Determination of sex from radiographic measurements of the humerus by discriminant function analysis in Saudi Population, Qassim region, KSA
Fahad Al Shehri(1) and Khaled E. A. Soliman(2,3)
(1) Assistant Prof. of Radiology, Vice Dean of Medical Affairs, Medicine College, Qassim University, Medical Director, Suleiman Al Habib Hospital, Al Qassim, KSA
(2)Basic Medical Sciences Department, Pathology & Forensic Medicine Unit, Unaizah College of Medicine, Qassim University, Al Qassim KSA
(3)Forensic Medicine & Clinical Toxicology Department, Sohag College of Medicine, Sohag University, Egypt. dr_ksayed@yahoo.com
Abstract
Background: Diagnosis of sex from skeleton or individual bone plays an important role in identifying unknown bodies, parts of bodies or skeletal remains for forensic purposes. This study aims to examine the applicability of the measurements taken from the humerus to assess sex, and to contribute to establishing discriminant function equations for Saudi populations for medico legal applications. Material & Methods: Archived x-ray radiographs of humerus for patients (216 males & 171 females) who attended the orthopedic clinics at Suleiman Al-Habib Hospital, Qassim region, KSA in the period from January 2011 to December 2013 were reviewed and analyzed. Five dimensions, including maximum length, vertical head diameter, diameter of head + greater tubercle, right-left diameter at midshaft, and epicondylar breadth were taken and subjected to Univariate and multivariate discriminant function analysis. Results: The studied radiographic dimensions of the humerus indicate that there are significant differences (p < 0.05) between the males and females measurements while the difference between right and left measurements was not significant. The findings revealed that the proximal part of the humerus has greater diagnostic accuracy than distal and middle parts. Accuracy of correct classification varies between 68.0% (epicondylar breadth) and 90.4% (vertical head diameter) for univariate analyses. When the multivariate analyses were conducted, three functions were produced, with the accuracy of ranging between 88.4% and 94.3%. Conclusion: These findings suggested that the dimensions of the humerus, especially the measurements taken from the proximal parts, could be used successfully for sex diagnosis.
Keywords: Sex determination, Forensic anthropology, humerus, discriminant function analysis
Introduction
Sex determination is a vital part of the medico-legal system but can be difficult in cases where the integrity of the body has been compromised. [1, 2] In adult skeleton, sex determination is usually the first step of identification process as subsequent methods for age and stature estimation are sex dependent. Marked sexual dimorphism is exhibited by several bones and they are suitable for sexing skeletons with high accuracy. The reliability of sex determination depends on the completeness of the remains and the degree of sexual dimorphism inherent in the population. The two most sexually dimorphic elements of the skeleton are pelvis and skull. [3] The sexing accuracy for various bones reported in literature is 80%-92% for cranium, 90% for skull and mandible, 95 – 98% for pelvis and 80% for long bones. [4] Larger limb bones may provide clear evidence of sex particularly if other individuals of same race and of both sexes are available for comparison.[3] Metric methods for sex determination are based on taking measurements of various dimensions of skeletal material. These methods are more easily repeatable than morphological methods because they rely on standardized osteometric points. In addition, metric methods are more objective than non-metric methods, because osteometric landmarks tend to be easier to find on a consistent basis and their assessment is not based on judgment against a scale of expression. Another strength of metric analysis is that simple measurements can be transformed into indices, thus eliminating the bias in using size itself as a sex indicator.[5] Populations have different morphological and metric manifestations in both the sexes.[6] Therefore, it is necessary to have population specific standards from skeletal collections.[7,8]
Sexual dimorphism in the humerus has been studied intensively and standards have been obtained for several different ethnic groups; in South Africa[9], India[10], Japan[4, 11], Germany[12], North Americans[13], Guatemala[14] and Greece.[1] As the extent of sexual variation in Saudi populations from humerus measurements has so far not been quantified by discriminant function analysis, the present study aimed to examine the applicability of the measurements taken from the humerus to assess sex and create a sex estimation technique using radiographic measurements of the humerus, derived from Qassim region populations, KSA.
