You are on page 1of 7

Forensic Science International: Reports 2 (2020) 100077

Contents lists available at ScienceDirect

Forensic Science International: Reports


journal homepage: www.elsevier.com/locate/fsir

Forensic Anthropology

Sex estimation using the proximal end of the femur on a modern


Chilean sample
David Carvallo a, * , Rodrigo Retamal b
a
Independent researcher
b
Department of Anthropology, University of Chile. Ignacio Carrera Pinto 1045, Santiago, Chile

A R T I C L E I N F O A B S T R A C T

Keywords: Sex estimation is fundamental for the identification of skeletonized human remains in forensic contexts. The
Sex estimation methods most often used are based on pelvis and skull morphology evaluation. When these elements are not
Proximal femur available due to poor preservation, other bone elements, such as the femur, can be used given its good preservation
Chile
and marked sexual dimorphism. Particularly, the proximal end of the femur is the most dimorphic region of this
Forensic anthropology population data
bone, so its evaluation is especially useful for sex estimation. A set of new sex estimation models were elaborated
utilizing 8 metric variables from the proximal end of the femur in a modern Chilean sample of 270 individuals (200
individuals for training, 70 individuals for validation) from the Santiago Modern Osteological Collection. Sex
estimation models were calculated using logistic regression. Results showed that the models based on the femoral
neck measurements are the best sex estimators, with up to 92.9% (univariable models) and 95.7% (multivariable
models) of overall accuracy and low sex bias. The models proposed here are the most accurate reported to date for
modern Chilean population and can contribute toward the identification of human remains in cases of Human Rights
violations committed during the Civil-Military Dictatorship in Chile (1973–1990).

Introduction contexts [5,8,10–12,14–17]. Although all dimensions of the femur can be


used for sex estimation, the proximal end is particularly useful since this
The estimation of biological sex is a fundamental step toward the region concentrates most of the dimorphism of the bone. This is because it
positive identification of human remains because a) it rules out one sex has the most direct biomechanical relationship with the pelvis [3,8,13,18]
from further consideration and b) it is required to estimate other and houses a group of dimorphic muscle attachments that are relevant in
components of the biological profile [1–6]. The pelvis and the skull are the body weight transmission [12].
the most commonly evaluated elements for sex estimation [3,7]. However, The osteological estimation of sex relies on the assessment of the
the preservation of these elements is frequently affected by taphonomic and sexual dimorphism of the human skeleton, which is determined by the
anthropic processes [8–11] which prevent its evaluation for the estimation interaction between genetic and environmental factors. Since this
of sex. In these cases, other elements of the skeleton must be utilized. interaction varies across populations, the osteological estimation of sex
In the post-cranium, the long bones stand out because they have high is considered population-specific [3,5,10,17,19–21]. As consequence,
levels of sexual dimorphism, arguably higher than the skull [2]. sex-predictive models should be elaborated from a representative sample
Additionally, long bones report better preservation than both pelvis and of each population to avoid loss of accuracy [5,9–12,15–17,19–23].
skull given their robusticity and tubular structure [9]. Finally, the Forensic anthropology in Chile is mostly related to Human Rights
morphology of the long bones allows to easily obtain morphometric violations committed during the Civil-Military Dictatorship (1973–1990)
measurements that can be used for sex estimation [12]. Among the long [1,4,6]. Only a small percentage of victims have been returned to their
bones, the femur has some notable advantages for sex estimation that make families, whereas most victims are still missing or unable to identify as the
it preferable over other elements. It is particularly dimorphic because of its location of most of the burials is still officially unknown [Intriago,
biomechanical relationship with the pelvis, which is the most sexually personal communication]. In the cases were a burial has been found, it is
dimorphic element of the human skeleton given its obstetric adaptation on common the presence of commingled, highly fragmented and poorly
females [3,8,13]. Besides, the femur is one of the best-preserved elements preserved bone elements due to anthropic action, taphonomic processes
of the human skeleton since it is heavy, robust, resistant, and is surrounded and perimortem trauma. In these contexts, skull and pelvis are
by a large muscle mass. Because of this, is frequently recovered in forensic particularly affected [4,24].These circumstances difficult the evaluation

* Corresponding author.
E-mail address: david.carvallo@ug.uchile.cl (D. Carvallo).

