Edinburgh Research Explorer Long-term effect of resistant starch on cancer risk in carriers of hereditary colorectal cancer

Background Observational studies report that higher intake of dietary ﬁ bre (a heterogeneous mix including nonstarch polysaccharides and resistant starches) is associated with reduced risk of colorectal cancer, but no randomised trials with prevention of colorectal cancer as a primary endpoint have been done. We assessed the eﬀ ect of resistant starch on the incidence of colorectal cancer. Methods In the CAPP2 study, individuals with Lynch syndrome were randomly assigned in a two-by-two factorial design to receive 600 mg aspirin or aspirin placebo or 30 g resistant starch or starch placebo, for up to 4 years. Randomisation was done with a block size of 16. Post-intervention, patients entered into double-blind follow-up; participants and investigators were masked to treatment allocation. The primary endpoint for this analysis was development of colorectal cancer in participants randomly assigned to resistant starch or resistant-starch placebo with both intention-to-treat and per-protocol analyses. This study is registered, ISRCTN 59521990. Findings 463 patients were randomly assigned to receive resistant starch and 455 to receive resistant-starch placebo. At a median follow-up 52·7 months (IQR 28·9–78·4), 53 participants developed 61 primary colorectal cancers (27 of 463 participants randomly assigned to resistant starch, 26 of 455 participants assigned to resistant-starch placebo). Intention-to-treat analysis of time to ﬁ rst colorectal cancer showed a hazard ratio (HR) of 1·40 (95% CI 0·78–2·56; p=0·26) and Poisson regression accounting for multiple primary events gave an incidence rate ratio (IRR) of 1·15 (95% CI 0·66–2·00; p=0·61). For those completing 2 years of intervention, per-protocol analysis yielded a HR of 1·09 (0·55–2·19, p=0·80) and an IRR of 0·98 (0·51–1·88, p=0·95). No information on adverse events was gathered during post-intervention follow-up. Interpretation Dietary supplementation with resistant starch does not emulate the apparently protective eﬀ ect of diets rich in dietary ﬁ


