
was quite low. Here, a novel thalassemia–iron deciency discrimination (TID) predictive score is proposed, which
demonstrated a sensitivity of 90.4% the specicity of 78.7%, the positive predictive value of 92.0 %, the negative
predictive value of 75.0%, and the accuracy of 87.2%.
Conclusion: e proposed TID predictive score is a novel uncomplicated formulation which oers high sensitivity
for all thalassemia subtypes prediction.
Keywords: Iron deciency anemia; microcytic anemia; predictive score; thalassemia (Siriraj Med J 2022; 74: 256-265)
Corresponding author: Weerapat Owattanapanich
E-mail: weerapato36733@gmail.com
Received 27 January 2022 Revised 18 February 2022 Accepted 23 February 2022
ORCID ID: https://orcid.org/0000-0002-1262-2005
http://dx.doi.org/10.33192/Smj.2022.32
INTRODUCTION
According to the World Health Organization, iron
deciency anemia (IDA) is the major cause of nutritional
anemia worldwide.
1
e incidence of IDA in ai women
of reproductive age was reported to be 28.7%, 30.2%, and
31.8%, in 2013, 2014, and 2015, respectively.
1
Another
study reported an anemia rate of 21% in educated young
ai women, with the two most prevalent causes among
those cases being thalassemia (28%) and IDA (21%).
2
Patients with IDA and thalassemia may both
present with microcytic anemia (defined as a mean
corpuscular volume (MCV) < 80 fL), which should be
further investigated to distinguish between these two
entities due to their dierent treatment approaches.
Iron supplementation and the correction of occult blood
loss remain the standard treatments for IDA. On the
other hand, certain types of thalassemia diseases, such
as hemoglobin (Hb) E/β-thalassemia and homozygous
Khorwanichakij et al.
Volume 74, No.4: 2022 Siriraj Medical Journal
https://he02.tci-thaijo.org/index.php/sirirajmedj/index
257
Original Article
SMJ
β-thalassemia, require regular blood transfusion and iron
chelation to prevent iron deposition in various organs,
which could lead to multiple organ dysfunction, i.e.,
liver cirrhosis, endocrinopathies, and heart failure.
3
Detailed evaluation to conrm IDA involves an
iron study test, consisting of measuring the levels of
serum ferritin, serum iron, total iron-binding capacity
(TIBC), and transferrin saturation (TSAT). If the serum
ferritin value is ≤30 ng/mL or the TSAT value is <16%,
IDA diagnosis can be conrmed with high sensitivity
and high specicity.
4
Meanwhile, Hb typing is employed
for the diagnosis of thalassemia.
5
However, in developing
countries, some primary hospitals have limited resources
to manage iron studies and Hb typing test. As a result, to
make a diagnosis, blood samples have to be transferred
to a comprehensive laboratory center, which can be a
time-consuming process. As such, other available tools
to initially discriminate these conditions could be of
value. For thalassemia diagnosis, a range of associated
factors can be assessed to develop a thalassemia predictive
score.
According to the extensive literature review we
performed, several formulas exist for thalassemia prediction
among microcytic anemic patients, including the Red
Blood Cell Count (RBC), Red Cell Distribution Width
(RDW), Red Cell Distribution Width Index (RDWI),
Green and King formula, Srivastava formula, Mentzer
formula, Ehsani formula, Ricerca formula, England
and Fraser formula, Sirdah formula, and Shine and Lal
formula.
6-14
e reported sensitivity and specicity of these
formulas range from 40% to 100%.
6-18
Another formula,
the 11T score is an interesting formula that combines
11 other formulas to calculate its score, providing a
higher discrimination ability.
15-18
In a previous study that
attempted to validate this score among a ai population,
the 11T score showed a sensitivity of 82.1% and specicity
of 91.7% for thalassemia prediction. However, it should
be noted that only β-thalassemia subtype was included
in previous studies. In addition, IDA in those trials was
diagnosed when serum ferritin was <10 ng/mL.
18
In the present study, we aimed to validate the sensitivity
assessment of all the available formulas for their ability to
dierentiate between IDA and all thalassemia subtypes.
In addition, we propose a novel formula to improve
the sensitivity of thalassemia screening. In addition,
to increase the diagnostic sensitivity in our study, the
diagnosis of IDA could be established when ferritin was
<30 ng/mL and TSAT was <16%.
4
MATERIALS AND METHODS
Study design and population
We conducted a 5-year, retrospective, single-
center, cohort study on microcytic anemia patients
diagnosed between June 1, 2015, and September 30,
2020, at Chaophraya Yommarat Hospital, Suphanburi,
ailand. e inclusion criteria were: (1) patients aged
15 years old or older, and (2) patients with microcytic
anemia. (3) patients who had the result of iron study
in the IDA group and Hb typing and/or PCR in the
thalassemia group. e exclusion criteria were patients
receiving erythropoiesis-stimulating agents or receiving
iron supplementation before blood testing. We categorized
patients into 2 groups by dierent timeframes; a group
for internal validation using patients during June 2015
- August 2017 and a group for calculation score using
patients during September 2017 - September 2020.
e study was approved for registration in the ai
Clinical Trial Registry with the identication number
TCTR20210725003.