http://doi.org/10.1016/j.fsir.2020.100077
Received 24 July 2019; Received in revised form 30 January 2020; Accepted 2 February 2020
Available o
2665-9107/© 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
D. Carvallo, R. Retamal FSIR 2 (2020) 100077

of osteological remains for sex estimation, thus the elaboration of sex Table 1
estimation models based on other skeletal elements that can be applied in Composition of the training sample by age and sex, following the classification
fragmentary contexts is crucial for Chilean scenario. However, the proximal proposed by Buikstra & Ubelaker [7].
end of the femur has been poorly evaluated for Chilean population [1,4]. Category Age range (years) Sex Frequency
Therefore, the goal of this study is to elaborate sexestimation models from the
Young Adult 20 – 34 Female 18
proximal end of the femur that can be applicable to forensic contexts Male 15
involving modern Chilean population. Middle Adult 35 – 49 Female 27
Male 38
Old Adult 50+ Female 55
Materials and methods
Male 47

For this study, the Santiago Modern Osteological Collection was employed
(Department of Anthropology, University of Chile). This collection is
composed of Chilean individuals deceased between 1950 and 1973 and
have documented sex, age-at-death and cause of death [25,26]. A training
sample of 200 adult individuals between 20–96 years and equal sex usually perform better than univariable models because they recover more
representation was randomly selected (Table 1). The age ranges are not variation related to sexual dimorphism, while univariable models are more
equally represented. However, previous studies have shown that age do not sex-biased and less accurate because they rely on absolute differences of only
have a significant influence over sex estimation from the proximal femur one measurement [13,30,38]. In contrast, multivariable models require
[16,22,27,28]. better preservation, limiting the potential application in fragmentary
A set of 8 measurements from the proximal end of the femur were contexts, while univariable models can be applied where poor preservation
selected from previous studies. The measurements definitions and prevents the evaluation of more than one trait [5,14,22,38,39]. Given the
references are summarized in Table 2, and in Figs. 1–3 are illustrated costs and benefits of each approach, it is important to have both univariable
graphically. Measurements were performed using a digital caliper. Intra and multivariable models for sex estimation.
and inter-observer measurements errors were estimated using Lin’s The models classify the individual as female if the logit value is
Concordance Correlation Coefficient (rc) [29], employing a random negative and as male if the logit value is positive. The accuracy of the
subsample of 20 individuals (10 females, 10 males) for both errors. models obtained was then assessed employing a validation sample of 70
Bilateral asymmetry was assessed using a paired t-test on a random individuals (35 females, 35 males) and calculating the sensitivity,
subsample of 20 individuals with equal sex representation. Descriptive specificity, sex bias and overall accuracy. The sensitivity was defined as
statistics, including the mean, standard deviation and 95% confidence the proportion of male individuals correctly classified and the
intervals of the mean, were obtained for each measurement. Normality specificity as the proportion of female individuals correctly classified
was evaluated through Shapiro-Wilk test and statistically significant [5]. Sex bias was calculated as the difference between sensitivity and
differences in the means between sexes were assessed using Student’s specificity. The overall accuracy represents the proportion of
t-test and Mann-Whitney-Wilcoxon tests. individuals of both sexes correctly classified over the total of cases.
Sex-predictive models were created using logistic regression. While Sex estimation models were proposed for modern Chilean population
discriminant analysis function (DFA) is the statistical approach most if they meet the following criteria: a) Statistical significance of all
commonly employed for sex estimation, logistic regression is a better variables included in the model; b) Overall accuracy equal or superior to
alternative due to the ability to calculate robust models with no 80%, according to the literature convention [3,28]; c) Sex bias equal or
assumptions for the underlying data, including normality, homoscedas- inferior to 10%; and d) higher overall accuracy than other models with
ticity and multicollinearity [36]. This last point is very important since fewer variables. The models that failed to meet these criteria were
the evidence suggest that the relationship between osteometric measure- discarded.
ments and the sex of an individual is in fact not linear [28]. For this reason, Based on the results obtained, 38 models were calculated because
logistic regression models can explain better than DFA the relationship models with more than 2 variables were not expected to meet the criteria
between variables and therefore are more accurate even when the a) and d). The sex estimation models were named according to the
assumptions for DFA has been satisfied, as previous forensic studies have measurements included, following the order shown in Table 2, and are
shown [5,28,30,37]. identified in italics to avoid confusion with the measurements. The level
A total of 38 sex-predictive models were elaborated, including 8 of significance was set as a = 0.05 for all tests. Statistical analyses were
univariable models and 30 multivariable models. Multivariable models conducted with R version 3.4.4

Table 2
Definitions and references of the measurements evaluated.
Measurement Definition References

FNAL Distance between the base of the greater trochanter and the apex of the femoral head [5,11,21,27,30]

FNW Minimum supero-inferior diameter of the femoral neck [5,10,11,16,18,19,21,31,32,33]

