Longitudinal changes in physical function and physical activity in older adults

Longitudinal changes in physical function and physical activity in older adults Abstract Background physical function (PF) and physical activity (PA) both decline as adults age and have been linked to negative outcomes, including dementia, depression and cardiovascular diseases. Although declines in each are associated with numerous negative outcomes, the longitudinal relationship between these two measures is unclear. Objective to examine the dynamic, bidirectional associations between declines in PF and PA. Design prospective cohort. Setting the Monongahela–Youghiogheny Healthy Aging Team (MYHAT) study. Subjects about 1,404 men and women, 76.96 ± 7.2 years, 62.4% female and 95.2% white. Methods over nine annual assessment cycles, PF was evaluated via the timed Up-and-Go task and PA via a self-reported questionnaire. Piecewise latent growth models examined bidirectional associations between PA and PF to determine whether the initial values (intercept) or early slope (cycles 1–5) (in either PF or PA) predicted later slope (cycles 5–9) (in either PF or PA). Results initial PF significantly predicted early (standardised β= −0.10, P < 0.001) and later (standardised β= −0.09, P = 0.01) PA slopes. Initial PA significantly predicted later (standardised β = −0.09, P = 0.04) but not early PF slope. Associations were independent of baseline memory test scores, baseline cognitive status, later cognitive status and age. Early physical function slope neither predicts later PA slope nor did early PA slope predict later PF slope (both P values >0.10). Conclusions the relationship between PF and PA is bidirectional, with PF more consistently predicting declines of PA, both in the short- and long-term. Intervening on PF impairments may improve PA engagement, which could in turn promote PF and translate to beneficial effects on cognitive function, cardiovascular health and mood. ageing, physical function, physical activity, older people Introduction With a life expectancy of older adults rising, there has been increasing emphasis on maximising the quality of life, including maintenance of physical function (PF) and physical activity (PA) [1]. PF is often characterised by measures of gait speed or gait slowing over time [2–5]. PA is less consistently defined, especially in older adults, with some studies using either pedometers or accelerometers [6, 7], self-reported total sedentary time [8, 9], or time spent walking, strength training, or conducting balance or stretching exercises [10–13]. Notably, walking is often the only form of PA reported in older adults [7, 10]. Gait slowing, a common and important problem in older adults, reflects multi-system impairments and often precedes the onset of physical disability, falls and cognitive impairment [2–5, 14, 15]. Similarly, PA declines in intensity and duration with age [7], and predicts reduced cardiovascular health [9], a faster decline in cognitive function [8], impaired mood, increased social isolation [16], increased mortality and a greater risk for diagnosis of any new disease [17]. PF contributes significantly to the associations between PA and mortality and disease, indicating a close interplay between PF and PA [17]. Despite a wealth of research to promote PA [6, 10, 11, 18] and to some extent PF [12, 13] in older adults, the joint association has rarely been examined. Declining PF likely lowers PA engagement but declining PA may lessen conditioning, in turn reducing PF. Thus, interventions to improve PF could augment effects of concurrent interventions to promote PA, and vice versa, transferring individuals from a vicious cycle of lower PF leading to lower PA, to a virtuous cycle where higher PF contributes to higher PA. Maintaining higher PF late in life is associated with lower risk of mortality and disability [5, 19]. Similarly, increased PA is associated with better cardiovascular health [9], cognitive function [8], mood and social support [16]. Interventions promoting PA can also promote PF [12, 20–22]. Few studies have examined the association of PF predicting PA. We recently showed declining walking speed over 4 years among high-functioning, community-dwelling older adults significantly predicted declining PA engagement over the subsequent 5 years but PA did not predict walking speed [23]. Here, we aim to extend this line of inquiry to a less highly selected and more representative samples of older adults with a greater range of co-morbidities, by testing the bidirectional associations between PF and PA over time. Methods Study population The Monongahela–Youghiogheny Healthy Aging Team (MYHAT) is a random population sample drawn from publicly available voter registration lists for a small town region of Pennsylvania (USA) [24]. Sampling, recruitment and cohort characteristics have been previously reported [24]. Briefly, participants were 65 years or older. Exclusion criteria were either residing in a long-term care facility at recruitment, substantial sensory or cognitive impairment (age-education-adjusted Mini-Mental State Examination (MMSE) <21), or decisional incapacity [24]. Participants were enroled between 2006 and 2008 (cycle 1), and followed for eight subsequent annual visits, with cycle 9 occurring between 2014 and 2016. This study was approved by the Institutional Review Board at the University of Pittsburgh and all participants provided written informed consent. Analytic cohort Our analytic sample was defined, a priori, as those who had baseline measures of PF and PA, plus at least two follow-up measures of each variable. Of the original 1,982 MYHAT participants, 578 were unable to perform the PF test, or had missing baseline PF data, or missing PF data from >6 subsequent visits. In addition, 472 participants, not mutually exclusive from those missing PF data had missing PA data (baseline or >6 subsequent visits). Consequently, our analytic sample consisted of 1,404 participants. Physical function PF was assessed with the timed Up-and-Go task (TUG) test [25], measured in seconds. Participants were instructed to stand up from a chair, if possible without using arms for support, to walk 10 feet forward, 10 feet back at their usual pace and return to the seated position. Physical activity PA was self-reported via a questionnaire which assessed the number of days per week, and average minutes per day of walking a self-determined level of moderate-intensity (see Appendix 1 in the supplementary data, available at Age and Ageing online). Total minutes per week were computed from total days per week and total minutes per day. Covariates At baseline, self-reported age, sex, race (white or non-white) and education (≤high school (vs) >high school) were collected. Self-reported weight (pounds) and height (inches) were recorded at cycle 1, and used to calculate body mass index (weight(lb)/[height(in)]2 × 703, BMI). Self-reported