Individualizing Care for Ovarian Cancer Patients Using Big data
A partir de données portant respectivement sur 1 251 et 1 525 patientes atteintes d'un cancer de l'ovaire de stade avancé, ces deux méta-analyses évaluent les performances de divers modèles, basés sur l'expression de gènes, pour prédire la survie des patientes
Over the last 10 years, numerous studies have attempted to determine a prognostic molecular signature for ovarian cancer, with little success. In this issue of the Journal, two articles leverage the wealth of data produced from gene expression microarray array studies to assess gene expression signatures for predicting outcomes in patients with ovarian cancer. Waldron et al. (1) present a framework for assessing and validating previously published prognostic gene signatures for ovarian cancer outcomes, noting that “most previously published models demonstrated lower accuracy in new independent datasets, compared with the validation sets presented in their publication.” Notably, Waldron and colleagues (1) undertook the monumental task of collecting, curating, and combining all publicly available ovarian cancer gene expression studies and deposited the resulting data in the curatedOvarianData database (2), an incredibly valuable resource for future ovarian cancer genomics studies. Using these data and a similar meta-analytic framework, Riester and colleagues developed gene signatures to predict overall survival and debulking status (optimal vs suboptimal) for late-stage ovarian cancer (3).
The prognostic models developed by Riester et al. (3) outperformed previously identified gene expression signatures; additionally their model for debulking status was further validated in an independent study and was able to classify 92.8% of patients correctly into high- and low-risk groups for suboptimal debulking. However, with the number of …
Journal of the National Cancer Institute , éditorial, 2014