AB Distance between the most lateral point on the articular margin of the femoral head (A) and the most medial point on the greater
trochanter (B)
AC Distance between the most lateral point on the articular margin of the head (A) and the most postero-medial point on the lesser [15,16,34,35]
trochanter (C)
BC Distance between the most medial point on the greater trochanter (B) and the most postero-medial point on the lesser trochanter (C)

GTFC Distance between the most lateral point on the greater trochanter (GT) and the most superior point on the fovea capitis (FC)
GTLT Distance between the most lateral point on the greater trochanter (GT) and the most superior point on the facet of the lesser
trochanter (LT) [8,10]
LTFC Distance between the most superior point on the facet of the lesser trochanter (LT) and the most superior point on the fovea capitis
(FC)

2
D. Carvallo, R. Retamal FSIR 2 (2020) 100077

Fig. 1. Anterior view of the proximal end of the femur showing FNAL and FNW measurements.

Fig. 2. Posterior view of the proximal end of the femur showing AB, AC and BC measurements.

3
D. Carvallo, R. Retamal FSIR 2 (2020) 100077

Fig. 3. Posterior view of the proximal end of the femur showing GTFC, GTLT and LTFC measurements.

Table 3 Univariable models achieved an overall accuracy from 81.4% (GTLT) to


Intra and inter-observer errors for the measurements evaluated. All values were 92.9% (FNAL and GTFC), while multivariable models reported an overall
statistically significant. accuracy between 82.9% (AB+BC and AC+GTLT) and 95.7% (FNW
Measurement Intra-observer error Inter-observer error +GTFC). Four multivariable models (GTFC+GTLT, BC+GTFC, AB
+GTFC and FNW+GTFC) showed higher overall accuracy than the
FNAL 0.997 0.992
FNW 0.993 0.994 highest overall accuracy observed in univariable models.
AB 0.963 0.987 In univariable models, sensitivity ranged between 82.9% (GTLT) and
AC 0.927 0.964 97.1% (FNAL and GTFC), while multivariable models reported a
BC 0.925 0.977
GTFC 0.999 0.997
GTLT 0.907 0.984
LTFC 0.991 0.992 Table 4
Bilateral asymmetry for the measurements evaluated. S.D. = Standard deviation.

Results Measurement Side Mean S.D. p-value (paired t-test)

Left 89.408 6.509


FNAL 0.823
Table 3 shows the intra and inter-observer errors results for each Right 89.272 6.484

measurement. All measurements reported high concordance between


Left 29.555 3.145
observations, with the measurements FNAL, FNW, GTFC and LTFC FNW 0.362
Right 29.768 3.295
showing almost perfect agreement between observations. Table 4 shows
the results of bilateral asymmetry. There is not statistically significant Left 26.222 3.982
AB 0.407
Right 26.771 3.754
difference between both sides in any measurement except GTLT, which
showed higher mean in the right side compared to the left side. Left 50.174 6.489
Table 5 summarizes the descriptive statistics of the measurements AC 0.265
Right 51.711 4.927
segregated by sex. All measurements showed that males had higher means
compared to females, and there is no overlap of the confidence intervals Left 55.635 6.047
BC 0.471
Right 56.200 6.349
between sexes. Tests conducted indicated that there are statistically
significant differences between sexes in all measurements. Left 90.409 6.483
Following the exclusion criteria, 19 sex-predictive models were GTFC 0.143
Right 89.656 6.425
selected from the 38 logistic regressions performed. Table 6 summarized
Left 53.366 5.008
the equations, sensitivity, specificity, sex bias and overall accuracy of the GTLT 0.022
Right 54.605 4.936
models proposed for sex estimation on modern Chilean population from
the proximal end of the femur. Left 69.078 6.389
LTFC 0.538
The results showed that the multivariable models had higher overall Right 68.733 6.184
accuracy and sex-specific accuracies compared to univariable models.

4
D. Carvallo, R. Retamal FSIR 2 (2020) 100077

Table 5
Descriptive statistics for the measurements evaluated. S.D. = Standard Deviation; C.I. = Confidence Interval; M-W-W = Mann-Whitney-Wilcoxon test.
Measurement Sex Mean S.D. C.I. (95%) p-value (t-test/M-W-W test)

