Study Population
We drew our study population from GIDER, a colonoscopy-based longitudinal cohort at the Massachusetts General Hospital (MGH). Participants older than 18 undergoing a screening or surveillance colonoscopy at MGH were invited to complete a comprehensive dietary, lifestyle, and medical history questionnaire before their colonoscopy. Patients with a history of gastrointestinal cancer, hereditary non-polyposis colorectal cancer, familial adenomatous polyposis, inflammatory bowel disease, or known bleeding disorders were excluded from enrollment in the cohort. In addition, patients who had used oral or intravenous antibiotics in the 2 months prior to their procedure were excluded. The study was approved by the Institutional Review Board at the Partners Human Research Committee. All recruitments were carried out in accordance with the Institutional Review Board regulations, with participants signing an informed consent form prior to data and sample collection.
Evaluation of Dietary Intake
We collected data on participants’ dietary intake through the administered baseline questionnaire. Specifically, we used the Semiquantitative Food Frequency Questionnaire (SFFQ) to define categories of dietary intake and aggregated individual foods into broader food groups (Supplementary Table 1). Prior to the colonoscopy, participants were asked to report their weekly frequency of consumption of red meat, processed meats, white meat, shellfish, fish, dairy, starches, fruit, and vegetables on an eight-category scale (Never, 1 per week, 2-4 per week, 5-6 per week, 1 per day, 2-3 per day, 4-5 per day, 6+ per day). We focused our analysis on the consumption of fruit (apples, raisins, bananas, oranges, strawberries, blueberries, etc.), vegetables (salad, tomatoes, onions, greens, carrots, peppers, etc.), red meat (beef, hamburger, pork, lamb), and processed meats (sandwich meat, ham, salami, bologna, sausage, bacon, “hotdogs”). We examined the frequency of participants’ responses in each of the eight categories for these foods and combined the individual categories into tertiles of weekly consumption. We grouped fruit and vegetables intake into less than five times per week, five to seven times per week, and greater than once per day and further categorized red and processed meat intake into never or not in the last week, once per week, and greater than or equal to 2 times per week.
In addition to our brief dietary questionnaire, a subset of participants also completed the previously validated SFFQ12. In a validation study of 97 participants who completed both the short dietary questionnaire and SFFQ, the intraclass correlation coefficients for weekly consumption of fruit and vegetables were 0.55 and 0.52, respectively.
Other variables
Participants’ body mass index (BMI) was calculated from their reported height and weight at baseline. We assessed physical activity by asking participants about the average time spent per week in the last year on various recreational activities using a previously validated physical activity questionnaire13. We then assigned a metabolic equivalent task (MET) to each activity based on previously established guidelines14 and determined the average MET-hour/week for all activities. Smoking history was defined as current smokers or those who have smoked more than 100 cigarettes in the past and regular NSAID use was defined as greater than 2 tablets per week consistent with prior analyses15,16. Bowel movement frequency was evaluated with the question, “How frequently do you have a bowel movement?” Participants were also asked about their dietary preferences with respect to red meat and were grouped into three patterns: Western standard diet (≥3 times/month), low-red-meat diet (<3 times/month), and no-red-meat-diet.
Outcome of interest
Our primary outcome of interest was the detection of colonic diverticulosis during the colonoscopy. A study coordinator obtained information on the presence and location of diverticula from the endoscopist at the end of each procedure.
Statistical Analyses
Participants with missing data on dietary intake, BMI, and physical activity were excluded from all analyses (n=161). Specifically, 549 participants had no missing information on dietary or lifestyle data and were therefore eligible for our analyses. There were no significant differences in age (61.3 vs 61.4 years, P=0.954), sex (46.6% vs 47.2% male, P=0.898), or prevalence of diverticulosis (44.6% vs 47.2%, P=0.563) between eligible participants and those excluded due to missing data. We used Poisson regression modeling with a robust error variance to calculate the prevalence ratios (PRs) and 95% confidence intervals (CIs) while adjusting for age, sex, smoking, BMI, physical activity, dietary pattern, regular NSAID use, and number of bowel movements per day. We elected to use Poisson regression instead of logistic regression because the outcome of interest, diverticulosis, was common17. Additionally, we did attempt to use a log-binomial model, which uses a binomial distribution and have been shown to be slightly less biased than Poisson regression18, but were unable to due to issues with model convergence. In addition, we evaluated for effect modification by age, BMI, smoking history, and red meat intake on associations between fruit and vegetables consumption and prevalence of diverticulosis by including cross-product terms of these potential risk factors and vegetables and fruit intakes in the multivariable models. All P-values were 2-sided and P < 0.05 were considered statistically significant. We used R version 3.2.0 for all analyses.