Can a Banana a Day Keep the Cancer Away in Patients with Lynch Syndrome?

In a remarkable effort, the CAPP-2 investigators conducted a multi-continent randomized controlled trial in approximately 1,000 patients with Lynch syndrome using a two-by-two factorial design to test a combination of 600 mg of aspirin and 30 g of resistant starch (RS) per day, for up to 4 years with incidence of colorectal cancer as the primary endpoint and incidence of other Lynch syndrome-related cancers as the secondary endpoint. The CAPP-2 trial is the largest and most important interventional study performed to date in Lynch syndrome, as it established the only preventive indication that we have for the Lynch syndrome community, which is the daily use of aspirin for Lynch syndrome–related cancer prevention. In the initial CAPP-2 publication, reporting on a mean follow-up of 29 months, aspirin did not protect against colorectal cancer (1), although a subsequent per-protocol analysis at a mean follow-up of 56 months demonstrated a solid protective effect of aspirin on the incidence of colorectal cancer (2) that persisted at long-term follow-up of at least 10 years (20 years for some study sites; ref. 3). In contrast to aspirin, RS at a median follow-up of 53 months did not show a reduction in the incidence of colorectal cancer (4).

In this issue of Cancer Prevention Research, the CAPP-2 investigators now present a secondary analysis evaluating the effect of RS on the incidence of colorectal cancer, and other Lynch syndrome–related cancers in trial participants at 10-year follow-up for all participating sites, and at 20-year follow-up for sites in England, Wales, and Finland, where national cancer registry data were available for cross-referencing in order to improve the robustness of endpoint capture (5). Adherence to the intervention was carefully measured through manual counting of returned starch sachets by study personnel, with consumption of at least 1,400 sachets defined as per protocol. In this context, the CAPP-2 investigators observed that daily use of RS did not reduce the incidence of colorectal cancer [HR = 0.92; 95% confidence interval (CI), 0.62–1.34; P = 0.63] but it did reduce the incidence of other Lynch syndrome–related tumors in both the intention-to-treat [HR = 0.54 (0.33–0.86), P = 0.010], and per-protocol analysis [HR = 0.42 (0.22–0.81), P = 0.009]. Notably, there was no effect of RS on the incidence of non–Lynch syndrome–related cancers.

While the results of this study are intriguing, there are some limitations to be considered. First, a long-term follow-up study such as this is expected to lose participants over time, and that is exactly what happened in the study by Mathers and colleagues, where the number of evaluable participants decreased by approximately 40% over 20 years (Fig. 1). Second, the concordance of non–colorectal cancer Lynch syndrome–related cancer diagnoses between the CAPP-2 study database and the Finnish national registry was only 55%, creating doubts about accurate capture of non–colorectal cancer Lynch syndrome–related cancer diagnoses in the two arms, thus suggesting that the 20-year follow-up results be considered cautiously. Third, investigators did not control for inflation of the family-wise error rate introduced by testing multiple hypotheses. Finally, and perhaps most importantly, the interpretation of a beneficial effect of an intervention on a secondary endpoint, given its failure to demonstrate an effect on the primary endpoint, is unclear and results should be viewed in an “exploratory light” (6).

Despite above limitations, the results of the study are highly provocative for a few reasons, including biological plausibility. Of all the extracolonic Lynch syndrome cancers, RS had the strongest impact on the incidence of upper-gastrointestinal (GI) cancers (stomach, duodenum, pancreas, and bile duct). This observation is consistent with its presumed mechanism of action, where its metabolism by gut bacteria produces downstream antioncogenic metabolites like the short chain fatty acid butyrate, which has shown a variety of in vitro effects on apoptosis, proliferation, angiogenesis, and inflammation (7), and it may also have immunomodulatory properties (8). Further, there was no interaction between the intake of RS and aspirin, thus suggesting independent cancer prevention mechanisms.

Upon closer inspection of the results, the protective effect of RS was observed predominantly among female participants (Table 1), and in those older than 53 years. In another trial performed by the same cooperative group in patients with familial adenomatous polyposis (FAP), RS did not affect polyp number or size (9). However, the participants in the FAP trial (CAPP-1) were approximately 20 years younger than the Lynch syndrome population in CAPP-2. While one cannot draw definitive conclusions from these subgroup analyses, as neither trial was designed to detect them, one can speculate that the chemopreventive effects of RS may be sex-, age-, and disease-specific. This fact could be linked to the differential role of the microbiome in these diseases. In fact, RS has been shown to have variable effects (increased or decreased) on production of downstream metabolites like butyrate depending on the individual microbiome (10, 11). Thus, there is an added layer of complexity in predicting the effects of RS in that prior knowledge of a person's microbiome may be needed. With advances in technology, using one's microbiome to predict efficacy of RS may become a reality. Finally, another striking observation by Mathers and colleagues is that short-term treatment (RS intake for ∼2 years) provided protection. The exact mechanisms remain unclear, but it appears that even brief use of RS was able to intercept oncogenic mechanisms that may include effects on stem cells as well as ongoing cancer immunosurveillance. This observation warrants further research as it suggests that, although cancer development in patients with Lynch syndrome is a dynamic process, the trajectory can be altered by a short-term intervention without the need to expose patients with Lynch syndrome to side effects associated with the long-term use of cancer prevention therapies.

Based on the findings of Mathers and colleagues (5), are we ready to systematically recommend RS to patients with Lynch syndrome for prevention of upper GI cancers? The short answer is “no” or “not yet”, primarily because the trial was not designed to address this question. However, the RS results of the CAPP-2 trial should stimulate further investigations into the mechanistic cancer preventive effects of RS on the incidence of upper GI cancers in the context of mismatch repair-deficient carcinogenesis. Although phase III clinical trials like CAPP-2 are resource-intensive and challenging to conduct, the next step should be a multi-site randomized controlled trial of RS in patients with Lynch syndrome with the primary endpoint of prevention of upper GI cancers. Further understanding of factors such as age and sex on the human microbiome will allow enrichment of participants in future trials who will benefit the most from the effects of this intervention. The study's intervention was a sachet of 30 g of an intentionally manufactured and packaged RS, but until future trials are completed, as the authors suggest, eating one green-tipped banana per day (that contains the equivalent of 30 g of RS) may be an easy ask for our patients with Lynch syndrome.

E. Vilar reports grants and personal fees from Janssen Research and Development; and personal fees from Recursion Pharma outside the submitted work. No disclosures were reported by the other authors.

A. Bansal: Conceptualization, supervision, writing–original draft, writing–review and editing. R. Montgomery: Conceptualization, formal analysis, writing–review and editing. E. Vilar: Conceptualization, supervision, writing–original draft, writing–review and editing.

This work was supported by a gift from the Feinberg Family Foundation to E. Vilar and P30 CA016672 (NIH/NCI) to The University of Texas MD Anderson Cancer Center Core Support Grant to E. Vilar This work was supported by the NIH/NCI Cancer Center Support Grant (grant no. P30 CA168524) to A. Bansal.