Female 82.894 4.572 81.986 - 83.801


FNAL <0.001
Male 95.060 4.749 94.118 - 96.003
Female 27.311 1.862 26.942 - 27.681
FNW <0.001
Male 31.789 2.152 31.362 - 32.216
Female 25.153 3.981 24.363 - 25.943
AB <0.001
Male 28.275 4.903 27.302 - 29.248
Female 46.606 4.339 45.745 - 47.467
AC <0.001
Male 51.869 5.671 50.744 - 52.994
Female 51.462 3.880 50.692 - 52.232
BC <0.001
Male 57.330 4.337 56.469 - 58.190
Female 83.588 4.700 82.655 - 84.520
GTFC <0.001
Male 96.268 4.830 95.309 - 97.226
Female 48.592 4.017 47.794 - 49.389
GTLT <0.001
Male 55.155 3.516 54.457 - 55.853
Female 63.051 4.627 62.132 - 63.969
LTFC <0.001
Male 70.746 5.456 69.664 - 71.829

Table 6
Proposed models for sex estimation on modern Chilean population from the proximal end of the femur.
Model Equation Sensitivity Specificity Sex Bias Overall Accuracy

FNAL 57.513 + 0.646*FNAL 97.1% 88.6% 8.5% 92.9%


FNW 32.727 + 1.115*FNW 85.7% 85.7% 0% 85.7%
GTFC 57.525 + 0.639*GTFC 97.1% 88.6% 8.5% 92.9%
GTLT 24.376 + 0.469*GTLT 82.9% 80.0% 2.9% 81.4%
LTFC 20.255 + 0.304*LTFC 85.7% 80.0% 5.7% 82.9%
FNAL + FNW 61.222 + 0.495*FNAL + 0.580*FNW 94.3% 91.4% 2.9% 92.9%
FNAL + BC 62.143 + 0.609*FNAL + 0.144*BC 94.3% 91.4% 2.9% 92.9%
FNAL + GTLT 71.124 + 0.594*FNAL + 0.345*GTLT 94.3% 91.4% 2.9% 92.9%
FNW + AC 35.963 + 1.064*FNW + 0.098*AC 88.6% 88.6% 0% 88.6%
FNW + BC 39.301 + 0.996*FNW + 0.186*BC 88.6% 91.4% 2.8% 90.0%
FNW + GTFC 64.519 + 0.606*FNW + 0.518*GTFC 100% 91.4% 8.6% 95.7%
FNW + LTFC 38.981 + 0.962*FNW + 0.161*LTFC 91.4% 88.6% 2.8% 90.0%
AB + BC 20.233 + 0.089*AB + 0.329*BC 85.7% 80.0% 5.7% 82.9%
AB + GTFC 58.458 + -0.173*AB + 0.700*GTFC 97.1% 91.4% 5.7% 94.3%
AB + GTLT 28.009 + 0.136*AB + 0.467*GTLT 91.4% 85.7% 5.7% 88.6%
AC + GTLT 31.291 + 0.165*AC + 0.445*GTLT 85.7% 80.0% 5.7% 82.9%
BC + GTFC 64.325 + 0.157*BC + 0.618*GTFC 97.1% 91.4% 5.7% 94.3%
GTFC + GTLT 66.268 + 0.562*GTFC + 0.297*GTLT 97.1% 91.4% 5.7% 94.3%
GTLT + LTFC 38.731 + 0.429*GTLT + 0.244*LTFC 91.4% 82.9% 8.5% 87.1%

sensitivity from 85.7% (AB+BC and AC+GTLT) to 100% (FNW+GTFC). the previous studies that evaluated the potential of these measurements for
The specificity ranged from 80% (GTLT and LTFC) to 88.6% (FNAL and the estimation of sex [5,8,10,11,15,16,18,19,21,27,30–35].
GTFC) in univariable models, and between 80% (AB+BC and AC+GTLT) All the models proposed showed low sex bias and high sensitivity and
and 91.4% (FNAL+FNW, FNAL+BC, FNAL+GTLT, FNW+BC, FNW specificity. Most of the models achieved higher accuracy for males than
+GTFC, AB+GTFC, BC+GTFC and GTFC+GTLT) for multivariable females, pattern commonly reported in the literature [5]. Given than in
models. The model FNW+GTFC showed both the highest sensitivity and this study a sample with equal representation for both sexes was used, this
specificity. The sex bias did not exceed 8.6% in any of the proposed pattern can be explained by the fact that, from a linear perspective, there
models. Except for FNW+BC, all models obtained higher accuracy for are more females with male characteristics than males with female
males than females. characteristics [28].
The present study is the first to deliver both univariable and
Discussion multivariable sex estimation models from the femur for Chilean
population. The best univariable models were FNAL and GTFC, with an
In the present study, we evaluated 8 measurements from the proximal overall accuracy of 92.9%. Previous studies conducted on the same
end of the femur on a modern Chilean sample and elaborated sex osteological collection have reported an overall accuracy up to 90.5% from
estimation models based on these measurements. the scapular glenoid cavity height [4]. Therefore, the measurements FNAL
The evaluation of the intra and inter-observer error showed high and GTFC are the most accurate individual sex estimators reported for
concordance between observations for all measurements, indicating that modern Chilean population. A general increased in overall accuracy was
they are repeatable and the models are replicable. All measurements observed in the multivariable models, with most achieving overall
showed symmetry between both sides except for GTLT for which the right accuracies over 90%. The best multivariable models were AB + GTFC,
side have higher mean than the left side. Considering this, the results here BC + GTFC and GTFC + GTLT with 94.3% overall accuracy, and FNW +
presented do not support the use of models that include GTLT on right GTFC with 100% of males correctly classified, 91.5% of females correctly
femora. classified and 95.7% of overall accuracy, making it the best sex estimation
The results of this study indicate that all the proximal femur measure- model described to the date for modern Chilean population.
ments evaluated in the Chilean sample have significant sexual dimorphism, The findings of this study are in concordance with the literature
with higher values in males compared to females. This is in concordance with establishing that the long bones, particularly the femur, are reliable sex

