Parts of this chapter have been reproduced from Hull et al.1 © 2018 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license (https://creativecommons.org/licenses/by/4.0/).
Summary of colorectal adenoma findings
The seAFOod polyp-prevention trial found no evidence of an effect of either EPA or aspirin on the primary end point of the proportion of individuals with one or more colorectal adenomas at the 12-month surveillance colonoscopy (the ADRa) in patients deemed ‘high risk’ in the English BCSP. Overall, ≈60% of trial participants had at least one colorectal adenoma at surveillance colonoscopy. The trial was designed to detect an absolute difference between groups of 10 percentage points, or a relative difference of 18%, assuming an ADRa of 60% in the double placebo group. The 95% CIs for the primary analyses excluded a difference of this magnitude on ADRa for both treatments. Moreover, the primary outcome data were robust to sensitivity analyses.
Secondary analyses of the effects of EPA and aspirin on colorectal adenoma number provided evidence of chemopreventive activity of both agents. Aspirin was effective at reducing the total number of colorectal adenomas per participant, but there was no evidence of an effect on total MAP by EPA. Other secondary analyses suggested that there are colorectal adenoma subtype- and site-selective effects of EPA and aspirin. Participants randomised to EPA had a reduced number (MAP) and ADRa for conventional dysplastic colorectal adenomas in the left colon and rectum compared with those randomised to placebo. Participants randomised to aspirin had a reduced number of adenomas in the right colon, particularly for serrated adenomas, and also a reduced risk of conventional colorectal adenomas. An exploratory analysis found no evidence of any effect of either EPA or aspirin on recurrent colorectal adenoma size.
Colorectal adenoma end points
The ADRa was chosen as the primary end point, on historical grounds that previous ‘sporadic’ polyp-prevention trials had all used this binary measure of the presence of any colorectal adenoma at surveillance colonoscopy as the primary measure of colorectal adenoma risk.49 We expected the high ADRa at the 1-year surveillance colonoscopy of ‘high-risk’ individuals in the English BCSP compared with previous polyp-prevention trials.98 The high ADRa in the seAFOod trial population of obese, male-predominant ‘polyp formers’ will also have been driven by the uniformly excellent colonoscopy quality in the BCSP, in which the ADRa figures prominently as an individual QA measure.38,95,97 It is noteworthy that those previous aspirin RCTs with the highest placebo ADRa {47.1% [Aspirin and Folate Polyp-Prevention Study (AFPPS)114] and 53.4% [APACC115]} reported the smallest (statistically insignificant) risk reduction associated with aspirin (risk ratio 0.88 and 0.95, respectively) compared with the other two aspirin RCTs,49 despite the overwhelming evidence for a CRC chemoprevention effect of aspirin.43–48 Therefore, the use of the ADRa as a biomarker of chemoprevention efficacy in screening cohorts undergoing high-quality colonoscopic assessment with a high ADRa must be questioned.
By contrast, previous chemoprevention RCTs in FAP patients with a large colorectal adenoma burden have reported efficacy based on colorectal adenoma number and size.55,67,116 Moreover, three of the four previous aspirin ‘sporadic’ polyp-prevention trials have reported colorectal adenoma multiplicity as a secondary outcome.115,117,118 Therefore, adenoma number was stipulated as a secondary end point in the seAFOod trial. Reduction in colorectal adenoma multiplicity is widely accepted as an indicator of anti-CRC activity in pre-clinical studies.61 More recently, adenoma number has gained credence as a biomarker in ‘sporadic’ colorectal adenoma prevention RCTs,39 driven by increasing use of MAP as an outcome measure in colonoscopy QA studies.38 Moreover, colorectal adenoma number predicts future CRC incidence and mortality in population-based observational studies.40–42 We suggest that subsequent chemoprevention RCTs utilise MAP as the primary measure of chemopreventative efficacy, especially in ‘high-risk’ study populations with a high ADRa, in which the ADRa will be a relatively insensitive measure of colorectal adenoma risk. The recent polyp-prevention RCT of metformin did indeed employ co-primary end points of colorectal adenoma incidence (ADRa) and number (MAP).39
Site and adenoma subtype specificity of the chemopreventive effects of aspirin and eicosapentaenoic acid
All three aspirin RCTs, which reported colorectal adenoma number as a secondary outcome,115,117,118 demonstrated a significant reduction in total MAP in the aspirin group compared with the placebo group. The APACC trial reported total MAP data at the 12-month colonoscopic follow-up, similar to the seAFOod trial.115,119 Coupled with the known CRC-preventative properties of aspirin demonstrated during observational follow-up of large-scale aspirin RCTs,46–48 our data (and those from the previous polyp-prevention trials)115,117,118 justify colorectal adenoma number as a clinically relevant biomarker of CRC risk in CRC chemoprevention trials, in addition to the ADRa. The similar magnitude decrease in colorectal adenoma number that we observed for aspirin in the seAFOod trial compared with the previous aspirin RCTs suggests that the reductions in colorectal adenoma number that we have reported are likely to be genuine and be clinically meaningful for CRC risk reduction. However, we do acknowledge that the number of secondary comparisons that have been made increases the possibility that findings could be spurious.
