Evaluating the Performance of Non-invasive Tests for Colorectal Cancer: Statistical Considerations.

Abstract

Like all interventions in health care, novel non-invasive tests for colorectal cancer should be properly evaluated before they can be recommended. Such evaluations should be performed in well-designed studies, of which the results can serve as the evidence base for recommendations. While the methods for evaluating novel tests have been slower to develop, there is now a solid base for developing suggestions and even strong recommendations for their assessment. These include advice for study design, analysis of results, and clear and informative reporting. These considerations should be guided by the intended use of the new test and its role in the testing pathway. We distinguish between tests proposed as a replacement for an existing test, as a triage test, before an existing test, or as an add-on test, after an existing test. We recommend the definition of explicit, a priori defined, minimally acceptable performance criteria for the intended use of the new test and rigorous statistical hypothesis testing, to prevent "spin" in the interpretation of the findings.