5
D. Carvallo, R. Retamal FSIR 2 (2020) 100077

estimators and support the position that this element should not be victims died during the Chile protests occurring since October 2019
considered as an alternative to pelvis and skull but instead as a first line of until the present. Burned and calcinated bodies have been found, which
evidence, comparable to the coxae and preferable over other elements may prevent the identification of the individuals based on DNA, so it is
like the cranium [2,13]. important to have osteological methods for identify these remains.
The performance of the models can be explained by the nature of the However, since sexual dimorphism may change through time due to
sexual dimorphism of the measurements evaluated. The combined selective secular change [17,20,26], it is essential to validate the models
forces of bipedalism and reproduction are visible in the angle and length of proposed here in a contemporary Chilean sample. Similar validations
the femoral neck [8]. Since FNAL and GTFC are both metric approaches of the must be done for its application on other Latin American population,
length of the femoral neck, it was expectable that these measurements would since recent studies had proved that these can differ significantly in
be reliable sex estimators. In the Chilean sample obtained for this study, FNAL terms of skeletal dimorphism expression despite being frequently
showed a 92.9% of overall accuracy in both univariable and multivariable grouped together as a homogeneous group [40].
models, which is higher than the 76% – 83.51% reported in Turkish,
American and Portuguese samples [5,11,21,30]. GTFC, on the other hand, Conclusions
was more consistent with the literature, with an overall accuracy for
univariable and multivariable models comparable to the around 90% of In the present study, sex estimation models based on osteometric
success reported for American and Spanish samples [8,10]. measurements from the proximal end of the femur were elaborated
FNW was the second-best sex estimator after FNAL and GTFC, with 85.7% using logistic regression on data from a modern Chilean sample. The
of overall accuracy in the univariable model and over 90% in multivariable proposed models achieved an overall accuracy between 81.4% and
models along with other measurements. The width of the femoral neck is 95.7%. Particularly, the results showed that the femoral neck
determined primarily by body weight [33], which is significantly influenced measurements are the best sex estimators among the variables
by biological sexbut also by many other factors [15]. In this way, FNW is a less evaluated. The models presented here are the most accurate reported
direct sex estimator compared to FNAL and GTFC. This can explain in part the to date for Chilean population. These models could be employed as a
high population variability in FNW sexual dimorphism, with previous studies novel tool for the identification of human remains in the contexts of
reporting lower accuracies in African, Iranian and Portuguese samples Human Rights violations committed during the Civic-Military Dicta-
[5,11,18,33] and higher accuracies in American, Guatemalan, Spanish and torship in Chile (1973–1990). Validation of these models on proper
Turkish samples [10,16,19,21,31,32]. samples may extend its utility to contemporary Chilean and Latin
The results showed that the femoral neck measurements, length and American populations.
width, are the best sex estimators among the measurements evaluated.
Accordingly, the best sex-predictive model was FNW + GTFC, which Fundings
incorporates both femoral neck dimensions.
When compared, GTLT and LTFC were less accurate than the femoral This research did not receive any grant from funding agencies in the
neck measurements. As individual estimators, these measurements public, commercial or not-for-profit sectors.
reported an overall accuracy of 82.9% and 85.7%, respectively, while
reaching up to 90% (LTFC) and 94.3% (GTLT) along with other
Declaration of Competing Interest
measurements in multivariable models. This is possibly because GTLT
