Data extraction and management
Data extraction will be done independently by two review authors. Disagreement will be resolved through discussion or by consulting a third review author. The data extraction form is inspired by the work of Moons et al. and adapted for our purpose [19]. The following data will be extracted: source of data (study design, study dates, sample size, number of events per candidate prognostic factor, exploratory/confirmatory approach, follow-up length), participant characteristics (setting, inclusion and exclusion criteria, recruitment method, sociodemographic characteristics, details of treatment received, clinical tumor characteristics), candidate prognostic factors (number and type of prognostic factor, definition, method and timing of measurement of candidate prognostic factors, all unadjusted (simple) and adjusted (multivariable) associations reported between the prognostic factor and the outcome, with details on any adjustment factors that are used), outcomes (type of outcome, definition, measurement and timing of outcome assessment, time of outcome occurrence), and missing data (number per patient/per candidate prognostic factor, handling of missing data). The data extraction sheet will be pilot-tested to ensure its performance.
We will classify included studies according to the authors’ objectives, study design, definition of variables, and analysis (exploratory versus confirmatory intent). This information will be considered when interpreting the strength of the available evidence.
We will not consider individual patient data as the collection of such data is not feasible. Instead, we will contact the authors of the original publication in case of unclear or missing data.
If we identify several publications of the same study, we will consider the most recent and comprehensive one in the analysis.
All references will be managed and stored in EndNote X7.