|No.||Study||Types of bones||Sample population, number of subjects [male/female]||Methods||Results||Outcomes|
|1||Padovan et al. (2019)||C1 (atlas)||Brazil (110/81)||
• Four measurements were made in each atlas from samples of skeletal remains in the museum.|
• The unpaired t test was used to test sexual dimorphism between the measurements.
• The stepwise-forward Wald method was used to obtain a logistic regression.
• The unpaired t test showed significant sex differences for all measurements (p < 0.05)|
• Out of four variables, only two were selected by establishing the best model equation, i.e. anterior diameter of atlas and maximum transverse diameter of atlas.
|• The atlas showed 81.2% accuracy in sex estimation.|
|2||Marlow and Pastor (2011)||C2||
153 known-sex individuals
• Using Wescott’s eight proposed sexually dimorphic dimensions plus an additional dimension width of the vertebral foramen (WVF), from samples of the museum collection of skeletal remains.|
• The paired sample t test was
used to determine significant differences between sexes.
• Utilize discriminant function analysis (DFA) to evaluate its efficacy for sex estimation.
• The stepwise technique chose a subset of factors based on squared partial correlation and the significance level from an analysis of covariance with the greatest number of separation capacity.
• The discriminant function by stepwise procedure, generated from the C2 vertebra measurements had the option to appraise sex with an accuracy of 83.3% with the option to arrange males and females with equivalent precision.|
• By stepwise selection, maximum breadth across superior facets (SFB), maximum sagittal length (XSL), the width of the vertebral foramen (WVF), and dense sagittal diameter (DSD) form the discriminant functions.
• The dimension SFB is the best discriminator of sex.
• Testing the five discriminant functions equations by the Wescott’s method resulted in an accuracy classification of 76.99%.
• The function of the present study has been tested with an overall classification accuracy of 83.3% (p < 0.001).
|• The sex estimation method from Wescott’s measurements of the C2 vertebra displayed significant discrimination between males and females with 83.3% accuracy.|
|3||Bethard and Seet (2013)||C2||American||
• Following the Wescott’s measurements, five dimensions were measured from the skeletal remains collections.|
• The Wescott’s five discriminant functions were utilised to estimate sex in each case.
• The five discriminant functions by the Wescott’s method were highly replicable, that yielded an overall accuracy of 80% or greater.|
• The five discriminant functions by Wescott and Marlow & Pastor showed accuracy rates of 80% and more.
|• The second cervical vertebra can provide a wealth of information in sex estimation.|
|4||Gama et al. (2015)||C2||
First sample = (99/91) (training set).
Second sample = (24/23) (testing set)
• Acquired 13 measurements from the skeletal collections by sliding calipers with an approximation of 0.5 mm.|
• Performed a t-test (two-tailed) to analyse the differences in measurements between males and females.
• Using logistic regression model to construct estimation models.
• The most dimorphic dimensions were the LMA (11.18%) and DSMC (10.6%).|
• The most predictive variables were LMA, DSMC, CMA and LMFS (right side).
• The resulting model identified sex in 89.7% of cases in the training set, whilst sex was accurately identified in 86.7% of cases in the testing set.
|• The second cervical vertebra was useful for sex estimation with accuracies that ranged from 86.7 to 89.7%.|
|5||Torimitsu et al. (2016)||C2||Japanese (112/112)||
• Nine measurements were collected from cadavers by PMCT scanning and subsequent forensic autopsy was done.|
• ANOVA was utilised to examine mean differences between sex groups.
• Univariate discriminant function analysis (DFA) was performed on each variable to generate a formula for sex estimation.
• Measurements of DMFS and LMA on the C2 vertebra achieved expected cross-validated accuracies of 83.5% and 83.1%, respectively.|
• There were four variables (AS, DSD, DSMC, and DTMC), with correct prediction rates of 80%.
• A five-variable model with 92.9% accuracy rate was achieved.
|• CT scan of C2 vertebra showed good estimation of sex with high accuracy rate.|
|6||Kaeswaren and Hackman (2019)||C2-C7||
• Using 25 human cadavers, three morphometric traits were measured for each cervical vertebra.|
• A total of 150 cervical vertebrae were sampled.
• An independent (two-sampled) t test was conducted to establish sexual difference for all three measurements.
• Stepwise discriminant analysis was done to select the most dimorphic variable for sex estimation.
• Discriminant functions utilised all three vertebral measurements i.e. the vertebral body height (CHT), anterior-posterior diameter of the vertebral foramen (CAP) and transverse diameter of the vertebral foramen (CTR) for sex estimation, with 77.3% - 100% accuracy for each vertebra.|
• Sex was estimated with an accuracy of above 80% for each vertebra with C2 giving the most accurate sex estimation of 100% in all four combinations of vertebral measurements.