and LTFC do not constitute a metric approach to a specific biomechanical
None.
parameter of the proximal femur like FNAL, FNW or GTFC. Instead, these
measurements seek to represent the overall morphology of the proximal
Acknowledgements
epiphysis [8], and therefore are indirect approximations of the sexual
dimorphism of the region.
The authors would like to thank Franco Castellani for the
The variables of the “Purkait’s triangle” (AB, AC and BC) showed the
photographic material presented here and Dr. Mauricio Hernández
lowest performance among the measurements evaluated, with none of the
and Dr. Eduardo Machicado for English revision. We thank Marisol
univariable models reaching acceptable levels of overall accuracy and/or
Intriago for the information on the identification process of the Pinochet’s
sex bias. Only multivariable models that included variables from Purkait’s
dictatorship victims, carried out by the Santiago Medico-Legal Service.
triangle along with other measurements meet the established criteria.
We also thank the Department of Anthropology of the University of Chile
This poor performance is consistent with the results reported by previous
for the access to the Modern Santiago Osteological Collection.
authors for these measurements in Indian, American, Greek and Spanish
samples, with comparable accuracies below the 80% [15,16,34,35]. A
possible explanation for this generally low accuracy may be found in the References
biomechanical properties of the entheses located in the proximal end of
[1] A.H. Ross, M.J. Manneschi, New identification criteria for the Chilean population:
the femur. Given the difference in size between male and female femora,
estimation of sex and stature, Forensic Sci. Int. 204 (2011) 206, doi:http://dx.doi.org/
greater distances between the points of Purkait’s triangle are expected. 10.1016/j.forsciint.2010.07.028 e1-3.
However, points B and C are located in entheses, which tend to be more [2] M.K. Spradley, R.L. Jantz, Sex estimation in forensic anthropology: skull versus
prominent in males since they have greater muscle robusticity. This postcranial elements, J. Forensic Sci. 56 (2011) 289–296, doi:http://dx.doi.org/
10.1111/j.1556-4029.2010.01635.x.
greater prominence of the muscle attachments have a counterproductive [3] A.M. Christensen, N.V. Passalacqua, E.G. Bartelink, Forensic Anthropology: Current
effect, shortening the distances between the points and thus reducing the Methods and Practice, Elsevier, San Diego, CA, 2014, doi:http://dx.doi.org/10.1016/
difference with the females. This makes Purkait’s triangle less sexually C2013-0-09760-5.
[4] C. Garrido-Varas, T. Thompson, A. Campbell, Parámetros métricos para la
dimorphic and the measurements AB, AC and BC less reliable as sex determinación de sexo en restos esqueletales chilenos modernos, Chungar 46 (2014)
estimators than other measurements evaluated. 285–294, doi:http://dx.doi.org/10.4067/S0717-73562014000200009.
It is well established that sexual dimorphism in the human skeleton [5] F. Curate, J. Coelho, D. Gonçalves, C. Coelho, M. Ferreira, D. Navega, E. Cunha, A
method for sex estimation using the proximal femur, Forensic Sci. Int. 266 (2016) 579,
is population-specific. Because of this, the elaboration of sex-predictive doi:http://dx.doi.org/10.1016/j.forsciint.2016.06.011z e1-7.
models requires the use of a representative sample of the target [6] T.R. Peckmann, C. Logar, S. Meek, Sex estimation from the scapula in a contemporary
population [3,5,9–12,15–17,19–23]. In the present study, a modern Chilean population, Sci. Justice 56 (2016) 357–363, doi:http://dx.doi.org/10.1016/j.
scijus.2016.05.003.
Chilean sample was employed in order to contribute to the identifica-
[7] J.E. Buikstra, D.H. Ubelaker, Standards for Data Collection From Human Skeletal
tion of victims of the Civil-Military Dictatorship (1973–1990). Remains, Arkansas Archeological Survey Research Series No. 44 Fayetteville, AR,
Additionally, the models presented here may contribute to identifying (1994) , doi:http://dx.doi.org/10.1002/ajhb.1310070519.

