Volume 74, No.5: 2022 Siriraj Medical Journal

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

330

Napakorn Sangchay, M.D., Ph.D.*, Veronika Dzetkuličová, Ph.D.**, Micol Zuppello, Ph.D.***, Jirapa Chetsawang,

M.D., Ph.D.*

*Department of Anatomy, Faculty of Medicine, Siriraj Hospital, Mahidol University, Bangkok, ailand, **Department of Anatomy, Faculty of

Medicine, Masaryk University, Czech Republic, ***Institute of Clinical Sciences, College of Medical and Dental Sciences, University of Birmingham,

Edgbaston, Birmingham, UK.

Consideration of Accuracy and Observational

Error Analysis in Pelvic Sex Assessment: A Study

in a Thai Cadaveric Human Population

ABSTRACT

Objective: In situations where skeletal human remains are recovered, pelvic bone morphology has been demonstrated

to have an essential role in forensic sex identication. Determination of sex is one of the four pillars used to

construct a biological prole of unidentied skeletal remains. Such analysis has mainly been conned to direct

visual inspection or morphometric analysis of pelvic elements available. is study evaluates the identication

accuracy and classication error established based on a morphometric sex determination of this bone either by

direct observation or digital image analysis.

Materials and Methods: We used morphometric analysis of human pelvic bone from modern ai samples to

clarify the eect of variation in pelvic morphometric parameters on prediction accuracy. A total number of 408

pelvic bones (Male, n=249 and Female, n=159) were examined. Pelvic morphometric variables were measured in

multiple regions for each bone.

Results: We found statistically signicant dierences in the pelvic morphometric parameters measured between

the two sexes with considerably accurate classication and unavoidable errors by all means of analytical assessment.

Conclusion: Our ndings suggest that it is not only variation of pelvic morphometric parameters between the two

sexes in this population, but also the selection of analytical approach that can impact prediction accuracy and thus

may contribute to the eect on the determination of sex. Ethical approval was not required for this study.

Keywords: Morphometric analysis; forensic anthropology; sex estimation; technical error of measurement (Siriraj

Med J 2022; 74: 330-339)

Corresponding author: Jirapa Chetsawang

E-mail: napakorn.sac@mahidol.ac.th

Received 3 February 2022 Revised 31 March 2022 Accepted 2 April 2022

ORCID ID: https://orcid.org/0000-0002-7776-6456

http://dx.doi.org/10.33192/Smj.2022.40

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

Forensic examiners must establish a biological

profile when identifying unknown human remains,

including sex, age, stature, and ancestry. is information

can be compared with antemortem records and other

information contributing to the identication process.

Analysis of skeletal remains should be organized promptly

to assess sex. Skeletal sex estimation is crucial for forensic

anthropologists and forensic osteologists in developing

a biological profile since sex assessment serves as a

foundation for developing other aspects of a biological

prole.

1-4

As part of a signicant step to establishing a

biological prole and personal identication, this process

requires experience and needs accurate decision-making.

Sangchay et al.

Volume 74, No.5: 2022 Siriraj Medical Journal

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

331

Original Article

SMJ

e manifestations of sex characteristics within bones

are dierent in both sexes under genetic inuence and

hormonal regulations.

They make up their skeletal

components and these distinctive skeletal traits can be

used to dierentiate males from females.

6,7

A reliable

and less subjective technique for assessing sex should be

conducted and documented to emphasise the identication

performed.

Accurate and valid assessments of sex are used

in field and laboratory settings for determining sex

from skeletal remains.

Many studies have focused on

gross skeletal features, utilizing cranial or postcranial

bones.

9-11

In general, the selection of pelvic bone from

unidentied human remains is preferable to estimate

sex because of a high level of certainty and validity due

to the sexual dimorphism demonstrated within this

bone.

12,13

Studies have examined the utility of human

pelvic bone to estimate sex, and in many cases, these lack

intra- and inter-observer comparisons of observational

error.

14,15

e estimation of sex can be established by

visual assessment based on observing sexual dimorphic

dierences.

is method to estimate sex from theos

pubisevaluates the dierent degrees of morphological

traits to dierentiate between the two sexes. It can be

performed by using various morphologic characteristics,

such as the subpubic concavity and the medial aspect of

the ischio-pubic ramus.



Alternatively, several morphological features of

the pelvic complex can indicate the biological sex of

individuals, including the size and shape of the os pubis,

greater sciatic notch, obturator foramen, the existence

or absence of the preauricular sulcus and evidence of

parturition scars. However, sex estimation utilizing

gross pelvic morphology requires complete or nearly

complete skeletal elements. When the pelvic complex

is less damaged, this method is achievable. However,

morphological assessment becomes more challenging

when these bones appear fragmented or severely damaged

by taphonomic factors. For these reasons, an alternative

method for sex estimation utilizing digital images of

the specic characteristics from pelvic bone should be

considered.

Despite the acceptable accuracy, an estimation of

sex utilizing gross morphology has several limitations.

This method creates both intra- and inter-observer

errors causing variation between observers and is non-

reproducible. e decision-based observational method

is highly subjective. It can be especially problematic

when it appears to be undetermined or unclassied.

Accordingly, sex estimation using pelvic morphometrics

by direct measurements has been introduced to increase

the certainty and accuracy.

is method of sex estimation from the pelvis is

problematic because it is regionally dependent.

