Systematic review of brain metastases prognostic indices
George Rodrigues1, Glenn Bauman2, David Palma2
1Department of Radiation Oncology, London Regional Cancer Program, London Health Sciences Centre, London, Ontario, Canada; Department of Epidemiology and Biostatistics, University of Western Ontario, London, Ontario, Canada.
Purpose:
A variety of prognostic indices for patients with brain metastases have been published in the literature, to guide clinical decision-making and clinical trial stratification. The purpose of this investigation is to perform a systematic review of all primary and validation reports of such prognostic systems. An assessment of index operating characteristics and misclassification rates was performed to assist in highlighting the advantages and disadvantages of competing systems.
Methods And Materials:
A systematic review of the English language literature regarding primary and validation brain metastases prognostic indices was performed according to PRISMA guidelines. Clinical, treatment, statistical, and prognostic index classification details were abstracted and organized into tables. Receiver operator characteristic curves were created from available Kaplan-Meier curves using a novel digitization procedure. From these curves, various operating characteristics such as positive predictive value (PPV), negative predictive value (NPV), accuracy (ACC), likelihood ratio (LR), and area under the curve (AUC) were calculated. Additionally, the major misclassification rate (MMR), defined as good or poor risk patients misclassified into the opposite group, was calculated for all available receiver operator characteristic curves.
Results:
A total of 9 prognostic systems have been published in the medical literature. In terms of the poor prognostic group, observed ranges are as follow: for PPV (0.25-0.72), NPV (0.72-0.97), ACC (0.57-0.95), LR (1.54-16.4), AUC (0.64-0.90), and MMR (0.02-0.39). Similarly, ranges of PPV (0.52-0.96), NPV (0.31-0.77), ACC (0.41-0.74), LR (1.69-20), AUC (0.64-0.89), and MMR (0.00-0.19) were observed for the good prognostic group.
Conclusions:
Operating characteristic and major misclassification analyses of all available prognostic index information demonstrated a range of results. As the ideal prognostic index has not yet been defined, further research into alternative approaches is warranted. Information contained within this report can serve as a benchmark for future investigations of existing and proposed prognostic indices.