history of medical diagnoses determined participant’s history of stroke or transient ischaemic attack (TIA), type 1 or type 2 diabetes, hypertension, heart attack or myocardial infarction (MI) and heart failure. Participants self-reported any painful gait (yes/no), smoking status (ever (vs) never and consumption of alcoholic beverages (current drinker consuming ≥1 alcoholic drinks per week (vs) never drinker or current drinker, drinking ≤1 drink per week). The MMSE [26] was administered at each cycle and scored on a scale of 0–30. A thorough cognitive evaluation was conducted for all participants using the Clinical Dementia Rating (CDR) scale [27, 28], participants were scored as having no dementia (CDR = 0), mild cognitive impairment (MCI; CDR = 0.5) and mild, moderate or severe dementia (CDR ≥ 1). Statistical analyses Raw mean standard deviation (SD) PF and PA variables were visually examined to see how they changed over time and tested for linear changes over time with the Shapiro–Wilk test. The associations between baseline characteristics and early and late PF and PA slopes were investigated using t-tests or analysis of variance (ANOVA) for categorical variables and Pearson’s correlations for continuous variables. Our primary analytic approach investigated longitudinal, bidirectional associations between PF and PA, using piecewise latent growth curve modelling [29], with the bootstrapping method (5,000 re-samples) for standard errors (SE) to account for asymmetric abnormalities in both distributions. This approach estimated an initial value (intercept of PF or PA at cycle 1), plus one slope of ‘early’ PA or PF (cycles 1–5) and one slope of ‘late’ PA or PF (cycles 5–9). By defining both early and late slopes to include cycle 5, it allowed cycle 5 to serve as a hinge, and the slope before and after cycle 5 to be different, without ignoring change occurring between cycles 4 and 5. Initial values and slopes were estimated in parallel, allowing simultaneous examination of all possible bidirectional associations. All models were adjusted for initial values of the outcome. A secondary analytic approach was specified a priori; the goal of this analysis was to, as closely as possible, match the analytic approach used in our previous work, which contained three-time points of PF and PA [23], to determine if results were consistent. This approach estimated raw PA and PF change scores, using only data from cycles 1, 5 and 9, to define ‘early change’ (cycles 5–1) and ‘late change’ (cycles 9–5). Change scores were computed as latent variables in piecewise latent growth curve models, testing all possible associations between PF and PA, and estimating SE using 5,000 bootstrapped re-samples. Baseline PF and PA were used as covariates in the models. For primary and secondary analytic approaches, models were first constructed without covariates and then with covariate adjustment. All analyses included 1,404 participants; missing follow-up data were handled using the maximum likelihood estimator (see Appendix 2 in the supplementary data, available at Age and Ageing online). Sensitivity analyses were conducted with varying degrees of missing data to examine how missing visits may have affected associations, and with upper limits placed on PA, to determine if extreme values of PA were influencing associations were also assessed. Levels of statistical significance were defined a priori as P values ≤0.05. Analyses were performed with Mplus, 7.3 [30] and SAS version 9.4 (SAS Institute Inc., Cary, NC). Results Of the 1,404 participants in our analytic sample, 32.3% and 45.1% had all nine assessments of PF and PA, respectively, compared to 13.0% and 10.9% with only three assessments of each. The number of participants at each cycle with complete data and the mean (SD) are reported in Appendix 3, available in Age and Ageing online. Overall, PF and PA both decreased, linearly, over time, as exemplified by an increase in time to complete the TUG task and less time spent walking, respectively (see Appendix 3 in the supplementary data, available at Age and Ageing online). Average participant baseline age was 76.96 ± 7.2 years, 62.4% were female, 95.2% were white and 43.4% had greater than a high school education. Average MMSE score at baseline was 27.84 ± 1.99 points, six participants (0.4%) had dementia (CDR ≥1) and 338 (24.1%) had MCI (CDR = 0.5) (Table 1). Compared to participants in the analytic sample, those excluded from analyses differed significantly on age, education and other factors indicating those excluded were less healthy (see Appendix 3 in the supplementary data, available at Age and Ageing online). Table 1. Descriptive characteristics, and associations with early and late PA and PF slopes. Participant characteristics Analytic samplea Early PA slopeb Late PA slopeb Early PF slopeb Late PF slopeb Cycle 1 n = 1,404 (Cycles 1–5) (Cycles 5–9) (Cycles 1–5) (Cycles 5–9) Age 76.96 (7.2) 0.01 0.05* 0.25** −0.07* Sex, female 876 (62.4%) −0.64 −1.41 −0.27 0.23 Race, white 1337 (95.2%) −1.52 0.52 −0.12 1.25 Education, >HS 609 (43.38%) 0.21 0.31 9.46** 1.72 MMSE score 27.84 (1.99) 0.05 0.01 0.01 0.02 CDR score, MCI (score = 0.5)# 338 (24.1%) 0.24 0.08 16.48** 3.44* Painful gait, present 158 (11.3%) −0.80 0.20 0.13 0.47 ≥1 Alcohol drinks/week 324 (23.1%) −0.30 0.26 1.26 0.34 Ever smoker 734 (52.3%) −0.64 −0.37 2.04* 0.32 BMI 28.06 (5.40) 0.05 0.06* 0.02 0.05 Stroke/TIA, present 153 (10.9%) 0.52 −0.66 −2.71* 0.75 Myocardial infarction, present 192 (13.7%) 0.52 0.65 −1.88 −0.61 Hypertension, present 900 (64.1%) 0.53 0.52 −2.06* −1.31 Diabetes, present 289 (20.6%) −0.30 −0.97 −1.38 −0.59 Heart failure, present 99 (7.1%) 2.28* −1.16 −1.83 0.04 TUG time, sec 12.75 (4.06) −0.01 0.04 −0.02 −0.23** PA, min/week 66.45 (115.12) −0.56** −0.33** −0.04 −0.04 Cycle 9 TUG time, sec 14.40 (4.31) 0.03 −0.02 0.24** 0.67** PA, min/week 33.61 (73.43) 0.12* 0.26** −0.07* −0.14** MMSE score 28.07 (3.86) 0.06 0.06 −0.04 −0.04 Worsening CDR score, present 120 (8.6%) 1.79 1.10 −3.76** −3.59** Participant characteristics Analytic samplea Early PA slopeb Late PA slopeb Early PF slopeb Late PF slopeb Cycle 1 n = 1,404 (Cycles 1–5) (Cycles 5–9) (Cycles 1–5) (Cycles 5–9) Age 76.96 (7.2) 0.01 0.05* 0.25** −0.07* Sex, female 876 (62.4%) −0.64 −1.41 −0.27 0.23 Race, white 1337 (95.2%) −1.52 0.52 −0.12 1.25 Education, >HS 609 (43.38%) 0.21 0.31 9.46** 1.72 MMSE score 27.84 (1.99) 0.05 0.01 0.01 0.02 CDR score, MCI (score = 0.5)# 338 (24.1%) 0.24 0.08 16.48** 3.44* Painful gait, present 158 (11.3%) −0.80 0.20 0.13 0.47 ≥1 Alcohol drinks/week 324 (23.1%) −0.30 0.26 1.26 0.34 Ever smoker 734 (52.3%) −0.64 −0.37 2.04* 0.32 BMI 28.06 (5.40) 0.05 0.06* 0.02 0.05 Stroke/TIA, present 153 (10.9%) 0.52 −0.66 −2.71* 0.75 Myocardial infarction, present 192 (13.7%) 0.52 0.65 −1.88 −0.61 Hypertension, present 900 (64.1%) 0.53 0.52 −2.06* −1.31 Diabetes, present 289 (20.6%) −0.30 −0.97 −1.38 −0.59 Heart failure, present 99 (7.1%) 2.28* −1.16 −1.83 0.04 TUG time, sec 12.75 (4.06) −0.01 0.04 −0.02 −0.23** PA, min/week 66.45 (115.12) −0.56** −0.33** −0.04 −0.04 Cycle 9 TUG time, sec 14.40 (4.31) 0.03 −0.02 0.24** 0.67** PA, min/week 33.61 (73.43) 0.12* 0.26** −0.07* −0.14** MMSE score 28.07 (3.86) 0.06 0.06 −0.04 −0.04 Worsening CDR score, present 120 (8.6%) 1.79 1.10 −3.76** −3.59** aMean (SD) presented for continuous and N(%) presented for categorical variables. bT-tests (t-statistics shown) or ANOVA (F-statistics shown) used for mean comparisons for two- and three-level categorical variables, respectively; Pearson’s correlations reported for continuous variables. *Indicates P value ≤0.05. **Indicates P value ≤0.001. #CDR score indicating cognitively normal n = 1,060 (75.5%) and CDR score indicating dementia n = 6 (0.04%). Table 1. Descriptive characteristics, and associations with early and late PA and PF slopes. Participant characteristics Analytic samplea Early PA slopeb Late PA slopeb Early PF slopeb Late PF slopeb Cycle 1 n = 1,404 (Cycles 1–5) (Cycles 5–9) (Cycles 1–5) (Cycles 5–9) Age 76.96 (7.2) 0.01 0.05* 0.25** −0.07* Sex, female 876 (62.4%) −0.64 −1.41 −0.27 0.23 Race, white 1337 (95.2%) −1.52 0.52 −0.12 1.25 Education, >HS 609 (43.38%) 0.21 0.31 9.46** 1.72 MMSE score 27.84 (1.99) 0.05 0.01 0.01 0.02 CDR score, MCI (score = 0.5)# 338 (24.1%) 0.24 0.08 16.48** 3.44* Painful gait, present 158 (11.3%) −0.80 0.20 0.13 0.47 ≥1 Alcohol drinks/week 324 (23.1%) −0.30 0.26 1.26 0.34 Ever smoker 734 (52.3%) −0.64 −0.37 2.04* 0.32 BMI 28.06 (5.40) 0.05 0.06* 0.02 0.05 Stroke/TIA, present 153 (10.9%) 0.52 −0.66 −2.71* 0.75 Myocardial infarction, present 192 (13.7%) 0.52 0.65 −1.88 −0.61 Hypertension, present 900 (64.1%) 0.53 0.52 −2.06* −1.31 Diabetes, present 289 (20.6%) −0.30 −0.97 −1.38 −0.59 Heart failure, present 99 (7.1%) 2.28* −1.16 −1.83 0.04 TUG time, sec 12.75 (4.06) −0.01 0.04 −0.02 −0.23** PA, min/week 66.45 (115.12) −0.56** −0.33** −0.04 −0.04 Cycle 9 TUG time, sec 14.40 (4.31) 0.03 −0.02 0.24** 0.67** PA, min/week 33.61 (73.43) 0.12* 0.26** −0.07* −0.14** MMSE score 28.07 (3.86) 0.06 0.06 −0.04 −0.04 Worsening CDR score, present 120 (8.6%) 1.79 1.10 −3.76** −3.59** Participant characteristics Analytic samplea Early PA slopeb Late PA slopeb Early PF slopeb Late PF slopeb Cycle 1 n = 1,404 (Cycles 1–5) (Cycles 5–9) (Cycles 1–5) (Cycles 5–9) Age 76.96 (7.2) 0.01 0.05* 0.25** −0.07* Sex, female 876 (62.4%) −0.64 −1.41 −0.27 0.23 Race, white 1337 (95.2%) −1.52 0.52 −0.12 1.25 Education, >HS 609 (43.38%) 0.21 0.31 9.46** 1.72 MMSE score 27.84 (1.99) 0.05 0.01 0.01 0.02 CDR score, MCI (score = 0.5)# 338 (24.1%) 0.24 0.08 16.48** 3.44* Painful gait, present 158 (11.3%) −0.80 0.20 0.13 0.47 ≥1 Alcohol drinks/week 324 (23.1%) −0.30 0.26 1.26 0.34 Ever smoker 734 (52.3%) −0.64 −0.37 2.04* 0.32 BMI 28.06 (5.40) 0.05 0.06* 0.02 0.05 Stroke/TIA, present 153 (10.9%) 0.52 −0.66 −2.71* 0.75 Myocardial infarction, present 192 (13.7%) 0.52 0.65 −1.88 −0.61 Hypertension, present 900 (64.1%) 0.53 0.52 −2.06* −1.31 Diabetes, present 289 (20.6%) −0.30 −0.97 −1.38 −0.59 Heart failure, present 99 (7.1%) 2.28* −1.16 −1.83 0.04 TUG time, sec 12.75 (4.06) −0.01 0.04 −0.02 −0.23** PA, min/week 66.45 (115.12) −0.56** −0.33** −0.04 −0.04 Cycle 9 TUG time, sec 14.40 (4.31) 0.03 −0.02 0.24** 0.67** PA, min/week 33.61 (73.43) 0.12* 0.26** −0.07* −0.14** MMSE score 28.07 (3.86) 0.06 0.06 −0.04 −0.04 Worsening CDR score, present 120 (8.6%) 1.79 1.10 −3.76** −3.59** aMean (SD) presented for continuous and N(%) presented for categorical variables. bT-tests (t-statistics shown) or ANOVA (F-statistics shown) used for mean comparisons for two- and three-level categorical variables, respectively; Pearson’s correlations reported for continuous variables. *Indicates P value ≤0.05. **Indicates P value ≤0.001. #CDR score indicating cognitively normal n = 1,060 (75.5%) and CDR score indicating dementia n = 6 (0.04%). Dementia status was strongly associated with PF but not with PA slopes (Table 1). Age was positively associated with early PF slope but negatively associated with late PF slope. There were no associations between age and PA slopes (Table 1). Smoking and cardiovascular risk factors (i.e. history of stroke/TIA, MI, hypertension and heart failure) were associated with early but not with later PF slope. Only BMI and heart failure were associated with PA slope (Table 1). Compared to early slopes, late PF slopes did not differ (Table 2); when the absolute change variables were computed, late PF slopes were significantly steeper (Table 2). Late PA slopes and change scores were not statistically different from early slopes or change scores (Table 2). Table 2. Mean (SD) PF and PA slope values and change score values for early (cycles 1–5) and later (cycles 5–9) portions of the study. Variable Mean (SD) Variance (SE) Comparison of early and late changes Early PF slopea 0.30 (0.64) 0.63 (0.04) Early (vs) late: Late PF slopea 0.37 (0.65) 0.90 (0.07) Chi-squared test (1) = 1.53, P = 0.22 Early PA slopea −4.55 (17.67) 491.62 (29.98) Early (vs) late: Late PA slopea −4.99 (14.53) 366.33 (23.37) Chi-squared test (1) = 0.14, P = 0.71 Early PF changeb 0.65 (3.44) 14.70 (0.65) Early (vs) late: Late PF changeb 1.73 (2.73) 16.52 (1.08) Chi-squared test (1) = 22.4, P < 0.001 Early PA changeb −14.29 (98.19) 11175.24 (473.61) Early (vs) late: Late PA changeb −21.32 (71.07) 7701.41 (391.53) Chi-squared test (1) = 1.91, P = 0.17 Variable Mean (SD) Variance (SE) Comparison of early and late changes Early PF slopea 0.30 (0.64) 0.63 (0.04) Early (vs) late: Late PF slopea 0.37 (0.65) 0.90 (0.07) Chi-squared test (1) = 1.53, P = 0.22 Early PA slopea −4.55 (17.67) 491.62 (29.98) Early (vs) late: Late PA slopea −4.99 (14.53) 366.33 (23.37) Chi-squared test (1) = 0.14, P = 0.71 Early PF changeb 0.65 (3.44) 14.70 (0.65) Early (vs) late: Late PF changeb 1.73 (2.73) 16.52 (1.08) Chi-squared test (1) = 22.4, P < 0.001 Early PA changeb −14.29 (98.19) 11175.24 (473.61) Early (vs) late: Late PA changeb −21.32 (71.07) 7701.41 (391.53) Chi-squared test (1) = 1.91, P = 0.17 aValues reflect estimated annual change, in minutes per week for PA and seconds for PF. bValues reflect overall cumulative change during time period of 5 years. Table 2. Mean (SD) PF and PA slope values and change score values for early (cycles 1–5) and later (cycles 5–9) portions of the study. Variable Mean (SD) Variance (SE) Comparison of early and late changes Early PF slopea 0.30 (0.64) 0.63 (0.04) Early (vs) late: Late PF slopea 0.37 (0.65) 0.90 (0.07) Chi-squared test (1) = 1.53, P = 0.22 Early PA slopea −4.55 (17.67) 491.62 (29.98) Early (vs) late: Late PA slopea −4.99 (14.53) 366.33 (23.37) Chi-squared test (1) = 0.14, P = 0.71 Early PF changeb 0.65 (3.44) 14.70 (0.65) Early (vs) late: Late PF changeb 1.73 (2.73) 16.52 (1.08) Chi-squared test (1) = 22.4, P < 0.001 Early PA changeb −14.29 (98.19) 11175.24 (473.61) Early (vs) late: Late PA changeb −21.32 (71.07) 7701.41 (391.53) Chi-squared test (1) = 1.91, P = 0.17 Variable Mean (SD) Variance (SE) Comparison of early and late changes Early PF slopea 0.30 (0.64) 0.63 (0.04) Early (vs) late: Late PF slopea 0.37 (0.65) 0.90 (0.07) Chi-squared test (1) = 1.53, P = 0.22 Early PA slopea −4.55 (17.67) 491.62 (29.98) Early (vs) late: Late PA slopea −4.99 (14.53) 366.33 (23.37) Chi-squared test (1) = 0.14, P = 0.71 Early PF changeb 0.65 (3.44) 14.70 (0.65) Early (vs) late: Late PF changeb 1.73 (2.73) 16.52 (1.08) Chi-squared test (1) = 22.4, P < 0.001 Early PA changeb −14.29 (98.19) 11175.24 (473.61) Early (vs) late: Late PA changeb −21.32 (71.07) 7701.41 (391.53) Chi-squared test (1) = 1.91, P = 0.17 aValues reflect estimated annual change, in minutes per week for PA and seconds for PF. bValues reflect overall cumulative change during time period of 5 years. Initial PF significantly predicted early and later PA slopes and these associations remained significant after adjustment for baseline age, MMSE and CDR scores (Table 3 and see Appendix 7 in the supplementary data, available at Age and Ageing online). Early PF slope did not significantly predict later PA slope (Table 3). Initial PA significantly predicted later, but not early PF slope (Table 3). Early PA slope was not a significant predictor of later PF slope (Table 3). The root mean square error or approximation and the comparative fit index (Table 3) indicate that the unadjusted and adjusted models had acceptable fit to the data. Table 3. Associations of initial and early slopes with later slopes for PF and PA. Unadjusted Adjusteda Standardised β (SE) P value Standardised β (SE) P value Predictive associations Initial PF value → Early PA slope −0.10 (0.03) <0.001 −0.09 (0.03) 0.003 Initial PA value → Early PF slope −0.03 (0.04) 0.47 −0.02 (0.03) 0.66 Initial PF value → Later PA slope −0.09 (0.03) 0.01 −0.07 (0.04) 0.05 Initial PA value → Later PF slope −0.09 (0.04) 0.04 −0.09 (0.03) 0.003 Early PF slope → Later PA slope −0.03 (0.04) 0.44 −0.02 (0.04) 0.69 Early PA slope → Later PF Slope −0.04 (0.04) 0.39 −0.03 (0.04) 0.45 Residual correlations Initial PF value ←→ Initial PA value −0.16 (0.03) <0.001 −0.13 (0.03) <0.001 Early PF slope ←→ Early PA slope −0.12 (0.03) <0.001 −0.11 (0.03) <0.001 Late PF slope ←→ Late PA slope −0.13 (0.04) <0.001 −0.12 (0.04) 0.002 Model fit Root mean square error of approximation (90% confidence Interval) 0.066 (0.062, 0.070) 0.058 (0.055, 0.061) Comparative fit index 0.95 0.94 Unadjusted Adjusteda Standardised β (SE) P value Standardised β (SE) P value Predictive associations Initial PF value → Early PA slope −0.10 (0.03) <0.001 −0.09 (0.03) 0.003 Initial PA value → Early PF slope −0.03 (0.04) 0.47 −0.02 (0.03) 0.66 Initial PF value → Later PA slope −0.09 (0.03) 0.01 −0.07 (0.04) 0.05 Initial PA value → Later PF slope −0.09 (0.04) 0.04 −0.09 (0.03) 0.003 Early PF slope → Later PA slope −0.03 (0.04) 0.44 −0.02 (0.04) 0.69 Early PA slope → Later PF Slope −0.04 (0.04) 0.39 −0.03 (0.04) 0.45 Residual correlations Initial PF value ←→ Initial PA value −0.16 (0.03) <0.001 −0.13 (0.03) <0.001 Early PF slope ←→ Early PA slope −0.12 (0.03) <0.001 −0.11 (0.03) <0.001 Late PF slope ←→ Late PA slope −0.13 (0.04) <0.001 −0.12 (0.04) 0.002 Model fit Root mean square error of approximation (90% confidence Interval) 0.066 (0.062, 0.070) 0.058 (0.055, 0.061) Comparative fit index 0.95 0.94 Notes: Initial = estimated intercept at year 1; Early slope = linear slope from year 1 to year 5 (using time points 1–5); Late slope = linear slope from year 5 to year 9 (using time points 5–9). aAll paths adjusted for baseline age, baseline MMSE, baseline CDR and baseline hypertension. Late slopes are additionally adjusted for year 9 CDR. Table 3. Associations of initial and early slopes with later slopes for PF and PA. Unadjusted Adjusteda Standardised β (SE) P value Standardised β (SE) P value Predictive associations Initial PF value → Early PA slope −0.10 (0.03) <0.001 −0.09 (0.03) 0.003 Initial PA value → Early PF slope −0.03 (0.04) 0.47 −0.02 (0.03) 0.66 Initial PF value → Later PA slope −0.09 (0.03) 0.01 −0.07 (0.04) 0.05 Initial PA value → Later PF slope −0.09 (0.04) 0.04 −0.09 (0.03) 0.003 Early PF slope → Later PA slope −0.03 (0.04) 0.44 −0.02 (0.04) 0.69 Early PA slope → Later PF Slope −0.04 (0.04) 0.39 −0.03 (0.04) 0.45 Residual correlations Initial PF value ←→ Initial PA value −0.16 (0.03) <0.001 −0.13 (0.03) <0.001 Early PF slope ←→ Early PA slope −0.12 (0.03) <0.001 −0.11 (0.03) <0.001 Late PF slope ←→ Late PA slope −0.13 (0.04) <0.001 −0.12 (0.04) 0.002 Model fit Root mean square error of approximation (90% confidence Interval) 0.066 (0.062, 0.070) 0.058 (0.055, 0.061) Comparative fit index 0.95 0.94 Unadjusted Adjusteda Standardised β (SE) P value Standardised β (SE) P value Predictive associations Initial PF value → Early PA slope −0.10 (0.03) <0.001 −0.09 (0.03) 0.003 Initial PA value → Early PF slope −0.03 (0.04) 0.47 −0.02 (0.03) 0.66 Initial PF value → Later PA slope −0.09 (0.03) 0.01 −0.07 (0.04) 0.05 Initial PA value → Later PF slope −0.09 (0.04) 0.04 −0.09 (0.03) 0.003 Early PF slope → Later PA slope −0.03 (0.04) 0.44 −0.02 (0.04) 0.69 Early PA slope → Later PF Slope −0.04 (0.04) 0.39 −0.03 (0.04) 0.45 Residual correlations Initial PF value ←→ Initial PA value −0.16 (0.03) <0.001 −0.13 (0.03) <0.001 Early PF slope ←→ Early PA slope −0.12 (0.03) <0.001 −0.11 (0.03) <0.001 Late PF slope ←→ Late PA slope −0.13 (0.04) <0.001 −0.12 (0.04) 0.002 Model fit Root mean square error of approximation (90% confidence Interval) 0.066 (0.062, 0.070) 0.058 (0.055, 0.061) Comparative fit index 0.95 0.94 Notes: Initial = estimated intercept at year 1; Early slope = linear slope from year 1 to year 5 (using time points 1–5); Late slope = linear slope from year 5 to year 9 (using time points 5–9). aAll paths adjusted for baseline age, baseline MMSE, baseline CDR and baseline hypertension. Late slopes are additionally adjusted for year 9 CDR. Secondary analyses yielded similar results. Baseline PF significantly predicted early and later PA changes (see Appendix 5 in the supplementary data, available at Age and Ageing online). These associations remained significant after adjustment for baseline age, MMSE score, CDR score and history of hypertension (see Appendix 5 in the supplementary data, available at Age and Ageing online). Baseline PA also significantly predicted early and later PF changes, and adjustment for covariates did not attenuate these associations (see Appendix 5 in the supplementary data, available at Age and Ageing online). Early PF change did not significantly predict later PA change; similarly, early PA change did not predict later PF change (see Appendix 5 in the supplementary data, available at Age and Ageing online). Sensitivity analyses with varying degrees of missing data and excluding extreme values of PA showed results were highly consistent, with the exception of a sample limited