19,20

Complete

pelvic elements are required when determining the sex

of unknown skeletal remains using either the non-metric

method or metric analysis. is means that the degree of

certainty in establishing sex from severely damaged pelves

due to taphonomic causes is reduced. Consideration of an

alternative pelvic landmark to establish sex identication

is essential, especially when encountering fragmented

pelvic bones. A morphologic analysis is mainly focused

on the pubic region. Such analysis relies on either the

presence or absence of morphological traits or the degree

of expression.

When comparing anterior and posterior

regions of the pelvic bone, it is clear that the pubis, being

located anteriorly, has a high possibility of being exposed

to postmortem changes. As a result, it can be impossible

to establish sex. In terms of pelvic anthropometric sex

assessment, this method relies on measuring observational

characteristics. It utilizes individual measurements or

combinations of measurements to dierentiate the two

sexes. e accuracy rate of prediction depends on the

selection of skeletal landmarks utilized in this analytical

method. e posterior region of the pelvic bone is more

robust than the anterior border, and measurements of

skeletal landmarks from this area are achievable.

Metric sex estimation using the pelvic bone is more

precise than visual morphological analyses and can

exceed an accuracy rate of 90%. However, this approach

requires a complete or nearly complete pelvic complex to

assess all the landmarks proposed by current literature.

e ischio-pubic index provided the most accurate sex

estimation of 96.5%.

Supportive results of the ischio-

pubic index as a sex indicator from pubic measurements

were analyzed and achieved 90% accuracy.



It is necessary to consider that either the non-metric

or metric analytical methods using pelvic elements depend

upon identiable morphologic features, which in turn

depend upon the bones being intact or nearly intact.

Currently, sex estimation methods are mainly established

from the pelvis samples from specic (black and white

American) skeletal collections, whereas morphological

traits for Asian and other ethnic groups are limited.

25,26

Several authors suggested that population-specific

databases and classication analysis for sex estimation

in heterogeneous populations are crucially required.



is research will focus on the prediction accuracy

and intra- and inter-observer observational errors from

three dierent sex estimation methods, including estimating

sex from gross morphological sex characteristics, digital

images, and a measured analytical approach in a modern

ai population utilizing human pelvic dry bones.

Volume 74, No.5: 2022 Siriraj Medical Journal

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

332

MATERIALS AND METHODS

e study was conducted on 408 human pelvic bone

samples from the Siriraj Bone collection, Department of

Anatomy, Faculty of Medicine Siriraj hospital, Mahidol

University. For each bone, three gross morphologies

based on the Phenice method, including the presence of

ventral arc, inferior pubic ramus and subpubic concavity,

were used to determine sex. Subsequently, three digital

images were taken from each bone to determine an inter-

observer accuracy and error of sex assessment using the

Phenice criteria. Pelvic morphometric parameters (Fig

1) were measured from each bone. e usefulness of

iliac associated bony landmarks for assessing sex was

examined. is was done by analysing sex discriminant

functions for sex indicators of those measurements taken

from prominent posterior foci to the anterior iliac bony

landmarks. e parameters measured in this method are

located on the anterior and posterior border of the ilium,

including ASIS, AIIS, PSIS, and PIIS. Additional parameters

were measured from those two iliac borders to other

skeletal landmarks: the pubic tubercle and ischial spine.

Pelvic landmarks and morphometric measurements were

described as shown in Table 1. e mean and standard

deviations were calculated for each parameter. Statistical

analyses using paired student’s t-tests were performed

to evaluate the dierences between each group. e null

hypothesis was rejected where the dierence between

groups was 0. A p-value of <0.05 was interpreted as

being statistically signicant. Inferential statistics were

used to determine discriminant function analysis (total

analytical sample) and evaluate the probability of sex

prediction and classication accuracy for those variables

that revealed statistically signicant dierences. e best

discriminant functions were compared and selected

based on positive prediction accuracy.

Observational error and reliability analysis

Intra-observer error of measurement evaluation

(Table 3).

Two measures were used in this observational

study to investigate intra-observer measurement error

rates between two separate measurements taken by the

same investigator. e measurements were taken by

the researcher who did this study, who rst measured

morphometric characteristics from pooled human pelvic

samples, then took a second measurement. e interval

between the rst and second measurements was two weeks.

To quantify intra-observer error, researchers measured

pelvic bone-related data. Six variables were measured:

PL, IL, vertical and horizontal Acetabular diameters, and

obturator foramen diameters. e statistical dierence

between the two measurements was analysed using the

student t-test.

Inter-observer error of measurement evaluation

(Table 3).

Two observers conducted two sets of measurements

independently to investigate an inter-observer technical error

of measurement between two independent investigators.

e researcher who conducted this current investigation

and analysed bone samples was the rst witness. Each of

the six parameters was assessed. e second observer was a

ten-year anatomical academic sta member who evaluated

the pelvic parameters using the same pelvic samples. e

objective of this study was to quantify the inter-observer

measurement error and the repeatability. e instructions

and descriptions for data collection were supplied to the

second observer. Before conducting the inter-observer

inquiry, the observers were given denitions for six criteria.

A brief session was provided to measure the variables with

the anthropometric tools. For non-bias considerations,

Fig 1. Diagrams illustrating pelvic morphometric landmarks and dimensional measurements.