Instrument and evaluation parameters
All blood samples were collected by using 3-ml
dipotassium ethylenediaminetetraacetic acid tubes
(K
2
EDTA) for a complete blood count (CBC) test and
analyzed within 2 hours aer taking the samples by
Mindray BC-6200 automated blood counter (Mindray
Bio‐Medical Electronics Co., Ltd, Shenzhen, China).
is device used impedance technology to count and
size RBC and platelet (PLT) together with cyanotic-free
colorimetric method for Hb. MCV and % RDW were
calculated based on the RBC histogram. In addition, mean
corpuscular hemoglobin (MCH) and mean corpuscular
hemoglobin concentration (MCHC) was also calculated
from RBC, Hb, and hematocrit parameters. e patient’s
demographic data and initial laboratory results were collected.
Patients with microcytic anemia were classied into two
groups: the IDA group and the thalassemia group. In the
thalassemia group, three included thalassemia disease
subtypes were as follows: α-thalassemia, β–thalassemia
disease, and α- combined β-thalassemia disease.
Study size consideration
At least 200 microcytic anemia cases were required
to validate the formulas and develop a novel predictive
score. Furthermore, 150 patients (40 patients with IDA and
110 patients with thalassemia) were separately assigned
for an internal validation of this score.
Handing of continuous predictors
e proposed predictive score was developed followed
by the predictive model study Risk of Bias Assessment
Tool (PROBAST). Four red blood cell parameters were
incorporated for this predictive score calculation including
MCH, RDW, RBC and PLT.
Volume 74, No.4: 2022 Siriraj Medical Journal
https://he02.tci-thaijo.org/index.php/sirirajmedj/index
258
Terminology
Anemia is dened by a hemoglobin (Hb) level <
13/dL in males or Hb < 12 g/dL in females.
19
Anemia
with small red blood cells (MCV < 80 fL) is termed
microcytic anemia.
4
A diagnosis of IDA is established if
a patient has microcytic anemia with serum ferritin < 30
ng/mL and transferrin saturation < 16%.
4
e 11T score
is a summary score from 11 formulas, comprising RBC
(×10
12
/L), RDW, RDWI (RDW × MCV/RBC), Green
and King formula (MCV2 × RDW/Hb × 100), Srivastava
formula (MCH/RBC), Mentzer formula (MCV/RBC),
Ehsani formula [MCV - (10 × RBC)], Ricerca formula
(RDW/RBC), England and Fraser formula [MCV - RBC
- (5 × Hb) - 3.4], Sirdah formula [MCV - RBC - (3 ×
Hb)], and Shine and Lal formula (MCV2 × MCH/100).
18
Statistical analysis
PASW Statistics for Windows, version 18.0 (SPSS
Inc., Chicago, IL, USA) was applied for the data analyses.
e patients’ demographic and clinical characteristics were
summarized descriptively by causes of microcytic anemia.
Continuous variables were reported as the mean±standard
deviation for normally distributed continuous variables,
and the median with interquartile ranges (Q1, Q3) for
nonnormally distributed continuous variables. Categorical
variables were reported as the frequency and percentage
and were compared using Fisher’s exact test or chi-square
test. Continuous variables were compared using the
Student’s t-test or Mann–Whiney U test. e univariate
and multivariate predictors of thalassemia were estimated
using Cox proportional hazards analysis (backward
stepwise method) and presented as an odds ratio (OR)
and 95% condence interval (CI). e receiver operating
characteristic (ROC) curve for the cuto score and for
thalassemia diagnosis was presented as the area under the
curve (AUC), accuracy, sensitivity, specicity, positive
predictive value (PPV), and negative predictive value
(NPV). For all the tests performed, a two-tailed p-value
< 0.05 was considered to be statistically signicant. e
calibration belt model was used for model calibration.
e model was attended by the Hosmer –Lemeshow χ
2
goodness-of-t test.
Ethics approval and consent to participate
is study was approved by the Ethics Committee for
Research in Human Subjects at Chaophraya Yommarat
Hospital, Suphanburi, ailand. All procedures followed
were in accordance with the ethical standards of the
responsible committee on human experimentation and
with the Helsinki Declaration of 1975, as revised in 2008.
Informed consent was waived due to a retrospective
study.
RESULTS
Baseline patient characteristics
In total, 227 microcytic anemic patients were included
in this study. Approximately three-quarters (73.1%)
were diagnosed with thalassemia disease, including
Hb E/β-thalassemia, homozygous β-thalassemia, Hb
H disease, Hb H/CS disease, AE Bart’s disease, and EF
Bart’s disease, whereas 61 patients (26.9%) were IDA.