Introduction
The incidence of colorectal cancer rises steeply with age and risk is aff ected strongly by environmental factors including adiposity, physical activity, and habitual diet. 1 A systematic review of the epidemiological literature shows convincing evidence that higher intakes of red meat, processed meat, alcoholic drinks (men only), and increases in body fat increase the risk of colorectal cancer, whereas greater physical activity reduces risk. Additionally, the evidence was assessed as probable that alcoholic drinks (women only) increase risk whereas foods containing dietary fi bre and garlic, milk, and calcium reduce risk of colorectal cancer. 1 A recent metaanalysis of 21 prospective studies showed a signifi cant, dose-dependent protective eff ect of dietary fi bre intake against colorectal cancer (relative risk [RR] per 10 g a day increase in dietary fi bre intake 0·90, 95% CI 0·86-0·94). 2 Dietary fi bre is a food-based measure that attempts to estimate the heterogeneous mix of carbohydrates (nonstarch polysaccharides, starches, and oligosaccharides) that escape digestion in the small bowel and fl ow to the large bowel where they exert a wide range of physiological eff ects due, in part, to the bacterial fermentation of these carbohydrates to yield biologically active short-chain fatty acids. 3 An inverse association exists between starch intake and risk of colorectal cancer, 4 which could be due to resistant starch (ie, the dietary starch and starch degradation products that escape digestion in the small intestine of healthy individuals). 5 Resistant starch reduced colonic neoplasia in several carcinogen-treated rat studies 6,7 but high intakes of this starch increased intestinal tumorigenesis in the genetically driven Apc1638N mouse model. 8 The antineoplastic eff ect of resistant starch is believed to result largely from the fermentation endproducts short-chain fatty acids and, in particular, butyrate. 9,10 In addition to inhibition of tumourcell proliferation, butyrate might reduce risk of colorectal cancer by enhancing the apoptotic response to DNA damage. 7,[11][12][13] Preliminary evidence shows that butyrate might have more potent antineoplastic eff ects on colon cancer cells with dysfunction of the DNA mismatch repair gene MLH1. 14 Few studies of the eff ects of resistant starch on colorectal carcinogenesis in human beings have been done and most have used putative biomarkers of colorectal-cancer risk as endpoints. Early studies showed that resistant-starch feeding for short periods (typically 4 weeks) reduced faecal excretion of cytotoxic secondary bile acids 15,16 and in some, but not all, studies reduced mucosal-cell proliferation. 16 Short-term (2-4 weeks) supplementation with resistant starch in 65 patients with colorectal cancer signifi cantly reduced (p=0·028) the proportion of mitotic cells in the top half of the colonic crypt and produced diff erential eff ects on expression of key cell-cycle regulatory genes (CDK4 and GADD45A) in tumour tissue. 17 The duration of resistant-starch feeding required to provoke changes in biomarkers of colorectalcancer risk is not known but is probably greater than 1 week. 18 In the Colorectal Adenoma/carcinoma Prevention Programme (CAPP) 1 study in young people with familial adenomatous polyposis, supplementation with resistant starch (1:1 blend of raw potato starch and high amylose maize starch [Hylon V11]) for a median intervention period of 17 months reduced crypt length and cell proliferation but no eff ect was detected on polyp count in the rectum and sigmoid colon (RR 1·05, 95% CI 0·73-1·49). 19 Lynch syndrome (also known as hereditary nonpolyposis colon cancer) is the most common monogenic predisposition to colorectal cancer with most patients carrying pathological DNA mismatch repair gene variants. In the CAPP2 study done over 6 years, 937 people with Lynch syndrome from 43 international centres commenced intervention with aspirin, resistant starch, or both in a two-by-two factorial design. 20,21 After intervention (mean 29 months [SD 13·7]) there was no evidence that either agent infl uenced development of colonic neoplasia with most lesions being adenomas. 20 However, the original design of the CAPP2 study included double-blind post-intervention follow-up for at least 10 years and a recent analysis (mean 55·5 months [SD 31·1]), showed that 600 mg aspirin a day for a mean of 25 months [13·4]) halved cancer incidence in carriers of hereditary colorectal cancer. 21 We assessed the eff ect of resistant starch on the incidence of colorectal cancer, the primary CAPP2 outcome, and on other Lynch syndrome cancers as secondary outcomes. The baseline population diff ers from our fi rst report, which was confi ned to those with an exit colonoscopy. 20

Trial design and participants
Details of the design and conduct of the CAPP2 study have been previously published. 20,21 Recruitment was from January, 1999, to March, 2005. Participants were required to have a proven germline mutation in a mismatch repair gene or a personal and family medical history commensurate with a diagnosis of Lynch syndrome. 83% of all participants had a proven germline mutation.
Ethics committee approval was obtained from all centres from which participants were recruited.

Randomisation and masking
In this two-by-two factorial trial, randomisation was undertaken centrally by DTB with the SAS (version 6) uniform random number generator to randomise each participant separately for aspirin and for starch. 20,21 We used block randomisation by geographical region so that after 16 randomisations four patients were in each of the four treatment combinations. Masking to treatment allocation was done with coded packaging, which did not reveal the nature of the contents (intervention or placebo).

Procedures
Patients who were randomly assigned to receive resistant starch were given Novelose 240 and Novelose 330 (National Starch and Chemical Company, Bridgewater, NJ, USA) in one-to-one blend daily with recommended administration in two separate doses, while those assigned to receive resistant-starch placebo were given Amioca waxy starch daily with recommended administration as per resistant starch to maintain masking, for up to 4 years. Novelose 240 and Novelose 330 are fermented in the large bowel to increase concentrations of butyrate and other short-chain fatty acids. 22 The additional digestible starch in the placebo (30 g a day) is nutritionally unimportant compared with typical dietary intakes of 150-350 g a day. 23 The period of intervention lasted a mean of 29 months (median 25·3 months [IQR 23·7-33·5]) at which point we reported eff ects of the interventions (aspirin and resistant starch) on colorectal neoplasia (most lesions being adenomas). 20 The original design of the CAPP2 study included double-blind post-intervention follow-up for at least 10 years to investigate eff ects on cancer risk. All CAPP2 participants were under regular colonic surveillance organised through their local clinical genetics service. Predominantly follow-up was via the routine annual surveillance. A few centres followed up every 2 years. The genetic centres were asked to return information on cancer history to the CAPP2 offi ce after these routine examinations. Follow-up continued as long as possible; this analysis refl ects the most recent followup data available from all participants.
The current analysis focuses on the follow-up of those participants randomly assigned to resistant starch or resistant-starch placebo who started the CAPP2 study from their date of entry into CAPP2 until the last known date for which the local clinical centre had information about their status regarding cancer diagnosis. However, in most cases, this timepoint corresponded with the date of last attendance at the clinic responsible for their surveillance.
In this analysis, we included (1) patients with Lynch syndrome cancers, which were recorded in the earlier report; 20 (2) all cancers that occurred in patients for whom an exit colonoscopy was not recorded in the initial report, thus excluding them from the statistical analysis in our fi rst report; 20 and (3) cancers that occurred subsequent to exit from the intervention phase. No information on adverse events was gathered during post-intervention follow-up. Details of adverse events reported during the intervention phase of the study are provided as an appendix to a study published by Burn and colleagues. 20

