Key PointsQuestion
Is cycling associated with risk of all-cause and cardiovascular disease mortality among persons with diabetes?
Findings
In this cohort study of 7459 persons with diabetes, cycling was associated with at least a 24% lower all-cause mortality rate when compared with noncyclists, independent of other physical activity and putative confounders. Taking up cycling over a 5-year period was associated with at least a 35% lower risk of all-cause mortality when compared with consistent noncyclists.
Meaning
Cycling could be encouraged as an activity for persons with diabetes to lower the risk of premature mortality.
Importance
Premature death from all causes and cardiovascular disease (CVD) causes is higher among persons with diabetes.
Objective
To investigate the association between time spent cycling and all-cause and CVD mortality among persons with diabetes, as well as to evaluate the association between change in time spent cycling and risk of all-cause and CVD mortality.
Design, Setting, and Participants
This prospective cohort study included 7459 adults with diabetes from the European Prospective Investigation into Cancer and Nutrition study. Questionnaires regarding medical history, sociodemographic, and lifestyle information were administered in 10 Western European countries from 1992 through 2000 (baseline examination) and at a second examination 5 years after baseline. A total of 5423 participants with diabetes completed both examinations. The final updated primary analysis was conducted on November 13, 2020.
Exposures
The primary exposure was self-reported time spent cycling per week at the baseline examination. The secondary exposure was change in cycling status from baseline to the second examination.
Main Outcomes and Measures
The primary and secondary outcomes were all-cause and CVD mortality, respectively, adjusted for other physical activity modalities, diabetes duration, and sociodemographic and lifestyle factors.
Results
Of the 7459 adults with diabetes included in the analysis, the mean (SD) age was 55.9 (7.7) years, and 3924 (52.6%) were female. During 110 944 person-years of follow-up, 1673 deaths from all causes were registered. Compared with the reference group of people who reported no cycling at baseline (0 min/wk), the multivariable-adjusted hazard ratios for all-cause mortality were 0.78 (95% CI, 0.61-0.99), 0.76 (95% CI, 0.65-0.88), 0.68 (95% CI, 0.57-0.82), and 0.76 (95% CI, 0.63-0.91) for cycling 1 to 59, 60 to 149, 150 to 299, and 300 or more min/wk, respectively. In an analysis of change in time spent cycling with 57 802 person-years of follow-up, a total of 975 deaths from all causes were recorded. Compared with people who reported no cycling at both examinations, the multivariable-adjusted hazard ratios for all-cause mortality were 0.90 (95% CI, 0.71-1.14) in those who cycled and then stopped, 0.65 (95% CI, 0.46-0.92) in initial noncyclists who started cycling, and 0.65 (95% CI, 0.53-0.80) for people who reported cycling at both examinations. Similar results were observed for CVD mortality.
Conclusion and Relevance
In this cohort study, cycling was associated with lower all-cause and CVD mortality risk among people with diabetes independent of practicing other types of physical activity. Participants who took up cycling between the baseline and second examination had a considerably lower risk of both all-cause and CVD mortality compared with consistent noncyclists.
Premature death from all causes and cardiovascular disease (CVD) is higher among people with diabetes.1 Regular physical activity is a critical behavioral target in the management of diabetes,2 but only structured exercise, in contrast with advice only, has been shown to improve CVD risk factors.3-6 Thus, it is necessary to investigate the influence of engagement in specific unstructured physical activities on mortality in this patient population.
Cohort studies in populations with diabetes have reported inverse associations between overall physical activity, leisure-time physical activity (LTPA), and walking with all-cause and CVD mortality.7,8 However, associations with walking have been inconsistent, likely because only moderate-intensity walking appears to be associated with a reduced risk of all-cause and CVD mortality.9 Meeting the physical activity recommendations both in terms of total physical activity volume as well as intensity is a major challenge, especially in people with diabetes.10-12 Because lack of time is often quoted as a barrier, incorporating activities into everyday life may be an effective strategy. Cycling is a potential activity to replace motorized transport for short-to-medium distance trips (eg, during commuting to work without a substantial effect on time use). As moderate-to-high intensities are reached during cycling at self-selected paces in adults,13-16 cycling could decrease the risk of premature mortality. It may also be a feasible strategy because cycling is one of the preferred activities in people with type 2 diabetes.17,18 It is well established that there is a strong inverse association between cycling and the risk of all-cause and cause-specific mortality, such as CVD, in the general population.19-21 However, to our knowledge, there are no studies that have examined the role of cycling in preventing premature mortality in people with diabetes.