Sangchay et al.

Volume 74, No.5: 2022 Siriraj Medical Journal

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

333

Original Article

SMJ

the measurement was carried out without knowing the

sample’s demographic characteristics. e statistical

dierence between the two observers was investigated

using a paired student t-test. e intra-observer error

rates for two measures and the inter-observer error rates

for two observers were calculated for the conventional

morphometric measurement. Repeatability between

observations/observers was assessed, calculating the

technical error of measurement (TEM), relative technical

error of measurement (rTEM) and coecient of reliability

(R). ese values indicate the repeatability between two

observations.

TEM is calculated as follows,

TEM=√

(ΣD

(D = dierence between the two observers or measurements

1 and 2, and N = the total number of tested samples.

Subsequently, the relative TEM (rTEM) was calculated

using the formula as shown below:

rTEM= (

TEM

/ mean ) × 100

An agreement threshold between observations is accepted

at a 5% cut o value.

is study assesses the coecient of

reliability (R) to determine a repeatability in anthropometric

measurement.

It was calculated using the formula as

shown below:

R= 1 − (

(TatalTEM)2

/SD

)

e value of coecient of reliability (R) ranges from

0 to 1. Levels of reliability coecient (R) are accepted

when the values were > 0.95.

e relationship between pelvic morphometric

variables and known sex were examined using discriminant

function analysis. e entire dataset was utilised at this

stage of the analysis (N = 408). Ten discriminant functions

were created (Table 4) shows a summary of all functions

and parameters.

RESULTS

Descriptive statistics for the pelvic measurements are

summarized in Table 2, including group means, standard

deviations, variances and minimum and maximum values.

Means of all pelvic measurements were signicantly

dierent between males and females (p < 0.05), except

for obturator foramen transverse diameter (OFTD).

Table 3 shows the summary of descriptive statistics and

intra- and inter-observer technical error of measurement

values and the reliability coecient obtained from all

six measurements. Among the six variables, the results

TABLE 1. Description of pelvic landmarks and morphometric

measurements.

Pelvic morphometric landmarks and

descriptions

VAD Vertical Acetabular Diameter

HAD Horizontal Acetabular Diameter

MAD Mean Acetabular diameter

OFLD Obturator Foramen Longitudinal Diameter

OFTD Obturator Foramen Transverse Diameter

MOFD Mean Obturator Foramen diameter

OF index Obturator Foramen index

PL Pubic Length

IL Ischial Length

P/I index Pubic/Ischial length index

A Ischial spine

B Posterior inferior iliac spine (PIIS)

C Posterior superior iliac spine (PSIS)

D Highest point of greater sciatic notch

E Pubic tubercle

F Anterior inferior iliac spine (AIIS)

G Anterior superior iliac spine (ASIS)

AB Distance between ischial spine and PIIS

BC Distance between PSIS and PIIS

XD Maximal greater sciatic notch heigth

EF Distance between pubic tubercle and AIIS

EG Distance between pubic tubercle and ASIS

BG Distance between PIIS and ASIS

CG Distance between PSIS and ASIS

AG Distance between ischial spine and ASIS

BF Distance between PIIS and AIIS

AF Distance between ischial spine and AIIS

CF Distance between PSIS and AIIS

Volume 74, No.5: 2022 Siriraj Medical Journal

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

334

TABLE 2. Descriptive statistics of pelvic landmarks and morphometric variables (* represented no statistical dierence).

TABLE 3. Comparisons of intra- and inter-observer technical errors of measurement (TEM) and Reliability coecient (SD – standard deviation, Ab TEM – Absolute TEM,

rTEM – relative TEM, R – Reliability coecient).