In the thalassemic group, the mean patient age
was 42.1±20.5 years old. e mean Hb and MCV were
8±1.7 g/dL and 61.7±9.2 fL, respectively. e median
PLT count was 308,000/µL (range, 189,000-413,000/
µL). Among the IDA group, the mean patient age was
57.6±18.3 years old. e mean Hb was 6.1±1.7g/dL and
the mean MCV was 62.9±8.3 fL. e median PLT count
and serum ferritin were 384,000/µL (range, 263,000-
478,000/µL) and 16.7 ng/mL (range, 4-22.2 ng/mL),
respectively. Several factors were signicantly dierent
between the thalassemic and the IDA groups, such as age,
body mass index, Hb level, MCH, MCHC, red blood cell
distribution width (RDW), red blood cell counts, PLT
count, and iron proles. Table 1 displays the baseline
patient features and initial laboratory results of the
thalassemic and IDA patients.
Validation of the previous formulas predicting thalassemia
We analyzed the sensitivity, specificity, PPV,
NPV, and accuracy of each previous formula to predict
thalassemia, including RBC, % RDW, RDWI, Green and
King, Srivastava, Mentzer, Ehsani, Ricerca, England,
and Fraser, Shine and Lal, and 11T score, by using the
included patient’s data in this study. e sensitivity of
almost all the formulas ranged between 13.9% - 44%.
Only the RDW and Shine and Lal formulas yielded
high sensitivity (97.6%), but with low specicity results,
with gures of 0% and 3.3%, respectively. e specicity
of each formula varied between 0%-98.4%. e high
specicity of above 90% was found with several formulas,
including RDWI, Green and King, England and Fraser,
Sirdah, and 11T score; unfortunately, the sensitivity of
these formulas was quite low. e PPV of almost all the
formulas provided high results, which were above 90%.
In contrast, the NPV of all the formulas was as low as
approximately 30% (range, 0%-36.3%). e accuracy
of each formula varied between 36.1%-72.3%. Table 2
demonstrates the sensitivity, specicity, PPV, NPV, and
accuracy of each formula for predicting thalassemia.
Subgroup analysis of the formulas for predicting each
type of thalassemia
We performed a subgroup analysis of each thalassemia
subtype, including β-thalassemia disease, α-thalassemia
Khorwanichakij et al.
Original Article
SMJ
Volume 74, No.4: 2022 Siriraj Medical Journal
https://he02.tci-thaijo.org/index.php/sirirajmedj/index
259
TABLE 1. Baseline patient features and initial laboratory results of the thalassemic and iron deciency anemia patients.
Parameters Total Thalassemia β-thalassemia α-thalassemia α-thalassemia Irondeciency P-valueformultiplecomparisons
(N=227) (All) (1)(N=89) (2)(N=62) combinedwith anemia(0)
(N=166) (39.2%) (27.3%) β-thalassemia (N=61)
(73.1%) (3)(N=15) (26.9%) 1vs.0 2vs.0 3vs.0 Allvs.0
(6.6%)
Age(mean±SD) 46.3±21 42.1±20.5 42.8±20.4 41.5±21.5 40.9±17.1 57.6±18.3 <0.001 <0.001 0.002 <0.001
(years)
Sex(Male) 74(32.6%) 54(32.5%) 33(37.1%) 18(29%) 3(20%) 20(32.8%) 0.589 0.652 0.531 0.971
BMI 21.4±4.1 20.8±3.6 20.8±3.9 20.3±3.1 22.5±3.3 23.2±4.9 0.001 <0.001 0.602 0.001
Hemoglobintyping 181 (79.7%) 166(100%) 89(100%) 62(100%) 15(100%) 15(24.6) <0.001 <0.001 <0.001 <0.001
PCRfor 14(6.2%) 14(8.4%) 4(4.5%) 3(4.8%) 7(46.7%) 0(0%) 0.146 0.244 <0.001 0.024
α-thalassemia
Comorbidities 77(33.9%) 44(26.5%) 26(29.2%) 15(24.2%) 3(20%) 33(54.1%) 0.002 0.001 0.018 <0.001
