This is a prospective cohort study based on two major cohorts: the Swedish Mammography Cohort (SMC), and the Cohort of Swedish Men (COSM).
The Swedish Mammography Cohort (SMC)
SMC is a major prospective cohort consisting of 66,651 women in central Sweden born between 1914 and 1948. The cohort was established in 1987–1990. Women were invited by mail to participate in a mammography screening programme. The invitation included a questionnaire regarding diet and alcohol intake, parity, weight, height, education level and marital status. After excluding women with a previous diagnosis of cancer, the SMC included 61,433 women. In 1997, a second questionnaire was sent to those women still alive and living within the study area. Information about diet and alcohol intake, physical activity, medical history, height, weight, education level and lifestyle factors such as cigarette smoking history and use of some medications and dietary supplements was updated and completed. In the present study, the 1997 questionnaire was used as baseline (Table 1). In all, 39,984 of 56,030 women (71.4%) living in the study area responded to the follow-up questionnaire in 1997; 219 women were too sick to fill in the questionnaire and 548 declined to answer (Fig. 1) . Women with IBD, cancer, previously diagnosed DD (N = 478) or error in the registration were excluded, leaving a final cohort of 36,110 women.
The Cohort of Swedish Men (COSM)
COSM is a prospective cohort consisting of 48,850 Swedish men born in 1918–1952 and was established in 1997–1998. Men living in central Sweden answered a questionnaire regarding diet, smoking, alcohol intake, physical activity, dietary supplements, some medications, height, weight and education level. The response rate was 49%.
After excluding those with an incorrect or missing national registration number, cancer diagnosis (except non-melanoma skin cancer) before baseline or other missing data, the final cohort consisted of 45,906 men. Men with IBD, cancer or previously diagnosed DD (N = 334) and error in the registration were excluded, leaving a final cohort of 44,723 men.
The baseline characteristics of the two cohorts are presented in Tables 1 and 2.
Assessment of dietary fibre intake
Fibre consumption was assessed with a food frequency questionnaire. Participants in the two cohorts indicated their average consumption of 96 foods and beverages over the previous year. Participants could choose from eight predefined frequency categories ranging from never to three or more times per day. The amount of intake of fibres from cereals and fruits and vegetables, respectively, was estimated in grams per day for each patient. The patients were divided into four quartiles (Q1–Q4). Q1 contained patients who ate the lowest amount of fibres and Q4 contained patients who ate a larger amount of fibres. The amount of every quartile of fibre intake from cereals and fruit/vegetables was analysed separately. In a validation study in a subsample of 129 women from the SMC using a similar FFQ (including 60 foods), the Pearson correlation coefficients between the FFQ and four 1-week diet records (completed 3–4 months apart) ranged from 0.4 to 0.5 for fruit items, from 0.4 to 0.6 for vegetable items, 0.5 for hard whole grain rye bread, 0.5 for soft whole grain bread, 0.6 for porridge and 0.7 for cold breakfast cereals (Wolk unpublished data).
The validity of FFQs was assessed for foods, nutrients, dietary supplements, glycaemic index and glycaemic load by comparison with multiple 24-h recall interviews and/or diet records [19,20,21,22,23].
Follow-up and ascertainment of hospital admissions
Cohorts were linked to the National Patient Registry (NPR) and the Causes of Death Registry (CDR). The NPR was established in 1964, but became nationwide since 1987 and contains information of all hospital visits in Sweden. The register covers more than 99% of all hospital discharges in Sweden and has been shown to be valid for most diagnoses . CDR was established in 1961 and includes information about dates and causes of death; 100% of the deaths are reported within 30 days.
Patients in the cohorts, who had symptomatic diverticular disease (DD) and at least one admission in hospital during the study period from September 15, 1997 to December 31, 2005 for women and from January 1, 1998 to December 31, 2005 for men, were compared with healthy controls in the cohort. Only incident cases were included and patients were censored after diagnosis. Outcome variables were defined in accordance with the WHO International Classifications of Diseases (ICD-10): diverticular disease was defined by a primary diagnosis of K572-9.
Statistical analysis and confounders
Multivariable cox regression was used to investigate the association of dietary intake of different types of dietary fibres with the incidence of hospitalisation due to DD. Each man and woman contributed details of follow-up time from entry into the cohort to the date of a diverticular disease diagnosis, or date of death from any cause or December 31, 2005, whichever occurred first. The proportional hazards assumptions were checked (by Kaplan–Meier curves) and satisfied, and Cox proportional hazards regression was used to estimate relative risks (RR) with 95% confidence intervals (CI) using the PHREG procedure in SAS (version 9.1; SAS institute, Inc., Cary, North Carolina, USA ). Multivariable analyses were adjusted for age (5-year age groups), diabetes (yes/no), hypertension (yes/no), BMI (< 20, 20–25, > 25 kg/m2), physical activity (h/day), alcohol (g/day), use of steroids (ever/never), smoking (ever/never) and educational level. We tested for linear trend across categorical models by modelling the median of each fibre quartile as a semi continuous variable and including this variable in a multivariable model.
All p values shown are two-sided. p < 0.05 was considered statistically significant for all analyses. Regarding the factors adjusted for in the multivariable model, a missing value for each specific variable was used so that the individual would not be excluded from the analyses. Data are presented separately for men and women, since they were recruited from different cohorts. All authors had access to the study data and have reviewed and approved the final manuscript.
Ethical approval for this study was given by the local ethical committee of the Karolinska Institutet (2006/147-32). The study is reported according to the criteria set out in the Strengthening the Reporting of Observational studies in Epidemiology (STROBE) .