Statistical analysis
The analysis was designed to test the hypothesis that resistant starch would reduce the development of colorectal cancer (the primary outcome) and Lynch syndrome cancers (as a secondary outcome). The primary endpoint of this analysis was the development of colorectal cancer in participants, comparing the incidence in those randomly assigned to resistant starch with the incidence in those assigned to resistant-starch placebo.
The analysis was based on time to fi rst occurrence of colorectal cancer with life-table methods and Cox proportional hazards. The life-table analysis used the end of follow-up for each participant as (1) the time of fi rst colorectal cancer diagnosis, if aff ected, or (2) for those unaff ected, the last recorded contact date at which the clinical status of the participant was known. Analyses also included Cox proportional hazards models to estimate sex adjusted hazard ratios (HRs) and 95% CI and Kaplan-Meier curves to non-parametrically assess the outcome diff erences between the resistant starch and resistant-starch placebo interventions.
We used Poisson regression modelling to estimate incidence rate ratios (IRR) for the eff ect of resistant starch on the number of primary colorectal cancers diagnosed after randomisation, with the exposure time starting from randomisation until date of last known clinical status. All estimates were adjusted for duration of aspirin taken and sex. Analyses were done on an intention-to-treat basis (intervention assigned at randomisation) and on a per-protocol basis restricting analysis to those taking resistant starch for 2 years or more.
Since patients with Lynch syndrome are susceptible to primary cancers at multiple anatomical sites, a separate analysis addressed the eff ect of resistant starch on risk of all Lynch syndrome cancers that included all new cancers that occurred as a result of Lynch syndrome (ie, colorectal cancer, endometrial cancer, ovarian cancer, pancreatic cancer, and cancer of the brain, small bowel, gall bladder, ureter, stomach, and kidney). 24 We also compared the incidence of Lynch syndrome excluding cancers of the colorectum as a measure of the potential extra colonic eff ects of resistant starch.   All estimated eff ects of resistant starch were adjusted for sex and duration of aspirin taken, all p values were twosided, and all analyses were done with Stata (version 10).
This study is registered, ISRCTN 59521990.

Role of the funding source
The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The contracts associated with the donations from manufacturers required that they had access to the results before submission with up to 90 days for assessment. JCM, MM, DTB, and JB had access to the raw data. JCM, DTB, and JB took responsibility for the decision to submit for publication.