Hypertension 34(15%) 16(9.6%) 12(13.5%) 3(4.8%) 1(6.7%) 18(29.5%) 0.016 <0.001 0.097 <0.001
Diabetes 22(9.7%) 12(7.2%) 10(11.2%) 1(1.6%) 1(6.7%) 10(16.4%) 0.361 0.004 0.682 0.039
Dyslipidemia 15(6.6%) 6(3.6%) 5(5.6%) 0(0%) 1(6.7%) 9(14.8%) 0.059 0.001 0.676 0.005
CAD 11(4.8%) 10(6%) 4(4.5%) 5(8.1%) 1(6.7%) 1(1.6%) 0.649 2.07 0.358 0.296
CKD 11(4.8%) 9(5.4%) 8(9%) 1(1.6%) 0(0%) 2(3.3%) 0.202 0.619 1.00 0.732
Liver disease 8(3.5%) 7(4.2%) 5(5.6%) 1(1.6%) 1(6.7%) 1(1.6%) 0.402 1.00 0.358 0.686
Arthritis 7(3.1%) 6(3.6%) 4(4.5%) 2(3.2%) 0(0%) 1(1.6%) 0.649 1.00 1.00 0.678
Others 19(8.4%) 7(4.2%) 2(2.2%) 5(8.1%) 0(0%) 12(19.7%) <0.001 0.062 0.109 <0.001
Volume 74, No.4: 2022 Siriraj Medical Journal
https://he02.tci-thaijo.org/index.php/sirirajmedj/index
260
TABLE 1. Baseline patient features and initial laboratory results of the thalassemic and iron deciency anemia patients. (Continue)
Parameters Total Thalassemia β-thalassemia α-thalassemia α-thalassemia Irondeciency P-valueformultiplecomparisons
(N=227) (All) (1)(N=89) (2)(N=62) combinedwith anemia(0)
(N=166) (39.2%) (27.3%) β-thalassemia (N=61)
(73.1%) (3)(N=15) (26.9%) 1vs.0 2vs.0 3vs.0 Allvs.0
(6.6%)
Laboratory
CBC
Mean±SD
Hemoglobin (g/dl) 7.5±1.9 8±1.7 7.9±1.9 8±1.5 8.5±1.7 6.1±1.7 <0.001 <0.001 <0.001 <0.001
Hematocrit (%) 25±14.3 26.4±16.2 26.2±21.7 26.9±4.4 25.9±6.3 21.2±5 0.082 <0.001 0.003 0.015
MCV (fL) 62±8.9 61.7±9.2 61.3±8.9 63.7±9.5 56.1±7.3 62.9±8.3 0.286 0.605 0.005 0.404
MCH (pg) 18.8±3 19.2±2.8 19.9±3 18.6±2.2 17.3±2.4 17.7±3.3 <0.001 0.066 0.674 0.001
MCHC (g/dl) 31.3±14.7 32.5±17 32.3±2.5 33.2±27.8 30.9±2.5 28.0±2.2 <0.001 0.148 <0.001 0.042
RDW (%) 23.6±5.4 24.7±5.7 24.5±6.3 25.2±5 23.8±4.7 20.4±3 <0.001 <0.001 0.016 <0.001
RBC (x10
12
/L) 4±1 4.2±1 4±1 4.3±0.9 4.9±0.9 3.4±0.8 <0.001 <0.001 <0.001 <0.001
Median±IQR
WBC (cells/µL) 7,410 7,595 7,860 6,900 7,750 6,320 0.009 0.495 0.096 0.031
Platelet(/µL) 316,000 308,000 307,000 298,000 353,000 384,000 0.015 0.006 0.759 0.006
(224,000- (189,000- (172,000- (202,000- (298,000- (263,000-
436,000) 413,000) 409,000) 401,000) 470,000) 478,000)
Ironstudy
Median±IQR
serum ferritin 249 590 882.5 383.6 548 16.7 <0.001 <0.001 <0.001 <0.001
(ng/ml) (8.96-859) (259-1,427) (323-2387.5) (184-759) (268.7-1499) (4-22.2)
serum iron (µg/dl) 17 69 66 70.5 79 12 <0.001 <0.001 0.004 <0.001
(11-51) (44-104) (41-104) (47.0-89) (49.0-105) (10-15)
Transferrin 4.9 27.5 26.3 32.9 36 3.5 <0.001 <0.001 <0.001 <0.001
saturation (%) (3.2-24.2) (19.1-44) (16.6-46.8) (23-44) (23-38.6) (2.5-4.6)
Mean±SD
TIBC (µg/dl) 322.7±90.2 250.6±80.3 258.1±96.8 241.9±57.7 235.7±31.8 371.3±58.6 <0.001 <0.001 0.027 <0.001
Abbreviations: MCV = Mean corpuscular volume, MCH = Mean corpuscular hemoglobin, MCHC = Mean corpuscular hemoglobin concentration, RDW = Red blood cell
distribution width, RBC = Red blood cell, WBC = White blood cell, TIBC = Total iron binding capacity (µg/dL), TSAT = Transferin saturation (%).
Khorwanichakij et al.
Volume 74, No.4: 2022 Siriraj Medical Journal
https://he02.tci-thaijo.org/index.php/sirirajmedj/index
261
Original Article
SMJ
TABLE 2. Sensitivity, specicity, positive predictive value, negative predictive value, and accuracy of each formula
to predict thalassemia.
