The Effect of Butyrylated Starch on Bowel Polyps in Familial Adenomatous Polyposis: Results of a Randomized, Double-blind, Placebo-Controlled Crossover Trial.

Introduction
Colorectal cancer is the third most common cancer globally and the second leading cause of cancer-related deaths (1). The incidence and mortality of the disease vary geographically, with the highest incidence recorded in developed countries. Although rates have stabilized and are now decreasing in these countries (1), presentation of younger individuals with late-stage disease is increasingly common (2).
The risk of developing colorectal cancer can probably be reduced by modifying lifestyle and environmental factors (3), and dietary factors are now considered the major causative lifestyle factor globally (4). There is convincing evidence that higher intakes of dietary fiber are protective, with the consumption of whole grains associated with a reduced risk of colon cancer (5) whereas the consumption of refined grains associated with increased risk (6). This protection may result from the colonic production of the short-chain fatty acid butyrate during microbial fermentation of indigestible dietary fiber.
Butyrate is a potent histone deacetylase inhibitor (7) and modulates the expression of genes that regulate pathways that have inhibitory effects on tumorigenesis, including cellular proliferation, apoptosis, differentiation, DNA repair, and inflammation [see reviews (8, 9)].
Dietary butyrylated high-amylose maize starch (HAMSB) delivers significant quantities of butyrate to the large bowel of humans (10, 11). In the rectal mucosa of healthy humans, butyrate delivered by HAMSB has been shown to mitigate expression of oncogenic miRNA (12) and increased levels of mutagenic DNA adducts (13) induced by the consumption of a high red-meat diet. However, to date, there is no clinical evidence that butyrate protects against the initiation or growth of colorectal cancer or adenomas.
The aim of this clinical study [Australian FAP (AusFAP) study] was to determine if butyrate delivered by HAMSB reduces distal bowel polyp numbers or polyp growth in participants (PTs) with familial adenomatous polyposis (FAP). FAP is caused by mutations in the adenomatous polyposis coli (APC) gene and is characterized by the development of many colorectal adenomas. As mutations in the APC gene are also found in most nonfamilial colorectal cancer (14), it was hypothesized that any chemoprotective effect of butyrate observed in FAP PTs may also have relevance to nonfamilial colorectal cancer occurring in the normal risk population.