Materials and Methods
This study was conducted as a retrospective design by reviewing x-ray radiographs of humerus for patients who attended orthopedic clinic at Suleiman Al-Habib hospital, Qassim region, KSA in the period from January 2011 – January 2013. The patients’ reasons of admittance to the hospital were mainly non-serious traumas and pain. They had no disorders of bone metabolism and/or other developmental diseases. In total, 387 x-ray radiographs (216 males and 171females; 174 right and 213 left) were compiled from the archives. All the cases in this study were above 20 years of age. The ages of the individuals vary between 20 and 62 years, with an average of 38.3 (SD = 11.45) years. This study was approved by the Faculty Ethics Committee.
Conventional AP radiographs of the entire humerus in this study were taken at the radiology department of above mentioned hospital using “AXIOM Vertix MD Trauma Digital X Ray†(Siemens, Germany) machine with a 66–70 kV / 12.5–16 mAs intervals. The x-ray films was taken when the subject was standing, their elbows fully extended, hand was at supination, shoulders and elbow joints were positioned to allow the pictures to be taken from the AP direction. In addition, greater tubercle was visible in lateral profile, and medial and lateral epiconyles of the humerus were visible in parallel on the radiographs. A total of 5 humeral dimensions were taken of the upper arms and shoulders using a digital caliper on x-ray films (Figures 1-a & b). The variables and their measurement techniques are as follows:
Maximum length: The distance between the most superior point on the head of the humerus and the most inferior point on the trochlea.[15, 16]
Vertical diameter of head: Taken from the most inferior point on the edge of the articular surface of the bone across to the opposite side on x-ray films.[17]
Diameter of head + greater tubercle: The diameter taken from the most inferior point on the edge of the articular surface across to the most posteriorly protruding point of the greater tubercle of humerus. [15, 16]
Right-left diameter at midshaft: The maximum right-left diameter was found at midpoint of the shaft, as the measurements horizontal to the long axis of humerus.[15, 16]
Epicondylar breadth: Distance of the most laterally protruding point on the lateral condyle from the corresponding projection of the medial epicondyle.[16]
To ensure the reliability of measurements, each measure was taken three times and then the average was calculated. The radiographs with pathological conditions or broken bones were excluded from the study. This study was approved by the Ethical Research Committee of the college.
Statistical analysis
Data assessments were carried out by the Statistical Package for Social Sciences (SPSS 19.0). Basic descriptive statistics were computed and student's t-test was applied to evaluate differences between the two sexes and two sides. The discriminant function analyses were made by both univariate and multivariate techniques.
Figure (1): Landmarks of studied humeral dimensions
A: Maximum length, B: vertical head diameter, C: right-left diameter at Midshaft, D: Epicondylar breadth, E: Diameter of the head + greater tubrecle
Results
Descriptive statistics for the studied measurements are shown in table 1. All measurements were signiï¬cantly greater (P < 0.05) in males than females. The most conspicuous difference between the sexes is in maximum length of humerus, whereas the smallest difference is in the midshaft maximum diameter. No significant difference between right and left side measurements.
Table 2 presents the univariate discriminant function. The lowest Wilks’ lambda value is vertical head diameter and diameter of head + greater tubercle. The values of eigen value and canonical correlation of these two measurements are higher than those of other variables. The highest Wilk’s lambda value was found for the variable of maximum length of humerus. The correct sex allocation data derived from univariate discriminant functions were also supported in the above findings (Table 3). Our analyses showed that correct sex discrimination rate was 90.4% using vertical diameter of the head. Second and third most successful variables were diameter of the head + greater tubercle (86.3%) and right left diameter in midshaft (75.5%). Least successful variable for sex allocation was epicondylar breadth (68.0%).
