Study design
This study was conducted as a part of the Japan Epidemiology Collaboration on Occupational Health (J-ECOH) Study, an ongoing large-scale cohort study among workers from multiple industries in Japan. The details of the J-ECOH Study and this study cohort have been described elsewhere [17, 18]. This analysis included data from one of the participating industries in the J-ECOH Study (electrical machinery and apparatus manufacturing), where detailed information on physical activity has been collected as a part of periodic health check-ups since 2006.
The study protocol was approved by the Ethics Committee of the National Center for Global Health and Medicine, Japan. The purpose and procedure of the J-ECOH Study were announced by using posters. Participants were allowed to refuse the provision of their data to the study. This procedure conforms to the Japanese Ethical Guidelines for Epidemiological Research for observational studies that use existing data.
Participants
In Japan, employers are obliged to organize health check-ups for their employees at least once per year under the Industrial Safety and Health Act. A total of 23,248 workers aged 20 to 65 years received health check-ups (comprehensive type) between April 2006 and March 2007 (baseline period) and had data for serum creatinine. We excluded workers with CKD (defined as an estimated glomerular filtration rate [eGFR] of < 60 mL/min/1.73 m2 and/or proteinuria [+, 2+, or 3+ on dipstick [19]]; n = 1,510); with self-reported cancer (n = 179); with an eGFR of ≥200 mL/min/1.73m2 (due to possible measurement errors[20]; n = 5); with incomplete information on physical activity (n = 2,424); and with engagement in an unspecified activity only during leisure (n = 489). We further excluded workers who attended no subsequent health check-ups or who had no measurement of eGFR or proteinuria in a subsequent health check-up (n = 1,973). Finally, 17,331 participants (15,544 men and 1,787 women) were included in the analyses (Figure S1).
Assessment of physical activity during leisure, commuting, and work
Details of the information collected for leisure-time physical activity have been described in the literature [18, 21, 22]. Participants were asked to choose up to 3 activities among a list of 20 exercise or sports activities and the frequency (times per month) and duration of time per occasion (minutes) for each activity. If participants engaged in activities not listed in the questionnaire, they were instructed to choose an activity of similar intensity from the list. Of the 20 exercise or sports activities in the list, one activity named ‘‘Other’’ was not used for further analysis.
The metabolic equivalent (MET; 1 MET=1 kcal per h per kg of body weight) value for each activity was determined according to Ainsworth’s compendium of physical activities [23]. Of the 19 activities, 12 (walking not for work or commuting, walking fast not for work or commuting, golf practice, golf, baseball, softball, bike cycling, table tennis, pang pong, badminton, muscle strength training, and radio gymnastics) were classified as moderate activities (3 to 5.9 MET), and 7 (light jogging [approximately 6 min/km], jogging, swimming, soccer, tennis, aerobics, and jump rope) were classified as vigorous activities (≥6 MET). Leisure-time physical activity was defined as the product of intensity (MET) and duration of exercise (h), and the calculated MET-h per week of each individual was placed into one of the 4 categories—inactive (0 MET-h), low (>0 to <7.5 MET-h), moderate (≥7.5 to <16.5 MET-h), or high (≥16.5 MET-h)—which roughly accords with the classification of current physical activity guideline [24].
Occupational physical activity was assessed by the question To what extent is your work physically demanding? with the following response options: mostly sedentary, mostly standing, walking often, or fairly active. We combined the 2 categories in the middle (i.e., mostly standing and often walking) to increase the statistical power.
Commuting physical activity was assessed by the self-reported duration of walking to and from work (in minutes) and categorized as <20, 20 to <40, and ≥40 min for the analysis, which is similar to categories in other studies in Japan [18, 25].