to only n = 450 with complete data at all nine cycles where all results were non-significant (see Appendix 6 in the supplementary data, available at Age and Ageing online). Discussion In this study, PF and PA both changed over time and were related in complex ways. We found initial PF strongly predicted both early and late PA slopes. Initial PA also predicted late but not early PF slope. There were no significant associations between early PF predicting later PA or early PA predicting later PF suggesting the starting values of PA and PF were more robust indicators than early slopes of subsequent PA and PF changes. Results were independent of factors related to PF or PA slope, including baseline age, MMSE score or CDR score, hypertension and converting to worse CDR scores, suggesting while covariates were related to PF and PA, they did not fully attenuate the observed associations. These results were also supported in sensitivity analyses allowing varying degrees of missing data, and when placing restrictions on upper limits of PA. Together, these results suggest that the associations are not driven solely by those at the highest or lowest levels of function. Change score analyses largely supported our primary findings (see Appendix 5 in the supplementary data, available at Age and Ageing online) except baseline PA significantly predicted early PF change. Most likely this discrepancy arose because slope values reflected an estimated annual change in seconds for PF, whereas change scores reflected overall cumulative change over 5 years (Table 2). This difference resulted in much larger values in change score than in primary analyses. These findings validate previous results [23], by demonstrating the relationship persists in a less highly selected, more representative sample of older adults, and when PF and PA measures are collected annually as opposed to every 4–5 years. Other studies examining the relationships between PA and PF, in older adults, have yielded mixed results. A meta-analysis concluded high-intensity but not moderate- or low-intensity, exercise promoted faster-walking speed [22]. The LIFE study, a large-scale PA intervention found moderate-intensity, multi-component PA significantly reduced physical disability but did not promote PF [12]. Better understanding the nature of the relationship between PF and PA will highlight potential mechanisms for improving these measures. It may also identify pathways to reduce other health-related outcomes, such as major mobility disability, falls, hospitalisations, poor cardiovascular health and declining cognitive function. An important consideration when interpreting these results is the complex nature of both PF and PA, especially in older adults. For example, PF may be influenced by a large number of factors, including disorders and diseases that contribute to disuse, muscle loss and weight gain in older adults, which adds a complexity in understanding the true nature of this relationship. Some diseases of ageing may affect PF progress slowly without overt clinical symptoms until years after underlying pathology has begun, which makes measuring these contributing factors difficult. An example of this is the accumulation of white matter lesions with age, which are common among older adults [31] and are associated with slower gait [31, 32]. Similarly, PA is affected by the motivation to be physically active, depressive symptoms, cognitive health and socioeconomic factors (i.e. safe place to walk, money to buy shoes or exercise equipment, etc.). PA is also likely conditional on maintaining some sort of baseline PF level. Thus, it is important to consider as many of these complex factors in older adults as we continue to disentangle these complex relationships. This study has several strengths, including the representative nature of the cohort, the duration of time over which participants were followed and the large number of visits within that time. Several potential limitations should also be considered. PA was self-reported, not objectively measured and there may be a discrepancy between what participants believe they usually do, and what they actually perform. If most participants exaggerated their activity, our results may be inflated; if the reverse is the case, our results would be underestimations. Future studies should examine these associations with more objective measures like accelerometers. Furthermore, the only type of PA measured is self-reported time spent walking at a moderate-intensity. However, a large proportion of older adults reports low- or moderate-intensity walking as their only PA, even after interventions to increase PA [7, 10]. Participant survivor bias may have underestimated the associations meaning participants included in these analyses were healthier and likely at a lower risk for PF impairment because they were healthy enough to make it to every visit over nine cycles and a natural variation in PF is lost. We see this supported in sensitivity analyses limited to n = 450 with complete data, and as we noted, those excluded from our sample were less healthy than those included. This is also supported by older age being associated with a greater change in early PF slope but a slower change in late PF slope, indicating less variation in PF in later cycles where those surviving are likely more similar than those lost to follow-up. This study found a bidirectional relationship between initial PF and PA predicting later PA and PF. Thus, a person with difficulties in PF may decrease PA, and in turn, worsen in PF. Conversely, a person who declines in PA, perhaps due to inclement weather or poor social surroundings, may decline in PF, resulting in a struggle to return to previous PA levels, even when or if the barrier is removed. Of course, this is an over-simplification of all the physical, social and health-related challenges older adults face that may affect PF and PA, but the idea suggests helping older adults maintain PF and PA will work at keeping them both healthier and free of mobility impairments. Interventions to improve early PF may support increased PA, and vice versa, thus altering a vicious cycle of lower PF leading to lower PA, to a virtuous cycle of higher PF conveying higher PA. Such a shift could reduce associated risk factors, including impaired cardiovascular health [9], cognitive function [8], mood and increased social isolation [16]. More work is needed to determine if targeted interventions can effectively reduce the risk for PA decline and associated co-morbidities in older adults, and to determine if one factor more strongly predicts the other. Key points: Physical function predicts decline in physical activity, in the short and long-term. Physical activity predicted long-term decline in physical function. Interventions to improve Physical Function could support increased Physical Activity, and vice versa, thus altering a vicious cycle to a virtuous cycle. Supplementary data Supplementary data mentioned in the text are available to subscribers in Age and Ageing online. Acknowledgements The authors would like to acknowledge K.M. in MYHAT project coordination, and E.J. for compiling and providing the analytic datasets. All the authors meet the criteria for authorship stated in the Uniform Requirements for Manuscripts Submitted to Biomedical Journals. Conflicts of interest None. Funding This work was supported in part by the National Institute on Aging (Grant nos. R01 AG023651 and K24 AG022035). The sponsor had no role in the study or manuscript writing. References 1 Werner CA . The Older Population: 2010 Census Briefs . Washington, DC : U.S. Department of Commerce , 2012 . 