6
D. Carvallo, R. Retamal FSIR 2 (2020) 100077

[8] J. Albanese, G. Eklics, A. Tuck, A metric method for sex determination using the [25] C. Lemp, M. Balboa, R. Retamal, E. Aspillaga, Arqueología del depósito: manejo
proximal femur and fragmentary hipbone, J. Forensic Sci. 53 (2008) 1283–1288, doi: integral de las colecciones bioantropológicas en el, 12, Departamento de Antropología
http://dx.doi.org/10.1111/j.1556-4029.2008.00855.x. de la Universidad de Chile, Conserva, 2008, pp. 69–96. Available at: http://www.
[9] G. Soni, U. Dhall, S. Chhabra, Determination of sex from femur: discriminant analysis, patrimoniocultural.gob.cl/Recursos/Contenidos/Centro%20de%20Conservaci%
J. Anat. Soc. India 59 (2010) 216–221, doi:http://dx.doi.org/10.1016/S0003-2778 c3%b3n/archivos/DocAdjunto_1342.pdf.
(10)80029-2. [26] V. Abarca, Efectos De La Nutrición Sobre El Dimorfismo Sexual Expresado En La
[10] A. Clavero, M. Salicru, D. Turbon, Sex prediction from the femur and hip bone using a Estatura (SSD) De Una Muestra De Población Chilena Subactual, Facso Avaiable at:,
sample of CT images from a Spanish population, Int. J. Leg. Med. 129 (2015) 373–383, University of Chile, Santiago, Chile, 2011http://repositorio.uchile.cl/handle/2250/
doi:http://dx.doi.org/10.1007/s00414-014-1069-y. 106343.
[11] F. Curate, C. Umbelino, A. Perinha, C. Nogueira, A. Silva, E. Cunha, Sex determination [27] J.R. Center, T.V. Nguyen, N.A. Pocock, K.A. Noakes, P.J. Kelly, J.A. Eisman, P.N.
from the femur in Portuguese populations with classical and machine-learning Sambrook, Femoral neck axis length, height loss and risk of hip fracture in males and
classifiers, J. Forensic Leg. Med. 52 (2017) 75–81, doi:http://dx.doi.org/10.1016/j. females, Osteoporos. Int. 8 (1998) 75–81, doi:http://dx.doi.org/10.1007/
jflm.2017.08.011. s001980050051.
[12] D. Kim, D. Kwak, S.H. Han, Sex determination using discriminant analysis of the [28] P. du Jardin, J. Ponsaille, V. Alunni-Perret, G. Quatrehomme, A comparison between
medial and lateral condyles of the femur in Koreans, Forensic Sci. Int. 233 (2013) 121– neural network and other metric methods to determine sex from the upper femur in a
125, doi:http://dx.doi.org/10.1016/j.forsciint.2013.08.028. modern French population, Forensic Sci. Int. 192 (2009) 127, doi:http://dx.doi.org/
[13] J. Albanese, A metric method for sex determination using the hipbone and the femur, J. 10.1016/j.forsciint.2009.07.014 e1-6.
Forensic Sci. 48 (2003) 263–273 PMID: 12664981. [29] L. Lin, A concordance correlation coefficient to evaluate reproducibility, Biometrics 45
[14] E.F. Kranioti, N. Vorniotakis, C. Galiatsou, M.Y. Iscan, M. Michalodimitrakis, Sex (1989) 255–268, doi:http://dx.doi.org/10.2307/2532051.
identification and software development using digital femoral head radiographs, [30] R.A. Meeusen, A.M. Christensen, J.T. Hefner, The use of femoral neck Axis length to
Forensic Sci. Int. 189 (2009) 113, doi:http://dx.doi.org/10.1016/j. estimate sex and ancestry, J. Forensic Sci. (2015) 1300–1304, doi:http://dx.doi.org/
forsciint.2009.04.014 e1-7. 10.1111/1556-4029.12820.
[15] I. Anastopoulou, C. Eliopoulos, E. Valakos, S. Manolis, Application of Purkait’s triangle [31] R.M. Seidemann, C.M. Stojanowski, G.H. Doran, The use of the supero-inferior femoral
method on a skeletal population from southern Europe, Forensic Sci. Int. 245 (2014) neck diameter as a sex assessor, Am. J. Phys. Antropol. 107 (1998) 305–313, doi:http://
203, doi:http://dx.doi.org/10.1016/j.forsciint.2014.10.005 e1-4. dx.doi.org/10.1002/(SICI)1096-8644(199811)107:3<305::AID-AJPA7>3.0.CO;2-A.
[16] M. Djorojevic, C. Roldan, M. Botella, I. Alemán, Estimation of Purkait’s triangle [32] C.M. Stojanowski, R.M. Seidemann, A reevaluation of the sex prediction accuracy of
method and alternative models for sex assessment from the proximal femur in the the minimum supero-inferior femoral neck diameter for modern individuals, J.
Spanish population, Int. J. Leg. Med. 130 (2016) 245–251, doi:http://dx.doi.org/ Forensic Sci. 44 (1999) 1215–1218, doi:http://dx.doi.org/10.1520/JFS14589J.
10.1007/s00414-015-1201-7. [33] A. Mitra, B. Khadijeh, A.P. Vida, R.N. Ali, M. Farzaneh, V.F. Maryam, Y. Vahid, Sexing
^
[17] P. Guyomarc’h, J. Velemínská, P. Sedlak, M. Dobisíková, I. Svenkrtová,

