Application of the adaptive validation substudy design to colorectal cancer recurrence

Background: Among men and women diagnosed with colorectal cancer (CRC), 20–50% will develop a cancer recurrence. Cancer recurrences are not routinely captured by most population-based registries; however, linkage across Danish registries allows for the development of predictive models to detect recurrence. Successful application of such models in population-based settings requires validation against a gold standard to ensure the accuracy of recurrence identification. Objective: We apply a recently developed validation study design for prospectively collected validation data to validate predicted CRC recurrences against gold standard diagnoses from medical records in an actively followed cohort of CRC patients in Denmark. Methods: We use a Bayesian monitoring framework, traditionally used in clinical trials, to iteratively update classification parameters (positive and negative predictive values, and sensitivity and specificity) in an adaptive validation substudy design. This design allows determination of the sample size necessary to estimate the corresponding parameters and to identify when validation efforts can cease based on predefined criteria for parameter values and levels of precision. Results: Among 355 men and women diagnosed with CRC in Denmark and actively followed semi-annually, there were 63 recurrences diagnosed by active follow-up and 70 recurrences identified by a predictive algorithm. The adaptive validation design met stopping criteria for the classification parameters after 120 patients had their recurrence information validated. This stopping point yielded parameter estimates for the classification parameters similar to those obtained when the entire cohort was validated, with 66% less patients needed for the validation study. Conclusion: In this proof of concept application of the adaptive validation study design for outcome misclassification, we demonstrated the ability of the method to accurately determine when sufficient validation data have been collected. This method serves as a novel validation substudy design for prospectively collected data with simultaneous implementation of a validation study.