The primary aim of this cohort study was to investigate the associations between cycling and all-cause and CVD mortality among persons with diabetes. A secondary aim was to study the associations between change in cycling over a 5-year period and all-cause and CVD mortality.
This study was a prospective cohort study of people with diabetes at baseline in the European Prospective Investigation into Cancer and Nutrition (EPIC) study cohort.22 In EPIC, 23 centers in 10 Western European countries (France, Italy, Spain, United Kingdom, the Netherlands, Greece, Germany, Sweden, Denmark, and Norway) collected information on nutrition, lifestyle, anthropometry, and medical history from more than 521 234 participating men and women.22 Medical history, sociodemographic, and lifestyle information was assessed by questionnaires at baseline between 1992 and 1998 (baseline examination) and at a second examination between 1996 and 2011. Data were only available from 22 centers because data from Greece were not released for this study. The ethical review boards from the International Agency for Research on Cancer and all local participating centers approved this study. Informed consent was obtained from all participants.
From the entire EPIC population, people with diabetes at the baseline assessment were included in the present study. Diabetes was based on self-report on a single occasion or by a second source (at least 1), including repeated self-report, by a general physician, linkage to register or medical record at a later point, prescription of use of glucose-lowering or diabetes-related medication, or baseline glycated hemoglobin at least 6.0% (42 mmol/mol [to convert to proportion of total hemoglobin, multiply by 0.01]).7
Study procedures have been described in detail elsewhere.22 Briefly, height, weight, and waist circumference were measured using similar protocols across study centers.22 Body mass index was calculated as weight in kilograms divided by height in meters squared. Central obesity was defined according the International Diabetes Federation criteria.23 Diabetes duration was calculated as the time from self-reported age or calendar year of medical diagnosis to baseline.
Dietary intake, including alcohol consumption, was assessed by a questionnaire (quantitative, semiquantitative, or a combination) and 7-day or 14-day record, and individual energy and nutrient intake was based on the standardized EPIC nutrient database.22,24 Because the Mediterranean diet is associated with improved metabolic control and decreased risk of diabetes,25-27 it was included as a covariate expressed as the relative Mediterranean diet score.25,27,28 Dietary data were only available for the baseline examination.
Assessment of Physical Activity
Information about physical activity habits was obtained from a lifestyle questionnaire and included information about duration and frequency of leisure-time and occupational physical activity.22,29 Weekly time spent cycling to and/or from work and leisure time during winter and summer was averaged into a single variable of total annual cycling time and then categorized as 0, 1 to 59, 60 to 149, 150 to 299, and 300 or more min/wk. Change in total cycling from baseline to the second examination was categorized based on total time spent cycling at the 2 examinations: (1) noncycling: participants who reported zero minutes of cycling at both examinations; (2) stopped cycling: participants who reported cycling (any amount) at baseline but not at the second examination; (3) started cycling: participants who did not report cycling at baseline but did report cycling (any amount) at the second examination; or (4) maintained cycling: those who were consistent cyclists at both examinations.
The LTPA energy expenditure (without cycling included) metabolic equivalent of tasks in hours per week (MET-h/wk15) was calculated at both examinations as the sum of energy expenditures from the following activities: gardening, do-it-yourself activities, stair climbing, housework activities, walking, and sports. Because information on stair climbing was only available from 4 study centers at the second examination, this activity was not included in the second examination LTPA variable. Occupational physical activity was reported in categories of sedentary occupation, standing occupation, manual or heavy manual work, or nonworker.29
The primary and secondary outcomes were all-cause and CVD mortality, respectively. Any deaths due to CVD were coded according to the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, using the codes I00 to I99. Vital status and cause of death were obtained through record linkage with registries, physicians or hospitals, or direct follow-up.30 Follow-up end for the primary and secondary analyses was on March 26, 2015 (eTable 1 in Supplement 1).
A statistical analysis plan was developed (Supplement 2) and preregistered (NCT04171557) prior to commencing the analyses. The risks of all-cause and CVD mortality were computed as hazard ratios (HRs) with 95% CIs according to weekly time spent cycling at baseline estimated using stratified Cox proportional hazard regression models with age as the underlying time scale. Analyses were corrected for delayed entry. Participants were considered at risk from age at baseline examination in the primary analyses and from age at the second examination in the analyses of change in cycling. Participants leaving the study during follow-up owing to emigration or premature withdrawal were right censored at the age of emigration or withdrawal. A priori, participants with sufficient information to be included in the most extensive statistical models (models 3 and 2 for the primary and secondary analyses, respectively) were included in the primary models (models 2 and 1 for the primary and secondary analyses, respectively).