Variables Male (n=249) Female (n=159) P-value

Mean Std Dev Variance Minimum Maximum Mean Std Dev Variance Minimum Maximum

VAD 49.05 2.88 8.28 40.36 57.28 44.14 2.74 7.52 34.14 51.98 < 0.05

HAD 49.99 2.85 8.10 39.88 59.79 45.11 2.97 8.81 36.33 64.84 < 0.05

MAD 49.52 2.72 7.39 40.12 58.54 44.62 2.67 7.10 35.24 55.73 < 0.05

OFLD 50.61 3.69 13.64 39.08 63.63 46.62 3.51 12.30 38.57 58.74 < 0.05

OFTD 32.96 3.44 11.85 21.70 44.66 33.32 3.36 11.29 21.69 43.27 0.29*

MOFD 41.78 3.19 10.19 30.64 54.15 39.97 3.06 9.35 31.80 47.75 < 0.05

OF index 65.17 5.54 30.72 47.63 80.84 71.54 5.93 35.17 51.75 87.92 < 0.05

PL 70.44 5.69 32.39 54.88 86.55 68.10 5.52 30.44 55.92 81.69 < 0.05

IL 87.96 6.03 36.31 64.21 119.40 79.63 5.04 25.36 67.58 91.52 < 0.05

P/I index 80.23 6.00 36.04 67.18 99.94 85.67 6.74 45.48 69.54 105.49 < 0.05

AB 54.52 5.70 32.53 37.42 70.01 60.97 7.10 50.44 43.08 84.43 < 0.05

BC 36.79 5.35 28.67 21.30 51.11 32.64 4.71 22.21 18.24 45.41 < 0.05

AB+BC 91.31 7.68 58.97 72.78 112.68 93.61 8.79 77.20 70.23 129.18 < 0.05

XD 35.84 3.20 10.27 27.69 44.51 33.70 3.70 13.68 24.96 45.77 < 0.05

EF 85.77 6.07 36.90 61.24 105.79 83.40 6.31 39.83 69.66 99.79 < 0.05

EG 119.34 9.69 93.86 78.01 149.48 116.75 8.60 73.88 96.53 146.62 < 0.05

BG 13.20 0.75 0.56 10.60 15.50 12.65 0.85 0.72 10.20 14.90 < 0.05

CG 14.61 0.83 0.69 11.00 16.80 14.22 0.86 0.74 12.00 16.50 < 0.05

AG 13.16 0.74 0.54 11.00 15.50 12.30 0.70 0.49 10.50 14.50 < 0.05

BF 11.25 0.72 0.52 8.50 13.60 10.79 0.80 0.64 8.70 12.90 < 0.05

AF 10.11 0.69 0.48 7.80 13.50 9.20 0.62 0.39 7.50 11.00 < 0.05

CF 13.36 0.76 0.58 10.00 15.70 12.84 0.81 0.66 10.20 14.80 < 0.05

Intra-observer error Inter-observer error

Mean SD

p value Ab TEM rTEM R

Mean SD

p value Ab TEM rTEM R

Measurement1 Measurement2 Measurement1 Measurement2 Observer1 Observer2 Observer1 Observer2

VAD 47.137 47.301 3.704 3.596 0.112 1.473 3.120 0.837 47.137 47.288 3.704 3.490 0.200 1.68 3.558 0.782

HAD 48.090 48.191 3.748 3.701 0.044 0.71 1.475 0.964 48.090 48.185 3.748 3.201 0.392 1.57 3.261 0.797

OFLD 49.056 49.202 4.109 4.115 0.105 1.29 2.626 0.901 49.056 49.257 4.109 4.192 0.126 1.86 3.784 0.799

OFTD 33.096 33.179 3.411 3.385 0.282 1.39 3.289 0.832 33.096 33.167 3.411 3.260 0.569 1.78 5.372 0.715

PL 69.524 69.316 5.732 5.608 0.032 1.38 1.988 0.941 69.524 68.893 5.732 5.258 0.000 2.55 3.685 0.786

IL 84.714 84.708 6.963 6.766 0.961 1.66 1.960 0.941 84.714 84.647 6.963 6.554 0.649 2.11 2.492 0.903

Sangchay et al.

Original Article

SMJ

Volume 74, No.5: 2022 Siriraj Medical Journal

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

335

TABLE 4. Comparisons of Canonical discriminant function coecients for pelvic dimensions, discriminant classication function and classication matrix.

Canonical discriminant function coefcients for pelvic dimensions

Functions and Raw Standardized Total Group Discriminant Classication Function Classication Matrix

parameters Canonical Canonical Canonical Centroids

Coefcients Coefcients Structure

(Pooled

Within-Class)

Group OFLD OFTD Constant Group \ Predicted F M Total Percent Correct

Function1 OFLD 0.343 1.242 0.819 F -0.887 F 3.018 0.956 -86.266 F 90 69 159 56.604%

OFTD -0.238 -0.812 -0.089 M 0.566 M 3.516 0.61 -99.016 M 37 212 249 85.141%

Constant -8.933 Total 127 281 408 74.02%

Number of correct = 302

Group MAD MOFD Constant

Function2 MAD 0.393 1.062 0.994 F -1.12 F 5.024 2.117 -154.411 F 126 33 159 79.245%

MOFD -0.051 -0.161 0.407 M 0.715 M 5.746 2.023 -184.549 M 26 223 249 89.558%

Constant -16.63 Total 152 256 408 85.539%

Number of correct = 349

Group PL IL PII Constant

Function3 PL 0.149 0.836 0.335 F -0.926 F -248.775 207.126 210.282 -8,783.987 F 115 44 159 72.327%

IL 0.034 0.19 0.98 M 0.591 M -248.549 207.177 210.034 -8,783.306 M 32 217 249 87.149%

PII -0.164 -1.031 -0.652 Total 147 261 408 81.373%

Constant 0.282 Number of correct = 332

Group VAD HAD MAD Constant

Function4 VAD 110.349 311.844 0.97 F -1.109 F 3.719 3.594 -1.188 -136.813 F 124 35 159 77.987%

HAD 110.33 319.357 0.964 M 0.708 M 4.242 4.016 -1.453 -168.813 M 27 222 249 89.157%

MAD -220.308 -594.476 1. Total 151 257 408 84.804%

Constant -17.645 Number of correct = 346

Group AB BC XD Constant

Function5 AB 0.118 0.741 0.77 F 0.894 F 1.423 1.098 2.554 -104.329 F 103 56 159 64.78%

BC -0.102 -0.522 -0.633 M -0.571 M 1.25 1.248 2.746 -106.251 M 37 212 249 85.141%

XD -0.132 -0.448 -0.504 Total 140 268 408 77.206%

Constant 1.475 Number of correct = 315

Group BG CG AG Constant

Function6 BG 0.331 0.26 0.61 F -0.768 F 4.569 9.442 13.432 -178.746 F 62 96 158 39.241%