J. Bružek, based on measurements of the femoral head parameters on pelvic radiographs, J.
Impact of secular trends on sex assessment evaluated through femoral dimensions of Forensic Leg. Med. 23 (2014) 70–75, doi:http://dx.doi.org/10.1016/j.
the Czech population, Forensic Sci. Int. 262 (2016) 284, doi:http://dx.doi.org/ jflm.2014.01.004.
10.1016/j.forsciint.2016.02.042 e1-6. [34] R. Purkait, Triangle identified at the proximal end of femur: a new sex determinant,
[18] S.A. Asala, M.A. Bidmos, M.R. Dayal, Discriminant function sexing of fragmentary Forensic Sci. Int. 147 (2005) 135–139, doi:http://dx.doi.org/10.1016/j.
femur of South African blacks, Forensic Sci. Int. 145 (2004) 25–29, doi:http://dx.doi. forsciint.2004.08.005.
org/10.1016/j.forsciint.2004.03.010. [35] R. Brown, D. Ubelaker, M. Schanfield, Evaluation of Purkait’s triangle method for
[19] L. Frutos, Brief communication: sex determination accuracy of the minimum supero- determining sexual dimorphism, J. Forensic Sci. 52 (2007) 553–556, doi:http://dx.
inferior femoral neck diameter in a contemporary rural Guatemalan population, Am. doi.org/10.1111/j.1556-4029.2007.00423.x.
J. Phys. Anthropol. 122 (2003) 123–126, doi:http://dx.doi.org/10.1002/ajpa.10227. [36] D.W. Hosmer, S. lemeshow, R.X. Sturdivant, Applied Logistic Regression, John Wiley
[20] V. Abarca, Desarrollo De Nueva Fórmula Para Estimar Estatura En Población Chilena & Sons, Inc., Hoboken, NJ, 2013, doi:http://dx.doi.org/10.1002/0471722146.
Adulta, a Partir Del Largo Del Fémur URI:, INTA, Santiago, Chile, 2013http:// [37] F. Curate, A. Albuquerque, I. Ferreira, E. Cunha, Sex estimation with the total area of
repositorio.uchile.cl/handle/2250/145652. the proximal femur: a densitometric approach, Forensic Sci. Int. 275 (2017) 110–116,
[21] O. Gulhan, K. Harrison, A. Kiris, A new computer-tomography-based method of sex doi:http://dx.doi.org/10.1016/j.forsciint.2017.02.035.
estimation: development of Turkish population-specific standards, Forensic Sci. Int. [38] V. Alunni, Pd. Jardin, L. Nogueira, L. Buchet, G. Quatrehomme, Comparing
(2015) 2–8, doi:http://dx.doi.org/10.1016/j.forsciint.2015.07.015. discriminant analysis and neural network for the determination of sex using femur
[22] V. Alunni-Perret, P. Staccini, G. Quatrehomme, Sex determination from the distal part head measurements, Forensic Sci. Int. 253 (2015) 81–87, doi:http://dx.doi.org/
of the femur in a French contemporary population, Forensic Sci. Int. 175 (2008) 113– 10.1016/j.forsciint.2015.05.023.
117, doi:http://dx.doi.org/10.1016/j.forsciint.2007.05.018. [39] J. Boldsen, G. Milner, S. Boldsen, Sex estimation from modern American humeri and
[23] M.K. Spradley, B.E. Anderson, M.L. Tise, Postcranial sex estimation criteria for femora, accounting for sample variance structure, Am. J. Phys. Anthropol. 158 (2015)
mexican hispanics, J. Forensic Sci. 60 (2014) S27–S31, doi:http://dx.doi.org/ 745–750, doi:http://dx.doi.org/10.1002/ajpa.22812.
10.1111/1556-4029.12624. [40] L. O’Bright, T.R. Peckmann, S. Meek, Is "Latin American" population-specific?
[24] C. Garrido-Varas, M. Intriago, Managing commingled remains from mass graves: Testing sex discriminant functions from the Mexican tibia on a Chilean sample,
considerations, implications and recommendations from a human rights case in Chile, Forensic Sci. Int. 287 (2018) 223, doi:http://dx.doi.org/10.1016/j.
Forensic Sci. Int. 219 (2012) e19–24, doi:http://dx.doi.org/10.1016/j. forsciint.2018.03.046 e1-7.
forsciint.2011.11.035.

You might also like