Formula Cutoff Thalassemia Iron Sensitivity Specicity PPV NPV Accuracy
n(%) deciency (95%CI) (95%CI) (95%CI) (95%CI) (95%CI)
anemia
n (%)
RBC(x10
12
/L) ≥5 36(97.3) 1(2.7) 21.7% 98.4% 97.3% 31.6% 42.3%
<5 130(68.4) 60(31.6) (15.7-28.7) (91.2-100) (85.8-99.9) (25-38.7) (35.9-48.7)
RDW(%) ≥14 162(72.6) 61(27.4) 97.6% 0% 72.6% 0% 71.4%
<14 4(100) 0(0) (93.9-99.3) (0-5.9) (66.3-78.4) (0-60.2) (65.48-77.25)
RDWI[6] <220 23(92) 2(8) 13.9% 96.7% 92% 29.2% 36.1%
≥220 143(70.8) 59(29.2) (9-20.1) (88.7-99.6) (74-99) (23-36) (29.9-42.4)
GreenandKing[7] <72 27(96.4) 1(3.6) 16.3% 98.4% 96.4% 30.2% 38.3%
≥72 139(69.8) 60(30.2) (11-22.8) (91.2-100) (81.7-99.9) (23.9-37) (32-44.7)
Srivastava[8] <3.8 56(84.8) 10(15.2) 33.7% 83.6% 84.8% 31.7% 47.1%
≥3.8 110(68.3) 51(31.7) (26.6-41.5) (71.9-91.8) (73.9-92.5) (24.6-39.5) (40.6-53.6)
Mentzer[9] <13 65(90.3) 7(9.7) 39.2% 88.5% 90.3% 34.8% 52.4%
≥13 101(65.2) 54(34.8) (31.7-47) (77.8-95.3) (81-96) (27.4-42.9) (45.9-58.9)
Ehsani[10] <15 73(90.1) 8(9.9) 44% 86.9% 90.1% 36.3% 55.5%
≥15 93(63.7) 53(36.3) (36.3-51.9) (75.8-94.2) (81.5-95.6) (28.5-44.7) (49-62)
Ricerca[11] <4.4 35(81.4) 8(18.6) 21.1% 86.9% 81.4% 28.8% 38.8%
≥4.4 131(71.2) 53(28.8) (15.1-28.1) (75.8-94.2) (66.6-91.6) (22.4-35.9) (32.4-45.1)
England[12] <0 24(96) 1(4) 14.5% 98.4% 96% 29.7% 37%
andFraser ≥0 142(70.3) 60(29.7) (9.5-20.7) (91.2-100) (79.6-99.9) (23.5-36.5) (30.7-43.3)
Sirdah[13] <27 63(94) 4(6) 38% 93.4% 94% 35.6% 52.9%
≥27 103(64.4) 57(35.6) (30.5-45.8) (84.1-98.2) (85.4-98.3) (28.2-43.6) (46.4-59.4)
ShineandLal[14] <1530 162(73.3) 59(26.7) 97.6% 3.3% 73.3% 33.3% 72.3%
≥1530 4(66.7) 2(33.3) (93.9-99.3) (0.4-11.3) (67-79) (4.3-77.7) (66.4-78.1)
11Tscore[16] ≥7 40(97.6) 1(2.4) 24.1% 98.4% 97.6% 32.3% 44.1%
<7 126(67.7) 60(32.3) (17.8-31.3) (91.2-100) (87.1-99.9) (25.6-39.5) (37.6-50.5)
11Tscore(cutoff5) ≥5 64(88.9) 8(11.1) 38.6 86.9 88.9 34.2 51.5
<5 102(65.8) 53(34.2) (31.1-46.4) (75.8-94.2) (79.3-95.1) (26.8-42.2) (45.0-58.0)
11Tscore(cutoff6) ≥6 58(95.1) 3(4.9) 34.9 95.1 95.1 34.9 51.1
<6 108(65.1) 58(34.9) (27.7-42.7) (86.3-99.0) (86.3-99.0) (27.7-42.7) (44.6-57.6)
11Tscore(cutoff8) ≥8 30(96.8) 1(3.2) 18.1 98.4 96.8 30.6 36.7
<8 136(69.4) 60(30.6) (12.5-24.8) (91.2-100.0) (83.3-99.9) (24.2-37.6) (33.3-46.0)
11Tscore(cutoff9) ≥9 20(95.2) 1(4.8) 12.0 98.4 95.2 29.1 35.2
<9 146(70.9) 60(29.1) (7.5-18.0) (91.2-100.0) (76.2-99.9) (23.0-35.8) (29.0-41.5)
Abbreviations: PPV = Positive predictive value, NPV = Negative predictive value, RBC = Red blood cell count interval, RDW = Red blood
cell distribution width, RDWI = Red blood cell distribution width index.
Khorwanichakij et al.