With increased understanding that different colorectal carcinogenesis pathways exist, improved insight into differential effects of chemopreventive agents on distinct colorectal adenoma types is of paramount importance. Chemopreventive activity of EPA against conventional colorectal adenomas (but not serrated lesions), based on both MAP and ADRa outcomes, is consistent with the known efficacy of the same dose of EPA-FFA in FAP patients67 with rectal adenomas, which are conventional, wholly dysplastic lesions occurring on a background of germline mutation of one APC allele and tumour initiation after loss of heterozygosity of the other APC allele.120 By contrast, we report that EPA has no chemopreventive activity against serrated adenomas. Based on secondary colorectal adenoma number analysis, a possible signal that EPA use may actually increase serrated lesion risk requires careful examination in future studies of ω-3 PUFA supplementation/dietary intake and CRC risk, in which there is stratification based on location and the different molecular subtypes of CRC.
By contrast, the effect of aspirin on proximal colorectal adenoma multiplicity was partly explained by the strong preventative activity against serrated lesions. The seAFOod trial data add to an existing body of evidence that the reduction in CRC incidence and mortality by aspirin is explained by a dominant effect on proximal CRC, with less preventative efficacy against distal CRC.47,48 Post hoc analysis of data from the AFPPS trial114 has revealed that risk of right-sided (but not left-sided) serrated polyps was lower in those allocated aspirin than in those allocated placebo.121 Moreover, a meta-analysis of observational studies of lifestyle factors and serrated polyp risk has reported that serrated polyp risk is decreased in aspirin users (relative risk 0.81, 95% CI 0.67 to 0.99).122
The seAFOod trial data suggest that aspirin also has some antineoplastic activity against conventional adenomas. However, in contrast with EPA, aspirin (100–600 mg daily) has not been demonstrated to have efficacy against rectosigmoid polyps in FAP RCTs,123,124 which is consistent with the dominant ‘right-sidedness’ of aspirin chemopreventative activity.47,48
Only two previous ‘sporadic’ polyp-prevention trials have reported an a priori analysis of colorectal adenoma risk related to location in the colorectum.114,125 In the RCT of the selective COX-2 inhibitor rofecoxib, there was no differential effect on ADRa dependent on location in either the right or left colon.125 In the AFPPS trial, no differential effects of aspirin on right or left ADRa were observed, similar to the seAFOod trial data.121 There has been no location-dependent analysis of colorectal adenoma number in a polyp-prevention trial prior to the seAFOod trial.
The seAFOod trial data, in combination with greater insight into the molecular pathogenesis of early stages of colorectal carcinogenesis,4 should prompt a paradigm shift in polyp-prevention trials driving data collection on colorectal adenoma subtype (conventional adenoma vs. serrated polyp) and tumour location (proximal/right vs. distal/left), leading to evaluation of stratified chemoprevention. It remains unclear whether or not the biology of the early stages of colorectal carcinogenesis during initiation and growth of a colorectal adenoma differs significantly such that rectal lesions should be classified separately from colonic neoplasia. At the present time, available evidence suggests that the molecular pathologies of colorectal adenomas from the rectum and distal colon do not differ significantly.126
Dose considerations
The seAFOod trial evaluated 300 mg of aspirin daily based on the random-effects meta-analysis of aspirin polyp-prevention RCTs that demonstrated a significant colorectal adenoma risk reduction at doses of 300–325 mg daily, particularly for advanced lesions (risk ratio 0.71, 95% CI 0.56 to 0.92 for advanced ADRa), albeit in subjects approximating to ‘intermediate-risk’ surveillance populations.49 Observational data on CRC risk suggest that lower daily doses of aspirin are likely to have ‘right-predominant’ effects on colorectal adenoma multiplicity similar to those reported herein.48
The FFA-equivalent dose of 2 g of EPA daily was based on the same dose of EPA-FFA that demonstrated efficacy in the proof-of-concept RCT in FAP patients.67 This is a relatively high daily amount of ω-3 PUFA, which would be difficult to provide using more widely available fish oil or re-esterified TG formulations.65 Restricted supply of the original EPA-FFA formulation, beyond TMG control, meant that we had to switch the IMP to an EPA-TG formulation and matching placebo during the recruitment phase of the trial. This, however, provided a unique (originally unplanned) opportunity to compare ‘bioavailability’ of two EPA formulations, with the same FFA dose, in the context of prolonged dosing in a RCT. The data suggest that there is no meaningful difference in tissue EPA incorporation, measured by RBC membrane and rectal mucosal EPA levels, between EPA-FFA and EPA-TG. Therefore, colonoscopic outcome data from the two populations receiving either EPA-FFA/placebo or EPA-TG/placebo were combined.
