Study population and design
This is a three repeated measure population-based cross-sectional study. From July 2019 through September 2020, we recruited 3 batches of community controls to compare their patterns of sleep disturbance and circadian rhythm with silicotic patients, and all participants were male. The community controls recruitment was conducted in collaboration with five non-governmental organizations and seven district council members, which located in different area of Hong Kong, including Kwun Tong, Kowloon City, Tsuen Wan, Sham Shui Po and Kwai Tsing Districts. As the recruitment of community subjects covered different periods of the COVID-19 outbreak, this study therefore provides a unique opportunity to investigate the dynamic changes in people’s physical activity and rest-activity circadian rhythm adversely impacted by the outbreak. As shown in Fig. 1, the first recruitment of 106 community subjects was carried out between 2 July 2019 and 1 August 2019, which was before the first positive COVID-19 case was reported in Hong Kong on 23 January 2020. The second recruitment of 66 community subjects were conducted between 23 June 2020 and 9 July 2020, just after the 2nd waves of COVID-19 which had lasted until May 2020 [2]. The third recruitment of 70 community subjects was performed between 15 September 2020 and 29 September 2020, which was during the late phase of the 3rd wave of COVID-19 according to the Hong Kong Government data. Overall, a total of 242 community-dwelling older male participants were included from 3 batches of recruitment covering the period before and during different periods of COVID-19. We excluded participants who have physician diagnosed mental health disorders or medical conditions that prevented them from completing the survey such as hearing problems.
Trained interviewers conducted a face-to-face interview with each participant using standardized questionnaires to obtain information on socio-demographic characteristics, tobacco smoking, alcohol drinking, history of medication uses, physical activity habit and occupational history. We also obtained participants’ anthropometric data through direct measurement including height, weight, and waist circumstance during the interview, where weight measurement was obtained by body composition monitor (TANITA corporation, BC-545 N) according to a standard protocol. Participants were requested to wear light clothes and without shoes during the measurement, and their rest-activity circadian rhythms were measured using actigraphy for seven consecutive 7 days (168 h). This study was approved by the Joint Chinese University of Hong Kong-New Territories East Cluster Clinical Research Ethics Committee, and all participants signed written informed consent forms before the survey was conducted.
Outcome measurements and procedure
Physical activity
We adopted the validated International Physical Activity Questionnaire (IPAQ)-short form to collect detailed information of participants’ physical activity habit, including the frequency and duration of vigorous-intensity and moderate-intensity physical activity, and walking during the past week. We further recorded the types of their physical activity such as hiking, weight lifting, jogging, and tai chi. We matched the physical activity information of our participants with metabolic equivalents (METS) levels obtained from the 2000 and 2011 compendium of physical activities [23, 24], and calculated their total weekly physical activity levels as duration (minutes) × frequency per week (days) × METS intensity. We further categorized their physical activity levels into low and high according to the median METS of our population. A METS refers to the resting metabolic rate, which is the amount of oxygen consumed while sitting at rest [25].
Rest-activity circadian rhythm
Each participant was requested to wear a GENEActiv Original (Activinsights Company, UK) device on his non-dominant wrist continuously for 7 days (168 h) with measurement frequency of 100 Hz and a sampling rate of 1 min. The assessment of circadian rhythm parameters had been described previously [26]. Briefly, the actigraphy detected and recorded individual movements in three mutually vertical axes (x, y, and z). A gravity-subtracted sum of vector magnitudes (SVM) was automatically calculated with data of these three axes and a formula defined by the manufacturer: SVM = [(x2 + y2 + z2)½ - 1 g] [27]. The SVM data were then imported into the Chronos-Fit program (v. 1.06) [28] to facilitate computing four rest-activity circadian rhythm parameters, namely midline statistic of rhythm (MESOR), amplitude, acrophase and percent rhythm by an extended cosine model [29]. MESOR refers to the adjusted mean levels of the rhythm. Amplitude refers to the magnitude of the rhythm cycle (i.e. the distance between the peak and MESOR). Acrophase refers to the time in the cycle of the daily peak rhythm; earlier time of peak rhythm suggests an advanced acrophase and later peak rhythm time suggests a delay acrophase. Percent rhythm refers to the percentage of variation in the data that is explained by the fitted model; a low percent rhythm suggests a dampened rhythm [30]. We also coded the SVM into one of three intensity categories: sedentary (SVM < 158.5), light (SVM = 158.5–261.8), moderate or vigorous activity (SVM > 261.8) according to previously validated cut-off for GENEActiv accelerometers among adults (mean aged =59.6 ± 5.5 years) [31].
Non-wearing time was determined by reviewing the SVM readings outputted from the GENEActiv software and the self-reported sleep log data collected from the interviewees. The non-wearing periods should present low and steady SVM readings, and we excluded these data from the calculation of the parameters. Only participants with a total length of wearing time more than 120 h (5/7 consecutive days of 168 h) that covered weekends records were included in the study. All rest-activity circadian rhythm parameters were stratified based on their median distribution, and categorized into low and high levels for MESOR (< 270.25/> 270.25), amplitude (< 132.56/> 132.56) and percent rhythm (< 18.09/> 18.09), and for acrophase it was classified as advanced and delayed acrophase (peak activity before 2:04 pm / after 2:04 pm).
Statistical analysis
We performed independent t-tests and chi-square test to compare the differences in basic sociodemographic characteristics, levels of physical activity and circadian rhymes for community subjects completed the interview during different periods of COVID-19 (i.e., before COVID-19, between the 2nd and 3rd waves of COVID-19, and during the late phase of the 3rd waves of COVID-19), using the category “before COVID-19” as the reference group. We performed multivariate unconditional logistic regression model to estimate the odds ratios (OR) and the 95% confidence intervals (95%CI) to examine the period effect of COVID-19 (i.e., before COVID-19, between the 2nd and 3rd waves of COVID-19, and during the late phase of the 3rd waves of COVID-19) on participants’ physical activity levels (METS/minutes) (low/high) and rest-activity circadian rhythm parameters [i.e. MESOR (low/high), amplitude (low/high), acrophase (advanced/delayed) and percent rhythm (low/high)] before and during different periods of the outbreak, using the high-level category of physical activity, MESOR, amplitude and percent rhythm, and advanced acrophase as the reference group. Potential confounders included in the multivariate logistic regression were age at recruitment (continuous), educational attainment (primary education or below, secondary education or above), employment status (retired, full-time/part-time job), and body mass index (BMI) [body weight (kg)/height2(m2)] [underweight/normal weight (BMI < 25), overweight, obesity (BMI > 25]. All statistical analyses were conducted with SPSS 26.0 for Windows (SPSS, Chicago, IL, USA), and a two-sided p-value of less than 0.05 was considered statistically significant.