• Twenty-five discriminant functions out of 37 discriminant functions were significant with sex predicting accuracy greater than 80%.
• The CTR and CHT were found to contribute greatly towards biological sex variation.
|• This study developed 25 multivariate discriminant functions that successfully classified individuals as male or female with accuracy greater than 80%.|
|7.||Rozendaal et al. (2020)||C1–C7||Europeans (157/138)||
• Samples were taken from the skeletal cemetery collections|
• Three measurements were taken from each of the seven cervical vertebrae: Maximum cervical vertebrae body height (CHT), cervical anterior-posterior diameter (CAP) and cervical transverse diameter (CTR).
• Discriminant functions were generated for each cervical vertebra, using all three measurements, to establish whether sex could be estimated from a single vertebra.
• Multivariate discriminant functions were produced using all seven cervical vertebrae to investigate whether a combination of vertebrae may be used for sex estimation.
• The functions that achieved predicted accuracies of 80% or greater, were cross-validated on independent samples of 32 individuals from the skeletal collections.
• Results indicated that CAP measurement did not demonstrate sexual dimorphism, whilst CHT and CTR demonstrated significant difference between males and females (p < 0.002) (except for CTR of C1 and C2).|
• Using combinations of all three measurements for sex estimation from a single vertebra, the accuracies ranged from 66.9 to 74% for males and 70.2–79.5% for females.
• This study produced seven discriminant function equations using 20 measurements from all seven cervical vertebras, which achieved an overall accuracy rate of greater than 80%. The cross-validation test showed that among these functions, only four had achieved accuracies equal or greater than their predictive accuracies. The results indicated that C2 and C5 vertebrae were the most sexually dimorphic bones.
|• The discriminant function equations achieved accuracy rates of 84.5% for cervical vertebrae (used in combination) in the European population.|
|8||Yu et al. (2008)||T12||Korean (52/50)||
• Used 33 linear measurements and two ratios by CT scan.|
• An independent sample t test was performed to evaluate differences between the means of the parameters.
• The respective discriminant equations were calculated for sex estimation using univariate, bivariate and stepwise methods of discriminant function analysis.
• Accuracies of the discriminant equations were obtained by cross-validation procedure.
• Twenty-three single variables with significant sex differences among 35 traits contributed to correct sex classification rates ranging from 62.7 to 85.3%.|
• Three measurements on the vertebral body (sBDc, sBDcm, iBDcm, and iBDc) exhibited accuracies greater than 80%.
• The coronal diameter of the superior endplate of the vertebral body (sBDc), the ratio of the anterior to middle height of the body (Hm/Hp), and the length of the left mammillary process and pedicle (lM&PL) predicted sex with 90% accuracy by DFA.
|• The 12th thoracic vertebra was sexually dimorphic with 90% accuracy rate in Korean individuals.|
|9.||Hou et al. (2012)||T12||Chinese (78/63)||
• Using 30 measurements from CT scan samples.|
• The data were analysed by one-way analysis of variance (ANOVA).
• Univariate discriminant function analysis and stepwise discriminant function analysis were performed, respectively.
• A leave-one-out classification procedure was used to assess the validity of these functions.
• The accuracy of sex classification was between 56.4% and 90.1%.|
• Variables such as sBDs, sBDsm, sVL, sBDc, iBDs, iBDsm, iVL, iBDc, mBDs, mBDc and BHp displayed 80% accuracy.
• The iVL had the highest accuracy rate of 90.1%.
• By stepwise discriminant function analysis, an equation with four variables, i.e. three linear measurements (superior maximum sagittal diameter of vertebral body endplate (sBDsm), inferior length of the whole vertebra (iVL), the distance between superior articular processes (sAD), and one ratio (the ratio of anterior to posterior height of the vertebral body (BHa/BHp) were obtained with 94.2% accuracy.
|• The 12th thoracic vertebra was found to be sexually dimorphic in the North easterners in China with an accuracy rate of 94.2%.|
|10.||Amores et al. (2014)||C7 and T12||
• Using eight measurements from the skeletal collection samples.|
• T-test was used to compare the data between sexes and evaluate the homogeneity of variance (F test).
• The effectiveness of the measurements was analysed by discriminant function analysis.
• The discriminant capacity of the selected variables was then evaluated using cross-validation procedure.
• For C7, discriminant function analysis selected the length of vertebral foramen (LVF), length of inferior surface of vertebra (LIVB), and width of inferior surface of vertebral body (WIVB) as having the highest discriminant power.|
• For T12, the length of inferior surface of the vertebral body (LIBV) was selected for its discriminant capacity.