A crude model (model 1) was fitted with categories of cycling as exposure (0 [reference], 1-59, 60-149, 150-299, and ≥300 min/wk), adjusted for sex and age (years), and further stratified by study center to adjust for confounding of this variable. The proportional hazards assumption for cycling was met within each stratum. Model 1 was further adjusted for attained educational level (no formal education, primary school, technical or professional school, secondary school, or university degree), smoking status (never smoker, former smoker, or current smoker), diabetes duration (years), adherence to the Mediterranean diet (relative Mediterranean diet score: low, 0-6 points; medium, 7-10 points; or high, 11-18 points),25,27 total energy intake (quartiles of kcal/d),24 physical activity excluding cycling (quartiles of LTPA energy expenditure), and occupational physical activity (sedentary occupation, standing occupation, manual or heavy manual work, nonworker, or unknown status) (model 2 [main model]). Finally, history of prevalent stroke (yes or no), previous myocardial infarction (yes or no), prevalent cancer (yes or no), hypertension (yes or no), hyperlipidemia (yes or no), and central obesity (yes or no) at baseline were added as covariates (model 3). Effect modification by sex and diabetes duration (≥5 vs <5 years) was evaluated statistically using the likelihood-ratio test by comparing model 2, adding a multiplicative interaction term for sex or diabetes duration and cycling with a model including only main effects. Because several covariates did not meet the proportional hazard assumption in the multivariable models 2 and 3, we computed an extended Cox regression analysis in which we stratified by study center and energy intake in the analyses with all-cause mortality as the outcome. In the models with CVD mortality as the outcome, we stratified by study center, educational level, and LTPA (excluding cycling). We conducted a range of preplanned sensitivity analyses for the primary model (model 2) specified in the statistical analysis plan (Supplement 2) to investigate the association of residual confounding and reverse causality (excluding all deaths and CVD deaths within the first 2 years following the baseline examination).
In the preplanned secondary analysis, associations between all-cause and CVD mortality and change in cycling (from baseline to second examination) were investigated. The associations were initially adjusted for sex and age at the second examination and stratified by study center. A multivariable model was fitted additionally adjusting for educational level at baseline, smoking status at both examinations, diabetes duration at the time of second examination, LTPA (excluding cycling) at both examinations, and occupational physical activity at the second examination. The multivariable analyses were stratified by study center, baseline occupational physical activity, adherence to the relative Mediterranean diet, and total energy intake at baseline.
Because the associations between baseline and cycling and all-cause or CVD mortality were nonlinear, we computed (post hoc) restricted cubic splines of the respective associations with knot placements at the 10th, 50th, and 90th percentiles, as recommended by Harrell.31 These analyses were repeated using the average of baseline and second examinations. Because of large amounts of zero values in the cycling variable, the percentiles were computed excluding zero. Moreover, 10-year adjusted (standardized) cumulative mortality according to cycling at baseline or change in cycling status, consistent with the primary models, were estimated post hoc. Flexible parametric survival models32 with additional postestimation of adjusted differences (95% CIs) in 10-year cumulative mortality comparing 0 min/wk of cycling at baseline and/or at the second examination to higher levels of cycling or stopping/starting/maintaining cycling were used. Also, to assess the influence of unmeasured confounding, the E-value was calculated based on our estimated HR for all-cause and CVD mortality.33 All analyses were conducted using STATA IC, version 16.1 (StataCorp), with 2-sided α = .05 considered statistically significant.
Of the 492 763 participants enrolled in the EPIC cohort, 10 995 had diabetes at the baseline examination. The analytic sample consisted of 7459 participants, of whom 4701 (63.0%) had confirmed diabetes and 4699 (63.0%) were noncyclists. Among the sample, the mean (SD) age was 55.9 (7.7) years, the mean (SD) diabetes duration was 7.7 (8.1) years, and 3924 (52.6%) were female (Table 1). The main reason for exclusion from the analytic sample was missing covariate or exposure information (eFigure 1 in Supplement 1). Reasons for missing data are reported in eTable 2 in Supplement 1.