Materials and Methods

Study design
The design, procedures, and management of this clinical trial have been described in detail elsewhere (15). The complete AusFAP protocol is given in Supplementary Materials S1.
AusFAP was a multicenter study carried out in five public and private referral hospitals in Australia. Volunteers with medically diagnosed FAP and with a history of polyp detection at any previous surveillance scope were recruited if they met the study inclusion/exclusion criteria. PTs with intact colons (IC), ileorectal anastomoses (IRA), or ileal pouch anal anastomoses (IPAA) were recruited from January 16, 2014, to February 24, 2017 (Table 1). Written informed consent was obtained from each participant prior to recruitment, and the study was conducted in accordance with the Declaration of Helsinki.
A crossover design was chosen because the condition of individuals with FAP would be relatively stable over the study time; treatment effects of increased colonic butyrate were considered reversible and short-lived; the between-participant variability in the primary endpoint was known to be high; and fewer PTs are required for crossover trials.
This randomized, double-blind placebo-controlled crossover trial had three experimental phases, each of 6-month duration (Fig. 1). The investigational product was HAMSB (manufactured by Ingredion), and the control treatment was digestible low-amylose maize starch (LAMS; Melogel; manufactured by Ingredion). The endoscopies were video recorded to enable later assessment. Mucosa and polyp biopsies were collected during each examination.
During the baseline colonoscopy, two areas of mucosa of 36 mm diameter (as measured using open biopsy forceps) and at least 50 mm apart were chosen for tattooing. The center point of each area was tattooed with Spot Ex Endoscopic Tattoo and the distance from the anal verge carefully measured. The first site (tattoo one) was placed in an area that either contained no polyps or was cleared of polyps at baseline and during subsequent scopes. This allowed the number of polyps initiated during each intervention period to be assessed. The second tattoo area (tattoo two) contained small polyps if present in the PT at the baseline examination. These polyps were left in situ during the study, if deemed safe by the endoscopist, to enable any change in the size of preexisting polyps to be determined. Retention of small polyps in FAP surveillance is the standard of care in the participating clinical practices. Further details of the tattooing procedure are provided elsewhere (15) and in the study protocol (Supplementary Materials S1).
During the clinical component of the study, the exclusion criteria were changed to enable PTs to consume anti-inflammatory drugs for ≤10 days without exclusion from the per-protocol (PP) population. This was due to the high frequency of the use of these medications and the likelihood that courses of short duration would not affect study endpoints. The protocol was also changed to allow PTs who could not consume 20 g twice daily to ingest 10 g twice daily. This lower dose still significantly increased fecal butyrate compared with baseline levels in FAP PTs in a pilot study (16). The results of the pilot study also informed our decision that >50% intake of HAMSB be deemed compliant. Two PTs included in the AusFAP analyses consumed the lower dose for most of the trial.
The study was approved by the human research ethics committees of Southern Health (Human Research Ethics Committee reference number/11/SHB/13; Melbourne, Victoria), St Vincent’s Hospital (Sydney, NSW), and Royal Brisbane and Women’s Hospital (Brisbane, Queensland). Before initiation, the study was registered in a publicly accessible database [https://www.anzctr.org.au/ (RRID: SCR_002967) number: ACTRN 12612000804886]. All authors had access to the study data and reviewed and approved the final report.
The representativeness of study PTs is shown in Supplementary Table S1.

Study endpoints
The primary endpoint was the number of polyps throughout the large bowel. Secondary endpoints included the number of small (<2.4 mm), medium (2.4–9 mm), and large (>9 mm) polyps in the large bowel and the total and number of small, medium, and large polyps in tattoo areas 1 and 2. Exploratory endpoints included the intake of dietary fiber of FAP PTs, the gastrointestinal function and quality of life index [GIQLI (10)], and the fecal SCFA concentrations of a subgroup of FAP PTs. Safety endpoints include the number and type of adverse events (AE) and serious AEs experienced by PTs during the study.
For detail of the video assessments, refer to the study protocol published previously (15). Briefly, each video was assessed by the chief investigator A (CIA; assessor 3) and an independent second gastroenterologist (other) from a team of seven reviewers to determine the number and size of polyps in the bowel and the sum totaled for analysis. Assessment of the polyp burden in the tattoos was also undertaken by the CIA and an independent second gastroenterologist. Videos were randomly assigned to the seven reviewers with all videos from individual PTs assigned to the same second assessor. The baseline scope was identifiable because of tattoo placement, but assessors were blinded to the sequence of scopes and treatments the PTs received.

Statistical analyses

Power calculations
Power calculations were originally based on a parallel study (17) and estimated that the trial would require 120 PTs (n = 60/group). AusFAP has a crossover design, and a reviewing statistician recalculated the power calculation based on the outcomes of a more recent crossover trial (18). These authors found a 23% difference (P = 0.046) between the number of FAP PTs with reduced small polyps (<2 mm diameter) in the treatment group compared with the placebo group. However, this was not the primary endpoint used in the AusFAP study. The reviewing statistician nevertheless calculated that using a full crossover analysis independent of carryover with a two-sided comparison (85% power; 23% difference in numbers of polyps), 64 PTs (n = 32/group) were required. Although the target sample size was 64, recruitment was ceased when 72 patients had been enrolled to allow for withdrawals during this long study.

