Participants
The British Regional Heart Study (BRHS) is a prospective cohort of 7735 men recruited from a single local primary care centre in 24 British towns in 1978-80 (age 40-59 years). In 2010-2012, all surviving cohort members resident in the UK (n = 3137) were invited to attend a further physical examination including measurements of weight and height and to participate in a study of objectively measured physical activity. The National Research Ethics Service (NRES) Committee for London provided ethical approval. Participants provided informed written consent to the investigation, which was performed in accordance with the Declaration of Helsinki [13].
Objective physical activity assessment
Procedures for distribution and wearing
Participants were invited to attend an assessment by study nurses at their local primary care centre. All men who attended were asked to wear an Actigraph GT3x accelerometer (Pensacola, Florida) over the right hip on an elasticated belt for 7 days, during waking hours, removing it for bathing, swimming or showering and returning the device by post.
Data processing
Actigraph accelerometers record “counts” and steps, which both depend upon the frequency and intensity of the raw acceleration [14]. Accelerometer data were processed using standard methods, as described previously [9]. In brief, raw data from movements registering on the vertical axis were integrated into 60 s epochs; therefore counts per minute (CPM) were derived. Non-wear time was identified and excluded using the R package “Physical Activity” [15], based on (i) periods of continuous zero activity lasting more than 90 min or (ii) periods of zero activity lasting more than 90 min broken only by non-zero counts lasting up to 2 min, provided no activity counts were detected during both the 30 min before and after that interval [9]. Valid wear days were defined as ≥600 min wear time, and participants with at least 3 valid days were included in analyses, a conventional requirement for estimating usual PA level [16].
Derived variables
The number of minutes per day spent in SB, LIPA and MVPA was categorised using count-based intensity threshold values of counts per minute developed for older adults [9, 11, 17]: <100 CPM for sedentary behaviour (<1.5 Metabolic Equivalent of Task, MET), 100-1040 for light activity (1.5-3 MET) and >1040 for MVPA,(≥3 MET). The cut-point of 1040 CPM was calibrated to identify moderate intensity activities (≥3 MET) in a sample of older adults [11], but we also investigated the more widely used cut-point of 1952 CPM which was calibrated to identify moderate intensity activities (≥3 MET) in middle-aged adults [17]. Two further summary measures of SB and MVPA were calculated: number of sedentary bouts of at least 1 h (a period of 60 or more consecutive minutes where the accelerometer registers <100 CPM) and MVPA bouts of at least 10 min (a period of 10 or more consecutive minutes where the accelerometer registers more than 1040 CPM).
Log diary and questionnaire data
Participants completed a log diary, detailing when the accelerometer was put on and taken off during the seven days of wear. During the first 3 days the men were also asked to report the type of activity (e.g. housework, gardening, preparing meals, watching TV) that they did during each hour of the day. Participants’ log diaries were checked and matched against accelerometer data to verify the date on which they started wearing the accelerometer. Age and season were derived from the first wear day. Season was categorised as summer (Jun-Aug), autumn (Sep-Oct), winter (Nov-Feb), and spring (Mar-May). Men were asked “do you have any difficulties getting about outdoors?” which was grouped as “none”, “slight” and “moderate, severe and unable to do”. Men reported a medical diagnosis of any of the following chronic conditions; heart attack, heart failure, angina, diabetes, stroke, osteoporosis, claudication, Parkinson’s disease and chronic kidney disease. Men were classified as having vision problems if they had one or more of glaucoma, macular degeneration or cataract, as in advanced age these are primary causes of visual dysfunction [18, 19]. Men scoring > =2 on the 4-item Geriatric Depression score were classified as depressed [20]. Cigarette smoking was self-reported. Participants completed the Lubben scale of social isolation which asks about interactions with family members and with friends, men scoring <12 were classed as at risk of social isolation [21]. Participants reported which forms of transport they used regularly (car, public transport, dial a ride, walk or cycle), those who reported regular walking or cycling were classified as using active transport. The social class classification was based on the longest-held occupation of subjects reported at study entry in 1978–1980. Participants’ occupations were grouped as non-manual or manual.
Statistical methods
Analyses were carried out using STATA/SE 13 [22] and MLwiN Version 2.02 [23]. To give a general overview of the within day variation of total PA, counts per minute and steps were plotted against hour of day. The main outcome variables were the proportions (percentages) of the day spent in (1) sedentary behaviour, (2) light PA and (3) moderate to vigorous PA. Each outcome was calculated according to hour of the day, with the number of minutes that the accelerometer was worn in that hour used as the denominator. Due to sparse data in early morning and late evening, we examined the mean activity counts per hour between 7.00 am and 10.59 pm. Only hours with ≥ 45 valid wear minutes were included. A first descriptive analysis was undertaken: for 1329 men with complete data, the percentage of time spent in SB, LIPA and MVPA was plotted against hour of day. In order to explore whether the diurnal patterns were modified by selected variables, the data were stratified by (i) age group (<75, 75-79, ≥80 years) (ii) mobility limitations (none, slight, moderate/severe/unable to do) (iii) number of chronic conditions (none, 1-2, ≥3) (iv) BMI category (<25, 25-30 and ≥30 kg/m2) (v) depression (depressed vs not) (vi) smoking status (current smoker vs not) (vii) social isolation (at risk of isolation vs not at risk) (viii) social class (manual vs non-manual) (ix) use of public transport (walk/cycle vs car/public transport) (x) vision problems (yes vs no) (xi) season (winter vs summer) and (xii) weekend vs weekday.
The distributions of each outcome were investigated: percentage of MVPA distribution was highly positively skewed as reported in previous studies [24, 25]. MVPA data were highly over-dispersed with variance 5 to 6 times higher than the means within each period of the day, so negative binomial model were used to investigate which factors were related to the percentage of time spent in MVPA and the results were reported as rate ratios (RRs) [26]. A RR is as multiplicative factor: any deviation from 1 indicates a percent difference in the outcome relative to the respective reference category (baseline) in the exposure variable. Linear multilevel regression models were used to investigate which factors were related to the percentage of time spent in LIPA and SB (normally distributed). Beta coefficients were reported to estimate the difference in time spent in SB and LIPA between the categories of each explanatory variable against the reference. In all multilevel models Level 1 was period of the day [morning (7 am-12.59 pm), afternoon (1 pm-6.59 pm) and evening (7 pm-10.59 pm)] and Level 2 was the individual. Each period (morning, afternoon and evening) had a minimum of 2 valid hours of wear time.
Two level random-intercept and random-slope models were used with adjustments for age, season, region and part of the day, and one additional covariable at a time (mobility limitations, number of chronic conditions, BMI, depression, smoking status, social isolation, social class, use of public transport, and vision problems). Next, fully adjusted models were run using all explanatory variables together. The random slope in the models allowed us to test the hypothesis that the changes in PA levels over the day varied between different men. The estimated slopes over the course of day were reported as mean differences between afternoon and evening vs morning (baseline).
For variables which were significantly associated with sedentary behaviour, light and MVPA levels, interactions were fitted to test whether the associations differed according to period of the day (morning, afternoon, evening). An overall Wald test for interaction between the categories of those explanatory variables and period of the day (morning, afternoon and evening) was used.