2 Guralnik J , Ferrucci L , Pieper CF et al. . Lower extremity function and subsequent disability: consistency across studies, predictive models, and value of gait speed alone compared with the Short Physical Performance Battery . J Gerontol 2000 ; 55A : M221 – 31 . Google Scholar CrossRef Search ADS 3 Guralnik JM , Ferrucci L , Simonsick EM , Salive ME , Wallace RB . Lower-extremity function in persons over the age of 70 years as a predictor of subsequent disability . N Engl J Med 1995 ; 332 : 556 – 61 . Google Scholar CrossRef Search ADS PubMed 4 Hardy SE , Perera S , Roumani YF , Chandler JM , Studenski SA . Improvement in usual gait speed predicts better survival in older adults . J Am Geriatr Soc 2007 ; 55 : 1727 – 34 . Google Scholar CrossRef Search ADS PubMed 5 Studenski S , Perera S , Patel K et al. . Gait speed and survival in older adults . JAMA 2011 ; 305 : 50 – 8 . Google Scholar CrossRef Search ADS PubMed 6 Baker G , Gray SR , Wright A et al. . The effect of a pedometer-based community walking intervention ‘Walking for Wellbeing in the West’ on physical activity levels and health outcomes: a 12-week randomized controlled trial . Int J Behav Nutr Phys Act 2008 ; 5 : 44 . Google Scholar CrossRef Search ADS PubMed 7 Berkemeyer K , Wijndaele K , White T et al. . The descriptive epidemiology of accelerometer-measured physical activity in older adults . Int J Behav Nutr Phys Act 2016 ; 13 : 2 . Google Scholar CrossRef Search ADS PubMed 8 Falck RS , Davis JC , Liu-Ambrose T . What is the association between sedentary behaviour and cognitive function? A systematic review . Br J Sports Med 2017 ; 51 : 800 – 11 . Google Scholar CrossRef Search ADS PubMed 9 Wilmot EG , Edwardson CL , Achana FA et al. . Sedentary time in adults and the association with diabetes, cardiovascular disease and death: systematic review and meta-analysis . Diabetologia 2012 ; 55 : 2895 – 2905 . Google Scholar CrossRef Search ADS PubMed 10 Kimura M , Moriyasu A , Kumagai S et al. . Community-based intervention to improve dietary habits and promote physical activity among older adults: a cluster randomized trial . BMC Geriatr 2013 ; 13 : 8 . Google Scholar CrossRef Search ADS PubMed 11 Mutrie M , Doolin O , Fitzsimons C et al. . Increasing older adults’ walking through primary care: results of a pilot randomized controlled trial . Fam Pract 2012 ; 29 : 633 – 42 . Google Scholar CrossRef Search ADS PubMed 12 Pahor M , Guralnik JM , Ambrosius WT et al. . Effect of structured physical activity on prevention of major mobility disability in older adults: the LIFE study randomized clinical trial . JAMA 2014 ; 311 : 2387 – 96 . Google Scholar CrossRef Search ADS PubMed 13 Sink KM , Espeland MA , Castro CM et al. . Effect of a 24-month physical activity intervention vs health education on cognitive outcomes in sedentary older adults: the LIFE randomized trial . JAMA 2015 ; 314 : 781 – 90 . Google Scholar CrossRef Search ADS PubMed 14 Buracchio T , Dodge HH , Howieson D , Wasserman D , Kaye J . The trajectory of gait speed preceding mild cognitive impairment . Arch Neurol 2010 ; 67 : 980 – 6 . Google Scholar CrossRef Search ADS PubMed 15 Quan M , Xun P , Chen C et al. . Walking pace and the risk of cognitive decline and dementia in elderly populations: a meta-analysis of prospective cohort studies . J Gerontol A Biol Sci Med Sci 2017 ; 72 : 266 – 70 . Google Scholar CrossRef Search ADS PubMed 16 Bauman A , Merom D , Bull FC , Buchner DM , Fiatarone Singh MA . Updating the evidence for physical activity: summative reviews of the epidemiological evidence, prevalence, and interventions to promote ‘active aging’ . Gerontologist 2016 ; 56 ( Suppl 2 ): S268 – 280 . Google Scholar CrossRef Search ADS PubMed 17 Fox KR , Ku PW , Hillsdon M et al. . Objectively assessed physical activity and lower limb function and prospective associations with mortality and newly diagnosed disease in UK older adults: an OPAL four-year follow-up study . Age Ageing 2015 ; 44 : 261 – 8 . Google Scholar CrossRef Search ADS PubMed 18 Macmillan F , Fitzsimons C , Black K et al. . West End Walkers 65+: a randomised controlled trial of a primary care-based walking intervention for older adults: study rationale and design . BMC Public Health 2011 ; 11 : 120 . Google Scholar CrossRef Search ADS PubMed 19 Perera S , Patel KV , Rosano C et al. . Gait speed predicts incident disability: a pooled analysis . J Gerontol A Biol Sci Med Sci 2016 ; 71 : 63 – 71 . Google Scholar CrossRef Search ADS PubMed 20 Brach JS , Lowry K , Perera S et al. . Improving motor control in walking: a randomized clinical trial in older adults with subclinical walking difficulty . Arch Phys Med Rehabil 2015 ; 96 : 388 – 94 . Google Scholar CrossRef Search ADS PubMed 21 Brach JS , Van Swearingen JM , Perera S , Wert DM , Studenski S . Motor learning versus standard walking exercise in older adults with subclinical gait dysfunction: a randomized clinical trial . J Am Geriatr Soc 2013 ; 61 : 1879 – 86 . Google Scholar CrossRef Search ADS PubMed 22 Lopopolo RB , Greco M , Sullivan D , Craik RL , Mangione KK . Effect of therapeutic exercise on gait speed in community-dwelling elderly people: a meta-analysis . Phys Ther 2006 ; 86 : 520 – 40 . Google Scholar PubMed 23 Best JR , Liu-Ambrose T , Metti AL et al. . Longitudinal associations between walking speed and amount of time spent walking over a 9-year period in older women and men . J Gerontol A Biol Sci Med Sci 2017 Jun 22 [Epub ahead of print]. 24 Ganguli M , Snitz B , Vander Bilt J , Chang CC . How much do depressive symptoms affect cognition at the population level? The Monongahela–Youghiogheny Healthy Aging Team (MYHAT) study . Int J Geriatr Psychiatry 2009 ; 24 : 1277 – 84 . Google Scholar CrossRef Search ADS PubMed 25 Mathias S , Nayak US , Isaacs B . Balance in elderly patients: the ‘get-up and go’ test . Arch Phys Med Rehabil 1986 ; 67 : 387 – 9 . Google Scholar PubMed 26 Folstein MF , Folstein SE , McHugh PR . ‘Mini-mental state’. A practical method for grading the cognitive state of patients for the clinician . J Psychiatr Res 1975 ; 12 : 189 – 98 . Google Scholar CrossRef Search ADS PubMed 27 Hughes CP , Berg L , Danziger WL , Coben LA , Martin RL . A new clinical scale for the staging of dementia . Br J Psychiatry 1982 ; 140 : 566 – 72 . Google Scholar CrossRef Search ADS PubMed 28 Morris JC . The Clinical Dementia Rating (CDR): current version and scoring rules . Neurology 1993 ; 43 : 2412 – 4 . Google Scholar CrossRef Search ADS PubMed 29 Selig JP , Little TD . Autoregressive and cross-lagged panel analysis for longitudinal data . New York : The Guilford Press , 2012 . 