Concordance analyses
Two independent readers assessed the global total polyp number and sizes [small (<2.4 mm), medium (2.4–9 mm), and large polyps (>9 mm)] for each video. Prior to analysis, a nonparametric Bland–Altman analysis (19) was used to assess reader agreement between the counts. The intraclass correlation coefficients (ICC) from the readers were calculated (20) and indicated moderate agreement between reader CIA and the other readers’ counts for global [ICC: 0.69; 95% confidence interval (CI), 0.62–0.75] and small polyps [ICC: 0.53; 95% CI, 0.43–0.61] and poor agreement for medium [ICC: 0.15; 95% CI, 0.09–0.2] and large polyps [ICC: 0.37; 95% CI, 0.15–0.56)]. ICCs were estimated using R’s iccCounts package version 1.1.2 (https://CRAN.R-project.org/package=iccCounts). Based on this concordance analysis, more than 65% of videos assessed had percentage differences outside the limits of agreement (2.5th and 97.5th percentiles of the paired differences). Hence, only the assessments for the CIA (reader 3), the only reader who assessed all videos, were used in the study analysis.
Concordance analysis was also used to assess reader agreement of the tattoo assessments. Less than 48% of videos had percentage differences outside the limits of agreement. Based on this analysis, the assessments of both reader 3 and the seven other assessors (other) were used in the tattoo analyses.

General statistical description
Statistical analyses followed a prespecified statistical analysis plan described in detail in Supplementary Materials S2, except for Table 2, which details a post hoc analysis not included in the statistical analysis plan.1. Polyp counts

Briefly, generalized linear mixed effects models (GLMM) suitable for count data (Poisson, negative binomial, and zero-inflated Poisson) were used to estimate the effect of HAMSB on the primary and secondary outcomes. GLMMs take into consideration the missing data.
The analysis of the intention-to-treat (ITT) population, which included all PTs who were randomized, was considered the primary analysis. A PP analysis was also performed, consisting of all study PTs who were randomized, achieved >50% treatment compliance, and did not experience major protocol violations.
Analyses were conducted using Stata/SE for Windows version 17 (64-bit x86-64) or higher or R software version 4.3.1 or higher. Count regression modeling was performed using R’s glmmTMB package (RRID: SCR_025512; ref. 21).
The global polyp counts after 6 months of ingesting LAMS and HAMSB were also analyzed using Prism version 10.5.0 for macOS (GraphPad Software, https://www.graphpad.com; RRID: SCR_002798). The data for PTs with global polyp counts for each treatment (n = 40) were combined irrespective of treatment sequence and the counts log-transformed and then analyzed using a paired two-tailed t test.2. Dietary and fecal measurements and GIQLI

Linear mixed effects models were used to estimate the difference between mean nutrient intakes of fiber and calcium and total GIQLI scores for HAMSB compared with LAMS. These included time (baseline, weeks 12, 26, 38, 52, and 78), treatment, stratification factors (age and surgery type), and a random individual effect to account for the correlation between total GIQLI scores from the same individual.

Results

Study PTs
Of 76 PTs assessed for eligibility, 72 were randomized to treatment with 64 commencing the dietary interventions and 49 completing the study (CONSORT diagram, Supplementary Fig. S1). At week 26 for period 1, there were 52 PTs in the ITT population assessed with complete data for the global, small, medium, and large polyp counts throughout the large bowel, 41 PTs at week 52 for period 2, and 43 at week 78 for period 3. There were similar numbers in each treatment group at each timepoint. The baseline characteristics of PTs by treatment sequence are shown in Table 1.
Descriptive analysis tables summarizing the values used as the response in the regression models are shown in Supplementary Tables S2–S5.