• The discriminant analysis yielded five functions, i.e. four for the 7th cervical and one for the 12th thoracic with 80% accuracy rate.
• The length of the vertebral bodies of the 7th cervical and 12th thoracic vertebrae offered the highest discriminant power for sex estimation.|
• The C7 and T12 vertebrae showed a higher accuracy rate of approximately 80%.
|11.||El Dine and El Shafei 2015||T12 and L1||
• Using 24 linear measurements and four ratios from the images of multi-slice computed tomography (MSCT)|
• t test was used to establish the difference between sexes.
• Unstandardized coefficient.
• Linear regression analysis was performed, in which each vertebral measurement was analysed for sex estimation.
• About 14 out of 24 linear measurements showed significant sex differences using T12 vertebra (predictive accuracy ranged from 49% to 85.5%), with three variables, i.e. lower endplate depth (EPDl), upper endplate width (EPWu), and superior vertebral length (VLs) as having more than 80% predictive accuracy rate.|
• By using the L1 vertebra, only seven linear measurements and one ratio were sexually dimorphic (predictive accuracy ranged from 47% to 79%), and only the upper endplate depth (EPDu) showed an accuracy above 75% (79%).
• The accuracy of the T12 vertebra was 93.1%.
• The accuracy of the L1 vertebra was 68%.
• With a combination of T12 and L1 vertebrae, only five variables were used in the equation that predicted sex with high level of accuracy (96.3%).
• The study did not analyse the discriminant equation by cross-validation procedure.
|• The T12 vertebra demonstrated a better sex estimation than L1 in the Egyptians. The accuracy increased when T12 and L1 were used in combination as sex predictors.|
|12.||Zheng et al. 2012||L1||China (113/97)||
• About 29 linear measurements were taken from 3D CT models, and five aspect ratios were calculated from linear measurements.|
• All measurements were significant by stepwise discriminant analysis (p < 0.01).
• Cross-validation of the discriminant function equations was performed to test the accuracy of the discriminant functions.
• About 25 traits demonstrated significant sexual dimorphism ranged from 57.1 to 86.6%. (p < 0.01)|
• EPWu showed the highest predictive accuracy.
• Discriminant functions were upper endplate width (EPWu), left pedicle height (PHI), and middle endplate depth (EPDm) with predicted sex accuracy of 88.6%.
|• The L1 vertebra may be used for sex estimation with 88.6% accuracy by discriminant function analysis.|
|13.||Ostrofsky and Churchill 2015||L1-L5||South Africa (47/51)||
• Samples comprised skeletal collection, and 11 measurements were taken to the nearest 0.1 mm with digital calipers.|
• To compare male and female sample means, t test was applied (for variables normally distributed) and a nonparametric Wilcoxon ranks sum test was used (for data not normally distributed)
• Each variable that showed significant sex difference was subjected to univariate discriminant function analysis (DFA) to test its effectiveness for sex estimation. (p < 0.01)
• The leave-one-out cross-validation procedure was conducted to test the accuracy of the discriminant functions.
• The highest accuracy was obtained by the measurements of the vertebral body.|
• Four variables of L1 and L2 vertebrae showed accuracies over 80%, i.e. vertebral body superior and dorso-ventral (BSDVD) and transverse diameters (BSTD).
• The discriminant functions predicted sex with accuracies over 80% for L1–L4, with the highest accuracy for L1 (87.1%).
|• The lumbar vertebrae exhibited the greatest degree of sexual dimorphism, which may be used for sex estimation.|
|14.||Ramadan et al. 2017||L1||Egyptian (61/62)||
• About 15 linear measurements of L1 were taken by MSCT.|
• An independent t test was applied to compare between different sexes.
• Correlation analysis was done followed by discriminant function analysis.
• Descriptive statistics showed significant differences between sexes for all measurements except the length of vertebral foramen (LVF). The upper endplate width (EPWu) showed the highest accuracy.|
• Additionally, sex could be predicted from L1 by discriminant analysis at an accuracy of 84.6%.
• The study did not analyse the discriminant equation for cross-validation procedure.
|• Sex could be estimated from L1 at 84.6% accuracy.|
|15.||Decker et al. (2019)||L1-L5||US-North American (76/78)||
• About 36 measurements were taken from abdominal CT scan comprising 30 linear measurements on vertebral body wedging angle, and five aspect ratios for each vertebra.|
• A stepwise analysis method used the measurements to produce individual discriminant equations for L1 through to L5 and in combination studies of the lumbar vertebrae, and all accuracies were obtained by cross-validation process.
• L1 vertebra had 21 out of 29 significant measurements, with a predictive accuracy of 83.1%.|
• L2 vertebra had 23 out of 29 significant measurements, with a predictive accuracy of 81.8%.