30 Muthén LK , Muthén B . Mplus: Statistical analysis with latent variables , 7th edition . Los Angeles, CA : Muthén & Muthén , 2012 . 31 Smith EE , O’Donnell M , Dagenais G et al. . Early cerebral small vessel disease and brain volume, cognition, and gait . Ann Neurol 2015 ; 77 : 251 – 61 . Google Scholar CrossRef Search ADS PubMed 32 Bolandzadeh N , Liu-Ambrose T , Aizenstein H et al. . Pathways linking regional hyperintensities in the brain and slower gait . NeuroImage 2014 ; 99C : 7 – 13 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of the British Geriatrics Society. All rights reserved. For permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Age and Ageing Oxford University Press

Longitudinal changes in physical function and physical activity in older adults

Loading next page...
 
/lp/ou_press/longitudinal-changes-in-physical-function-and-physical-activity-in-MLvxlBmYcN
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the British Geriatrics Society. All rights reserved. For permissions, please email: journals.permissions@oup.com
ISSN
0002-0729
eISSN
1468-2834
D.O.I.
10.1093/ageing/afy025
Publisher site
See Article on Publisher Site

Abstract

Abstract Background physical function (PF) and physical activity (PA) both decline as adults age and have been linked to negative outcomes, including dementia, depression and cardiovascular diseases. Although declines in each are associated with numerous negative outcomes, the longitudinal relationship between these two measures is unclear. Objective to examine the dynamic, bidirectional associations between declines in PF and PA. Design prospective cohort. Setting the Monongahela–Youghiogheny Healthy Aging Team (MYHAT) study. Subjects about 1,404 men and women, 76.96 ± 7.2 years, 62.4% female and 95.2% white. Methods over nine annual assessment cycles, PF was evaluated via the timed Up-and-Go task and PA via a self-reported questionnaire. Piecewise latent growth models examined bidirectional associations between PA and PF to determine whether the initial values (intercept) or early slope (cycles 1–5) (in either PF or PA) predicted later slope (cycles 5–9) (in either PF or PA). Results initial PF significantly predicted early (standardised β= −0.10, P < 0.001) and later (standardised β= −0.09, P = 0.01) PA slopes. Initial PA significantly predicted later (standardised β = −0.09, P = 0.04) but not early PF slope. Associations were independent of baseline memory test scores, baseline cognitive status, later cognitive status and age. Early physical function slope neither predicts later PA slope nor did early PA slope predict later PF slope (both P values >0.10). Conclusions the relationship between PF and PA is bidirectional, with PF more consistently predicting declines of PA, both in the short- and long-term. Intervening on PF impairments may improve PA engagement, which could in turn promote PF and translate to beneficial effects on cognitive function, cardiovascular health and mood. ageing, physical function, physical activity, older people Introduction With a life expectancy of older adults rising, there has been increasing emphasis on maximising the quality of life, including maintenance of physical function (PF) and physical activity (PA) [1]. PF is often characterised by measures of gait speed or gait slowing over time [2–5]. PA is less consistently defined, especially in older adults, with some studies using either pedometers or accelerometers [6, 7], self-reported total sedentary time [8, 9], or time spent walking, strength training, or conducting balance or stretching exercises [10–13]. Notably, walking is often the only form of PA reported in older adults [7, 10]. Gait slowing, a common and important problem in older adults, reflects multi-system impairments and often precedes the onset of physical disability, falls and cognitive impairment [2–5, 14, 15]. Similarly, PA declines in intensity and duration with age [7], and predicts reduced cardiovascular health [9], a faster decline in cognitive function [8], impaired mood, increased social isolation [16], increased mortality and a greater risk for diagnosis of any new disease [17]. PF contributes significantly to the associations between PA and mortality and disease, indicating a close interplay between PF and PA [17]. Despite a wealth of research to promote PA [6, 10, 11, 18] and to some extent PF [12, 13] in older adults, the joint association has rarely been examined. Declining PF likely lowers PA engagement but declining PA may lessen conditioning, in turn reducing PF. Thus, interventions to improve PF could augment effects of concurrent interventions to promote PA, and vice versa, transferring individuals from a vicious cycle of lower PF leading to lower PA, to a virtuous cycle where higher PF contributes to higher PA. Maintaining higher PF late in life is associated with lower risk of mortality and disability [5, 19]. Similarly, increased PA is associated with better cardiovascular health [9], cognitive function [8], mood and social support [16]. Interventions promoting PA can also promote PF [12, 20–22]. Few studies have examined the association of PF predicting PA. We recently showed declining walking speed over 4 years among high-functioning, community-dwelling older adults significantly predicted declining PA engagement over the subsequent 5 years but PA did not predict walking speed [23]. Here, we aim to extend this line of inquiry to a less highly selected and more representative samples of older adults with a greater range of co-morbidities, by testing the bidirectional associations between PF and PA over time. Methods Study population The Monongahela–Youghiogheny Healthy Aging Team (MYHAT) is a random population sample drawn from publicly available voter registration lists for a small town region of Pennsylvania (USA) [24]. Sampling, recruitment and cohort characteristics have been previously reported [24]. Briefly, participants were 65 years or older. Exclusion criteria were either residing in a long-term care facility at recruitment, substantial sensory or cognitive impairment (age-education-adjusted Mini-Mental State Examination (MMSE) <21), or decisional incapacity [24]. Participants were enroled between 2006 and 2008 (cycle 1), and followed for eight subsequent annual visits, with cycle 9 occurring between 2014 and 2016. This study was approved by the Institutional Review Board at the University of Pittsburgh and all participants provided written informed consent. Analytic cohort Our analytic sample was defined, a