CG -0.662 -0.557 0.413 M 0.488 M 4.985 8.61 15.39 -197.064 M 38 211 249 84.739%

AG 1.559 1.125 0.955 Total 100 307 407 67.076%

Constant -14.726 Number of correct = 273

Group AF BF CF Constant

Function7 AF 1.82 1.211 0.976 F -0.865 F 6.594 4.758 14.186 -147.09 F 108 51 159 67.925%

BF -0.325 -0.244 0.502 M 0.552 M 9.174 4.297 13.933 -163.599 M 39 210 249 84.337%

CF -0.178 -0.139 0.54 Total 147 261 408 77.941%

Constant -11.803 Number of correct = 318

Group MAD MOFD OFI PII Constant

Function8 MAD -0.297 -0.8 -0.911 F 1.327 F 5.824 1.166 1.574 1.958 -293.442 F 136 23 159 85.535%

MOFD -0.008 -0.024 -0.373 M -0.847 M 6.469 1.182 1.408 1.87 -305.777 M 20 229 249 91.968%

OFI 0.077 0.437 0.659 Total 156 252 408 89.461%

PII 0.04 0.255 0.534 Number of correct = 365

Constant 5.913

Group AB BC EF EG Constant

Function9 AB -0.138 -0.87 -0.769 F -0.895 F 0.922 0.841 1.387 0.359 -120.523 F 109 50 159 68.553%

BC 0.092 0.47 0.633 M 0.572 M 0.719 0.975 1.48 0.378 -123.615 M 32 217 249 87.149%

EF 0.064 0.393 0.317 Total 141 267 408 79.902%

EG 0.013 0.138 0.257 Number of correct = 326

Constant -2.27

Group VAD HAD OFLD OFTD PL IL Constant

Function10 VAD 0.161 0.456 0.889 F -1.328 F 1.524 2.012 1.317 0.481 0.737 0.606 -166.936 F 136 23 159 85.535%

HAD 0.12 0.347 0.873 M 0.848 M 1.875 2.273 1.546 0.148 0.64 0.717 -198.467 M 19 230 249 92.369%

OFLD 0.105 0.381 0.651 Total 155 253 408 89.706%

OFTD -0.153 -0.522 -0.071 Number of correct = 366

PL -0.044 -0.25 0.274

IL 0.051 0.29 0.801

Constant -14.728

Volume 74, No.5: 2022 Siriraj Medical Journal

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

336

demonstrated statistically significant differences in

HAD and PL between the two measurements (p-value

of < 0.05). HAD demonstrated intra-observer error

rates with R value > 0.95. is nding indicates high

repeatability with only the HAD not exceeding the 5%

acceptance threshold. In terms of inter-observer error

analysis, statistical analysis demonstrated that only PL

examined by the two observers was statistically signicantly

dierent (p-value of < 0.05). e six pelvic parameters

demonstrated inter-observer error rates with R value <

0.95. ese ndings indicate low repeatability with all

measured variables not exceeding the 5% acceptance

threshold.

When accuracy for estimating sex by direct observation

using Phenice’s characteristic was compared with sex

assessment using Phenice’s pelvic morphologies digital

images, 91.42% accuracy was achieved from sex estimation

based on gross pelvic examination (373/408). Similar

gures were obtained from two observers (90.32%, 84/93

and 87.09%, 81/93) using pelvic morphologies digital

images approach based on Phenice’s characteristics with

the mean accuracy of 88.71%.

Table 4 shows the standard, structure and

unstandardized coecients, the group centroids, results,

and the discriminant function analysis results from the

study data. For all of the ten discriminant functions

produced for sex determination, sex was correctly assessed

with an accuracy between 67.07% and 89.46%. Using the

discriminant function analysis of pelvic dimensions, the

lowest accuracy in the prediction was observed from

Function 6 when using a combination of BG, CG and

AG as a sex predictor with an accuracy of 67.07%. e

best discriminator between the sexes with the highest

predicted accuracy was achieved using VAD, HAD,

OFLD, OFTD, PL and IL with an accuracy of 89.71%.

For the function using only acetabular-related

variables, average accuracy was 84.8%; using only obturator

foramen-related variables classication accuracy was

74.02%. Function 5, where sciatic notch measurements

were assessed, performed an average accuracy of 81.37%.

Functions established from distances measuring ASIS

and AIIS to the posterior pelvic landmark achieved an

accuracy of 67.08% and 77.94%, respectively.

DISCUSSION

Skeletal sex estimation can be assessed by visual

observation of morphological variation in the bones because

of sexual dimorphism.

Skeletal size and robusticity,

attributed by extrinsic factors such as biomechanics,

interact with bones, and intrinsic factors including genetic

and sex hormones, are the preferable sex indicator.

Cranial and postcranial skeletal elements exhibit sexual

dimorphism observed in adult skeletons.

Non-metric

analysis of pelvic morphology, along with metric methods,

is the most reliable sex indicator for adult skeletons.

Within the pelvis, the morphology of the pubic region

is thought to establish the most reliable indicators for

sex estimation. Existence of the ventral arc, the subpubic

Sangchay et al.

