Sex and height estimation using percutaneous ulna and tibia length in a Ghanaian population: New data and a test of published equations

Whether one population-specific model for sex and height estimation can be extrapolated to another population has been a source of controversy. This study sought to develop and compare models for sex and height estimation from the percutaneous lengths of the ulna and tibia in a Ghanaian population. The study was cross-sectional from January to June 2021 at the University for Development Studies, Tamale. There were 191 (male=89, female=102) participants between 18 and 30 years of age. The standing height, ulna length (UL) and tibial length (TL) were measured following recommended anthropometric techniques. The sample was randomly assigned to training (60%) and holdout (40%) samples. Discriminant models for sex and linear regression models for height estimations were formulated using the training sample. The new models and other population-specific models were tested on the holdout sample for reliability using the Bland-Altmann method, Cohen’s d for height and cross-validation for sex estimations. The observed and the estimated height of males and females using UL (bias, d: male=2.75, 0.46; female= 0.73, 0.13), TL (bias, d: male= 2.74, 0.42; female= 1.50, 0.28) or UL+TL (bias, d: male= 2.76, 0.42; female= 0.86, 0.14) were not statistically different. The average sex estimation accuracy from the holdout sample was better in the multivariable UL+TL (82.9%) than in the univariable UL (76.3%) or TL (55.3%). Models based on UL [bias: 0.50 (95%CI: − 8.10 to 9.09), d: 0.07] from Kumasi-Ghana and TL [bias: − 0.01 (95%CI: − 10.37 to 10.34), d: 0.00] from the Amhara Region-Ethiopia were most reliable for estimating male’s height. For female height estimation, models based on UL [bias: − 0.62 (95%CI: − 12.08 to 10.83), d: 01.2] from Tamale-Ghana and TL [bias: − 3.72 (95%CI: − 16.81 to 9.37), d: 0.76] from the Amhara Region-Ethiopia were most reliable. However, the other models deviated in the range of − 17.09 cm [Khasi tribe India (UL)] to 0.54 cm [Kumasi-Ghana (UL)] in male height estimation and − 9.86 cm [Barcelona-Spain (TL)] to 1.05 cm [Kumasi-Ghana (UL)] in estimating female height. Cohen’s d was more precise than the Bland-Altmann method in assessing the reliability of height estimation models. Height can be estimated using UL, TL or UL+TL, however, it is recommended to use UL+TL for sex estimation in a Ghanaian population. Although the use of population-specific models (UL and TL) for sex and height estimations is recommended, other models outside the target population may be applicable. Additionally, when measuring the reliability of height estimation models (UL and TL), Cohen’s d should be preferred to the Bland-Altmann method.