Mechanistic considerations
We have not yet addressed the mechanistic basis for differential chemopreventive activity of EPA along the colorectum. Conventional dysplastic ‘sporadic’ colorectal adenomas are found along the length of the colon and can exhibit molecular features compatible with CIN, MSI and CIMP+ pathways, but with proportionally more CIN+ lesions in the distal colon.4,23 Therefore, differential activity of EPA against the distinct molecular CRC subtypes should be investigated, particularly in view of the absence of any efficacy signal against serrated lesions that are predominantly CIMP+ and often exhibit features of MSI.4,21,23
We cannot rule out that differences in mucosal EPA levels explain colorectal site selectivity of the modest chemopreventive activity of EPA. There are no human data describing ω-3 PUFA levels in different regions of the colorectum. However, rodent data suggest that there is no gradient in mucosal EPA exposure along the colon.127 One clinical study has measured ω-3 PUFA levels in the sigmoid colon (but not the rectum).128 However, data from this Japanese study cannot be compared with the rectal mucosal PUFA levels measured in the seAFOod trial because of profound differences in baseline characteristics of the two study populations, including dietary ω-3 PUFA intake.128
We have recently reported that mixed ω-3 PUFA supplementation is associated with an increase in short-chain fatty acid (SCFA)-producing bacterial genera such as Lactobacillus and Bifidobacteria in faecal samples.129 Although SCFA levels are believed to be higher in the caecum and ascending colon than in the distal colon,130 the SCFA receptor free fatty acid receptor (FFAR) 2, also known as GPR43, is expressed at higher levels in the left colon.131 Therefore, a valid hypothesis is that differential antineoplastic activity of EPA may relate to increased antineoplastic SCFA-FFAR2 signalling in the distal versus proximal colon.132
Both EPA and aspirin inhibit the COX enzymes COX-1 and COX-2.57 Independent reports have described how expression of COX-2, which is believed to play a critical role in intestinal tumorigenesis,54 is higher in distal colonic neoplasms than in the proximal colonic lesions.133,134 Differential expression and activity of COX-1 in colorectal adenomas in different parts of the colorectum has not been studied. Differences in gene expression in non-neoplastic mucosa and baseline colorectal adenoma characteristics that may predict response to EPA and/or aspirin deserve further investigation.
The mechanism of the CRC-preventative activity of aspirin remains unclear despite intense scrutiny over several years.44,45 Postulated mechanisms of action include COX-dependent and COX-independent activity, either directly on colorectal epithelial cells or indirectly via the antiplatelet activity of aspirin.45 More recently, subversion of the host anti-tumour immune response by COX-dependent PGE2 has been reported,135 which is abrogated by aspirin, thus driving more effective host anti-tumour immunosurveillance. This mechanism of action may explain the preferential antineoplastic activity of aspirin against proximal colonic neoplasms, which are more likely to exhibit a higher neo-antigenic load (and hence potential to stimulate a host anti-tumour immune response) as a result of defective DNA MMR in MSI tumours.136
Combination colorectal cancer chemoprevention
The individual effects of EPA and aspirin on colorectal adenoma number beg the question of the antineoplastic activity of combined treatment, which, given the single agent effects apparent on the ‘at-the-margins’ analysis, one might expect to be apparent for total colorectal adenoma number. The trial was powered to be able to detect only a major interaction between EPA and aspirin with an ‘inside-the-table’ analysis of this 2 × 2 factorial RCT. Comparison between the four treatment groups highlighted that total and left-sided colorectal adenoma multiplicity was lower in the combined EPA and aspirin treatment group, consistent with efficacy of both agents on left-sided, conventional colorectal adenomas. A key objective of future work, in order to translate the seAFOod trial findings into clinical application, will be to apply precision/personalised medicine principles to pose the question regarding which individuals might gain most from chemoprevention with one or both agents, based on baseline colorectal adenoma characteristics and/or other mucosal biomarkers.