Volume 74, No.4: 2022 Siriraj Medical Journal
https://he02.tci-thaijo.org/index.php/sirirajmedj/index
262
disease, and β-thalassemia combined with α-thalassemia
disease. In the subgroup of the β-thalassemia group, the
results were not signicantly dierent from in the full
analysis. Similarly, the results remained similar to the full
analysis in the α-thalassemia disease group. However,
when we validated the formulas in the β-thalassemia
combined with α-thalassemia disease patients, the NPV
and the accuracy of almost all the formulas were better
than in the full analysis, with gures ranging between
79.0%-93.0%, while the sensitivity and specicity were
not dierent from in the full analysis.
Proposed novel thalassemia–iron deciency discrimination
(TID) predictive score
Because the validation of each previous formula
was imperfect, we attempted to determine the signicant
factors to dierentiate between thalassemic and IDA
patients. We found that MCH, % RDW, RBC, and PLT
were signicant factors related to thalassemia. erefore,
the predictive score is calculated by using these factors
as the following:
y = (-4.643) + (2.273 if MCH 17 to 20) + (3.888 if
MCH > 20) + (2.025 if RDW 21 to 25) + (4.986 if RDW
> 25) + (0.485 if RBC 3.5 to 4.5) + (4.787 if RBC > 4.5) +
(0.785 if PLT < 265,000) + (1 if PLT 265,000 to 400,000)
Subsequently the TID predictive score was simplied
by multiplying with 2 as the following:
y = (-9) + (5 if MCH 17 to 20) + (8 if MCH > 20)
+ (4 if RDW 21 to 25) + (10 if RDW > 25) + (1 if RBC
3.5 to 4.5) + (10 if RBC > 4.5) + (2 if PLT < 265,000) +
(2 if PLT 265,000 to 400,000) (Fig 1)
We used the ROC analysis for the TID predictive
score. e most appropriate cuto level for predicting
thalassemia was ≥2, in which the AUC was 0.93 (95%
CI: 0.890 - 0.969; Fig 2). e sensitivity and specicity
of the score were 90.4% and 78.7%, respectively (Table 3).
Internal validation of the TID predictive score
e split 150 sample proles were utilized for the
internal validation study of the TID predictive score. e
AUC was 0.88 (95% CI: 0.815 - 0.947). e sensitivity
to predict thalassemia from the internal validation was
96.4%, with the specicity and the accuracy of 50.0%
and 84.0%, respectively. ere was a non-statistically
signicant dierence in AUCs between the predictive
TID predictive score creation and the internal validation
(P-value = 0.805).
DISCUSSION
e most common cause of microcytic anemia
in developed countries is IDA. However, thalassemia
should not be overlooked in patients with microcytic
anemia, especially in Asian populations.
20-21
In Southeast
Asian subjects, the prevalence of α-thalassemia among
anemic patients is 20%–30%, and β-thalassemia is about
3%–9%.
20-21
A CBC is a worthwhile initial investigation
that can be performed in every hospital. Although CBC
is a simple test, it gives an instant result, but it cannot
totally dierentiate the cause of microcytic anemia.
Hence, conrmation tests, such as iron study, Hb typing,
and PCR for α-thalassemia, are still mandatory for a
denitive diagnosis. However, such conrmation tests
are invariably more sophisticated, quite expensive, and
can take several days to several weeks to get the results
back. So these are not always practical in developing
countries with limited resources.
Several formulas have been developed to predict
thalassemia and used as a screening tool for thalassemia
diagnosis. For example, predictive formulas, such as
Green and King, Srivastava, and Mentzer have shown a
sensitivity of 87.7%–93.8% and specicity of 82.5%–95%
according to the previous results.
7-9,15
e 11Tscore was
developed to improve the sensitivity and specicity for
improving thalassemia diagnosis.
15
It is composed of 11
predictive formulas.
15-18
A previous study from France
found it had a sensitivity of 85.7% and specicity of 97.5%,
15
while the study from ailand reported a sensitivity of
82.1% and specicity of 91.7%.
18
However, the 11Tscore
has been applied for predicting only the β-thalassemia
subtype.
15-18
Another study reported that the Jayabose
RDW index, the Green and King formula, and the Janel
11T score are good formulas to dierentiate thalassemia
trait from IDA among their population.
22
Moreover,
the serum ferritin cuto value from previous studies
for IDA diagnosis varied between <10 ng/ml and <16
ng/ml.
16-18
Currently, the denition for IDA diagnosis is
serum ferritin < 30 ng/ml and TSAT < 16 %.
4
therefore,
we dened IDA according to this recent suggestion in
this study.
In our study, we validated all the previous formulas
with all subtypes of thalassemia, including α-thalassemia
disease, β-thalassemia disease, α-thalassemia combined
with β-thalassemia disease, and IDA patients. In contrast
to the previous results, the sensitivity and specicity
to predict thalassemia disease among these included
patients were not high. Therefore, we proposed a
novel TID predictive score composed of 4 red blood
cell indices, namely % RDW, RBC, and PLT count.
e TID predictive score had a sensitivity of 90.4%
and specicity of 78.7% to dierentiate all thalassemia
subtypes from IDA. Although, the specicity from the
internal validation was insignicantly lower compared
Khorwanichakij et al.