Acetabular morphometric parameters

Male VAD

Femal e VAD

Male HAD

Femal e HA D

Male MAD

Femal e MA D

100

Obturator foramen morphometric parameters

Male OFLD

Female OFLD

Male OFTD

Female OFTD

Male MOFD

Female MO FD

Male OF index

Female OF index

100

120

140

Pubic length, ischial length and P/I index

Male P L

Female PL

Male IL

Female IL

Male P /I in dex

Female P/I in dex

100

120

140

160

Pelvic morphometric measurements

Male AB

Female AB

Male BC

Female BC

Male AB+BC

Female AB +BC

Male XD

Female X D

Male EF

Female EF

Male EG

Female EG

Pelvic morphometric measurements

Male BG

Female BG

Male CG

Female CG

Male AG

Female AG

Male BF

Female BF

Male AF

Female AF

Male CF

Female CF

Figure 2. Comparison of mean acetabular

morphometrics (A: VAD, HAD and MAD-

mm), obturator foramen morphometrics (B:

OFLD, OFTD, MOFD and OF index-mm)

pubic length, ischial length and P/I index (C:

PL, IL and P/I index-mm), pelvic

morphometrics (D: AB, BC, AB+BC, XD, EF

and EG-mm and E: BG, CG, AG, BF, AF and

CF-cm) between male and female, x

designated mean value.

Fig 2. Comparisons of mean acetabular

morphometrics (A: VAD, HAD and MAD-

mm), obturator foramen morphometrics (B:

OFLD, OFTD, MOFD and OF index-mm)

pubic length, ischial length and P/I index (C:

PL, IL and P/I index-mm), pelvic morphometrics

(D: AB, BC, AB+BC, XD, EF and EG-mm and

E: BG, CG, AG, BF, AF and CF-cm) between

male and female, x designated mean value.

Volume 74, No.5: 2022 Siriraj Medical Journal

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

337

Original Article

SMJ

concavity, and the morphology of the medial surface

of the ischiopubic ramus was introduced by Phenice as

an accurate method for sex estimation. Phenice’s pelvic

morphological traits have been further validated with

improved accurate prediction and reduced classication

errors.

15,24,33,34

The morphologies of the acetabulum, the size

of the obturator foramen, and the pubic and ischial

lengths can reveal sex dierences. In addition, studies

report other morphological features that are possible

for sex indicators within the pelvic bone, including

the acetabulum and the greater sciatic notch.

ese

anatomical regions have a durability that potentially

endures destructive processes more readily than the pubis.

e use of morphometric measurement incorporated with

stepwise-selected discriminant functions can yield more

accurate classication than observational morphometric

approaches. Nonetheless, analysis using a digital image is

achievable. is study found that this method can provide

less accuracy than direct observational sex dierentiation.

While most adult skeletons demonstrate sexually

dimorphic characteristics, the validity of sex assessment

is also inuenced by other factors: population variations,

age, and pathological and taphonomic changes. e

expression of sexual dimorphism is variable among

and across populations. erefore, it is necessary to

consider comparative data from a specic population when

applying these techniques in forensic circumstances.

Sexual dimorphism and the dierences between the

sexes vary in other populations. In addition, dierent

methodologies, including morphological and metric

features, may analyze skeletal remains for sex estimation.

Skeletal dierences in morphological attributes vary by

shape, morphological traits, and relative size between the

sexes. Methods based on the shape and size of the pelvis

and the presence or absence of pelvic characteristics are

favoured. Other morphological features may represent

sex dierences. However, they are usually less accurate

than methods using distinct morphological features

with substantial sexual dimorphism and observations.

Instruments, standards, appropriate analytical soware

and a combination of measurements and multivariate

approaches can increase reliability in sex evaluations,

although a single measurement may provide reasonably

reliable sex estimation.

It is well acknowledged that intra- and inter-observer

error caused by visual or metric variables can lead to

disparities in sexual dimorphism evaluation and sex

determination. Current studies, however, have shown

that visual assessment is associated with significant

levels of inter-observer error due to imprecise variable

denitions, substantial reliance on previous observer

experience, and the seriation process employed to sort

the individuals. is error measurement analysis reveals

that geometric morphometrics produces good intra- and

inter-observer agreement.

Pelvic morphologies can be dierentiated consistently

even among observers with no prior experience. However,

those with obscure morphologies are exceedingly dicult

to determine consistently. Our ndings show that even

observers familiar with pelvic analysis using sexual

dimorphism have an inconsistent interpretation of

coordinative landmarks for assessing sex; even though

the presence of the ventral arc is the best sex indicator

for females, determination of sex from digital images still

establishes various degrees of prediction accuracy among

observers. is shows that descriptive anatomical landmarks

may be misinterpreted in the ischiopubic regions.

24,35

It is worth describing the quantitative methodologies

utilizing the pelvic characteristics which show sexual

dimorphism variance. ese methodologies are relevant

and applicable to every forensic practitioner. e eect

of sex estimation utilizing pubic length, ischial length

and P/I index on classication accuracy is critical. e

description of anatomical landmarks for these measurements,

especially the pubic length, is diverse among current

literature. In this study, measurement of pubis length

and ischial length between the two observers revealed

lower reliability than results from an intra-observer error

of measurement, indicating a low reproducibility.

Additionally, the denition of a specic landmark

being measured within the acetabulum is identical to a

problematic issue occurring in the ischiopubic region.