Use of the NHS Bowel Cancer Screening Programme for a polyp-prevention trial
The rationale for the study of a ‘high-risk’ population undergoing 1-year surveillance colonoscopy within the English BCSP was based on data from previous aspirin polyp-prevention trials that suggested that colorectal adenoma risk reduction associated with aspirin use was already evident (and exceeded the risk reduction at 3–4 years) at 1 year, combined with the high ADRa.49 The relatively short follow-up duration raises the question of how far colorectal adenoma outcomes at 12 months simply reflect missed colorectal neoplasia from the screening examination. However, BCSP data98,100 and other reports137,138 have confirmed that the colorectal adenoma yield at 12 months from baseline assessment is significantly higher than the yield from a procedure within 3 months of the index assessment, implying that de novo colorectal adenoma growth contributes significantly to colorectal adenoma ‘recurrence’ rather than solely ‘missed lesions’. In reality, colorectal adenomas detected at a 12-month endoscopic procedure will almost certainly represent a mixed population of ‘new’ and ‘missed’ lesions, even after the highest-quality colonoscopic examination, but this does not negate the role of the colorectal adenoma as a biomarker of chemoprevention efficacy (related to regression of existing colorectal adenomas, as well as inhibition of new growth), as accepted in proof-of-concept FAP RCTs.37
The low yield of advanced colorectal adenomas in the seAFOod trial is probably explained by the short surveillance interval and high-quality colonoscopic colorectal adenoma clearance achieved in the BCSP, but also the rather stringent ‘advanced’ lesion definition, which required complete (100%) ‘villousness’ of conventional dysplastic adenomas, unlike previous observational and intervention studies that have included lesions with > 25% ‘villousness’ in the definition of ‘advanced’ colorectal adenoma.32,139
Using the English BCSP as a vehicle for the seAFOod trial provided several advantageous features to the design and performance of the trial, including strict, protocol-driven colonoscopy screening and surveillance procedures, uniform colonoscopy, and histopathology performance and reporting.95–97 In particular, strict timing of the 12-month surveillance colonoscopy in the seAFOod trial was a major advantage over previous ‘sporadic’ polyp-prevention trials undertaken in non-screening settings49,125,141 and was essential for interpretable PUFA biomarker data at 12 months. The trial population had a strong male predominance, mirroring BCSP practice,95 which limits the generalisability of the data to women. The high prevalence of overweight and obesity in the trial population probably indicates the major contribution of excess body weight to early-stage colorectal carcinogenesis,140 and is highly relevant to general CRC chemoprevention in non-screening populations given the high prevalence of overweight and obesity in many areas of the world.142
A large number (nearly 4000) of individuals stratified as ‘high risk’ after BCSP screening colonoscopy were identified and screened by participating trial sites, as expected from BCSP data pertaining to the prevalent screening round, which suggested that ‘high-risk’ features were detected in 9.8–10.3% of screening colonoscopies (Public Health England, 2009, personal communication).95 However, since the seAFOod trial opened for recruitment, subsequent incident screening rounds have detected less colorectal neoplasia and more recent BCSP data demonstrate that ‘high-risk’ features make up only 8.3% of screening colonoscopy outcomes (Public Health England, 2009, personal communication). The reduction in incidence of ‘high-risk’ colorectal neoplasia probably contributed to the slower than expected recruitment during the trial.
Only 18% of ‘high-risk’ individuals were randomised. This is a higher screen success rate than in several previous polyp-prevention trials that reported a trial screening success rate,141,143 but lower than we anticipated. Exclusion criteria included concomitant use of anticoagulants or antiplatelet agents, including existing aspirin use (19%). In addition, a higher than expected proportion of patients (31%) declined to be screened for and participate in the trial, despite the fact that the trial design did not entail an additional colonoscopy and that venepuncture for blood sampling was optional. Trial acceptability was lower than expected from a cohort of patients who had already engaged fully with a multistage screening process and who had been informed that they had ‘high-risk’ colorectal neoplasia necessitating surveillance colonoscopy in 12 months’ time. Formal data pertaining to reasons for declining participation were not collected on the basis that any individual did not need to provide a reason for his or her decision. However, an informal review showed that reasons for declining participation at screening included concern about risk associated with the IMP (particularly aspirin) and unwillingness to start medication when previously not taking any regular drugs, but also, conversely, unwillingness to add further medication to an existing drug regimen.
