Mathematical prognostic models for predicting survival in ovarian cancer patients – A meta-analysis of validation studies for prognostic indices and nomograms

Background: Decisions about curative or palliative therapies in ovarian cancer patients rely substantially on the assessment of prognosis and future events. For various malignancies, prognostic models have shown superiority in comparison to conventional staging systems.

Methods: We performed a systematic literature review and meta-analysis of studies between 2005 and 2015 that validated prognostic models of survival in ovarian cancer patients. Inclusion criteria were internal or external validation studies which presented discrimination and/or calibration measures to assess the performance of the individual model. Exclusion criteria were development studies and studies which described independent prognostic factors without including these in a new prognosis model.

Results: We identified 32 studies which validated internally or externally the prognostic index or a nomogram in ovarian cancer patients. The concordance (c) indices of these models ranged between 0.60 and 0.74 compared to c indices of 0.54 to 0.62 for FIGO stage.

Conclusion: The assessed prognostic models appeared to be superior when compared with conventional FIGO staging for prognosis assessment. Some models have the potential to significantly improve patient stratification in clinical trials and patient counselling.