In addition to univariate analyses, multivariate methods were also applied. For this, three ways of techniques were used, Table 4. First, all of the variables obtained were calculated using stepwise regression analysis and then the best result was chosen, and displayed as Function 1. Second, the measurements of the humerus were analyzed two or three times and the most accurate equation was displayed as Function 2. Third, all measurements were included in the analysis to construct Function 3. It can be argued that the most accurate equation is Function 1, based on the coefficients of Wilks’ lambda. In this framework, the second most accurate equation is Function 3. Using the leave-one-out method, the functions developed were compared in terms of correct allocation of sex (Table 5). The most successful one is the Function 1 with the accuracy of 94.3% correct allocation followed by function 3 and 2 (92.2% & 88.4%).
In order to determine the sex of an individual, the discriminant score is calculated by multiplying each dimension by its unstandardized coefficient. These are added to along with the constant. A score greater than the sectioning point indicate male, and smaller female.
Table (1): Descriptive statistics of radiographic measurements (mm) of humerus
Sex
Measurement Male Female
Rthumerus Left humerus Rthumerus Left humerus
Maximum length of humerus 334.40±17.65* 332.59±15.71* 320.154±8.06 320.01±14.29
Vertical diameter of humerus head 50.90 ± 2.90 * 49.75 ± 2.15 * 43.57 ± 1.93 43.97 ± 2.30
Diameter of head + greater tubercle 55.90 ± 2.90 * 54.47 ± 2.15 * 49.28 ± 1.93 49.50 ± 2.30
Right-left diameter 24.42 ± 2.82 * 23.58 ± 2.30 * 20.58 ± 1.79 20.69 ±1.85
Epicondylar breadth 61.82 ± 5.34* 61.12 ± 4.54* 57.70 ± 4.73 56.96 ± 4.12
* p ≤ 0.05 (for corresponding male–female values).
Table (2): Univariate discriminant function analysis
Unstandardized coefficient Constant Wilk's lambda Eigen
Value Group centroid Sectioning point F-statistic
Maximum length of humerus 0.414 -22.768 0.829 0.207 Male = 0.404
Female = -0.510 – 0.053 25.702*
Vertical diameter of humerus head 0.810 -20.202 0.343 1.914 Male = 1.228
Female = -1.551 -0.1615 5.870 *
Diameter of head + greater tubercle 0.777 -22.472 0.397 1.519 Male = 1.094
Female = – 1.382 – 0.144 5.870 *
Right-left diameter 0.608 -10.592 0.631 0.585 Male = 0.679
Female = -0.857 – 0.089 54.986 *
Epicondylar breadth 0.427 -12.776 0.818 0.223 Male = 0.419
Female = -0.529 – 0.055 1.866 *
*p< 0.05
Table (3): Percentage of correct group membership for univariate analysis
Males Females Total
N % N % n %
Maximum length of humerus 168 77.8% 111 64.9% 279 72.1%
Vertical diameter of humerus head 208 96.3% 142 83.0% 350 90.4%
Diameter of head + greater tubercle 192 88.9% 142 83.0% 334 86.3%
Right-left diameter. 144 66.7% 148 86.5% 292 75.5%
Epicondylar breadth 152 70.4% 95 55.6% 274 68.0%
Table (4): Canonical discriminant function coefficient for the dimensions of the humerus
Unstandardized coefficient Constant Wilk's lambda Eigen value Structure Matrix Group centroid Sectioning point
Function 1:
Vertical diameter of humerus head
Epicondylar breadth
6.182
-1.710
-19.075
0.272
2.646 0.850
0.290
Male = 1.444
Female = -1.824 – 0.19
Function 2
Diameter of head + greater tubercle
Maximum length of humerus
0.4531
-20.574 -20.202 0.393 1.543 0.992
0.366
Male = 1.103
Female = -1.393 – 0.145
Function 3
Maximum humerus length
Vertical diameter of humerus head
Diameter of head + greater tubercle
Rifgt-Left diameter
Epicondylar breadth
0.095
6.159
Omitted
-0.061
-1.874
-20.969 0.274 2.675
0.278
0.289
0.468
0.846
0.846 Male = 1.452
Female = -1.834 – 0.191
Table (5): Percentage of correct group membership for multivariate analysis
Males Females Total
N % N % n %
Function 1 212 98.1% 154 90.1% 366 94.3%
Function 2 200 92.6% 142 83.0% 342 88.4%
Function 3 208 96.3% 149 87.1% 357 92.2%
Discussion
Sex determination is a very important component of any human skeletal analysis. Sex discriminatory functions which are obtained for each bone are very useful in mass disasters and criminal cases of multiple human burials, where charred bodies and scattered, mixed or incomplete remains are recovered. Since the osteometric methods for the determination of sex from the skeleton are population specific, researchers from around the world have conducted studies to establish group specific standards of assessment. Many studies have set osteometric standards for sexual dimorphism.[8, 18-20] Research on sex estimation focused on skeletal material for skeletal biology and forensic cases reveals that the humerus and its measurements can be used for sex allocation.[1, 4, 9, 11, 21, 22]