37,38

All measurements should be repeatable and independent.

erefore, future research should incorporate an assessment

of the interobserver error in these dimensions to avoid

sex misinterpretation.

e integration of the iliac landmark measurement

and the assessment of pubic sex characteristics should

be considered when dealing with skeletal parts and

severely fragmented skeletal remains to improve sex

prediction accuracy. e range of prediction accuracy

from discriminant functions utilizing these measurements

in this study was 67.08% (Function6)-79.9% (Function9),

indicating that those measurements are potentially reliable

in estimating the sex, as highlighted in previous studies.

22,39,40

us, sex discrimination utilizing iliac measurement

can be conducted independently.

However, a correct

estimation between an isolated morphometric measurement

analysis and the morphological assessment showed that

the former established a less favourable outcome and

provided a lower prediction accuracy than the latter

Volume 74, No.5: 2022 Siriraj Medical Journal

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

338

method. Sex dimorphism of pelvic morphological traits

has been documented across dierent human populations,

and it is also considered as a sex-determining method.

is study shows that the utilization of iliac associated

morphological characteristic measurements is less reliable

when applied to sex assessment because the attribution

of sex dimorphism is demonstrated less within this bony

segment. As a result, these measurements require further

investigation and the expansion of this analytical method

into the dierent populations to validate the usefulness

of these sex indicators in a forensic application.

e eect of sample size is another aspect to consider

in this situation. e requirement of cross-validation study

using dierent populations is necessary to evaluate the

discriminant outcomes established from these present

classication functions. e use of approximately 400

individuals in this study can aect prediction accuracy

and increase technical errors. As a result, it is feasible

that increased samples will reduce undesirable errors

and improve eciency.

In summary, the consideration of factors aecting

prediction accuracy and classication errors for pelvic

sex estimation is critical because this bone is under

hormonal regulation rather than mechanical inuence.

ese functions may be benecial where the population

of origin of the unidentied skeleton is unknown. Further

research is required on cooperative group data from

other divergent populations, both in the pelvis region and

other skeletal parameters that express sexually dimorphic

characteristics.

CONCLUSION

Identifying sex from skeletal remains is a crucial

step in human identication. Although human pelvic

bone has been the most crucial determinant in this

process, only a few anthropologic methods can inuence

sex determination accuracy. Promoting alternative

morphological assessment methods, such as digital evaluation

and direct morphometric measurement, as well as direct

morphological interpretation from this bone, may be

shown to be the most accurate methods. It is critical to

recognize instruments that inuence correct prediction

in sex discrimination. e selection of an appropriate

analytical method is essential because it can aect the

whole process of forensic human identication. erefore,

it can minimize the possibility of misidentication.

Conicts of interest: None

Funding: None

Ethics/IRB approval: Yes, SIRB Protocol No. (Si 625/2020)

REFERENCES

1. Christensen AM, Crowder CM. Evidentiary standards for

forensic anthropology. J Forensic Sci. 2009;54(6):1211-6.

2. Christensen AM, Passalacqua NV. Introduction to Forensic

Anthropology. In: A Laboratory Manual for Forensic Anthropology.

London: Academic Press, 2018.

3. Christensen AM, Passalacqua NV, Bartelink EJ. Forensic

Anthropology: Current Methods and Practice. London: Academic

Press, 2014.

4. Krishan K, Chatterjee PM, Kanchan T, Kaur S, Baryah N, Singh

RK. A review of sex estimation techniques during examination

of skeletal remains in forensic anthropology casework. Forensic

Sci Int. 2016;261:165.e1-8.

5. Kirkham GR, Cartmell SH. Genes and Proteins Involved in

the Regulation of Osteogenesis. Genes Osteogenes. 2007;3(0):

1-22.

6. Durić M, Rakočević Z, Donić D. e reliability of sex determination

of skeletons from forensic context in the Balkans. Forensic Sci

Int. 2005;147(2-3):159-64.

7. White T, Folkens P. e Human Bone Manual. Oxford: Academic,

2005.

8. Spradley MK, Jantz RL. Sex estimation in forensic anthropology:

skull versus postcranial elements. J Forensic Sci. 2011;56(2):289-

96.

9. Guyomarc’h P, Bruzek J. Accuracy and reliability in sex

determination from skulls: A comparison of Fordisc® 3.0 and

the discriminant function analysis. Forensic Sci Int. 2011;208

(1-3):180.e1-6.

10. Bruzek J. A method for visual determination of sex, using the

human hip bone. Am J Phys Anthropol. 2002;117(2):157-68.

11. Sangvichien S, Boonkaew K, Chuncharunee A. Sex Determination

in ai Skulls by Using Craniometry: Multiple Logistic Regression

Analysis. Siriraj Med J. 2007;59:216-21.

12. Austin D, King RE. e Biological Prole of Unidentied

Human Remains in a Forensic Context. Acad Forensic Pathol.

2016;6(3):370-90.

13. Rösing FW, Graw M, Marré B, Ritz-Timme S, Rothschild MA,

Rötzscher K, et al. Recommendations for the forensic diagnosis

of sex and age from skeletons. Homo. 2007;58(1):75-89.

14. González PN, Bernal V, Ivan Perez S, Barrientos G. Analysis

of dimorphic structures of the human pelvis: its implications

for sex estimation in samples without reference collections. J

Archaeol Sci. 2007;34(10):1720-30.