Ascertainment of chronic kidney disease cases
CKD was assessed by using the data of annual health check-ups from baseline to March 2019 and defined as an eGFR of < 60 mL/min/1.73 m2 and/or proteinuria (+, 2+, or 3+ on dipstick) [19]. eGFR was calculated by using the following formula established by the working group of the Japanese CKD Initiative: eGFR (mL/min/1.73 m2) = 1.94 × (serum creatinine)−1.094 × (age)−0.287 × (0.739 if female) [26]. Serum creatinine was measured by the enzyme method with an autoanalyzer (Hitachi 7600, Japan). Proteinuria was tested by dipsticks and by using an autoanalyzer (Siemens Healthcare, Japan) and categorized as negative, ±, 1+, 2+, and 3+ (corresponding to protein levels of undetectable, trace, 30 mg/dL, 100 mg/dL, and ≥300 mg/dL, respectively). The date of check-up when CKD was first identified was the incidence date of CKD.
Covariates
The covariates included eGFR, age, sex, smoking status, alcohol consumption, job position, overtime work, shift work, commuting mode, and sleep duration, hypertension, diabetes, history of cardiovascular disease, dyslipidemia, hyperuricemia, and body mass index (BMI) at baseline. We refer to the online Supplementary Appendix 1 for data collection methods, which have been described in previous papers [18, 22, 27].
Statistical analysis
We calculated person-years of follow-up for each participant, from the date of the baseline health check-up to the date of health check-up when the CKD was first identified or the date of the last health check-up, whichever came first. We ran a Cox proportional hazards regression to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) for the association between leisure-time, occupational, and commuting physical activity and time to incident CKD. We adjusted for covariates in a stepwise manner.
Model 1 included baseline eGFR, age (continuous, years), and sex. Model 2 additionally adjusted for smoking status (never, former, or current), alcohol consumption (0, >0 to <2, or ≥2go/day), job position (high or low), overtime work (<45, 45 to <60, 60 to <80, 80 to <100, or ≥100 h/month), shift work (yes or no), primary commuting mode (walking, bicycling, train/bus, or car/motorbike), and sleep duration (<5, 5 to<6, 6 to<7, or ≥7 h per day). Model 3 further adjusted for potential mediators, such as hypertension, diabetes, history of cardiovascular disease, dyslipidemia, hyperuricemia, and BMI (<18.5, 18.5 to <25.0, 25.0 to <30.0, or ≥30.0 kg/m2). Model 4 adjusted for other types of physical activity, which were mutually adjusted.
The trend association between leisure-time physical activity and risk of CKD was assessed by assigning the median dose of leisure-time physical activity in each category and treating this variable as continuous. For the trend for occupational physical activity, we assigned a score of 1–3 to sedentary work, walking or standing during work, and fairly active during work, respectively. For commuting physical activity, we assigned 10, 30, and 50 min to increasing categories of walking to and from work (<20, 20 to <40, and ≥40 min). We also conducted sensitivity analyses and excluded participants with <2 years of follow-up term for the aforementioned major analyses.
To test the effect modification by hypertension, diabetes, obesity (BMI<25 or ≥25 kg/m2), baseline eGFR (60 to 89 or ≥90 mL/min/1.73 m2), occupational (sedentary work or not), and commuting physical activity (<20 or ≥20 min), we conducted subgroup analyses in a fully adjusted model (model 4). Those subgroup analyses were repeated between occupational or commuting physical activity and CKD onset. The proportional hazards assumption was examined using Schoenfeld residuals, and all covariates agreed with the proportional hypothesis except for age and baseline eGFR.
The proportion of missing data for each covariate was as follows: smoking status (0.4%), job position (3.3%), monthly overtime work (3.6%), shiftwork (2.9%), primary commuting mode (0.1%), sleep duration (0.2%), diabetes (0.2%), dyslipidemia (0.02%), and hyperuricemia (0.04%). We performed multiple imputations by using the chained equation method, with multinomial logistic regression for imputation of smoking status, monthly overtime work, primary commuting mode, and sleep duration; and logistic regression for job position, shiftwork, diabetes, dyslipidemia, and hyperuricemia. The imputation used all the variables involved in all the analytic models, including the outcome variables of time-to-event and event status. The 20 imputed data sets were generated, and the results were combined by using Rubin’s rules. All statistical analyses were performed with Stata/MP version 16.1 (Stata Corp., College Station, Texas). The statistical significance was set at a 2-sided P value of 0.05 for all analyses.