Primary and secondary endpoints

ITT analyses excluding carryover effects
Analysis of the global polyp count after PTs consumed 6 months of LAMS control compared with 6 months of HAMSB showed that ingestion of HAMSB did not significantly reduce the number of polyps when analyzed using a two-tailed paired t test on log-transformed data (33.63 ± 8.05 polyps for LAMS and 31.41 ± 7.92 polyps for HAMSB; mean ± SEM; P = 0.16; Table 2). The distribution and median of data at baseline, when LAMS and HAMSB were consumed, and after a 6-month washout period without treatment are shown in the Tukey boxplot in Supplementary Fig. S2. The number of polyps removed was taken into account, but no allowance was made for sequence of treatment, baseline polyp counts, age group (12–18, 19–45, and >45 years), or surgery type (IC, IRA, and IPAA) in either the t test or the graphed data.
GLMM analysis of the primary endpoint in the ITT cohort, excluding any carryover effect, showed no significant effect of HAMSB compared with LAMS on the mean count of global polyps in the colon [0.9 fold change (FC); 95% CI, 0.77–1.06; P = 0.218] after taking into account the number of polyps removed at prior colonoscopies, baseline polyp counts, age group (12–18, 19–45, and >45 years), and surgery type (IC, IRA, and IPAA; Table 3). Similarly, analysis of secondary endpoints showed that HAMSB did not affect the numbers of small (0.88 FC; 95% CI, 0.71–1.10; P = 0.267), medium (1.16 FC; 95% CI, 0.76–1.77; P = 0.486), or large polyps (0.85 FC, 95% CI, 0.39–1.86; P = 0.686) in the colon compared with LAMS. The 95% CIs are wide, indicating uncertainty about the true direction of any treatment effect, and the P values provide only weak evidence against the null hypothesis of no treatment difference.
In tattoo one, there were trends toward decreased mean total polyp counts (0.78 FC; 95% CI, 0.58–1.05; P = 0.106) and mean small polyp counts (0.72 FC; 95% CI, 0.50–1.03; P = 0.074) by HAMSB compared with LAMS (Table 4). The count for medium polyps was not affected (1.03 FC; 95% CI, 0.63–1.66; P = 0.914) by HAMSB. The mean total (0.97 FC; 95% CI, 0.78–1.22; P = 0.812), small (1.04 FC, 95% CI, 0.75–1.44; P = 0.806), and medium (1.33 FC; 95% CI, 0.83–2.13; P = 0.231) polyp counts in tattoo two were unaffected by HAMSB compared with LAMS (Table 5). There were insufficient large polyps for meaningful analyses in either tattoo. No statistically significant differences were observed for the secondary endpoint polyp counts (regardless of size) in tattoos one and two in the ITT cohort.
Sensitivity analyses were undertaken to determine whether removing certain factors affected the primary endpoints. Neither mean count nor the size of global polyps changed greatly if polyps of size 6 to 10 mm were excluded from the number of polyps excised at prior colonoscopy (Supplementary Table S6). Similarly, removal of videos with obscured footage and images with mammillations had little impact on the primary endpoint analysis (Supplementary Table S7).

PP analyses excluding carryover effects
The results of the analysis of the PP population, assuming no carryover effects for counts throughout the large bowel, were similar to those for the ITT population, except that the effects on global and small polyp numbers were slightly stronger (Table 3). The mean global count showed no effect of HAMSB compared with LAMS (0.87 FC; 95% CI, 0.73–1.05; P = 0.152), whereas the mean count of small polyps approached significance (0.79 FC; 95% CI, 0.62–1.0; P = 0.051). The number of medium (1.13 FC, 95% CI, 0.67–1.91; P = 0.655) and large polyps (0.85 FC; 95% CI, 0.14–5.27; P = 0.860) were unaffected by treatment.
Effects of HAMSB ingestion on polyp burden in tattoo one in the PP population (Table 4) were also similar to those in the ITT population. In tattoo two, the total mean polyp count was unaffected by HAMSB (0.92 FC; 95% CI, 0.7–1.21; P = 0.554) compared with LAMS, whereas there was a trend toward a reduction in the number of small polyps (0.74 FC; 95% CI, 0.53–1.02; P = 0.067) and a potential increase in the number of medium polyps (1.53 FC; 95% CI, 0.92–2.55; P = 0.103; Table 5). There were insufficient large polyps in either tattoo for meaningful analyses.
There were no statistically significant differences observed for the primary and secondary endpoints in the PP population. The P values consistently indicated weak evidence against the null hypothesis of no treatment difference, and the 95% CIs were wide, reflecting uncertainty about the true direction of any treatment effect.
The two sensitivity analyses undertaken to exclude polyps of medium size (6–10 mm) excised at prior colonoscopies and videos with tiny mammillations or obscured footage did not change the number or size of global polyps in the large bowel.

