Critical appraisal:Burn J, Bishop DT, Mecklin JP, Macrae F, Möslein G, Olschwang S, et al 2008 1

This appraisal has been completed.

Risk of bias assessment: Randomised Controlled Trial (Cochrane risk of bias tool)

Random sequence generation

Describe the method used to generate the allocation sequence in sufficient detail to allow an assessment of whether it should produce comparable groups.

Participants were randomly assigned separately for the two interventions (either assignment to aspirin or placebo or assignment to resistant starch or placebo), and half the participants were randomly assigned to receive either the active agent or placebo for each intervention. This randomization should have resulted in four groups of equal size for assignments to aspirin plus placebo, aspirin plus starch, starch plus placebo, and placebo plus placebo. For some persons, however, randomization to one intervention was precluded by a factor such as sensitivity to aspirin, and in such cases, randomization to the other intervention was permitted; this resulted in four additional (smaller) groups.

Randomization was computer-generated. The study centers were categorized into six geographic groups — Americas, southern Europe (the Iberian Peninsula and Italy), northern Europe, South Africa, the United Kingdom, and the Pacific Rim (Australia and Hong Kong). Block randomization (in blocks of 16) was performed separately for each group of centers to ensure balance across the four main study groups.

What was the risk of bias from the random sequence generation?

Low

Allocation concealment

Describe the method used to conceal the allocation sequence in sufficient detail to determine whether intervention allocations could have been foreseen in advance of or during, enrolment.

The investigators and patients were unaware of the study-group assignments. Dispensing and related records were managed by the Newcastle Hospitals NHS Foundation Trust Pharmacy. After colonoscopic examination at study entry, packs containing the study drugs were sent directly to the participants or local clinicians and thereafter were sent every 6 months, when compliance data were collected.

What was the risk of bias from the allocation concealment?

Low

Blinding

Describe all measures used, if any, to blind outcome assessors from knowledge of which intervention a participant received. Provide any information relating to whether the intended blinding was effective.

The investigators and patients were unaware

of the study-group assignments. Dispensing and
related records were managed by the Newcastle
Hospitals NHS Foundation Trust Pharmacy.

What was the risk of bias from the blinding of participants and personnel and outcome assessors?

Low

Incomplete outcome data

Describe the completeness of outcome data for each main outcome, including attrition and exlusions from the analysis. State whether attrition and exclusions were reported, the numbers in each intervention group (compared with total randomized participants), reasons for attrition/exclusions where reported, and any re-inclusions in analyses performed by the review authors.

We recruited 1009 eligible patients. The analysis excluded 263 participants (26%); 72 withdrew before the colonoscopic examination at study entry, and 191 withdrew before the follow-up colonoscopic examination. The remaining 746 participants were included in the analysis. There were no notable differences between the patients who were recruited and those who completed the study.

What was the risk of bias from incomplete outcome data?

Low

Selective outcome reporting

State how the possibility of selective outcome reporting was examined by the review authors and what was found.

Participants who underwent a follow-up colonoscopic examination were included in the analyses of treatment effects. The primary outcome was detection of at least one adenoma or colorectal carcinoma at follow-up; since intervals between colonoscopic examinations and the duration of participation in the study varied, the analysis considered time in the study in its entirety. Secondary outcomes were the detection of an adenoma only, colorectal cancer only, adenoma and colorectal cancer, and advanced adenoma or colorectal cancer. A neoplasm was classified as an advanced adenoma on the basis of one or more of the following features: a diameter of 1 cm or more, a villous or tubulovillous component, or high-grade dysplasia. Other secondary outcomes

were the neoplastic burden, defined as the sum of the maximum diameters of neoplasms detected by means of endoscopic examination during the study, and other cancers associated with the Lynch syndrome. The primary analysis compared the occurrence of each outcome involving a neoplasm according to whether the subject received resistant starch or aspirin. Secondary analyses were adjusted for age and sex. Adjustment for the presence or absence of bowel neoplasms before participation in the study, the number of colonoscopic examinations, or both with the use of generalized linear models gave results that were equivalent to those seen after adjustment for age and sex only; these results are not reported. Time-to-event analysis was used in the secondary analysis, with the number of months to detection of any neoplasia or an advanced adenoma or colorectal cancer as end points.

What was the risk of bias from selective outcome reporting? Assessments should be made for each main outcome (or class of outcomes).

Low