Clinical Trial Simulations Based on A Meta-Analysis of Studies In Patients With Locally Advanced and/or Metastatic Adenocarcinoma Pancreatic Cancer Receiving Gemcitabine (GEM) Alone or In Combination

To develop a model to: (1) describe median overall survival (mOS) in trials with locally advanced and/or metastatic adenocarcinoma pancreatic cancer treated with GEM alone and in combination; (2) simulate predictive distributions based on Von Hoff 20131and assess the probability of the results in a future study; and 3) compare predictions based on mOS and overall survival hazard ratios (OS-HR). A systematic review of randomized clinical trials with GEM alone or in combination was conducted. A linear mixed-effects model was fit to log-transformed mOS data, with an intercept reflecting GEM treatment alone, a between-trial random effect (ηi~N(0,τ2)) and residual error term (εij~N(0,σ2/Nij), i=study, j=treatment arm). Potential confounding or prognostic factors were tested as covariates. Drug-class combinations were simulated to produce model-based prediction distributions. Data consisted of 83 arms (40 studies;4813 patients) and first-line treatment across 21 drug-classes. The final model (abbreviated) is: LN(mOSij) = intercept+ θ1*platinumij+ θ2* taxaneij+ θ3*igf1r.inhibitorij+ …+ θn * immunomodulatorij+ asian.study+ ηi+ εijwhere drug-class and asian.study are indicators (0 or 1). Estimated mOS for GEM was 6.6 months with 95% CI: 6.3, 7.0. The model mOS extimates for drug classes in combination with GEM were significantly better than GEM alone for: Hypoxia activated prodrug 8.9 (7.2, 11.2), Anti folates 8.5 (7.8, 9.3), Taxane 8.3 (7.6, 9.0) and Platinum 7.8 (7.2, 8.5). Simulations showed that the model predicts the Von Hoff study well and results related to drug classes for mOS and OS-HR were comparable. This meta-analysis is useful: (1) as a repository for data exploration of current randomized trials; (2) in characterizing typical endpoints for designing randomized clinical trials; (3) to guide a default product profile for pancreatic cancer therapies and; (4) for informing decisions during the drug development process.