Effects of carryover in ITT and PP analyses
There were some carryover effects of HAMSB consumption into subsequent periods. This bias was more apparent in the analysis of PP (Supplementary Table S8) than that of ITT (Supplementary Table S9) population global polyp counts, strengthening any apparent treatment effect for the primary and secondary outcomes. Bias due to carryover effects similarly strengthened the possible effect of HAMSB treatment on the small polyp numbers in tattoo one for the PP (Supplementary Table S10) but not the ITT population (Supplementary Table S11). Polyp number or sizes in tattoo two seemed less affected by carryover in both ITT and PP populations (Supplementary Tables S12 and S13, respectively). Its impact on medium-sized polyps in tattoos one and two was variable in the PP population. Overall, ingestion of HAMSB in the previous period may have altered the treatment effect in the subsequent study period.
The 95% CIs and P values corresponding to the carryover effect estimates in Supplementary Tables S8–S13 are wide and contain a range of scenarios including a large reduction in the treatment effect to a large inflation of the treatment effect due to carryover. All P values indicate that there is weak evidence against the null hypothesis of no carryover effect.

Other secondary outcomes

GIQLI
There were 71 PTs with GIQLI data included in the ITT analysis. A higher score across the range 1 to 5 is associated with a better quality of life and fewer side effects (Supplementary Materials S3). The mean total GIQLI score while consuming HAMSB relative to LAMS did not differ significantly (mean difference: −0.28; 95% CI, −1.51 to 0.94; P = 0.647; Supplementary Table S14). Similar results were seen for the PP population with a GIQLI mean difference when consuming HAMSB relative to LAMS of 0.64 (95% CI, −1.03 to 2.31; P = 0.451).

Diet of the PTs
The means, IQRs, and ranges of the dietary intakes of nutrients for the ITT population were similar during baseline and the treatment periods of the study (Supplementary Table S15). The intakes of the PTs were comparable with the mean daily nutrient intakes for Australians 19 years and older (22), with the exception that the reported intake of alcohol was lower in the AusFAP cohort.
The mean differences in daily intakes of dietary fiber and calcium by the ITT population were significantly higher at weeks 12 and 37 than at baseline [Supplementary Table S16; fiber: 7.93 g (95% CI, 5.59–10.58; P < 0.001); 7.2 g (95% CI, 4.41–9.99; P < 0.001); calcium: 173.99 mg (95% CI, 48.53–299.46; P < 0.007) and 111.46 mg (95% CI, −20.79 to 243.7; P < 0.098) for weeks 12 and 37, respectively). About 40% of the PTs consumed their supplements in milk products, which may explain the higher calcium intakes during the study than at baseline.
The average intake of HAMSB ingested per day by >50% compliant PTs was 32.0 g, which provided 20.5 g of dietary fiber as resistant starch (RS). RS is dietary starch that resists digestion in the small intestine and is fermented in the large bowel. The average intake of LAMS per day by >50% compliant PTs was 32.8 g of LAMS/day, which contained 1.8 g of dietary fiber. When intakes from the interventions and diet were combined, PTs consumed an estimated 48.7 g fiber/day during HAMSB treatment and 27.9 g/day during LAMS treatment.

