Prognosis research and risk of bias

The interest in prognosis research has been steadily growing during the past few decades because of its impact on clinical decision making. However, since the methodology of prognosis research is still incompletely defined, the quality of published prognosis studies is largely unsatisfactory. Seven major domain for risk of bias in prognosis research have been identified, including study participation, attrition, selection of candidate predictors, outcome definition, confounding factors, analysis, and interpretation of results. The methodology for performing prognostic studies is currently aimed at avoiding such potential biases. Amongst methodologic requirements in prognosis research, the following should be considered most relevant: beforehand publication of the study protocol including the full statistical plan; inclusion of patients at a similar point along the course of the disease; rationale and biological plausibility of candidate predictors; complete information; control of overfitting and underfitting; adequate data handling and analysis; publication of the original data. Validation and analysis of the impact that prediction models have on patient management, are key steps for translation of prognosis research into clinical practice. Finally, transparent reporting of prognostic studies is essential for assessing reliability, applicability and generalizability of study results, and recommendations are now available for this aim.