Model-Based Meta-Analysis for the Efficacy and Safety of Paclitaxel in Cancer Patients
Jin Yan Jin (1), Dan Lu (1), Hanbin Li (2), Nancy Zhang (2), Russ Wada (2), Amita Joshi (1)
(1) Genentech Inc., South San Francisco, CA, USA; (2) Quantitative Solutions Inc., Menlo Park, CA, USA
Objectives: Meta-analysis of integrated literature data can help to maximize our learning from the past and to optimize ongoing drug development. This project aims to quantify the effect of paclitaxel (PAC) dose and regimen on efficacy and safety using data from published trials. To our knowledge, this is one of the earliest attempts of model-based meta-analysis of literature data in oncology.
Methods: A literature database with PK, efficacy, and safety of PAC mono-therapy in cancer patients was developed based on thorough literature search. Meta-analysis was conducted for the objective response (OR) and overall survival (OS) in patients with metastatic breast cancer (MBC) as efficacy endpoints, and the incidence of neutropenia in all patients as safety endpoint. Percent of patients with OR or with ≥Grade 2 neutropenia were modeled as a function of dose, regimen, and other covariates using logistic regression. Relative risk of OS was modeled as a function of dose, regimen and other covariates using proportional hazard model. A mixed-effects modeling approach accounting for inter-trial variability was used (S-Plus v6.2).
Results: The PAC database includes 49 trials with 95 arms and contains trial-level data for 4256 patients. Average PAC dose ranges from 44 to 130 mg/m2/wk with once-every-3-weeks (Q3W) or once-a-week (QW) regimen. Dose-response relationship was established for both efficacy (%OR, n=3070 in 29 MBC trials; or median OS, n=2749 in 15 MBC trials) and safety (%neutropenia, n=1886 in 24 trials). Model predicted that an increase of PAC dose from 60 to 90 mg/m2/wk QW (180 to 270 mg/m2 Q3W) may increase the %OR from 30.5% (25.3-36.2%, 95% CI) to 40.0% (34.9-44.8%), and median OS from 11.1 (7.2-17.8) months to 20.6 (13.1-32.2) months in MBC patients. Of special note, %OR and median OS best correlated with average PAC dose in mg/m2/wk regardless of Q3W or QW regimen, while %neutropenia best correlated with administered dose in mg/m2. These results implied PAC efficacy was driven by overall exposure, while safety was driven by Cmax (i.e. for the same total dose, QW may show the same efficacy with better tolerability relative to Q3W).
Conclusions: The effect of PAC dose and regimen on clinical efficacy and safet was quantified by model-based meta-analysis integrating literature data from multiple trials. These analyses can be used to guide trial design and interpretation for PAC as control agent or as combination therapy with new anti-cancer agents.