Volume 74, No.4: 2022 Siriraj Medical Journal
https://he02.tci-thaijo.org/index.php/sirirajmedj/index
263
Original Article
SMJ
Fig 1. e thalassemia–iron deciency discrimination (TID) predictive score and their values.
Fig 2. Receiver operating characteristic curve and the area under the curve for obtaining the cut o value for thalassemia prediction using
the TID predictive score (A) all thalassemia subtypes (B) β-thalassemia disease (C) α-thalassemia disease (D) α-thalassemia combined with
β-thalassemia
Volume 74, No.4: 2022 Siriraj Medical Journal
https://he02.tci-thaijo.org/index.php/sirirajmedj/index
264
to the gure from the predictive score generation, the
sensitivity remained satisfying. e TID predictive score
applies % RDW for calculation because in ailand CBC
report is practically presented with % RDW. However, a
recent study showed that absolute RDW is more specic
to dierentiate thalassemia from IDA in microcytic
anemia comparing with relative RDW.
23
This score might be beneficial for thalassemia
screening, whereby patients who have a score ≥2 can be
selected for further investigation to conrm thalassemia
disease. is could reduce unnecessary expenses from
over investigation, which would be especially important
in resource-limited countries. In other words, patients
who have a lower likelihood of having thalassemia as
assessed from the predictive score could be treated as
IDA while waiting for their iron study.
ere are some limitations of this study to note.
First, because this was a retrospective study, some
information may have been missing. Second, the TID
predictive score demonstrated a specicity of 50.0% from
the internal validation, some IDA patients who have high
TID predictive scores might experience a treatment delay
while awaiting for Hb typing result. ird, IDA patients
from this cohort had signicantly more severe anemia
than the thalassemia subjects. is factor might inuence
the sensitivity/specicity of discriminant formulas. Lastly,
in the subgroup analysis, the size of some subgroups is
quite small which leads to imprecise formula validation.
CONCLUSION
alassemia and IDA are the most common causes
of microcytic anemia. Here, a TID predictive score
was proposed that demonstrated higher sensitivity for
thalassemia prediction while remaining uncomplicated
to be applied due to its few involved parameters.
ACKNOWLEDGEMENTS
The authors are grateful to the Department of
Medicine, Chaophraya Yommarat Hospital, Suphanburi,
ailand and the Faculty of Medicine Siriraj Hospital,
Mahidol University, ailand for grant support. We
also thank Assoc.Prof. Preechaya Wongkrajang and Dr.
Ratikorn Anusorntanawat for their valuable consultation
and advice and thank Ms. Pattaraporn Tunsing and
Ms. Kemajira Karaketklang for the data collection and
statistical analyses.
Conicts of interest: e authors conrm that there
are no known conicts of interest associated with this
publication.
Funding: is study was funded by grants from the
Department of Medicine, Chaophraya Yommarat Hospital,
Suphanburi, ailand and the Faculty of Medicine Siriraj
Hospital, Mahidol University, Bangkok, ailand.
REFERENCES
1. WHO. Trends in anemia in women and children: 1995 to
2016; WHO 2017; Available from http://www.WHO.int/data/
gho/data/indicators/indicator-details/GHO/prevalence-of-
anemia-in-women-of-reprodutive-age.
2. Brimson S, Suwanwong Y, Brimson JM. Nutritional anemia
predominant form of anemia in educated young ai women.Ethn
Health. 2019;24(4):405-14. doi:10.1080/13557858.2017.1346
188
3. Taher AT, Saliba AN. Iron overload in thalassemia: dierent
organs at dierent rates.Hematology Am Soc Hematol Educ
Program. 2017;2017(1):265-71. doi:10.1182/asheducation-2017.1.265
4. Camaschella C. Iron-deficiency anemia.N Engl J Med.
2015;372(19):1832-43. doi:10.1056/NEJMra1401038
5. Brancaleoni V, Di Pierro E, Motta I, Cappellini MD. Laboratory
diagnosis of thalassemia.Int J Lab Hematol. 2016;38 Suppl
1:32-40. doi:10.1111/ijlh.12527
Khorwanichakij et al.
TABLE 3. Sensitivity, specicity, positive predictive value, negative predictive value, and accuracy of the TID predictive
score.
Logisticmodel Cutoff Thalassemia Iron Sensitivity Specicity PPV NPV Accuracy
n(%) deciency (95%CI) (95%CI) (95%CI) (95%CI) (95%CI)
anemia
n (%)
Logisticmodelscore ≥2 150(92) 13(8) 90.4% 78.7% 92.0% 75.0% 87.2%
<2 16(25) 48(75) (84.8-88.1) (66.3-88.1) (86.7-95.7) (62.6-85.0) (82.9-91.6)
Abbreviations: PPV = Positive predictive value, NPV = Negative predictive value.
