Sexual Dimorphism and a Cross-Validated Logistic Regression Model for Sex Estimation from Lumbar Vertebral Morphometry in a Central Indian Population

Abstract

Purpose: Lumbar vertebral morphometry has an established role in forensic sex estimation when the pelvis or skull are unavailable, but discriminant models are known to be population-specific. No sex-estimation model derived from lumbar vertebral morphometry has previously been reported for the Madhya Pradesh and Central Indian population. This study aimed to characterise sexual dimorphism across nine lumbar vertebral morphometric parameters and to develop a cross-validated logistic regression model for sex estimation in this population. Methods: Documented-sex dry adult lumbar vertebrae (61 males, 49 female; 110 total, 22 at each of L1-L5) from a departmental osteology collection were measured in triplicate for nine linear parameters. Sex differences were assessed using independent-samples t-tests with Levene's test and Cohen's d effect size. A binary logistic regression model was constructed using the parameters showing significant dimorphism, with performance evaluated by 10-fold stratified cross-validation and receiver operating characteristic (ROC) analysis. Results: Of nine parameters, only vertebral body height (male 29.14±1.49 mm vs. female 27.25±1.43 mm; t = 6.77, p<.001, d = 1.29) and transverse width (male 47.58±3.58 mm vs. female 46.02±3.83 mm; t = 2.19, p = .031, d = 0.42) differed significantly by sex; pedicle width, interpedicular distance, canal dimensions, and spinous process length did not (all p>0.45). A logistic regression model combining body height and transverse width classified sex with 83.6% cross-validated accuracy, 85.2% sensitivity, 81.6% specificity, and an area under the ROC curve of 0.904. Conclusion: Sexual dimorphism in the Central Indian adult lumbar vertebra is concentrated in the vertebral body rather than distributed across the vertebra as a whole, consistent with international forensic morphometric literature. The resulting logistic regression model provides the first population-specific tool for morphometric sex estimation from lumbar vertebral remains in this region.