Microsatellite Instability andBRAFandKRASMutations in Stage III Colon Cancer

Abstract

Microsatellite instability (MSI) and mutations of BRAF and KRAS have been investigated as candidate biomarkers of “independent” prognostic value in patients with colorectal cancer for some time.1,2 Depending on tumor stage and treatment, their prognostic value seems to vary, but even in trials with similar populations and treatment the prognostic value is still controversial. Whether and how these 3 biomarkers are associated with outcome in patients with stage III colon cancer treated with combination leucovorin-calcium, fluorouracil, and oxaliplatin (FOLFOX)based chemotherapy and how they influence each other are relevant research questions. Most of the current evidence supports the notion that stage-adjusted prognosis ismore favorable forMSI-high than formicrosatellite-stable tumors,with larger advantage for patients with stage II than stage III disease.3 A putative predictive role of MSI for fluorouracil-based adjuvant chemotherapy has been a more contentious issue with conflicting evidence.Manystudieshave investigated theprognosticvalue of KRAS and BRAF tumor mutations on retrospectively collected cohorts of patients with colorectal cancer of different MSI status, also with conflicting results—some reporting no prognostic value, others findingprognostic value either alone or concomitantlywithmutated TP53 or PIK3CA.KRAS tumor mutation status is consequently now widely recognized as a predictive marker of resistance to epidermal growth factor receptor–targeted antibodies in colorectal cancer. In this issue of JAMA Oncology, Taieb and colleagues4 report a post hoc analysis of the PETACC-8 trial. For some impressive 1800 patients they assessed the association of MSI, BRAF, and KRAS status and more classic prognostic factors such as age, tumor grade, T/N stage, and location with outcome in terms of disease-free survival and overall survival. The authors are to be commended for studying such an important question in a large trial population and for emphasizing the results of multivariable models that adjust for potential confounding between variables. Their most important finding is a significant interaction of BRAF and KRAS mutations and MSI status, which we had observed and reported, albeit from a different angle, in the PETACC-3 trial as well.4 That being said, we still want to point out a number of issues. All analyses in the BRAF-mutant tumors (10%) and in theMSI-high subgroup have low power to detect presence or absence of effects reliably. While often done correctly, high P values are sometimes interpreted negligently. For instance, they state that KRAS status was not prognostic inpatientswithMSI-high tumors.Thiswouldhavebeen better presented as “there was no evidence for a prognostic effect” or the like. With a confidence interval from approximately 0.3 to 3 for both disease-free survival and overall survival,we cannotdrawconclusions from thesedata, due to the lack of events and hence power as mentioned. Also, it might have been more advisable to compare BRAF-mutated and KRAS-mutated with double wild-type tumors instead of BRAF/KRAS-nonmutated tumors, respectively. Also, importantly, because the RAS status is incompletely assessed in these samples, and as acknowledged by the authors, KRAS mutations beyond exon 2 might change the results. The absence of these data limits the immediate clinical interpretation. Most questionably, the authors do not stratify for the treatment arms because “no difference was found between the 2 arms for efficacy analyses” and because no interaction was found between treatment and KRAS status. But interaction tests have low power with a high false-negative rate5 and, again, just because we do not find a significant interaction does not mean that it is not there. Importantly, the interaction of KRAS status and cetuximab therapy is well known and established a priori from pathophysiological principles; hence, stratified subgroup analyses are justified and should be presented with similarities to or differences from the unstratified analyses. We second the conclusion of Taieb and colleagues4 that in future adjuvant trialsMSI,BRAF, andKRAS statuswill need to be considered for stratification. This could also include tumor site (see theirprevious report6on this trial indicating thatKRAS mutation statuswasonlyprognostic indistal tumors, a finding that the authors now suggest is explained by the MSI status). With this many highly correlated factors to assess, a pooled analysiswithacarefullydesignedstatistical analysisplanof individualpatientdata (fullRASstatus,MSIstatus, sidedness,and so forth) on stage II and III colon cancer from asmany trials as possible would be timely and relevant. A first question to answeronexistingadjuvant trialdatasetswouldbewhether there is any interaction of these subgroups with therapy. Any markers that can help guide individualized toxicity and/or efficacy evaluations in this setting are welcome. Further stratificationof thediseasemayalsohelpassessmoreprecisely the effect of drugs that failed in the adjuvant setting. In terms of identification of purely prognostic subgroups in the disease, it will be interesting to confront these findings withgeneexpression–basedprognostic stratificationsof stage II and III colorectal cancer. Related article page 643 BRAF and KRASMutations in Stage III Colon Cancer Original Investigation Research