Ethics statement

Taiwan Biobank received ethics approval from the Institutional Review Board on Biomedical Science Research (IRB-BM), Academia Sinica, Taiwan, and from the Ethics and Governance Council (EGC) of Taiwan Biobank, Taiwan. Written informed consent was obtained from each participant in accordance with requirements and the principles of the Declaration of Helsinki. The present study has been approved by the research ethics committee of Taichung Veterans General Hospital (no. CE20023A) and Taiwan Biobank (no. TWBR10901-02).

Data source Taiwan Biobank Cohort

The Taiwan Biobank enrolled community-based cohort consisting of ~ 120,000 volunteers (as of April 30, 2020) between 30 and 76 years of age with no history of cancer. The Taiwan Biobank is governed by the Ethics and Governance Council (EGC) of the Taiwan Biobank and the Ministry of Health and Welfare, Taiwan.

Each participant completed a physical examination as well as a questionnaire that addressed personal information and lifestyle factors by a face to face interview. Regarding exercise, the questionnaire asks participants whether they have a regular exercise habit (at least 3 times per week, more than 30 min each). If the answer is yes, then the questionnaire asks to list the 3 most frequently performed exercises and their frequency and duration. Based on this information, we calculated metabolic equivalent of tasks (METs)/week of each participants using a conversion formula between types of exercise and the METs. The conversion formula in Taiwan Biobank is shown elsewhere [25]. Participants were classified into 3 groups based on their exercise status (those without regular exercise habit, those with regular exercise habit and intensity less than 20 METs/week, and those with regular exercise habit and intensity more than 20 METs/week).

Study population

The present study extracted data from Taiwan Biobank data cycles. Adults aged > 18 years, had complete information of dual-energy X-ray absorptiometry (DXA) measures on BMD, sleep hours, and components to calculate metabolic equivalent of tasks (METs) were eligible for inclusion. Participants without complete data of body mass index (BMI) were excluded from the primary cohort.

Assessment of sleep duration

Duration of sleep was captured by a single question in the Taiwan Biobank database: “In the past month, how many hours did you sleep on average on weekdays”. Participants were categorized into three groups according to their sleep hours by < 6.5 h/day, 6.5–8 h/day, and ≥ 8 h/day.

Ascertainment of osteoporosis

The study subjects’ BMD at the total femoral areas were measured through DXA scan (Hologic, Bedford, MA, USA). Quality control was routinely conducted on all DXA machines. We categorized the study cohort into with or without osteoporosis, defined as a BMD T-score equal to or less than 2.5 standard deviations (SD) at either the total femur, femoral neck or lumbar spine L1 ~ L4, using reference values from the young Caucasian populations of the same sex according to the WHO criteria.

Assessment of physical activity level

In this study, we used the definition proposed by Ford et al. and the US Department of Health and Human Services for physical activity, which is the sum of the product of the time spent weekly in each activity reported by the participants multiplied by the metabolic equivalent of task (MET) value for that activity, yielding a MET index [26]. One MET was the energy expenditure for every 1 kcal/kg of body weight per hour. Participants were classified into 3 groups based on their METs/week (0; less than 20 METs/week; and ≥ 20 METs/week) in accordance with a previous study [25].

Demography and covariates

Data of participants’ age and gender were obtained from standard questionnaires through in-person interviews conducted by trained interviewers.

Body mass index (BMI) value was obtained from the Taiwan Biobank examination measurements, calculated as body weight (kilograms) divided by height (meters squared). Body weight was measured using an electronic load cell scale, and standing height was measured with a fixed stadiometer. In addition, body fat percentage, waist circumference, hip circumference, systolic blood pressure (SBP) and diastolic blood pressure (DBP) were also extracted from the Taiwan Biobank examination data file.

Level of HbA1c, fasting plasma glucose, serum total cholesterol, triglycerides, high density lipoprotein-cholesterol (HDL-c), low density lipoprotein-cholesterol (LDL-c), GPT, ALT, serum creatinine and uric acid were obtained through the laboratory data.

Diabetes was defined by the question of “Other than during pregnancy, have you ever been told by a doctor or health professional that you have diabetes?” or HbA1c ≥ 6.5% or fasting glucose > 7 mmol/L in laboratory measurements.

GFR was estimated from re-calibrated serum creatinine using the 4-variable Modification of Diet in Renal Disease (MDRD) Study equation. Here we use the IDMS-traceable MDRD Study equation that uses standardized creatinine: GFR = 175 × (standardized serum creatinine)−1.154 × (age)−0.203 × 0.742(if the subject is a woman) × 1.212(if the subject is black). Participants were categorized in to having chronic kidney disease (CKD) with an estimated GFR < 60 ml/min/1.73 m2.

Statistical analysis

Continuous variables are presented in mean and standard deviation (SD). Categorical variables are presented in number and proportion. Analysis of variance (ANOVA) was used for continuous variables to examine the differences of baseline characteristics among the groups of different physical activity. Chi-square test was performed to examine differences in the proportions among the groups of different physical activity. Generalized linear model (GLM) was used to test the joint effects of sleep duration and physical activities on T-score of total femoral area. Logistic regression presented as odds ratio (OR) and 95% confidence interval (CI) was used to test the joint effects of sleep duration and physical activities on risk of osteoporosis. Trend test was also tested by GLM or Logistic regression. All reported p-values were bidirectionally < 0.05 denoting statistical significance.