Abstract:
This thesis investigated a number of aspects of the clinical utility and cost-effectiveness of clinical risk factors, online multivariate algorithms, DXA and a nail-based Raman spectroscopy test, BQT for fracture risk using archived nail samples, and questionnaire data from the Nurses’ Health Study which followed women for up to 23 years.
The results showed that the BQT in combination with CRFs improved the results over CRFs alone, using logistic regression and Cox’s proportional hazards analysis. The improvement seen was larger using the Cox model, indicating that time is an important factor.
The multivariate algorithms, FRAX and QFractureScores were compared in retrospective and cohort models and found to be predictive, but the relative performance of the two algorithms was highly dependent on the input data.
Reclassification is an exciting new approach to evaluating the addition of new biomarkers in multivariate algorithms and was found in the Nurses’ Health Study to provide better discrimination than AUC.
Cost-effectiveness analysis using Markov and decision tree approaches showed that the BQT with a low cut-off in combination with DXA was consistently on the cost-effective frontier, indicating that this new biomarker would be an integral part of any mass screening strategy.
In conclusion, it is clear that the use of the BQT can enhance the performance of clinical risk factors and, with further improvements, the combination may offer a cost-effective alternative to the use of DXA for mass screening in multivariate algorithms.
