Translating results from randomized clinical trials (RCTs) to individual patients in clinical practice is challenging, as treatment effects can vary substantially among individuals. Data from RCTs can be used for individualized treatment effect prediction, to identify patients who benefit from specific treatments. In this study, we developed and validated a prediction model for estimating absolute treatment effect of pemetrexed plus carboplatin versus single-agent pemetrexed in the second-line treatment of non-squamous non-small-cell lung cancer (NSCLC).Patients and methods
Using data of relapsed patients with advanced non-squamous NSCLC from the NVALT-7 trial, a Weibull model for prediction of gain in median progression-free survival (PFS) by pemetrexed–carboplatin was derived based on patient and tumor characteristics. The model was externally validated in the GOIRC 02-2006 trial. The applicability of the model for guiding clinical decision-making was evaluated using decision curve analysis.Results
A wide distribution of predicted gain in median PFS by pemetrexed–carboplatin over pemetrexed was found, with a median of 0.7 months (interquartile range: −0.1 to 1.5 months). Patients who benefited most included women, those with stage IV, high body mass index and/or adenocarcinoma. External validation showed satisfactory calibration and moderate discrimination (C-index: 0.61, 95% confidence interval 0.56–0.67). Decision curve analysis confirmed that the model adequately identified patients who benefit from pemetrexed–carboplatin, as prediction-based treatment led to improvement in net benefit with regard to PFS and overall survival when assuming a treatment threshold of 0–5 months gain in PFS, compared with other treatment strategies.Conclusions
The effects of pemetrexed–carboplatin can be predicted for individual patients based on routinely available patient and tumor characteristics. There is important heterogeneity in the effects on PFS of pemetrexed–carboplatin versus pemetrexed in pretreated patients with advanced non-squamous NSCLC. Individualized prediction of treatment effect could be used to guide shared decision-making by discriminating patients who benefit most, to improve clinical outcome.Clinical Trial numbers
NVALT-7: ISRCTN38269072 (ISRCTN registry), GOIRC 02-2006: NCT00786331 (clinicaltrials.gov).