Fecal parameters of the PTs
Fecal parameters were available from 12 PTs at baseline and 11 PTs when consuming HAMSB and LAMS. There were 22 observations from HAMSB PTs and 21 observations from LAMS PTs collected either at 4 and 26 weeks or 30 and 52 weeks, depending on each participant’s randomization sequence. Fecal pHs were lower when PTs consumed HAMSB compared with baseline or after LAMS [mean (range); pH 5.72 (5.03, 6.97); pH 6.43 (5.3, 7.96); and pH 6.32 (5.1, 7.5) respectively], and butyrate concentrations were higher [µmol/g; 37.12 (9.64, 90.3); 24.06 (2.04, 72.52); and 19.28 (2.44, 53.76), respectively; Supplementary Table S17]. These limited results suggest that the microbiota of the residual large bowel of FAP PTs can release esterified dietary butyrate.

AEs
The AEs are listed in Supplementary Tables S18 and S19. Mild-to-moderate gastrointestinal disorders were the most common AE, and these were frequent irrespective of dietary intervention. There were no significant differences in blood test result abnormalities between treatment sequences (Supplementary Table S20).

Protocol deviations
The majority of the protocol deviations were minor (88%), and none of the major deviations (12%) affected the safety of the PTs (Supplementary Table S21).

Discussion
This is the first clinical trial to investigate the effects of butyrate delivered by HAMSB on distal bowel polyp burden and growth in PTs with FAP.
There was no statistically significant effect of HAMSB on the global large bowel polyp number, although we found a trend for HAMSB to reduce the number of total and small polyps in tattoo one. There was no consistent effect on the number or size of polyps left in situ (tattoo two). There was evidence of carryover of treatment effects between periods, which were more pronounced in the PP than the ITT population. Carryover strengthened the apparent treatment effect on the global counts for polyps of any size throughout the large bowel and had mixed effects on secondary outcomes.
Although the effects of HAMSB on study endpoints were not statistically significant, there was a consistent trend suggesting that butyrate, delivered by HAMSB, may reduce the initiation of colonic polyps and hence contribute to the chemoprotective effect of dietary fiber against colorectal cancer. Butyrate is a pleiotropic chemical with a wide range of anticancer-like properties. As an histone deacetylase inhibitor (7), it has been shown to modulate intracellular signaling cascades that promote apoptotic (23) and anti-inflammatory pathways. Furthermore, butyrate can protect against tumorigenesis (24) by stimulating the differentiation and inhibiting the proliferation of colorectal cancer cells (25). The influence of large bowel HAMSB on mucosa and polyp apoptotic and proliferative pathways will be examined in future analyses of biopsies from the PTs in this study.
Butyrate can stimulate mucosal proliferation under starved or atrophic conditions (26). In this study, there was a trend toward increased numbers of medium-sized polyps in response to HAMSB among polyps left in situ, particularly in the PP population. Although this result was not statistically significant and may be an artifact of the small number of medium-sized polyps occurring in this trial, in other studies high intakes of fiber have been shown to be protective against advanced adenomatous polyp growth. In the Polyp Prevention Trial (27), intakes of dry beans in the top percentile reduced the recurrence of advanced adenomas, suggesting that high fiber intakes may be protective (OR = 0.35; 95% CI, 0.18–0.69; P = 0.001). Similarly, the Australian Polyp Prevention study (28) showed that supplementation with 25 g unprocessed wheat bran daily in combination with a low-fat diet protected patients from the development of large adenomas.
Other studies investigating the effects of RS on polyposis found no protective effect in PTs with FAP (29) and hereditary nonpolyposis colorectal cancer (30, 31). Burn and colleagues (29) supplemented PTs with a mixture of potato starch and high-amylose maize starch which provided an estimated total of 17 g RS/day. These authors found a trend of increased total crypt cell proliferation (95% CI, 0.94–1.73 28%; P = 0.12) in FAP PTs supplemented with RS compared with the non-RS group. In the latter studies (30, 31), PTs consumed 30 g of high-amylose starches/day for a mean of 25 months. This provided an estimated 9 g RS/day, which may have been insufficient for chemoprotection (32). Follow-up analyses undertaken at 10 and 20 years after the intervention in PTs showed that this level of RS intake afforded no long-term protection against colorectal cancer but may protect against non-colorectal cancers in Lynch syndrome, particularly those of the upper gastrointestinal tract (33). The average intake of HAMSB ingested per day by compliant PTs in AusFAP provided 20.5 g of RS to the large bowel in addition to high quantities of butyrate (50.2 ± 2.4 mmol butyrate/day; refs. 10, 11). Neither the high-amylose maize starch nor potato starch used in these studies contained significant levels of dietary fiber other than RS.
