Shortcomings in and preferred strategies for evaluating biomarkers as medical tests

There is an abundance of promising publications that advocate the use of biomarkers as medical tests for diagnostic, prognostic, screening, or other purposes. However, despite massive research efforts and technological advancements in the development, few biomarkers have been successfully implemented in clinical care. A lack of adequate and consistent methodology has been suggested as possible reasons for this translational failure.
The overall aim of this thesis was to identify and document the use of suboptimal methods in clinical evaluations of biomarker and to suggest preferred strategies to overcome these.
We found that the use of suboptimal and poor study designs and reporting practices, including misrepresentations and overinterpretations of study findings i.e. spin, were frequent in recent evaluations of ovarian cancer biomarkers, also despite the abundant tools, guidance and recommendations that exist within this area. We furthermore found that the terminology used for describing diagnostic accuracy study designs is very heterogenous and often confusingly used and, moreover, that distinguishing between a ‘pragmatic vs explanatory’ approach potentially could help to clarify important differences between study designs. We also performed evaluations of a prognostic and diagnostic biomarker, applying where possible preferred methods and terminology, to illustrate how performance measures should be examined, interpreted, and reported in such a way that they can adequately inform decision makers about their potential value.
In conclusion, if more researchers would use optimal study designs and report their studies in a complete and transparent manner, without spin, biomarker evaluations could be far more helpful.