The participants were followed for a mean (SD) of 14.9 (4.4) years (110 944 person-years) with 1673 deaths from all causes and 811 deaths attributable to CVD. A subset of participants (n = 5423) also completed the second examination and were included in the analysis of change in cycling. This analysis had a mean (SD) of 10.7 (4.3) years of follow-up, accumulating a total of 57 802 person-years with 975 deaths from all causes and 429 from CVD.
Baseline Cycling and All-Cause and CVD Mortality
The HR for all-cause mortality was lower for people reporting any cycling (>0 min/wk) when compared with noncyclists (model 1; Table 234). Cycling was also associated with a reduced risk of CVD mortality (Table 2). Adjusting for educational level, lifestyle risk factors, and diabetes duration did not materially change the association between cycling and all-cause mortality (model 2; Table 2). The 10-year cumulative mortality risk difference relative to 0 min/wk of cycling for ascending cycling categories (1-59, 60-149, 150-299, and ≥300 min/wk) were −1.9%, −2.0%, −2.7%, and −2.1% for all-cause mortality and −1.2%, −1.2%, −2.2%, and −1.0% for CVD mortality, respectively (eTable 3 in Supplement 1), with the lowest category of cycling consistently having the highest cumulative mortality (eFigure 2 in Supplement 1).
No statistically significant multiplicative interactions of sex or diabetes duration and cycling were observed for all-cause nor CVD mortality (eTable 4 in Supplement 1). Further adjustment for existing conditions and CVD risk factors only slightly attenuated the associations (model 3; Table 2). Sensitivity analyses investigating residual confounding by smoking, sports participation, self-reported diabetes, and reverse causality broadly confirmed the associations between cycling and both all-cause and CVD mortality (eTable 5 in Supplement 1), and E-values ranged from 1.66 to 2.91 in categories of baseline cycling (eTable 6 in Supplement 1).
The dose-response relationship with baseline cycling as a continuous variable for both all-cause and CVD mortality was modeled post hoc; for comparison, the association for LTPA (excluding cycling) is provided (Figure 134). This revealed a reversed J-shaped association between both outcomes and cycling, and a linear association for LTPA (excluding cycling) (HRs per 10 MET-h increase per week: 0.97 [95% CI, 0.95-0.98] and 0.96 [95% CI, 0.94-0.98] for all-cause and CVD mortality, respectively) (Figure 1). The distribution of cycling and LTPA among observations with and without events underlying these analyses are found in eFigure 3 in Supplement 1. The associations between average cycling at baseline and second examination approximated a curve-linear dose-response relationship (eFigure 4 and eTables 7 and 8 in Supplement 1).
Change in Cycling and All-Cause and CVD Mortality
For both all-cause and CVD mortality, HRs were 0.65 or less among participants who started or maintained cycling (Figure 2) relative to noncyclists (risk differences relative to noncyclists were −3.7% for all-cause mortality and −2.7% for CVD mortality; eTable 9 and eFigure 5 in Supplement 1). After excluding deaths within the first 2 years from the second examination, the associations were unchanged: HRs for all-cause mortality were 0.93 (95% CI, 0.72-1.19), 0.66 (95% CI, 0.45-0.95), and 0.64 (95% CI, 0.51-0.80) for people who stopped, started, or maintained cycling, respectively, compared with noncyclists), and corresponding HRs for CVD mortality were 1.08 (95% CI, 0.77-1.51), 0.56 (95% CI, 0.31-0.99), and 0.54 (95% CI, 0.39-0.75). E-values were more than 2 for the groups that started cycling and maintained cycling (eTable 6 in Supplement 1).
We observed that time spent cycling was associated with a lower risk of all-cause and CVD mortality in people with diabetes, independent of other physical activities and other putative confounders. However, the dose-response relationships were ambiguous.
