Bone is characterised by age-related morphological and histological changes. We have previously established
an automated method of recording bone morphometry and histology from entire transverse sections of
cortical bone. Our aim was to determine whether data acquired using this automated system were useful in
the prediction of age. Ninety-six specimens of human femoral middiaphysis were studied from subjects aged
21–92 y. Equations predicting specimen age were constructed using macroscopic data (total subperiosteal
area (TSPA), periosteal perimeter (PP), endosteal perimeter (EP), cortical bone area (CA) and moments of
area) and microscopic data (the number, size and diversity of pores and intracortical porosity) together with
sex, height and weight. Both TSPA and PP were independent predictors of age but the number of pores was
not a significant predictor of age in any equation. The age predicted by these equations was inaccurate by
more than 8 y in over half the subjects. We conclude that we could not predict age at a clinically acceptable
level using data from our automated system. This most likely reflects an insensitivity to regional age-related
changes in bone histology because we recorded data from each entire cortex. Automated bone measurement
according to cortical region might be more useful in the prediction of age. The inclusion of TSPA together
with PP as independent predictors of age raises the possibility that a future measure of periosteal shape at
the femoral diaphysis could also be helpful in the prediction of age. The accuracy reached with the relatively
simple methods described here is sufficient to encourage the development of image-analysis systems for the
automatic detection of more complex features.