• L3 vertebra had 25 out of 29 significant measurements, with a predictive accuracy of 85.1%.
• L4 vertebra had 24 out of 29 significant measurements with a predictive accuracy of 82.5%.
• L5 vertebra had 23 out of 29 significant measurements with a predictive accuracy of 81.2%.
• The discriminant function for the five lumbar vertebrae had an overall accuracy rate ranged between 81.2 and 85.1% for sex estimation, with the highest accuracy attained by L3 vertebra.
• The study presented that the L1–L5 vertebrae can be used for sex estimation with accuracies ranging from 81.2 to 85.1%|
• The accuracy rate rose to 92.2% when all five vertebrae were used in combination.
|16.||Azofra-Monge and Alemán Aguilera 2020||L1–L5||Spain (46/48)||
• Samples comprised identified adult individuals from the skeletal collections.|
• Thirty three linear measurements were taken with digital calipers in millimetre.
• Gender differences between means were analysed by student’s t test and non-parametric Mann-Whitney U test, and the binary logistic regression was generated.
• The equations with higher sex classification rates were selected and validation was performed on a random selection of 20% of the original sample.
• All measurements from the L1 and L2 vertebrae were greater in males than females, in which 11 variables were significant. (p < 0.05)|
• The most sexually dimorphic vertebrae were L1 and L2.
• The total width (TW) was the variable selected in the equations for L1, L3 and L4 vertebrae. TW is the maximum distance between the ends of the transverse processes.
|• The discriminant equations for sex estimation showed accuracy rates that ranged from 90.1 to 94.5% for L1, 85.4% to 89.4% for L2, 85.3% to 88.3% for L3, 85.3% to 88.2% for L4 and 80% to 85.3% for L5.|
|17.||Oura et al. (2018)||L4||
Ages 20 = (147/228)
Ages 30 = (147/228)
Ages 46 = (618/745)
• Samples were obtained from MRI scans.|
• There were only three measurements from the vertebral body of L4 (width, depth and heights) as they were achievable in forensic context.
• A binary logistic regression was used for sex estimation and their accuracy was analysed.
• All the measurements were greater in men than in women.|
• The multivariate regression analyses which included mean width, depth, and height of L4 vertebra, and yielded good sex accuracies in all age groups (86.4%, 87.7% and 82.8% at the ages of 20, 30 and 46, respectively).
• The classification accuracy for sex were consistently higher for females than males.
• The models based on the 46-year-old sample showed lower sex estimation accuracy than the corresponding 20- and 30-year-old sample models.
|• The width, depth and height of the L4 vertebral body were found to be useful for sex estimation.|
|18.||Suwanlikhid et al. (2020)||L1–L4||Thai (75/75)||
• Samples of this study were dry bones that went through digital image processing after being captured by a camera.|
• Nine variables comprising the area, major and minor axes in total area, cortical and trabecular areas in each lumbar vertebra.
• Quantitative variables and sex were analysed using discriminant function analysis.
• There were significant differences between sexes in most variables and males were generally greater than females in most measurement variables. (p < 0.01)|
• The upper endplate of the L1 vertebra had the most predictive precision (81.8%). Most of the upper lumbar vertebrae had higher precision than the lower lumbar vertebrae.
|• The lumbar vertebra can be used for stature and sex estimation in incomplete skeletal remains.|
|19.||Garoufi et al. (2020)||T1, T12, L1||European: Greek (119/96) and Danish (62/55)||
• The samples comprised two modern contemporary European populations, i.e. the Greek sample of dry bones, and Danish sample of virtual vertebrae derived from CT 3D models.|
• Ten measurements were applied in each vertebra (T1, T12, L1).
• The analysis of variance (ANOVA) was used to analyse sexual dimorphism in each variable, whilst the univariate discriminant function analysis (DFA) was performed to analyse sex estimation accuracy.
• ANOVA showed that almost all of the measurements in three vertebrae (T1, T12, L1) were sexually dimorphic and the effect sizes were similar in two populations (Greek and Danish). The measurements, i.e. aVBH in T12 vertebra in Greek sample and aVBH in L1 vertebra in Danish sample did not yield any difference for sex estimation.|
• The discriminant function equations showed acceptable accuracies from cross-validation process with T1 showing the highest classification accuracy. These equations were based on three measurements, i.e. IEPW, mVL and mTD with 89.9% accuracy (Danish sample) and pVBH, mVL and mTD with 88.8% accuracy (Greek sample).
• Among the three vertebrae, T1 was the most reliable for sex estimation in both populations reaching almost 90% cross-validated accuracy.|
• The maximum vertebral lengths (mVL) of the T1, T12, L1 vertebrae showed the highest degree of sexual dimorphism.