The screen failure rate varied significantly between the BCSP sites despite the strict, uniform inclusion criteria based on BCSP practice and an identical BCSP screen pathway used at all BCSP sites. Several patient (e.g. differential prevalence of comorbidities requiring antiplatelet/anticoagulant use in different regions of England) and research staff factors (e.g. research time available to PI and SSPrs, as well as CRN-funded RN support) contributed to marked differences in recruitment efficiency. In addition, BCSP screening colonoscopy activity varied widely across individual sites, thereby limiting the identification of ‘high-risk’ individuals at some sites. Recognition of the most-suitable BCSP sites for any similar polyp-prevention trial will be key to improving recruitment efficiency, particularly when ‘high-risk’ individuals remain a relatively fixed proportion (7–9%) of guaiac FOBt-positive patients. A general theme (but with notable exceptions) was that BCSP sites that utilised CRN-funded research staff best had a higher recruitment rate. This may reflect insufficient time that many BCSP staff have available for research and/or some SSPrs having limited understanding of clinical trials research and a protectionist approach towards patients; these need to be addressed prior to a future polyp-prevention trial set in the BCSP.
The switch from guaiac FOBt to FIT in the UK BCSP programmes may alter colorectal neoplasia incidence rates relevant to future trial recruitment projections.
Safety and tolerability of eicosapentaenoic acid and aspirin
Overall, safety and tolerability of both agents was excellent, with no excess of clinically significant bleeding events, even in the combined treatment group. However, we did mandate stopping IMP before and after any planned invasive procedures (including polypectomy at repeat endoscopy) in the unlikely event that the combined antiplatelet activity of EPA and aspirin increased bleeding risk. Excellent tolerability of EPA and aspirin contributed to high compliance levels and trial retention, with limited drop-out.
Eicosapentaenoic acid was associated with an excess of mild to moderate GI AEs, including diarrhoea. Although ω-3 PUFA products have an excellent safety profile and daily doses of up to 4 g are ‘generally recognized as safe’ (GRAS) by the US Food and Drug Administration (FDA), mild to moderate GI AEs are recognised, commonly diarrhoea, dyspepsia, eructation (burping) and an unpleasant ‘fishy’ taste.144 These GI symptoms often occur in the general population unrelated to dietary or ‘nutraceutical’ ω-3 PUFA intake.113 Therefore, we confirmed that the treatment groups were well matched for prevalent GI symptoms at baseline. Dose-related ω-3 PUFA-related GI symptoms can occur.65,144,145 However, an open clinical question has been whether or not prolonged dosing with the different forms of EPA and mixed ω-3 PUFAs (FFA vs. TG vs. EE) available in ‘nutraceutical’ preparations is associated with differential GI tolerability.65 The seAFOod trial data do not indicate any major difference in GI tolerability between the same FFA-equivalent dose of the EPA-FFA and EPA-TG formulations in the context of a 12-month intervention trial.
Mild to moderate GI AEs, particularly diarrhoea and abdominal pain, were more common in those allocated to EPA alone, as opposed to EPA plus aspirin. A valid hypothesis is that aspirin co-therapy is causally protective for ω-3 PUFA intolerance. This requires further investigation in the context of subsequent studies of combined antineoplastic activity of both agents.
Five episodes of AF were reported in three participants, all of whom were allocated active EPA alone. Several RCTs have investigated ω-3 PUFA supplementation for primary and secondary prevention of AF that are summarised in recent American Heart Association guidelines on ω-3 PUFA use and prevention of CVD.146 Overall, there is no supportive evidence for use of ω-3 PUFAs for AF prevention. There has been no signal in any previous RCT of increased AF risk in ω-3 PUFA users.146
An important component of the trial was dietary ω-3 PUFA analysis to exclude significant change in dietary ω-3 PUFA intake prompted by participant-facing trial information, prior to consent, that suggested possible anti-CRC activity of EPA. Using the established short-form EPIC FFQ,108 we confirmed that there was no evidence of increased fish intake during the intervention phase and that the four treatment groups were well matched on dietary fish intake at baseline and at the end of trial participation.
Biomarkers of omega-3 polyunsaturated fatty acid tissue exposure
The seAFOod trial biobank obtained excellent coverage considering the geographical spread of a large number of trial sites. There was widespread engagement from SSPrs and RNs at sites, with careful sample handling and storage at sites prior to temperature-controlled transport to the biobank in Bradford. The trial protocol outlines the lipid mediator analyses that were originally proposed.100 At the start of the trial, measurement of RBC PUFA levels was planned in only a small subset of participants. However, the capsule IMP switch necessitated the measurement of PUFA levels in all RBC and rectal mucosal samples to provide the strongest possible evidence that the trial cohorts exposed to EPA-FFA and EPA-TG gained equivalent exposure to EPA and could be combined for primary and secondary outcome analysis. Therefore, many planned measurements were not carried out, but these should be performed in the future if further funding can be obtained for biomarker analysis.
