Associations of sedentary behavior and physical activity with older adults’ physical function: an isotemporal substitution approach

Participants were drawn from a larger epidemiological study conducted in Matsudo city, Chiba Prefecture, Japan. Matsudo city is located 20 km north of Tokyo with approximately half a million residents in 2016. To recruit the participants, first a randomly-selected 3000 people aged 65 to 84 years from the registry of residential addresses were contacted by an invitation letter. Of these, 951 agreed to participate in the main study, and 349 took part in a sub-study, in which PA and physical function were assessed objectively. Participants received a book voucher as compensation. All participants provided written informed consent. The study was approved by the Waseda University Institutional Committee on Human Research (2013–265) and the Institutional Review Board of Chiba Prefectural University of Health Sciences (2012–042).

The participants’ PA was assessed using an accelerometer (Active style Pro HJA-350IT, Omron Healthcare, Kyoto, Japan). The detailed algorithm and validity of the accelerometer device have been described elsewhere [17‐19]. Briefly, the device (74 × 34 × 46 mm; 60 g) measures anteroposterior (x-axis), mediolateral (y-axis), and vertical (z-axis) acceleration signals. The integral of the absolute value of the accelerometer output during a 60-s interval was calculated and converted into the total energy expenditure. The device estimates the intensity of activity by METs using a built-in algorithm. The CSV data files from the accelerometer were downloaded by Omron health management software BI-LINK for PA professional edition version 1.0 and then the files were processed by custom software (Custom-written Macro program for compiling data). A previous study, in which METs for household and locomotive activities were calculated, reported a linear relationship between filtered synthetic accelerations with PA intensity [18]. Participants were asked to wear the accelerometer on their waist for at least 7 days —except when sleeping or during water-based activities (e.g., bathing, showering, and swimming). To be eligible, participants needed to wear the accelerometer for ≥4 days (including 1 weekend day), with at least 10 h/day of wear time each day [20]. Non-wear time was defined as at least consecutive 60 min of 0 counts per minute (cpm), with allowance for up to 2 min of some limited movement (<50 cpm) within those periods [20]. For those who met the inclusion criteria, the daily average time spent on SB (≤1.5 METs), LIPA (>1.5 to <3.0 METs) and MVPA (≥3.0 METs) were calculated. These MET levels have been used by previous studies examining functional decline among older adults [21, 22].

Physical function was assessed using five performance-based functional tests: hand grip strength (upper body strength), usual and maximum gait speeds (gait speed), timed up and go (TUG) (mobility), and one-legged stance with eyes open (balance). Hand grip strength was measured using a Smedley-type handgrip dynamometer (TKK5041, Takei Scientific Instruments Co., Ltd., Niigata, Japan). We chose this type of handgrip device (rather than the Jamar dynamometer), because it is the most widely used method for evaluating hand grip strength in Japanese health studies [23]. Participants stood with their arms hanging naturally at their sides holding the dynamometer with the grip size adjusted to a comfortable level. They were instructed to squeeze the handgrip as hard as possible. Participants performed one trial with the dominant hand (to the nearest 0.1 kg). Usual and maximum gait speeds were measured using an 11 m course. Participants started walking at their normal or maximum paces, and time to walk 5 m was measured, starting when the body trunk passed the 3-m mark and ending when the body trunk passed the 8-m mark. Gait speed was calculated as distance divided by walking time (m/s). Usual gait speed was measured twice, and the faster of the two results (to the nearest 0.1 m/s) was used. Maximum gait speed was measured once. TUG measured the time for participants to complete the following sequence: start from a seated position in an armless chair; stand up from the chair and walk as fast as possible toward a pole placed 3 m in front of the chair, turn around the pole; and return to the chair and sit. TUG was conducted twice, and the faster of the two results (to the nearest 0.1 s) was used. One-legged stance with eyes open was measured using a participant’s preferred leg. Participants raise one leg and stand as long as possible. They were timed until they lost their balance or reached the maximum of 60 s. Participants performed two trials, and the longer of the two results (to the nearest 0.1 s) was used.

The following individual-level variables were considered as potential confounders: age, gender, body mass index (BMI), the number of past illnesses, complications and comorbidity, smoking status, drinking status, living arrangement, and highest educational attainment. The BMI was calculated using objectively-measured height and weight.

Three multiple linear regression models including single-activity model, partition model, and IS model were used to examine the associations of SB, LIPA, and MVPA for each of the five items of physical function. We used 10 min as a unit for activity, thus the IS models examined the effect of replacing a 10-min of one activity with the same amount of another activity. This time unit was chosen for two reasons; this is the minimum amount of time in which activities should be accrued to meet current PA guidelines [7] and Japanese official PA guidelines for health promotion recommended trying to move for an additional 10 min a day for longer healthy life expectancy [24].

The single-activity model assessed each activity component separately (e.g., SB only), without taking into account the other activity types, but adjusting for total wear time and confounders. The model (in the case of SB) is expressed as follows:

Outcome variable = (b0) SB + (b3) total wear time + (b4) covariates.

The partition model examined all the behaviors simultaneously, without adjusting for total wear time. It is expressed as follows:

Outcome variable = (b0) SB + (b1) LIPA + (b2) MVPA + (b4) covariates.

In this model, the coefficient for one type of activity represents the effect of increasing this type of activity while holding the other activities constant. Since the total wear time is not included in the model (thus is not held constant), it represents the effects of adding rather than substituting an activity type.

The IS model estimates the effect of substituting one activity type with another for the same amount of time (e.g., replacing MVPA with SB, by removing SB from the model). The IS model (in the case of omitting SB from the model) is expressed as follows:

Outcome variable = (b1) LIPA + (b2) MVPA + (b3) total wear time + (b4) covariates.

The coefficients b1 and b2 in this model represent the effect of a 10-min substitution of SB with one of the activity types (LIPA or MVPA) while holding the other activity types and total wear time constant. For instance, b1 can be interpreted as the effect of replacing SB with LIPA for 10 min while holding MVPA and total wear time constant.

All the statistical contrasts were made at the 0.05 level of significant. Analyses were conducted using IBM SPSS Statistics 20.0 (IBM Japan Corp., Tokyo, Japan).