Results
Of the 937 eligible participants with Lynch syndrome who started the study, 463 were randomly assigned to receive resistant starch and 455 to receive resistant-starch placebo. The remaining 19 participants were randomly assigned to the aspirin or aspirin placebo intervention only and were excluded from this analysis (fi gure 1; appendix). There was no evidence of any interaction between the eff ects of aspirin and starch (Mantel-Haenszel χ² [1 degree of freedom] 0·02; p=0·9), however, for completeness, we report the eff ects of resistant starch adjusted for aspirin usage. More than half of participants were women (table 1). Long-term follow-up data were not available for 204 (22%) participants. Thus, for 714 participants, we report both on-trial information and longer follow-up information, whereas for 204 we report on-trial information only. Demographic data show that no diff erences occurred between those traced and not traced after the trial in this follow-up with respect to sex, randomisation category, or geographical location (data not shown). At the time of this analysis, eight (1%) participants were 10 years or longer from randomisation.
Median follow-up was 52·7 months (IQR 28·9-78·4; appendix). Since randomisation, 53 individuals-27 of those given resistant starch and 26 given resistant-starch placebo-developed 61 primary colorectal cancers ( Within 2 years of randomisation 10 10 More than 2 years from randomisation 17 16 Number of participants with other Lynch syndrome cancers* Since randomisation 17 26 Within 2 years of randomisation 6 9 More than 2 years from randomisation 11 17 Number of participants with one or more Lynch syndrome cancers (including colorectal cancer) Since randomisation 43 51 Within 2 years of randomisation 16 19 More than 2 years from randomisation 27 32 Number of participants with non-Lynch syndrome cancers 22 22 Data are number or median (IQR). *Two participants (one in resistant-starch placebo group and one in resistant starch group) had colorectal cancer and another had a Lynch syndrome cancer. These two participants are counted in the rows relating to both colorectal cancer and other Lynch syndrome cancers but in the row relating to all Lynch syndrome cancers these participants are counted only once.  We re-estimated the potential protective eff ect of resistant starch with a per-protocol analysis and obtained similar results. For this analysis, we defi ned 2 years intervention as consumption of 1400 packs (2×15 g a day); consumption was rounded down from a 2-year total (1461 packs) to allow for early scheduling of the exit colonoscopy, occasional missed dosage, or both. The HR for participants taking resistant starch for less than 2 years was 2·38 (95% CI 0·98-5·77; p=0·05), which suggested a possible adverse eff ect of shorter-term resistant starch treatment on incidence of colorectal cancer (table 3, fi gure 2). By contrast, the HR for participants taking resistant starch for 2 years or more was 1·09 (0·55-2·19; p=0·80) and the IRR was 0·98 (0·51-1·88; p=0·95; table 3, fi gure 2).
A secondary analysis assessed other Lynch syndrome cancers (ie, all Lynch syndrome cancers except colorectal cancer). Of the participants who developed cancer at a Lynch syndrome site other than the colorectum, 17 were randomly assigned to resistant starch and 26 to placebo (table 2; appendix). The HR for participants randomly assigned to resistant starch was 0·72 (95% CI 0·38-1·35; p=0·30; table 4, appendix) compared with the resistant-starch placebo group. Endometrial cancer was the most common non-colorectal cancer; 21 participants had endometrial cancer of whom ten were randomly assigned to resistant starch and 11 to resistant-starch placebo (appendix). Of note, no participants assigned to resistant starch had pancreatic or small-bowel cancer compared with fi ve cases of pancreatic cancer and three cases of small-bowel cancer in those randomly assigned to resistant-starch placebo. In view of the small numbers, this potential protective eff ect could be a chance observation. The per-protocol analysis showed that the HR for those allocated to resistant starch who took treatment for less than 2 years was 0·87 (95% CI 0·34-2·22; p=0·77) whereas for those taking resistant starch for 2 years or more the HR was 0·63 (0·28-1·40; p=0·26) with an IRR of 0·56 (0·26-1·23; p=0·15; table 4).
In analyses of all Lynch syndrome cancers including colorectal cancer, there was no evidence of a protective eff ect for resistant starch in either the per-protocol or intention-to-treat populations, nor when analysed by duration of treatment (table 5, fi gure 3). Cox proportional hazards models analysis by cumulative resistant starch consumption showed no evidence of a signifi cant doseresponse eff ect for colorectal cancer (p=0·56), noncolorectal Lynch syndrome cancers (p=0·21), or Lynch syndrome cancers overall (p=0·56; tables 3-5).
We have published a detailed analysis of the adenomas that were reported during the intervention phase of the CAPP2 study. 20 Where possible, details of adenoma development were also gathered by masked investigators in the post-intervention period. While incomplete, these data on 1068 colonoscopy reports showed no apparent eff ect of resistant starch on numbers of participants who developed adenomas subsequent to the intervention phase (ie, 114 reports of adenomas in each of the resistant starch and resistant-starch placebo groups from 558 and 488 colonoscopies, respectively [p=0·25]). No diff erence was seen in the number of colonoscopies in participants randomly assigned to resistant starch and to the corresponding placebo group (p=0·24; data not shown). Additionally, there was no evidence of an eff ect of resistant starch treatment on adenomas when the analysis was restricted to those on starch treatment for 2 years or more (p=0·25).  In view of the suggestion that the end product of resistant starch fermentation, butyrate, could have more potent antineoplastic eff ects on colon-cancer cells with dysfunction of the DNA mismatch repair gene MLH1, 14 the data were analysed according to the underlying mismatch repair gene defect and no eff ect was seen (appendix). 20 (32·8%) of the 61 colorectal cancers diagnosed in resistant starch or placebo groups were Dukes stage A, 25 (41·0%) were Dukes stage B, 12 (19·7%) were Dukes stage C and D, and four (6·6%) were unknown (appendix). 33 (54·1%) tumours were located in the ascending colon, transverse colon, and splenic fl exure, six (9·8%) in the descending colon, 13 (21·3%) in the sigmoid and rectum, and nine (14·8%) were unknown (appendix). No signifi cant diff erences in tumour staging (χ² [3 degrees of freedom] 6·73; p=0·08) and tumour location (χ² [3 degrees of freedom] 0·54; p=0·91) were seen between resistantstarch and resistant-starch placebo groups.