The existing literatures and the sex estimation equations they propose are derived from three different contexts of skeletal remains. The first group consists of human skeletal remains obtained from archaeological excavations. The sex determination of the specimens was estimated by using morphological or visual methods, and then, based on these materials, the sex determination formulae were developed.[23] In the second group, there are the studies focused on dry skeletons, with known sex.[1, 4, 9, 11, 12] The third group consists of radiographic films and the measurements taken from them.[22] The number of this type of study based on measurements taken from radiographic images is very limited. Among those methods mentioned above, the third one is the most reliable. In the first group, the sex of the skeletons could not be absolutely determined. In the second group, there is the risk of misplaced bones and/or records. On the other hand, authors pointed out that the reliability of the measurements taken from radiographs is fairly high.[24] This opinion was also supported by various studies based on the measurements taken from radiographs of computerized tomography and x-rays, such as on the cranium[25], the calcaneus[26], and the femur.[24, 27] In these studies, the sex was estimated with accuracy between 92.6% and 99.0%. For these reasons, we used x-ray films to establish the equation for sex determination based on the measurements of the humerus.
In the present study, all measurements were significantly greater in males than females. These findings are in conformity with the findings reported by Singh (1972),[10] Iscan et al., 1998,[11] Atamturk, (2010)[28] and Patil (2011).[8] The statistical significant difference of the measurements among males and females may be attributed to the early maturity of girls than boys; consequently, boys have more two years to develop physically. Our results suggest that the proximal portion of the humerus is a more successful indicator than the distal part. These results are in agreement with that obtained by other researchers.[1, 11, 12, 14, 28] Contrarily, other researchers found that the distal portion of the humerus is a more successful indicator than the proximal part[4, 9, 11] (Table 6). This can be explained by population and ethnic differences. Population differences in anthropological studies have been noted and it is well realized that they need to be studied separately.[29, 30]
By using multivariate analysis, the rate of correct classification in sex assessment from the measurements of the humerus in this study was ranged from 88.4% – 94.6 90%, with an accuracy of 94.6% for function 1 (vertical head diameter + epicondylar breadth). The results of the present study were compared with those of other studies[1, 9, 11, 14, 28] in respect to the accuracy of stepwise discriminant functions for sex identification derived from the humerus (Table 7). It can be said that among the stepwise discriminant functions, the classification accuracy is lowest for the equation derived from the material measured in China[11] and Turkey.[28] This can be explained by the population & ethnic differences as well the methodology used. Indeed, it is well known that circumferential measurements cannot be taken on radiographic film, only one-dimensional measurements such as length or width can be taken. On the other hand, when circumferential measurements were included to the equations for sex assessment, the precision of the formulae were increased.[4, 21, 23] For instance, in the equations derived from skeletal remains recovered from clandestine graves attributed to the recent armed conflict in Guatemala, circumferential and minimum midshaft diameter measurements were included.[14]
In comparison to the classification accuracies compiled from the literature, the capability of humerus in sex determination is quite similar to that of the main long bones of extremities. The accuracy was 91-95% in ulna [4, 12, 31], 90-97% in radius [4, 12,31, 32], 94-95% in tibia [4] and 91-95% in femur.[33, 34, 35] The findings of studies focused on other body parts except the long bones were evaluated. It can be said that the variability in correct classification of sex was higher than those of long bones, with the accuracy of 99% in cranium, [25] 96% in patella, [36] 95-97.7% in clavicle and scapula, [37] 87.6-95.7% in talus and calcaneus. [38]
Conclusion:
Considering all of the above findings, it was concluded that the measurement of the humerus is a relatively reliable method for sex assessment in anthropological studies and forensic applications.
