An evidence table (accessible at http://caonline.amcancersoc.org/content/vol56/issue3/) was organized to include the elements of study design. Ideally, the best study design would fulfill the following criteria: (1) be a randomized controlled trial or an observational cohort study of patients with adenoma(s) at baseline that were cleared by colonoscopy, after excluding people at high risk (such as familial syndromes); (2) consider all the candidate risk factors; (3) have sufficient follow-up time for adenomas to develop, with few dropouts; (4) have planned colonoscopic assessment for recurrence in all patients in the cohort; (5) have enough outcome events for reasonable statistical precision and sufficient statistical power to detect associations between baseline characteristics and adenoma outcomes; and (6) present the analyses that include adjustment for multiple risk factors and consider what the independent effects are.
The evidence table includes classification of the type of design (randomized controlled trials [RCTs] or observational cohort studies), the number at risk, the follow-up intervals recommended, and the time followed. The guideline developers also list the variables considered as risk factors and the effect of these factors on incidence of subsequent adenomas or on advanced neoplasia. The multivariate estimate of the relative risk is presented whenever available. The definition of an advanced neoplasia is given for each study and varies considerably by study. Summary comments on each study are also included.
Review of the evidence was confounded by variations in definitions, design of studies, timing and multiplicity of surveillance intervals, and quality of baseline colonoscopy. Due to these variations, the review of the literature cited was descriptive rather than a single summary value of risk (i.e., meta-analysis) for all studies. The literature cited is grouped by type of study design: (1) RCTs, where the surveillance interval is set and maintained as much as possible though eligibility requirements may vary; or (2) observational cohort studies, which are primarily registry studies with more passive recruitment for surveillance. The RCTs provide stronger evidence for the timing of follow-up examinations because those who received surveillance colonoscopy were not a special subset of all enrolled. As noted above, relative risks (RR) or odds ratios (OR) from multivariate analysis were presented in the evidence table whenever available. For two studies, the measure of risk was the standardized incidence ratio (SIR) with adjustment for age and sex rather than a relative risk. In one study, the hazard ratio (HR) is given as the measure of the effect. A descriptive graphical presentation was given with point estimates and confidence intervals for the relative risk for adenomas and advanced neoplasia by baseline adenoma characteristics of multiplicity, size, histology, high-grade dysplasia, and location. These descriptive plots (Figure 1 in the original guideline document) of the measure of the effect for various risk factors provide a summary of the number of studies reporting a measure of effect for a given risk factor and the consistency and magnitude of this factor on adenoma and advanced neoplasia recurrence. The review of evidence assessed the risk factors for adenomas as well as for advanced adenomas, but the discussion concentrated on the factors affecting advanced adenomas. The definition of advanced adenoma differs from study to study. The most encompassing definition included any adenoma >1.0 cm, any villous component (i.e., nontubular), or high-grade dysplasia, or invasive cancer.
Given the concern in detecting colorectal cancers at surveillance, the number of colorectal cancers detected by time under surveillance is cited whenever these data are included in the published study. Special characteristics of the study population and selection for the cohort were also noted in the evidence tables.