Discussion
To the best of our knowledge, our report is the fi rst randomised trial to assess the eff ectiveness of dietaryfi bre treatment (provided as a resistant-starch supplement) on carcinoma in human beings (panel). A recent meta-analysis of 21 prospective observational studies showed a signifi cant protective eff ect of dietaryfi bre int ake against colorectal cancer (RR per 10 g a day increase 0·90, 95% CI 0·86-0·94). 2 An earlier pooled analysis of prospective cohort studies 25 showed no eff ect of dietary fi bre on risk of colorectal cancer after adjusting for other dietary factors but that analysis included fewer studies and fewer cases of colorectal cancer and had a narrower geographical reach compared with the recent systematic review and meta-analysis by Aune and colleagues. 2 Previous randomised trials with occurrence of further adenoma as an outcome have not detected any benefi t of increased intake of dietary fi bre. These included studies with supplements of ispaghula husk 26 and cereal fi bre 27 and studies in which participants were counselled to increase intake of fi bre-rich foods. 28,29 The lack of benefi t associated with increased intake of dietary fi bre in these studies led to questions about the usefulness of adenoma recurrence as the primary surrogate endpoint in colorectal cancer chemoprevention research. 30,31 The CAPP2 study has provided experimental support for this scepticism since we reported no eff ect of aspirin treatment on neoplasia (principally adenomas) during the intervention phase of the study 20 but signifi cant protection against carcinoma in longer-term follow-up. 21 A limitation of the present study is that lack of resources has restricted longer-term follow-up of all CAPP2 participants. This loss to follow-up was concentrated in centres that had insuffi cient staff resources to allow long-term follow-up of all the study participants. Since randomisation to treatment was stratifi ed within geographical location, it is unlikely that any systematic bias exists in loss to follow-up. We examined the numbers of colorectal cancer that occurred during the intervention phase of the study and the followup phase, separately, and found no evidence to suggest that loss of cancer data due to failure to follow-up patients was diff erent between those randomly assigned to resistant starch or to placebo; there was also no eff ect of duration of taking resistant starch. Thus we conclude that there was no evidence for a protective eff ect of resistant starch even in those participants who were most compliant. We do not have data for intake of dietary fi bre for our study participants. However, geographical location is a crude surrogate for dietary and other lifestyle exposures. We randomised participants to treatment within geographical regions. We found that adjusting our analysis for region as well as for sex and aspirin duration gave a hazard ratio of 1·44 (95% CI 0·79-2·62),   A strength of the present study was that we used patients with Lynch syndrome due to a genetic defect in DNA mismatch repair as a sensitive model, which can show the eff ectiveness of chemoprevention agents (aspirin) on colorectal cancer and carcinoma at other sites. 21 The intention-to-treat analysis showed no eff ect of resistant starch on colorectal cancer or on other Lynch syndrome cancers. Per-protocol analysis suggested higher risk of colorectal cancer in those taking resistant starch for less than 2 years, but this eff ect is probably due to chance observation because no eff ect was seen in participants taking resistant starch for 2 years or more (table 3). We cannot exclude the possibility that resistant starch could have a small positive eff ect that we did not have the power to detect. A recent meta-analysis of the protective eff ects of dietary fi bre against colorectal cancer suggests an estimated 10% reduction in colorectal-cancer risk. 2 The 30 g a day source of resistant starch used in the CAPP2 study provides 13·2 g resistant starch, which reaches the large bowel and is classed as dietary fi bre. 20 On this basis, the resistant starch might have reduced colorectal risk by 8-18%, which is within the 95% CI of our intention-to-treat analysis (table 2). This lack of detectable eff ect of resistant starch on risk of colorectal cancer contrasts with the weight of observational studies showing a protective eff ect of higher intakes of dietary fi bre but is consistent with the fi ndings from randomised trials [26][27][28][29] of increased intake of dietary fi bre with occurrence of new colorectal adenoma as surrogate outcome. The eff ects of resistant starch in carriers of hereditary colorectal cancer could be diff erent from those in the general population and this remains to be tested. Additionally, observational studies might be aff ected by confounding or the apparent protection aff orded by higher intakes of dietary fi bre could refl ect the health benefi t of the dietary patterns (or whole lifestyles) adopted by such consumers rather than the specifi c antineoplastic eff ects of the carbohydrates measured as dietary fi bre. Evidence shows that the individual carbohydrates (and associated food components) quantifi ed as dietary fi bre have highly characteristic physicochemical properties and very diff erent eff ects on both fermentation in the large bowel and on function of the bowel mucosa. 3 As a consequence, emulation of the eff ects of naturally occurring dietary fi bre in plant-rich diets by a single type of polysaccharide, such as resistant starch, is unlikely. 1,32 We are under taking several secondary analyses, including investi gation of eff ects of smoking behaviour and of adiposity on risk of colorectal cancer in patients with Lynch syndrome in the CAPP2 study, which will be reported elsewhere.
In conclusion, we found no evidence that supplementation with 30 g a day of resistant starch aff ected development of colorectal cancer in carriers of hereditary colorectal cancer, although eff ects in the general population remain to be tested. Our study shows that supplementation with resistant starch does not emulate the apparently protective eff ect against colorectal cancer of diets rich in dietary fi bre, which has been shown in many, but not all, observational studies. From a public health perspective, eating more of a variety of food rich in dietary fi bre including wholegrains, vegetables, fruits, and pulses is a preferable strategy for reducing cancer risk.