Baseline RBC PUFA values in the seAFOod trial cohort were comparable with previous cross-sectional UK studies and data from other Western countries with relatively low population ω-3 PUFA levels and dietary fish intake.147 Despite self-reported compliance being excellent by capsule counting, the increase in RBC EPA levels observed in EPA users was highly variable, a phenomenon that has been seen in several previous ω-3 PUFA studies, in which individual RBC fatty acid profiles have been measured.88,129,148 The explanation for large variations in RBC membrane EPA levels, despite uniform dosing, remains unclear and requires further investigation in the large trial cohort. There was no evidence that COX inhibition by concurrent aspirin use is associated with higher RBC membrane or rectal mucosal EPA content.
To our knowledge, this is the first large-scale RCT evidence that oral administration of EPA results in incorporation of EPA in the rectum. There was a medium-strength correlation between the RBC % EPA value at 12 months and the rectal mucosal % EPA content. The RBC membrane EPA and DHA content is considered to be the best proxy biomarker of tissue ω-3 PUFA exposure.84,148 Further exploratory analysis of the seAFOod trial data is required to explore the predictive value of the rectal mucosal EPA level for individual EPA efficacy against left-sided colorectal neoplasia. Integration of dietary ω-3 PUFA intake, baseline PUFA levels and ‘nutraceutical’ EPA dosing is needed in a treatment-independent analysis of the relationship between EPA level and colorectal adenoma recurrence risk during surveillance. A similar analysis performed in the context of the WELCOME trial149 of ω-3 PUFA use in patients with non-alcoholic fatty liver disease provided new insights into the role of DHA in treatment of that condition.
There was no evidence of significant conversion of EPA to DHA in either RBC membranes or rectal mucosa, consistent with previous human data from EPA intervention trials.67,87 Therefore, the trial data do not support the concept of EPA as the ‘universal ω-3 PUFA donor’, which can substitute for mixed EPA-DHA preparations. The question of whether or not equivalent doses of EPA alone, DHA alone or a EPA-DHA mix have different antineoplastic activity in the intestine has been tested only in rodents, in which equivalent anti-CRC effects have been observed.61 Equivalent bioactivity of EPA and DHA in a CRC chemoprevention context cannot be assumed from the seAFOod trial; this will be a critical public-facing message from the trial to minimise incorrect assumptions about, and use of, mixed ω-3 PUFA preparations in the context of CRC chemoprevention. The majority of prescription and health supplement ω-3 PUFAs are a mix of EPA and DHA with a smaller proportion of other PUFAs.150 However, separate EPA formulations are licensed for treatment of severe hypertriglyceridaemia and are marketed as health supplements.144,151 The low cost and excellent safety/tolerability profile of EPA, combined with existing health claims regarding ω-3 PUFAs, means that, even with modest CRC chemopreventative activity, it may be adopted widely in a cancer-prevention setting by the public.
‘Contamination’ of the treatment groups by participant use of OTC formulations is an ever-present risk in placebo-controlled trials of agents that are available for purchase by the public. The seAFOod trial fatty acid analysis allows us to speculate about possible placebo contamination by ‘own’ ω-3 PUFA use by identifying those individuals who were randomised to placebo EPA who had a rise in RBC EPA level. Further analysis of concomitant changes in DHA levels is now required. Individual RBC ω-3 PUFA profiles suggest that placebo contamination, if it did occur, was very rare and did not interfere with the findings pertaining to EPA use.
Aspirin use entails a small, but significant, risk of serious harm from GI and intracranial bleeding.152 This has hampered translation of the significant body of evidence supporting the CRC chemopreventative properties of aspirin into routine clinical practice. The seAFOod trial data add to this body of evidence and emphasise the chemopreventative activity of aspirin against proximal and serrated neoplasms, which are widely considered to be more ‘missable’ at colonoscopy and contribute significantly to the burden of interval CRC.