Table (6): The best variable for sex determination based on the measurements of humerus according to the various authors
Population and reference Best variable Correct classification
Male % Female % Total %
This study Vertical head diameter 96.3% 83.0% 90.4%
Diameter of head + greater tubercle 88.9% 83.0% 86.3%
Turkey (Atamtrik et al., 2010) [28] Vertical head diameter 91.2% 94.9% 93.2%
Epicondylar diameter 76.0% 88.5% 83.0%
Greece (Kranioti and Michalodimitrakis, 2009) [1] Vertical head diameter 90.5% 89.3% 89.9%
Chinese (Ä°ÅŸcan et al., 1998) [11] Vertical head diameter 79.1% 82.1% 80.5%
Japanese (Ä°ÅŸcan et al., 1998) [11] Epicondylar breadth 93.2% 85.7% 89.9%
Thai (Ä°ÅŸcan et al., 1998) [11] Epicondylar breadth 91.4% 97.1% 93.3%
Guatemala (Frutos, 2005) [14] Maximum head diameter 93.5% 97.0% 95.5%
Africans whites (Steyn and Ä°ÅŸcan, 1999) [9] Epicondylar breadth 83.6% 95.8% 94.7%
Africans blacks (Steyn and Ä°ÅŸcan, 1999) [9] Vertical head diameter 93.0% 88.9% 94.7%
Table (7): Equation developed by various authors and their comparisons in respect of the percentage of correct classification of sex
Population and reference Function (stepwise) Correct classification
Male (%) Female (%) Total (%)
This study Vertical head diameter + epicondylar breadth 98.1% 90.1% 94.6%
Turkey (Atamtrik et al., 2010) [28] Vertical head diameter + maximum length 88.02% 91.4% 90.0%
Greece (Kranioti & Michalodimitrakis, 2009) [1] maximum length + vertical head diameter + minimum midshaft diameter + epicondylar breadth 92.8% 92.7% 92.9%
Chinese (Ä°ÅŸcan et al., 1998) [11] Maximum length + vertical head diameter + epicondylar breadth + midshaft diameter 85.4% 88.6% 86.8%
Japanese (Ä°ÅŸcan et al., 1998) [11] Vertical head diameter + minimum midshaft diameter + epicondylar breadth + midshaft diameter 95.5% 88.6% 97.0%
Thai (Ä°ÅŸcan et al., 1998) [11] Vertical head diameter + minimum midshaft diameter + epicondylar breadth 97.1% 97.1% 97.1%
Guatemala (Frutos, 2005) [14] Maximum head diameter + minimum midshaft diameter + epicondylar breadth 98.5% 97.8% 98.5%
Africans whites (Steyn and Ä°ÅŸcan, 1999) [9] Epicondylar breadth + vertical head diameter 89.1% 95.8% 92.5%
Africans blacks (Steyn and Ä°ÅŸcan, 1999) [9] Vertical head diameter + maximum length 95.1% 91.1% 93.1%