Fractures of long bones (e.g., shinbone, thighbone) typically healing by secondary fracture healing, which occurs when the ends of the bone are not perfectly connected. In contrast to primary (or direct) bone healing, during secondary bone healing a cartilaginous fracture callus forms around the injury and gradually ossifies over a period of four to five months. However, this process can occasionally become delayed or stalled, resulting in a delayed union or nonunion. These complications are associated with increased healthcare cost, loss of productivity, and considerable patient disability, resulting in 41% of nonunion patients unable to return to work within one year. To complicate matters, it is difficult to quantitatively track the progress of delayed unions and differentiae if they will eventually unite successfully or need further intervention.

Currently, bone healing progress is usually tracked using subjective measures such as visual assessments of X-rays, assuming that denser looking callus indicates a stiffer healing zone. While there are some techniques, predominantly RUST scoring, that try to quantify this, studies have shown that callus size and density do not independently predict fracture stiffness. Therefore, the overall objective of this research was to develop an objective, quantitative virtual mechanical test to accurately predict the fracture structural integrity in humans.

Starting with a pre-clinical cohort of ovine (sheep) models, we developed and validated a virtual torsional rigidity (VTR) test from down-sampled micro-CT scans. These virtual tests were strongly and significantly correlated (R² > 0.70, p < 0.0005) with the physical biomechanical tests of the ovine limbs. This development and validation would be impossible in a clinical setting where physical mechanical test cannot be performed and provided a promising pre-clinical and clinical tool for use in fracture analysis.

Translation to clinical application was relatively straight-forward due to steps to ensure similarities between the pre-clinical and clinical datasets. Once applied in a clinical setting, preliminary studies shows that the virtual mechanical test was a robust tool for investigating fracture healing and was able to predict healing time (R² > 0.383 & p < 0.005) where morphometric measures and RUST scores could not. Furthermore, the test was able to identify the only delayed union as a statistically outlier, reinforcing its utility to track fracture healing.

Most impressively, when the virtual mechanical test was applied in a clinical study to investigate the effects of bone healing comorbidities in 27 patients, it was able to show a large and significant difference between the two groups(22% mean difference, p = 0.004). This is remarkable in clinical studies where less sensitive assessments (with lower statistical power) are typically used, resulting in much larger required sample sizes consisting of many hundreds to thousands of patients.

The virtual mechanical test developed in this work shows promise to not only reduce the costs and difficulties of clinical studies, but to meaningfully impact patient care. This test would allow for better bone fracture assessment and care by reducing the risk of nonunion and ensuring patients can return to their pre-injured life as quickly as possible.