When combined with dietary sources, the supplements consumed by PTs in the present study resulted in total estimated intakes of 48.7 g/day fiber when HAMSB was consumed and 27.9 g/day fiber when LAMS was ingested compared with 19.6 g/day dietary fiber at baseline. In addition to acting as a substrate for SCFA production in the large bowel, dietary fiber may protect from colorectal cancer by reducing the concentration of intestinal carcinogens by increasing stool mass and reducing transit time (34), by suppressing secondary bile acid production (35) and proteolytic fermentation (36), and/or slowing the glycemic response (37).
Intake of alcohol is associated with increased colorectal cancer risk, whereas dietary fiber, calcium, magnesium, and phosphorus are inversely associated (38). The intakes of these nutrients by our FAP PTs were not significantly different across the study and would unlikely influence polyp count.
Although the current study did not examine the role of the gut microbiota in FAP polyp development, the aberrant composition of intestinal microorganisms has been associated with colorectal cancer (39, 40). The colonic mucosa of patients with FAP is covered with patchy bacterial biofilms composed predominantly of Escherichia coli and Bacteroides fragilis and is highly enriched with oncotoxins (colibactin clbB and Bacteroides fragilis toxin bft) compared with healthy individuals (41). There is strong experimental evidence supporting the carcinogenic potential of molecular subtypes of these two oncotoxin-producing bacteria (42, 43).
Okumura and colleagues (44) suggested that a relationship may exist between butyrate secreted by Porphyromonas spp invading colorectal cancer tissues and tumorigenesis as the butyrate increase coincides with tumor senescence and associated inflammation. However, Porphyromonas spp are normally part of the oral microbiome (45) and were not detected in fecal, polyp, or mucosal samples of FAP children (46) or feces of FAP adults (47).
No washout period between treatments was included in AusFAP, as the dietary interventions were relatively long, and there were concerns that PTs may be lost to follow-up. The effects of HAMSB were considered reversible and short-lived, and carryover effects were not anticipated as the life of an intestinal epithelial cell is only 4–5 days (48). Furthermore, our earlier studies showed that the concentration of butyrate in the feces changes quickly on ingestion of esterified butyrate (10). More recently, however, a carryover protective effect from HAMSB against the oncogenic effects of a high red-meat diet lasting at least 4 weeks was observed in the rectal mucosa of PTs (12). Inclusion of a substantial washout period should be considered in future crossover studies involving butyrylated starch.
In the present study, 45 of 72 enrolled patients had no confirmatory genotype information available. Our study population may therefore have contained some polyposes driven by variants in genes other than APC. However, our observations focused on the number and size of polyps in familial polyposis per se, irrespective of genetic etiology.
This study experienced several practical challenges. Polyp counts differed between endoscopists reviewing the same colonoscopy video, particularly when polyp numbers were very high, bowel preparation was suboptimal, or hyperplastic polyps were present. The implications of interobserver assessment variability on analysis should be carefully considered in future studies seeking to address similar questions. Accurate identification of the two tattoos was also difficult at times. A further potential limitation was inclusion of patients with FAP with IRA, IPAA, and IC in this study rather than limiting the cohort to PTs with IC only. These factors all contribute to the wide 95% CIs associated with the treatment studied here. Carryover effects associated with the HAMSB intervention were not anticipated and further cloud definitive interpretation of the present study results despite their potential to be advantageous clinically. Longer-term studies in a larger cohort of PTs with FAP and ICs or Lynch Syndrome are warranted, and studies using the pig model of FAP to investigate the analysis of early events in adenoma progression (49) may be justified.
Overall, we were encouraged by the consistency of trends observed, particularly as they relate to small polyps. In addition, HAMSB seems to have been well tolerated when taken twice daily for 6 months with no indication of side effects compared with other chemopreventative agents such as aspirin, sulindac, or celecoxib.
In conclusion, although ingestion of HAMSB had no significant effect on the primary endpoint, global polyp numbers, consumption tended to reduce the growth of small polyps without causing regression or growth of existing polyps. The carryover effect suggests that any impact may continue after supplementation has ended. Future analysis of biopsies from this study may reveal whether HAMSB affects the rates of apoptosis or proliferation in colonic biopsies. HAMSB consumed as a food ingredient, supplement, or as a medical food may have potential to reduce the risk of colorectal cancer in individuals with FAP and sporadic colorectal cancer in the wider community, in which low dietary fiber intake remains a major factor contributing to the high incidence of this disease.