15. Lesciotto KM, Doershuk LJ. Accuracy and Reliability of the

Klales etal. (2012) Morphoscopic Pelvic Sexing Method. J

Forensic Sci. 2018;63(1):214-20.

16. Phenice TW. A newly developed visual method of sexing the

os pubis. Am J Phys Anthropol. 1969;30(2):297-301.

17. Ali SHM, Omar N, Shae MS, Ismail NAN, Hadi H, Nor

FM. Sex estimation using subpubic angle from reconstructed three-

dimensional computed tomography pelvic model in a contemporary

Malaysian population. Anat Cell Biol. 2020;53(1):27-35.

18. Mahakkanukrauh P, Ruengdit S, Tun SM, Case DT, Sinthubua

A. Osteometric sex estimation from the os coxa in a ai

population. Forensic Sci Int. 2017;271:127.e1-127.e7.

19. Bytheway JA, Ross AH. A geometric morphometric approach

to sex determination of the human adult os coxa. J Forensic

Sci. 2010;55(4):859-864.

20. Gonzalez PN, Bernal V, Perez SI. Geometric morphometric

approach to sex estimation of human pelvis. Forensic Sci Int.

Sangchay et al.

Volume 74, No.5: 2022 Siriraj Medical Journal

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

339

Original Article

SMJ

2009;189(1–3):68-74.

21. Chanapa P, Kijkuokool P, Singsuwan P, Mahakkanukrauh P.

e most accurate sex determination using three morphologies

in the pubis. Int Med J. 2018;25(3):189-192.

22. Rmoutilová R, Dupej J, Velemínská J, Brůžek J. Geometric

morphometric and traditional methods for sex assessment

using the posterior ilium. Leg Med. 2017;26:52-61.

23. ieme F, Schull W. Sex determination from the skeleton.

Hum Biol an Int Rec Res. 1957;29(3):242-73.

24. Drew R. A review of the ischium-pubis index: Accuracy,

reliability, and common errors. Hum Biol. 2013;85(4):579-96.

25. Patriquin ML, Loth SR, Steyn M. Sexually dimorphic pelvic

morphology in South African whites and blacks. HOMO- J

Comp Hum Biol. 2003;53(3):255-62.

26. Patriquin ML, Steyn M, Loth SR. Metric analysis of sex dierences

in South African black and white pelves. Forensic Sci Int.

2005;147(2-3 SPEC.ISS.):119-27.

27. Sierp I, Henneberg M. e Diculty of Sexing Skeletons from

Unknown Populations. J Anthropol. 2015;2015:1-13.

28. Perini TA, de Oliveira GL, dos Santos Ornellas J, Palha de

Oliveira F. Technical error of measurement in anthropometry.

Rev Bras Med do Esporte. 2005;11(1):81-90.

29. Purkait R. Anthropometric landmarks: How reliable are they?

Anthropometric landmarks. Medico-Legal Updat. 2004;4(4):

133-40.

30. Ulijaszek SJ, Kerr DA. Anthropometric measurement error and

the assessment of nutritional status. Br J Nutr. 1999;82(3):165-

77.

31. Goto R, Mascie-Taylor CGN. Precision of measurement as a

component of human variation. J Physiol Anthropol. 2007;26(2):2

53-6.

32. Jeyashree T, Sangeetha S, Premavathy D. Quantitative and

qualitative morphometry of hip bone for determining sex.

Drug Invent Today. 2019;11(10):2590-2.

33. Ubelaker DH, Volk CG. A Test of the Phenice Method for the

Estimation of Sex. J Forensic Sci. 2002;47(1):19-24.

34. Kenyhercz MW, Klales AR, Stull KE, McCormick KA, Cole

SJ. Worldwide population variation in pelvic sexual dimorphism:

A validation and recalibration of the Klales et al. method.

Forensic Sci Int. 2017;277:259.e1-259.e8.

35. Antony M, Mohanraj KG. Sex determination using geometric

dimensions of greater sciatic notch and subpubic angle of

human pelvic bone: A morphometric study. Drug Invent

Today. 2019;12(10):2199-202.

36. Oikonomopoulou EK, Valakos E, Nikita E. Population-specicity

of sexual dimorphism in cranial and pelvic traits: evaluation

of existing and proposal of new functions for sex assessment

in a Greek assemblage. Int J Legal Med. 2017;131(6):1731-8.

37. James MacAluso P. Sex determination from the acetabulum:

Test of a possible non-population-specic discriminant function

equation. J Forensic Leg Med. 2010;17(6):348-51.

38. Mary DJ, Mohanraj KG. Morphometric analysis of acetabulum

and pubis of pelvic bone using acetabulopubic index. Drug

Invent Today. 2019;12(10):2206-8.

39. Kanabur V. Identication of the sex of human hip bone by

metric analysis of its anterior border. Biomed Res. 2012;23(2):211-

214.

40. Brůžek J, Santos F, Dutailly B, Murail P, Cunha E. Validation

and reliability of the sex estimation of the human os coxae

using freely available DSP2 soware for bioarchaeology and

forensic anthropology. Am J Phys Anthropol. 2017;164(2):440-

449.

41. Akshaya K, Mohanraj KG. Sex determination using the distance

between posterior inferior iliac spine and ischial spine in dry

human innominate bone. Drug Invent Today. 2019;11(9):2246-

2249.