Concluding remarks
In summary, the seAFOod polyp-prevention trial has demonstrated that the ω-3 PUFA EPA (2 g of FFA daily) and aspirin (300 mg daily) did not reduce colorectal adenoma risk (measured by the ADRa) at a 1-year surveillance colonoscopy in ‘high-risk’ individuals with colorectal neoplasia in the English BCSP. However, chemopreventative efficacy of both agents was observed, as measured by the reduction in colorectal adenoma number per patient at surveillance colonoscopy. The colorectal adenoma subtype- and location-dependent specificity of EPA and aspirin are consistent with previous observations. Existing data on CRC risk reduction by aspirin suggest that the colorectal adenoma risk reduction observed for both agents is likely to translate into a clinically meaningful decrease in long-term CRC risk. Both agents were safe and well tolerated in patients aged 55–75 years with ‘high-risk’ colorectal neoplasia, although EPA therapy alone was associated with a slight excess of mild to moderate GI AEs. There was no difference in tolerability of the two EPA formulations used in the trial.
Results in context
Despite several decades of research into the efficacy of several drug (e.g. aspirin, selective COX-2 inhibitors) and nutrient (e.g. folic acid, vitamin D, ω-3 PUFAs) interventions for chemoprevention of CRC, meaningful translation into clinical practice has yet to occur.153 The weight of the evidence that long-term (> 5 years) use of aspirin is associated with a ≈25% decrease in the risk of CRC incidence and mortality, particularly marked for proximal CRC (the predominant interval CRC type during colonoscopy screening and surveillance, as well as the CRC type not afforded protection by FS screening),46–48 makes the lack of ‘effectiveness’ trials driving clinical use of this agent even more surprising.
One reason for this lack has been the dependency on the polyp-prevention trial, using the colorectal adenoma as an accepted surrogate biomarker of CRC risk,49 in the absence of direct data regarding how a decrease in either colorectal adenoma incidence (ADRa) or number (MAP) translate into subsequent CRC risk reduction.
Historically, the ADRa has been favoured as the colorectal adenoma end point of choice in polyp-prevention trials. The ADRa may be considered the most relevant end point related to the use of a CRC chemoprevention agent to reduce the frequency of, or need for, endoscopic surveillance of patients with previous colorectal neoplasia, stratification for which is based largely on colorectal adenoma incidence.30–33 Whether or not the use of a CRC chemoprevention agent will lead to a significant reduction in the amount of surveillance colonoscopy performed is a question that can be addressed only by ‘effectiveness’ studies in which long-term use of a concurrent chemoprevention agent and surveillance colonoscopy requirement are observed using a ‘big-data’ linkage approach or perhaps a ‘point-of-care’ randomised trial.
Reduction in colorectal adenoma multiplicity (MAP) arguably has more biological meaning as a read-out of reduced colorectal neoplastic risk associated with a potential CRC chemoprevention agent, hence its widespread use in FAP trials and pre-clinical studies.37,61 This read-out is more suited to consideration of a chemoprevention agent for long-term CRC risk reduction, as opposed to more efficient use of colonoscopic surveillance for benign colorectal neoplasia. Consistent with use of colorectal adenoma number as a biomarker of future CRC risk in chemoprevention trials, observational data suggest that colorectal adenoma multiplicity at baseline endoscopy is associated with increased CRC incidence and mortality compared with the general population.40–42 Moreover, a pooled analysis of eight prospective colonoscopy surveillance studies reported that initial colorectal adenoma number at colonoscopy predicts subsequent advanced colorectal neoplasia (advanced colorectal adenoma and CRC combined), in a stepwise manner with increasing baseline colorectal adenoma number.139
The seAFOod trial data on colorectal adenoma number, as well as existing MAP data for aspirin,115,117,118 should focus more attention on the further validation of colorectal adenoma number as a colonoscopic end point, particularly given that colonoscopy performance and reporting has improved significantly, leading to better routine data quality on colorectal adenoma characteristics.
Strengths and limitations
Trial design
A key strength of the seAFOod trial was integration in the English BCSP, which continues colonoscopic surveillance in the BCSP, unlike in Scotland, where ongoing surveillance responsibility is passed to general endoscopy services. This provided several advantages to the trial, including:
- consistent high-quality colonoscopy performance and reporting above that available in general colonoscopy populations
- accurate screening colonoscopy data from the BCSP, with which to predict screening colonoscopy outcomes for recruitment projections
- uniform, protocol-driven care within the BCSP, particularly the small window for the 12-month surveillance colonoscopy in the BCSP, despite high procedural pressure in NHS endoscopy units
- use of a ‘high-risk’ surveillance cohort and 1-year colonoscopy outcomes, which minimised the intervention period compared with previous polyp-prevention trials,34–36,141,143 with consequent time and cost savings.