Supplementary Material
Supplementary Figure 1Consolidated Standards of Reporting Trials (CONSORT) diagram shows the flow of participants through the trial from enrolment to data analysis.

Supplementary Figure 2Tukey boxplot of large bowel polyp numbers

Supplementary Materials 1Study protocol

Supplementary Materials 2Statistical analysis plan

Supplementary Materials 3Gastrointestinal quality of life questionnaire questions

Supplementary Table 1Representativeness of study participants table

Supplementary Table 2Descriptive statistics for polyp counts of each size in the large bowel intention to treat population

Supplementary Table 3Descriptive statistics for polyp counts of each size in the large bowel in the per protocol population

Supplementary Table 4Descriptive statistics for polyp counts of each size in tattoo 1 and 2 in the intention to treat population

Supplementary Table 5Descriptive statistics for polyp counts of each size in tattoo 1 and 2 in the per protocol population

Supplementary Table 6Ratio of mean counts for effect of treatment, week and carry-over on global polyp burden in the large bowel excluding polyps 6-10 mm in size ITT population

Supplementary Table 7Ratio of mean counts for effect of treatment, week and carry-over on global polyp burden in the large bowel excluding participants with video issues ITT population

Supplementary Table 8Ratio of mean counts for effect of treatment, week and carry-over on global polyp burden in the large bowel PP population

Supplementary Table 9Ratio of mean counts for effect of treatment, week and carry-over on global polyp burden in the large bowel ITT population

Supplementary Table 10Ratio of mean counts for effect of treatment, week and carry-over on tattoo 1 polyp burden in the large bowel PP population

Supplementary Table 11Ratio of mean counts for effect of treatment, week and carry-over on tattoo 1 polyp burden in the large bowel ITT population

Supplementary Table 12Ratio of mean counts for effect of treatment, week and carry-over on tattoo 2 polyp burden in the large bowel ITT population

Supplementary Table 13Ratio of mean counts for effect of treatment, week and carry-over on tattoo 2 polyp burden in the large bowel PP population

Supplementary Table 14Mean difference for effects of treatment and time point on modified gastrointestinal quality of life index

Supplementary Table 15Dietary intake by ITT population during each intervention period compared to Australian daily macronutrient intake

Supplementary Table 16Mean differences for effects of treatment and time point on fiber and calcium intake after accounting for age group and surgery type in ITT population.

Supplementary Table 17Faecal parameters of participants while consuming HAMSB and LAMS

Supplementary Table 18Severity of adverse events by system organ class and treatment sequence

Supplementary Table 19Adverse events and their relationship to study interventions

Supplementary Table 20Clinically significant clinical laboratory evaluations and other assessments

Supplementary Table 21Summary of protocol deviations