The association between cycling and all-cause and CVD mortality in this study of person with diabetes was of the same magnitude and direction as observed in the healthy population.13,20,21,35-41 This is moreover in line with previous observations by Andersen et al,8 who observed that the associations between physical activity and all-cause mortality seemed to be similar between the healthy population and persons living with chronic disease, including cardiovascular, lung, and musculoskeletal diseases. However, their investigation did not include persons with diabetes nor did it specifically investigate the interaction between chronic disease state and associations between cycling and all-cause mortality.8 The lower risk of all-cause and CVD mortality associated with overall physical activity as well as walking among persons with diabetes is well established.7,9,42-48 The present investigation extends the level of evidence within this field by documenting that cycling and taking up cycling may offer specific health benefits in people with diabetes over and above other physical activities, including walking. Mixed-mode commuting (walking and/or cycling) has been associated with decreased mortality in persons with diabetes.49 However, the association was weaker as compared with our observations. Because physical activity intensity is important in mediating the health benefits from walking among people with type 2 diabetes,9,50 a lower intensity of physical activity, such as walking, when compared with cycling, may account for differences.15 The lower risk of all-cause and CVD mortality observed in consistent cyclists or persons initiating cycling may be mediated by improvements in aerobic fitness, which is associated with all-cause and CVD mortality.48,51
The shapes of the dose-response curves were ambiguous. The uptick in the point estimate at high volumes of cycling can be associated with an increased risk of fatal injuries with increased cycling (eg, in urban settings or in those with increased risk of CVD or respiratory diseases owing to exposure to air pollutants during cycling in settings with dense motorized traffic).52-54 In addition, the beneficial effects of physical activity on cardiovascular risk factors with increasing air pollutants may be attenuated.55-58 However, previous cohort studies36,59,60 have reported that levels of traffic-related air pollution did not modify the inverse association of outdoor physical activity with mortality or incidence of heart disease. Although air pollution during exercise may decrease lung function acutely,55,57,58,61 it seems that the benefits of physical activity on the risk of asthma and chronic obstructive pulmonary disease are maintained when performed in moderately polluted settings such as urban environments.62 Because cycling is associated with an increased risk of fatal injuries compared with passive transportation,52 this may also explain the small uptick in the point estimate for all-cause mortality risk with increased cycling, but it cannot explain the corresponding shape of the curve for CVD mortality. Of note, commuter cycling may increase the risk of injuries and hospital admissions compared with nonactive commuting in the general population.52 However, the health benefits of cycling may outweigh the increased risk of injuries owing to a decreased risk of morbidities in cyclists.63 While data on traumatic deaths due to cycling would have enabled a benefit-harm analysis in this study, those data were unfortunately not collected. Another explanation for the uptick may be associated with potential adverse effects of very high volumes of physical activity. While a recent study does not support this in the healthy population,64 it may constitute a concern in persons with diabetes. Participants who cycle many hours per week (eg, ≥5 h/wk) may do so at a higher intensity and thus accumulate a greater volume of physical activity compared with participants who cycle less. However, looking at the total volume of cycling (up to 75 MET-h/wk) in the context of the total volume of LTPA, it seems unlikely that this should explain the uptick because this is not observed for LTPA even at substantially higher volumes than observed with cycling. Finally, because the post hoc analyses using the average of the baseline and second examination as exposure to limit misclassification of cycling owing to change in cycling habits during follow-up to some extent attenuated the minor uptick observed when using the baseline exposure only, this suggests that the uptick might partly be a consequence of measurement error on the exposure variable. This is in line with previous observations.65-67
Limitations of the study include the observational design that limits causal inferences. We were unable to distinguish between type 1 and type 2 diabetes, but generally type 2 diabetes accounts for 90% of all diabetes in adults.68 To maximize the analytic sample, we chose to include both persons with self-reported diabetes and persons with diabetes confirmed through other sources, which increases the risk of misclassification. However, only few numerical differences were observed in the characteristics between persons with confirmed and self-reported diabetes (eTable 10 in Supplement 1), and the analysis restricted to those with confirmed diabetes only supported the overall findings. Although we adjusted the analyses for a range of potential confounders, these were mostly self-reported and thus prone to misclassification. Although slightly attenuated, the associations observed for all-cause and CVD mortality were confirmed in sensitivity analyses, when those who had ever smoked and people reporting engaging in any sports were excluded. This suggests that residual confounding by smoking and sports-related physical activity may be minor, although the 95% CIs for the latter were wide for CVD mortality. A concern may be confounding by concomitant pharmacological intervention. However, as pharmacological intervention intensifies with increasing diabetes duration,69 and as we consistently adjusted for diabetes duration, we may, to some extent, have addressed this issue. Because the prevalence of microvascular and macrovascular complications are highly prevalent among persons with diabetes,70 persons with a history of CVD at baseline were included in the primary analyses to increase the generalizability of the findings. While this may increase the risk of reverse causality, the 2-year benchmark sensitivity analysis did not suggest reverse causation. E-value calculations showed that to explain away the estimated associations of baseline and change in cycling with mortality with adjustment for the measured covariates, an unmeasured confounder would need to have an HR of at least 1.66 and 1.94 with both cycling and mortality, respectively. We decided a priori only to include participants with complete data for all statistical models, which could have introduced selection bias and limited generalizability. While rerunning the analyses with missing data statistically imputed for cycling and confounders confirmed the present findings, this may not fully overcome the biases. Finally, the results may not be generalizable to people using electric cycles.
