Reply to D.A. Palma et al and A. Addeo et al.

Abstract

We thank Palma et al for their insightful comments on our study of carboplatin and paclitaxel (CP) versus etoposide and cisplatin (EP) using the Veterans’ Affairs Central Cancer Registry. In our initial analysis, we did not include International Classification of Diseases, 9th revision, code 508.0 as radiation pneumonitis because it is most often a clinical diagnosis with no definitive diagnostic test. Furthermore, in our clinical experience, it is infrequently coded and largely underreported. We have since performed an analysis using this International Classification of Diseases, 9th revision, code and found that this adverse event did not differ between treatment arms and was infrequent overall. Specifically, it was reported in 6.3% of CP treatments and 6.8% of EP treatments (P .001, Pearson ). We agree with Palma et al that the lack of difference in our study, when compared with the individual patient data meta-analysis and randomized clinical trial (RCT), could be because of the patient populations studied. In response to Addeo and Comins, we agree that in the absence of a large, well-powered, prospective RCT, there will be no definitive answer as to either inferiority or superiority of CP versus EP. However, as is often the case in oncology, we must use the best available evidence to guide clinical decision making. We believe that our work adds to this evidence base. In particular, using multiple approaches (multiple regression, propensity score, and instrumental variables) to analyze rich clinical data can increase the usefulness of observational data analyses. We acknowledge the concerns of Addeo and Comins regarding retrospective analysis of observational data. We believe that our analysis comparing the outcomes of CP versus EP in a relatively uniform patient population, along with the limited availability of informative RCTs, establishes sufficient equipoise to justify the conduct of RCTs to settle this important question. In the meantime, our data suggest that after controlling for important clinical variables, patients treated with these two regimens have equivalent survival.