Systematic review
We searched PubMed for articles on use of dietary fi bre or resistant starch as a bowel cancer, hereditary colorectal cancer, and Lynch syndrome chemopreventive agent published up to Feb 10, 2012. Our search terms were "colorectal cancer" OR "bowel cancer" AND "prevention" AND "fi bre" with the search restricted to randomised trials. Additionally, we replaced "fi bre" with "fi ber" or with "starch". Systematic reviews by other investigators had revealed compelling evidence for the protective eff ect of dietary fi bre against sporadic bowel cancer in observational studies. Resistant starch, a dietary starch fraction that is not digested in the small bowel, is a specifi c form of dietary fi bre. Our search of published works showed that no previous randomised trials have tested the effi cacy of resistant starch in patients with Lynch syndrome with cancer as the primary endpoint.

Interpretation
We found no evidence that dietary supplementation with resistant starch aff ected risk of colorectal cancer in carriers of hereditary colorectal cancer. The lack of eff ect of resistant starch shows that this supplement does not emulate the apparently protective eff ect of diets rich in dietary fi bre against colorectal cancer, at least in patients with Lynch syndrome. The eff ect of resistant starch on risk of bowel cancer in the general population remains to be tested.

Figure 3: Time to fi rst Lynch syndrome cancer in participants randomly assigned to resistant starch versus those assigned to resistant-starch placebo
Kaplan-Meier analysis was restricted to participants who had taken the intervention for 2 years or more and the analysis was adjusted for sex. Each point on the plot shows the estimated cumulative incidence by years of follow-up together with the corresponding 95% CI. CAPP=Colorectal Adenoma/carcinoma Prevention Programme.