Results of this cohort study showed that engaging in cycling was associated with a lower risk of all-cause and CVD mortality among people with diabetes after considering other physical activities as well as other risk factors. Studies with repeated measurement of cycling and documentation of cycling-related accidents are needed to elaborate on the dose-response relationship between cycling and mortality.
Accepted for Publication: May 15, 2021.
Published Online: July 19, 2021. doi:10.1001/jamainternmed.2021.3836
Corresponding Author: Mathias Ried-Larsen, PhD, Centre for Physical Activity Research, Rigshospitalet, Blegdamsvej 9, Copenhagen 2100, Denmark (mathias.ried-larsen@regionh.dk).
Author Contributions: Drs Ried-Larsen and Rasmussen had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Ried-Larsen and Rasmussen contributed equally to this work.
Concept and design: Ried-Larsen, Rasmussen, Blond, Ardanaz, L. Andersen, Brage, Grøntved.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Ried-Larsen, Rasmussen, Steindorf, Brage.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Ried-Larsen, Rasmussen, Steindorf, Freisling, L. Andersen, Brage, Grøntved.
Obtained funding: Ried-Larsen, K. Overvad, Panico, Bergmann, Wareham.
Administrative, technical, or material support: Rasmussen, Katzke, Masala, Weiderpass, Bergmann, Schulze.
Supervision: T. Overvad, J. Andersen, Petersen, Verschuren, Ardanaz, L. Andersen, Wareham, Brage, Grøntved.
Conflict of Interest Disclosures: Dr Ried-Larsen reported personal fees from Novo Nordisk outside the submitted work. Prof Steindorf reported personal fees from Preventon and the Swiss Group for Clinical Cancer Research outside the submitted work. Prof Schulze reported grants from the German Federal Ministry of Education and Research, the German Cancer Aid, and the State of Brandenburg (82DZD00302) during the conduct of the study. No other disclosures were reported.
Funding/Support: This work was supported by the Health Research Fund of Instituto de Salud Carlos III; regional governments of Andalucía, Asturias, Basque Country, Murcia, and Navarra; and the Catalan Institute of Oncology. The Centre for Physical Activity Research is supported by a grant from TrygFonden. Dr Zamora-Ros was supported by the Miguel Servet program (CP15/00100) from the Institute of Health Carlos III (cofunded by the European Social Fund).
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: Where authors are identified as personnel of the International Agency for Research on Cancer/World Health Organization, the authors alone are responsible for the views expressed in this article and do not necessarily represent the decisions, policy, or views of the International Agency for Research on Cancer/World Health Organization.
Additional Contributions: We acknowledge and thank the participants as well as personnel at all European Prospective Investigation into Cancer and Nutrition study sites for their efforts. Participants were not compensated for their participation, and personnel were not compensated outside of their standard salaries.
Additional Information: Individual participant data that underlie the results reported in this article after deidentification (text, Tables, Figures, and Supplements) can be accessed by contacting the International Agency for Research on Cancer/World Health Organization. Analytic codes are available on request by contacting the corresponding author.
2.Colberg
SR, Sigal
RJ, Yardley
JE,
et al. Physical activity/exercise and diabetes: a position statement of the American Diabetes Association.
Diabetes Care. 2016;39(11):2065-2079. doi:
10.2337/dc16-1728PubMedGoogle ScholarCrossref 3.Umpierre
D, Ribeiro
PA, Kramer
CK,
et al. Physical activity advice only or structured exercise training and association with HbA1c levels in type 2 diabetes: a systematic review and meta-analysis.
JAMA. 2011;305(17):1790-1799. doi:
10.1001/jama.2011.576PubMedGoogle ScholarCrossref 4.Boulé
NG, Haddad
E, Kenny
GP, Wells
GA, Sigal
RJ. Effects of exercise on glycemic control and body mass in type 2 diabetes mellitus: a meta-analysis of controlled clinical trials.
JAMA. 2001;286(10):1218-1227. doi:
10.1001/jama.286.10.1218PubMedGoogle ScholarCrossref 6.Hayashino
Y, Jackson
JL, Fukumori
N, Nakamura
F, Fukuhara
S. Effects of supervised exercise on lipid profiles and blood pressure control in people with type 2 diabetes mellitus: a meta-analysis of randomized controlled trials.
