Review of methodological challenges in comparing the effectiveness of neoadjuvant chemotherapy versus primary debulking surgery for advanced ovarian cancer in the United States

摘要

Randomized trials outside the U.S. have found non-inferior survival for neoadjuvant chemotherapy (NACT) versus primary debulking surgery (PDS) for advanced ovarian cancer (AOC). However, these trials reported lower overall survival and lower rates of optimal debulking than U.S. studies, leading to questions about generalizability to U.S. practice, where aggressive debulking is more common. Consequently, comparative effectiveness in the U.S. remains controversial. We reviewed U.S. comparative effectiveness studies of NACT versus PDS for AOC. Here we describe methodological challenges, compare results to trials outside the U.S., and make suggestions for future research. We identified U.S. studies published in 2010 or later that evaluated the comparative effectiveness of NACT versus PDS on survival in AOC through a PubMed search. Two independent reviewers abstracted data from eligible articles. Nine of 230 articles were eligible for review. Methodological challenges included unmeasured confounders, heterogeneous treatment effects, treatment variations over time, and inconsistent measurement of treatment and survival. Whereas some limitations were unavoidable, several limitations noted across studies were avoidable, including conditioning on mediating factors and immortal time introduced by measuring survival beginning from diagnosis. Without trials in the U.S., non-randomized studies are an important source of evidence for the ideal treatment for AOC. However, several methodological challenges exist when assessing the comparative effectiveness of NACT versus PDS in a non-randomized setting. Future observational studies must ensure that treatment is consistent throughout the study period and that treatment groups are comparable. Rapidly-evolving oncology data networks may allow for identification of treatment intent and other important confounders.