Volume 74, No.4: 2022 Siriraj Medical Journal
https://he02.tci-thaijo.org/index.php/sirirajmedj/index
265
Original Article
SMJ
6. JayaboseS,GiavanelliJ,Levendoglu O, Sandoval C, Ozkayak F,
Visintainer P. Dierentiating iron deciency anemia from
thalassemia minor by using an RDW-based index.J Pediatr
Hematol Oncol. 1999;21:314.
7. Green R, King R. A new red cell discriminant incorporating
volume dispersion for dierentiating iron deciency anemia
from thalassemia minor.Blood Cells. 1989;15(3):481-95.
8. Srivastava PC, Bevington JM. Iron deciency and-or thalassaemia
trait.Lancet. 1973;1(7807):832. doi:10.1016/s0140-6736(73)
90637-5
9. Mentzer WC Jr. Dierentiation of iron deciency from thalassaemia
trait.Lancet. 1973;1(7808):882. doi:10.1016/s0140-6736(73)
91446-3
10. Ehsani M, Darvish A, Eslani A, Seighali F. A new formula for
dierentiation of iron deciency anemia (IDA) and thalassemia
trait (TT). Turk J Hematol 2005;22(Suppl): 268.
11. Ricerca BM, Storti S, d’Onofrio G, Mancini S, Vittori M,
Campisi S, et al. Dierentiation of iron deciency from thalassaemia
trait: a new approach.Haematologica. 1987;72(5):409-13.
12. EnglandJM,BainBJ,FraserPM.Dierentiation of iron deciency
from thalassemia trait by routine blood‐count.Lancet.1973;1:449-
52.
13. SirdahM,TaraziI,Al NajjarE,Al HaddadR.Evaluation of
the diagnostic reliability of dierent RBC indices and formulas
in the dierentiation of theβ‐thalassemia minor from iron
deciency in Palestinian population.Int J Lab Hematol2007;30:
324-30.
14. Shine I, Lal S. A strategy to detect beta‐thalassemia
minor.Lancet1977;1:692-4.
15. Janel A, Roszyk L, Rapatel C, Mareynat G, Berger MG, Francois
A, et al. Proposal of a score combining red blood cell indices for
early dierentiation of beta-thalassemia minor from iron
deciency anemia.Hematology. 2011;16(2):123-7. doi:10.11
79/102453311X12940641877849.
16. Chandra H, Shrivastava V, Chandra S, Rawat A, Nautiyal R.
Evaluation of Platelet and Red Blood Cell Parameters with
Proposal of Modied Score as Discriminating Guide for Iron
Deciency Anemia and β-alassemia Minor.J Clin Diagn
Res. 2016;10(5):EC31-EC34. doi:10.7860/JCDR/2016/17672.
7843
17. Plengsuree S, Punyamung M, Yanola J, Nanta S, Paiping K,
Maneewong K, et al. Red Cell Indices and Formulas Used in
Dierentiation of β-alassemia Trait from Iron Deciency
in ai Adults.Hemoglobin. 2015;39(4):235-9. doi:10.3109/
03630269.2015.1048352
18. Pornprasert S, ongsat C, Panyachadporn U. Evaluation
of Applying a Combination of Red Cell Indexes and Formulas
to Dierentiate β-alassemia Trait from Iron Deciency
Anemia in the ai Population.Hemoglobin. 2017;41(2):116-9.
doi:10.1080/03630269.2017.1323763
19. WHO. Haemoglobin concentrations for the diagnosis of
anaemia and assessment of severity. Vitamin and Mineral
Nutrition Information System. Geneva, World Health Organization,
2011 (WHO/NMH/NHD/MNM/11.1). Available from: http://
www.who.int/vmnis/indicators/haemoglobin. pdf, accessed
[date].
20. Goh LPW, Chong ETJ, Lee PC. Prevalence of Alpha (α)-
alassemia in Southeast Asia (2010-2020): A Meta-Analysis
Involving 83,674 Subjects.Int J Environ Res Public Health.
2020;17(20):7354. doi:10.3390/ijerph17207354
21. Kattamis A, Forni GL, Aydinok Y, Viprakasit V. Changing
patterns in the epidemiology of β-thalassemia.Eur J Haematol.
2020;105(6):692-703. doi:10.1111/ejh.13512
22. Urrechaga E, Homann JJML. Critical appraisal of discriminant
formulas for distinguishing thalassemia from iron deciency in
patients with microcytic anemia. Clin Chem Lab Med. 2017;
55(10):1582-91. doi:10.1515/cclm-2016-0856
23. Homann JJML, Urrechaga E. Role of RDW in mathematical
formulas aiding the differential diagnosis of microcytic
anemia.Scand J Clin Lab Invest. 2020;80(6):464-9. doi:10.10
80/00365513.2020.1774800