Diabetes Res Clin Pract. 2012;98(3):349-360. doi:
10.1016/j.diabres.2012.10.004PubMedGoogle ScholarCrossref 11.Morrato
EH, Hill
JO, Wyatt
HR, Ghushchyan
V, Sullivan
PW. Physical activity in U.S. adults with diabetes and at risk for developing diabetes, 2003.
Diabetes Care. 2007;30(2):203-209. doi:
10.2337/dc06-1128PubMedGoogle ScholarCrossref 12.van der Berg
JD, Stehouwer
CD, Bosma
H,
et al. Associations of total amount and patterns of sedentary behaviour with type 2 diabetes and the metabolic syndrome: The Maastricht Study.
Diabetologia. 2016;59(4):709-718. doi:
10.1007/s00125-015-3861-8PubMedGoogle ScholarCrossref 13.Blond
MB, Rosenkilde
M, Gram
AS,
et al. How does 6 months of active bike commuting or leisure-time exercise affect insulin sensitivity, cardiorespiratory fitness and intra-abdominal fat? a randomised controlled trial in individuals with overweight and obesity.
Br J Sports Med. 2019;53(18):1183-1192. doi:
10.1136/bjsports-2018-100036PubMedGoogle ScholarCrossref 19.Kelly
P, Kahlmeier
S, Götschi
T,
et al. Systematic review and meta-analysis of reduction in all-cause mortality from walking and cycling and shape of dose response relationship.
Int J Behav Nutr Phys Act. 2014;11:132. doi:
10.1186/s12966-014-0132-xPubMedGoogle ScholarCrossref 20.Nordengen
S, Andersen
LB, Solbraa
AK, Riiser
A. Cycling and cardiovascular disease risk factors including body composition, blood lipids and cardiorespiratory fitness analysed as continuous variables: part 2—systematic review with meta-analysis.
Br J Sports Med. 2019;53(14):879-885. doi:
10.1136/bjsports-2018-099778PubMedGoogle ScholarCrossref 21.Nordengen
S, Andersen
LB, Solbraa
AK, Riiser
A. Cycling is associated with a lower incidence of cardiovascular diseases and death: part 1—systematic review of cohort studies with meta-analysis.
Br J Sports Med. 2019;53(14):870-878. doi:
10.1136/bjsports-2018-099099PubMedGoogle ScholarCrossref 22.Riboli
E, Hunt
KJ, Slimani
N,
et al. European Prospective Investigation into Cancer and Nutrition (EPIC): study populations and data collection.
Public Health Nutr. 2002;5(6B):1113-1124. doi:
10.1079/PHN2002394PubMedGoogle ScholarCrossref 24.Slimani
N, Deharveng
G, Unwin
I,
et al. The EPIC nutrient database project (ENDB): a first attempt to standardize nutrient databases across the 10 European countries participating in the EPIC study.
Eur J Clin Nutr. 2007;61(9):1037-1056. doi:
10.1038/sj.ejcn.1602679PubMedGoogle ScholarCrossref 25.Romaguera
D, Guevara
M, Norat
T,
et al; InterAct Consortium. Mediterranean diet and type 2 diabetes risk in the European Prospective Investigation into Cancer and Nutrition (EPIC) study: the InterAct project.
Diabetes Care. 2011;34(9):1913-1918. doi:
10.2337/dc11-0891PubMedGoogle Scholar 27.Buckland
G, Agudo
A, Luján
L,
et al. Adherence to a Mediterranean diet and risk of gastric adenocarcinoma within the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort study.
Am J Clin Nutr. 2010;91(2):381-390. doi:
10.3945/ajcn.2009.28209PubMedGoogle ScholarCrossref 29.Wareham
NJ, Jakes
RW, Rennie
KL,
et al. Validity and repeatability of a simple index derived from the short physical activity questionnaire used in the European Prospective Investigation into Cancer and Nutrition (EPIC) study.
Public Health Nutr. 2003;6(4):407-413. doi:
10.1079/PHN2002439PubMedGoogle ScholarCrossref 31.Harrell
FE
Jr. Regression Modeling Strategies: With Applications to Linear Models, Logistic and Ordinal Regression, and Survival Analysis. Springer; 2001.
35.Grøntved
A, Rasmussen
MG, Blond
K, Østergaard
L, Andersen
ZJ, Møller
NC. Bicycling for transportation and recreation in cardiovascular disease prevention.
