Design and setting
A prospective cohort study design was used to investigate the described hypothesis. The study was based on data from the Danish “Diet, Cancer and Health” (DCH) cohort and from Danish registries. Participants in the DCH cohort were recruited between December 1993 and May 1997, and included a total of 57 053 participants (27 178 men and 29 875 women) aged between 50 and 64 years of age, with residence in the Copenhagen area or Aarhus area, born in Denmark and with no previous cancer diagnosis in the Danish Cancer Registry. At baseline, the participants completed a lifestyle questionnaire and a food frequency questionnaire (FFQ). Both questionnaires were interviewer-checked and validated regarding PA and diet(42, 43). A detailed description of the DCH cohort has been described elsewhere(44).
Participants, eligibility criteria and follow-up
Inclusion criteria: All men and women included in the DCH cohort with no diagnosis of CD or UC (diagnostic codes described in next section) before entry to the cohort and with information on PA were included in the analyses.
Follow-up: The participants were followed from the date of their first visit at the DCH study clinic until the date of diagnosis of CD or UC, date of death or emigration, or December 31, 2018, whichever came first.
Materials, data sources, and methods
Outcome diagnostic criteria: The outcome late onset IBD was defined by the criteria: 1) having a main diagnosis (A-diagnosis) registered in the Danish National Patient Registry (DNPR)(45) with the International Classification of Diseases (ICD) 8 and 10 codes for CD (563.00–563.09, 563.91 and K50 (including all sub-codes)) and UC (563.19, 563.99, 569.04, and K51 (including all sub-codes)) in years 1977-2018 from a department with a relevant area of specialization (Surgical Gastroenterology, Medical Gastroenterology, Internal medicine) and 2) the diagnosis was followed by at least one additional registration in the DNPR (inpatient or outpatient visit) related to the first diagnosis within 180 days. The date and year of the diagnosis were defined as the date and year of the first diagnosis registered in the DNPR.
Registry data: The Danish health registries included the DNPR and the Danish Civil Registration System (DCRS)(46). The DNPR was used to identify patients with IBD during follow-up. ICD-8 and ICD-10 codes from the DNPR was used to identify cases diagnosed before and after entry to the DCH cohort, and to calculate comorbidity in the cohort using the updated Charlson’s comorbidity index(47). Comorbidity was categorized as a binary variable (comorbidity=no/yes) to ensure enough power of the group with comorbidity. High completeness of IBD registration in the DNPR has previously been reported(94%), with an estimated positive predictive value of 97% for CD and 90% for UC(48). The DCRS was used to extract follow-up information on death and immigration. Data were linked by the unique identification number assigned to all residents in Denmark at birth or first immigration.
Exposure: The exposure was defined as 1) a binary indicator of exposure: being active/inactive, both for total intensity of PA and separated on the six different types of leisure time PA: walking, housework, gardening, do-it-yourself work, cycling and sports, and as 2) levels of intensity of total PA and time spend on the six types of activities. The intensity was measured as MET hours/week. The time spent was measured as hours/week.
The MET system is based on the understanding that all activities are assigned an intensity unit based on their rate of energy expenditure. One MET is defined as the energy expenditure at rest (the resting metabolic rate), which for the average adult is approximately 3.5 ml of O2/kg body weight/min. The intensities of different activities are calculated as the ratio between the associated metabolic rate for the specific activity and the resting metabolic rate(49).
Being active was defined as having an intensity level of ≥3 MET hours/week or as spending >0 hours/week on each type of activity. The cut point for the binary variable of MET hours/week was chosen to correspond to inactivity equivalent of <1 hour of walking at an average pace per week, consistent with prior studies(36). The levels of intensity and time spent were categorised in quartiles. The pre-defined variable for MET hours/week from the DCH dataset was used. The variable is further described below.
Questionnaire data: Information on leisure time PA was based on six questions covering the average number of hours per week spent the past year on the six types of leisure time PA during summer and winter, respectively. The MET hours/week variable was calculated by multiplying the MET value of each specific activity by duration and frequency of the activities. The following MET-values were used according to Ainsworth’s Compendium of Physical Activities(49, 50): walking 3.0, housework 3.0, gardening 4.0, do-it-yourself work 4.5, cycling 6.0, and sports 6.0.
Covariates and potential confounders: Based on the known and putative IBD risk factors and preventative factors, the analyses were adjusted for occupational PA, smoking, intake of fibre, fermented dairy products, red and processed meat, alcohol, HRT (only for women), NSAID, comorbidity, and also the demographic factors age and gender. These factors were expected to be possible confounders of the association between leisure time PA and risk of IBD.
Information on occupational PA was obtained from a question with five categories (sitting, standing, light manual work, heavy manual work, no occupation). Light and heavy manual work were combined in one category. Total energy intake was measured in mega joule (MJ) per day, alcohol consumption and intake of fibre, fermented dairy products and meat (red and processed meat) were measured in grams per day – all retrieved from the FFQ. A detailed description of the calculation of the dietary variables in the DCH study is described elsewhere(51). Smoking habits within the past year was defined as current, never or former. The questionnaire also gave information on the use of a pain-relieving drug, which was defined by the variable NSAID and assessed as >1 pill per month during the last year before baseline (yes/no). HRT was divided into the following categories: never, current, and former user.
Statistical analyses
To investigate the risk of and time to IBD event the Cox proportional hazards model with age as the underlying time scale was applied. Death, emigration and loss to follow-up were not considered as competing risks, and thus, were handled as censoring. The assumption of proportional hazards was checked by evaluating parallel curves of the cumulative hazard function on the log-scale. Furthermore, sensitivity analyses modelling time-varying effect of covariates that did not fulfil the proportional hazard assumption, were performed.
Hazard-ratios (HRs) and the corresponding 95% confidence intervals (95% CI) and p-values for IBD onset associated with participation in and levels of leisure time PA were estimated. All analyses were carried out according to the principle of complete-case-analysis(52) to ensure an equal number of participants in all analyses. The analyses were carried out in both a crude model and a model adjusted for baseline values of preventative factors and risk factors for IBD. In the dose-response analyses, inactive individuals were included by assigning indicator variables of being active/inactive, as the IBD risk among inactive individuals may deviate from the risk among active individuals. Since the lowest quartile group of gardening, do-it-yourself-work, cycling, and sports only included inactive people (0 hours/week), the second quartile group was used as a reference for these variables.
Furthermore, the analyses were stratified according to strata of age groups (50-59 and 60-64 years), BMI (<25 kg/m2 and ≥25 kg/m2), smoking (‘never smoker’ and ‘current/former smoker’), occupational PA (‘not active at work’, including sitting and not working, and ‘active at work’, including standing and manual work) and work status (not working/working), as these variables were assumed to interact with the effect of leisure time PA.
All analyses were carried out using Stata version 15 (53). For all tests, a P-value below 0.05 was considered statistically significant.
The study was not submitted to the Southern Denmark Ethics Committee, which is the local ethics committee. The study does not need approval from the Ethics committee or Institutional Review Board by Danish law: “Questionnaire studies and health science registry research projects must be reported to the scientific ethics committee system only if the project includes human biological material”. (the Act on Research Ethics Review of Health Research Projects (Danish: Lov om videnskabsetisk behandling af sundhedsvidenskabelige forskningsprojekter, Lov nr. 593 af 14. juni 2011, § 14, stk. 2)).(54)