Improved Outcomes in Metastatic Colon Cancer: Giving Credit Where Credit Is Due.
D. Sargent
DOI: https://doi.org/10.1001/jamaoncol.2015.1789
IF: 33.006
2015-09-01
JAMA Oncology
Abstract:For decades, effective treatment for metastatic colorectal cancer (mCRC) was an almost impenetrable obstacle, with a single therapeutic option—fluorouracil. Outcomes were highly consistent across time and across trials, with a median overall survival (OS) of approximately 1 year. The last 20 years have seen highly clinically relevant improvements in not only OS but also progression-free survival (PFS) and response rates. In the most recently reported large first-line randomized trials, median OS has reached or exceeded 30 months.1 Randomized trials, the approach through which we evaluate new therapies, have a fundamental, self-imposed limitation of addressing an isolated hypothesis. While trials test 1 element of a patient’s treatment, they are conducted in an environment that is in constant flux (thus the never-ceasing cautions regarding drawing conclusions from cross-trial comparisons). In mCRC, the last 20 years has brought advances in surgical techniques and approaches (increased use of metastasectomy), increased rates of screening (detecting disease earlier—not all stage IV disease is the same), improvements in imaging (earlier detection of recurrence after resection and adjuvant therapy), and various improvements in end-of-life care. These concurrent advances pose a considerable challenge when we try to identify the respective contributions of each to the greater than doubling in median survival that has been observed. In this issue, Jawed et al2 attempt to estimate what fraction of the improvement observed in mCRC outcomes in the last 2 decades may be attributed to novel treatments. They performed a literature-based systematic review of 96 published phase 2 and 3 trials, comparing improvements in OS with improvements in PFS and response rate (among many other analyses). The approach hinges on a key assumption, which is that improvements in PFS, more or less, should translate directly and predictably into improvements in OS—we consider the veracity of this assumption below. Their primary conclusion, that chemotherapy has only contributed partly to the overall gains in survival, is highly plausible and is consistent with other findings from observational data.3 That being said, as with any provocative analysis, a number of issues are present that may introduce bias into the findings. First, as recently shown by Shi et al,4 in modern mCRC trials, at the trial level, between-arm differences in PFS are not reliable predictors of differences in OS. The association between PFS and OS, as measured by the R2 coefficient, is almost identical (approximately 0.50) in the Shi et al4 and Jawed et al2 analyses. Interestingly, Jawed et al consider this correlation sufficient to justify an accurate prediction, but Shi et al conclude that the correlation is too low to allow PFS to be considered as a reliable surrogate end point. Second, subtleties in protocol-specific definitions of progression cannot be accounted for in a literature-based analysis, complicating interpretation. In addition, in practice (ie, outside of the trial setting), patients may be treated through RECIST progression with first-, second-, or later-line regimens, further challenging the relationship between RECIST-based PFS as measured in a study and the ultimate effect of an agent on OS. Most critically, in their analysis, Jawed et al2 discount the possibility that a novel agent could improve OS in settings other than first-line therapies, stating that gains in secondand laterline outcomes have been unimpressive (based on a cross-trial comparison). Multiple real-world observations challenge this conclusion. Because many patients discontinue first-line treatment without progression (owing to toxic effects or for other reasons), agents are often reused at the time of actual progression or even after progression. A variety of treatment strategies including therapy reduction and treatment holidays are in common use, through which agents may not affect (or even may reduce) traditional first-line PFS but still may influence OS. Reintroduction of oxaliplatin treatment likely has clinical benefit,5 as does continuation of bevacizumab through multiple lines of therapy,6 indicating that novel treatments can provide benefit beyond the first line. Perhaps most critically, in current practice patients do not receive all active agents (even those that have demonstrated first-line survival benefit) in firstline therapy. Specifically, there is a patient-specific choice between oxaliplatin or irinotecan in conjunction with a fluorouropyrimidine, and between an angiogenesis inhibitor and an epidermal growth factor receptor inhibitor in first line regimens, with the unused options reserved for later-line treatment (if appropriate). Each of these several factors may mask the true OS therapeutic benefit of an agent across the entire patient experience if only first-line, RECIST-based PFS is considered as a measure of the benefit that an agent may provide. Beyond these specific technical issues, when considered broadly, advances in supportive care, surgery, and imaging were all ongoing in the 2 decades prior to 1995, when mCRC median OS was highly stable at approximately 12 months. This suggests a constrained ability for these other factors to provide major OS advances in a way that could more than double median OS. The combination of all of these factors suggests that the Jawed et al2 analysis likely underestimates the actual proportion of OS improvement due to novel therapies. However, none of these limitations are sufficiently problematic as to invalidate their primary findings that new agents are not wholly responsible. So who deserves credit for the substantially improved outcomes for patients with first mCRC observed over the last 2 deRelated article Improved Outcomes in Metastatic Colon Cancer Invited Commentary