Another strength was the comprehensive seAFOod trial biobank with careful QA control. The biobank will be critical for further biomarker-driven, stratified analysis of the trial data. Availability of dietary and tissue biomarker data for the ω-3 PUFA intervention allowed interpretation of the EPA effect on colorectal outcomes, analysis of which could be confounded by dietary and OTC ω-3 PUFA exposure.
The small number of advanced colorectal adenomas found during the seAFOod trial is a limitation and probably relates to the short surveillance interval, as well as the excellent quality of the clearance screening colonoscopy. Previous polyp-prevention trials have reported an advanced ADRa of ≈10% in the placebo group of patient populations, approximating to combined ‘high-risk’ and ‘intermediate-risk’ BCSP patients with more heterogeneous follow-up (1–5 years).34,36,141,143 Detailed data on advanced ADRa and MAP at surveillance colonoscopy were not available from the BCSP when the seAFOod trial was designed but are now available from the seAFOod trial and other reports98 in order to determine the feasibility of using advanced neoplasia as a primary or main secondary outcome.
As the first chemoprevention CTIMP in the BCSP, the outcome and performance data from the seAFOod trial will be invaluable for any future polyp-prevention trial (see Chapter 5, Recommendations for research). The seAFOod trial has established the BCSP as a realistic vehicle for interventional research during the endoscopy phase of the patient pathway.
Trial performance
Failure to reach the recruitment target
Although the BCSP sites provided a large number of potential ‘high-risk’ trial participants at a constant rate, consistent with well-characterised colonoscopy outcomes after guaiac FOBt screening,95 recruitment of ‘high-risk’ individuals after screening colonoscopy (18%) was significantly lower than projected (60%) owing to several factors, which included:
- Exclusion because of increasing use of antiplatelet agents, such as clopidogrel, although concurrent aspirin use (19%) occurred almost as predicted from pilot data (20%).
- Exclusion because of the need for additional endoscopic assessment in the BCSP, which varied significantly across BCSP sites/different BCSP colonoscopists and had not been reported prior to trial design.
- Unwillingness of patients to take part in the trial. Acceptability of the trial was not tested in a pilot study, nor subjected to broader PPI assessment prior to opening, except for the lay input provided by the PPI representative in the application team (a BCSP colonoscopy patient) and the review inherent in both Research Ethics Service and BCSP Research Advisory Committee assessments. Assumptions about a high acceptability level for the trial were based on the necessary prior engagement of ‘high-risk’ individuals in the BCSP, the recent provision of a diagnosis that stratified potential participants who were at a higher risk of future CRC, and a short intervention period prior to a subsequent routine colonoscopy with no change to normal BCSP practice. Anecdotally, a major reason for unwillingness to join this CTIMP after the screening colonoscopy was patient relief at not having CRC (all patients are informed that there is a 1 in 10 chance of CRC at colonoscopy, prior to the procedure) and a strong desire to ‘forget about my colon for a while’, at least until surveillance colonoscopy.
- Limited research experience of some BCSP staff, coupled with reluctance to discuss research participation during routine clinical care episodes, may have contributed to a high screen failure rate. There was a lengthy ‘lag time’ at many sites between local R&D approval and FPFV, which contributed significantly to reduced overall recruitment.
Although recruitment reached only 83% (n = 709) of the target (853 randomised participants), we do not believe that the smaller sample size affected the primary results of the trial because the 95% CI for the risk difference in ADRa for both EPA and aspirin did not include –10%. Therefore, assuming no change in the point estimate, a larger sample size would tend only to decrease the width of the CI, thereby moving the lower 95% CI limit further from the hypothesised absolute risk difference of –10%.
Capsule investigational medicinal product switch
The need to switch capsule IMP during the trial was unforeseen, beyond the control of the TMG and very disruptive to recruitment for at least 12 months, especially for BCSP sites that exhausted IMP stock and had to suspend recruitment for several months.
The potential consequences of a switch in one of the IMPs for analysis of the trial results was given very careful consideration. After extensive discussions between the TMG, oversight committees and the funder (which also sought anonymous peer review), it was concluded that the proposed replacement active IMP (90% EPA-TG) was sufficiently similar to the original active IMP (EPA-FFA) in terms of content, FFA-equivalent dose and probable bioavailability to continue the trial, on the basis that colorectal adenoma outcome data from those who received EPA-FFA or EPA-TG could almost certainly be combined. This potential threat to trial integrity has now become a unique aspect of the trial whereby two different (2-g FFA dose equivalent) EPA formulations have been compared in a RCT, demonstrating no clear difference in either tolerability or EPA bioavailability (as measured by RBC and rectal mucosal EPA content) during dosing over several months.