Curr Cardiovasc Risk Rep. 2019;13(26). doi:
10.1007/s12170-019-0623-zGoogle Scholar 36.Andersen
ZJ, de Nazelle
A, Mendez
MA,
et al. A study of the combined effects of physical activity and air pollution on mortality in elderly urban residents: the Danish Diet, Cancer, and Health Cohort.
Environ Health Perspect. 2015;123(6):557-563. doi:
10.1289/ehp.1408698PubMedGoogle ScholarCrossref 39.Sahlqvist
S, Goodman
A, Simmons
RK,
et al. The association of cycling with all-cause, cardiovascular and cancer mortality: findings from the population-based EPIC-Norfolk cohort.
BMJ Open. 2013;3(11):e003797. doi:
10.1136/bmjopen-2013-003797PubMedGoogle Scholar 41.Schnohr
P, Marott
JL, Jensen
JS, Jensen
GB. Intensity versus duration of cycling, impact on all-cause and coronary heart disease mortality: the Copenhagen City Heart Study.
Eur J Prev Cardiol. 2012;19(1):73-80. doi:
10.1177/1741826710393196PubMedGoogle ScholarCrossref 50.Boulé
NG, Kenny
GP, Haddad
E, Wells
GA, Sigal
RJ. Meta-analysis of the effect of structured exercise training on cardiorespiratory fitness in type 2 diabetes mellitus.
Diabetologia. 2003;46(8):1071-1081. doi:
10.1007/s00125-003-1160-2PubMedGoogle Scholar 55.Sinharay
R, Gong
J, Barratt
B,
et al. Respiratory and cardiovascular responses to walking down a traffic-polluted road compared with walking in a traffic-free area in participants aged 60 years and older with chronic lung or heart disease and age-matched healthy controls: a randomised, crossover study.
Lancet. 2018;391(10118):339-349. doi:
10.1016/S0140-6736(17)32643-0PubMedGoogle ScholarCrossref 56.Kubesch
NJ, de Nazelle
A, Westerdahl
D,
et al. Respiratory and inflammatory responses to short-term exposure to traffic-related air pollution with and without moderate physical activity.
Occup Environ Med. 2015;72(4):284-293. doi:
10.1136/oemed-2014-102106PubMedGoogle ScholarCrossref 58.Kubesch
N, De Nazelle
A, Guerra
S,
et al. Arterial blood pressure responses to short-term exposure to low and high traffic-related air pollution with and without moderate physical activity.
Eur J Prev Cardiol. 2015;22(5):548-557. doi:
10.1177/2047487314555602PubMedGoogle ScholarCrossref 59.Kubesch
NJ, Therming Jørgensen
J, Hoffmann
B,
et al. Effects of leisure-time and transport-related physical activities on the risk of incident and recurrent myocardial infarction and interaction with traffic-related air pollution: a cohort study.
J Am Heart Assoc. 2018;7(15):e009554. doi:
10.1161/JAHA.118.009554PubMedGoogle Scholar 63.Welsh
C, Celis-Morales
CA, Ho
F,
et al. Association of injury related hospital admissions with commuting by bicycle in the UK: prospective population based study.
BMJ. 2020;368:m336. doi:
10.1136/bmj.m336PubMedGoogle Scholar 65.Lee
DH, Rezende
LFM, Ferrari
G,
et al. Physical activity and all-cause and cause-specific mortality: assessing the impact of reverse causation and measurement error in two large prospective cohorts.
Eur J Epidemiol. 2021;36(3):275-285. doi:
10.1007/s10654-020-00707-3PubMedGoogle ScholarCrossref 68.Saeedi
P, Petersohn
I, Salpea
P,
et al; IDF Diabetes Atlas Committee. Global and regional diabetes prevalence estimates for 2019 and projections for 2030 and 2045: results from the International Diabetes Federation Diabetes Atlas, 9th edition.
Diabetes Res Clin Pract. 2019;157:107843. doi:
10.1016/j.diabres.2019.107843PubMedGoogle Scholar 69.Espeland
MA, Glick
HA, Bertoni
A,
et al; Look AHEAD Research Group. Impact of an intensive lifestyle intervention on use and cost of medical services among overweight and obese adults with type 2 diabetes: the action for health in diabetes.
Diabetes Care. 2014;37(9):2548-2556. doi:
10.2337/dc14-0093PubMedGoogle ScholarCrossref