priori, as those who had baseline measures of PF and PA, plus at least two follow-up measures of each variable. Of the original 1,982 MYHAT participants, 578 were unable to perform the PF test, or had missing baseline PF data, or missing PF data from >6 subsequent visits. In addition, 472 participants, not mutually exclusive from those missing PF data had missing PA data (baseline or >6 subsequent visits). Consequently, our analytic sample consisted of 1,404 participants. Physical function PF was assessed with the timed Up-and-Go task (TUG) test [25], measured in seconds. Participants were instructed to stand up from a chair, if possible without using arms for support, to walk 10 feet forward, 10 feet back at their usual pace and return to the seated position. Physical activity PA was self-reported via a questionnaire which assessed the number of days per week, and average minutes per day of walking a self-determined level of moderate-intensity (see Appendix 1 in the supplementary data, available at Age and Ageing online). Total minutes per week were computed from total days per week and total minutes per day. Covariates At baseline, self-reported age, sex, race (white or non-white) and education (≤high school (vs) >high school) were collected. Self-reported weight (pounds) and height (inches) were recorded at cycle 1, and used to calculate body mass index (weight(lb)/[height(in)]2 × 703, BMI). Self-reported history of medical diagnoses determined participant’s history of stroke or transient ischaemic attack (TIA), type 1 or type 2 diabetes, hypertension, heart attack or myocardial infarction (MI) and heart failure. Participants self-reported any painful gait (yes/no), smoking status (ever (vs) never and consumption of alcoholic beverages (current drinker consuming ≥1 alcoholic drinks per week (vs) never drinker or current drinker, drinking ≤1 drink per week). The MMSE [26] was administered at each cycle and scored on a scale of 0–30. A thorough cognitive evaluation was conducted for all participants using the Clinical Dementia Rating (CDR) scale [27, 28], participants were scored as having no dementia (CDR = 0), mild cognitive impairment (MCI; CDR = 0.5) and mild, moderate or severe dementia (CDR ≥ 1). Statistical analyses Raw mean standard deviation (SD) PF and PA variables were visually examined to see how they changed over time and tested for linear changes over time with the Shapiro–Wilk test. The associations between baseline characteristics and early and late PF and PA slopes were investigated using t-tests or analysis of variance (ANOVA) for categorical variables and Pearson’s correlations for continuous variables. Our primary analytic approach investigated longitudinal, bidirectional associations between PF and PA, using piecewise latent growth curve modelling [29], with the bootstrapping method (5,000 re-samples) for standard errors (SE) to account for asymmetric abnormalities in both distributions. This approach estimated an initial value (intercept of PF or PA at cycle 1), plus one slope of ‘early’ PA or PF (cycles 1–5) and one slope of ‘late’ PA or PF (cycles 5–9). By defining both early and late slopes to include cycle 5, it allowed cycle 5 to serve as a hinge, and the slope before and after cycle 5 to be different, without ignoring change occurring between cycles 4 and 5. Initial values and slopes were estimated in parallel, allowing simultaneous examination of all possible bidirectional associations. All models were adjusted for initial values of the outcome. A secondary analytic approach was specified a priori; the goal of this analysis was to, as closely as possible, match the analytic approach used in our previous work, which contained three-time points of PF and PA [23], to determine if results were consistent. This approach estimated raw PA and PF change scores, using only data from cycles 1, 5 and 9, to define ‘early change’ (cycles 5–1) and ‘late change’ (cycles 9–5). Change scores were computed as latent variables in piecewise latent growth curve models, testing all possible associations between PF and PA, and estimating SE using 5,000 bootstrapped re-samples. Baseline PF and PA were used as covariates in the models. For primary and secondary analytic approaches, models were first constructed without covariates and then with covariate adjustment. All analyses included 1,404 participants; missing follow-up data were handled using the maximum likelihood estimator (see Appendix 2 in the supplementary data, available at Age and Ageing online). Sensitivity analyses were conducted with varying degrees of missing data to examine how missing visits may have affected associations, and with upper limits placed on PA, to determine if extreme values of PA were influencing associations were also assessed. Levels of statistical significance were defined a priori as P values ≤0.05. Analyses were performed with Mplus, 7.3 [30] and SAS version 9.4 (SAS Institute Inc., Cary, NC). Results Of the 1,404 participants in our analytic sample, 32.3% and 45.1% had all nine assessments of PF and PA, respectively, compared to 13.0% and 10.9% with only three assessments of each. The number of participants at each cycle with complete data and the mean (SD) are reported in Appendix 3, available in Age and Ageing online. Overall, PF and PA both decreased, linearly, over time, as exemplified by an increase in time to complete the TUG task and less time spent walking, respectively (see Appendix 3 in the supplementary data, available at Age and Ageing online). Average participant baseline age was 76.96 ± 7.2 years, 62.4% were female, 95.2% were white and 43.4% had greater than a high school education. Average MMSE score at baseline was 27.84 ± 1.99 points, six participants (0.4%) had dementia (CDR ≥1) and 338 (24.1%) had MCI (CDR = 0.5) (Table 1). Compared to participants in the analytic sample, those excluded from analyses differed significantly on age, education and other factors indicating those excluded were less healthy (see Appendix 3 in the supplementary data, available at Age and Ageing online). Table 1. Descriptive characteristics, and associations with early and late PA and PF slopes. Participant characteristics Analytic samplea Early PA slopeb Late PA slopeb Early PF slopeb Late PF slopeb Cycle 1 n = 1,404 (Cycles 1–5) (Cycles 5–9) (Cycles 1–5) (Cycles 5–9) Age 76.96 (7.2) 0.01 0.05* 0.25** −0.07* Sex, female 876 (62.4%) −0.64 −1.41 −0.27 0.23 Race, white 1337 (95.2%) −1.52 0.52 −0.12 1.25 Education, >HS 609 (43.38%) 0.21 0.31 9.46** 1.72 MMSE score 27.84 (1.99) 0.05 0.01 0.01 0.02 CDR score, MCI (score = 0.5)# 338 (24.1%) 0.24 0.08 16.48** 3.44* Painful gait, present 158 (11.3%) −0.80 0.20 0.13 0.47 ≥1 Alcohol drinks/week 324 (23.1%) −0.30 0.26 1.26 0.34 Ever smoker 734 (52.3%) −0.64 −0.37 2.04* 0.32 BMI 28.06 (5.40) 0.05 0.06* 0.02 0.05 Stroke/TIA, present 153 (10.9%) 0.52 −0.66 −2.71* 0.75 Myocardial infarction, present 192 (13.7%) 0.52 0.65 −1.88 −0.61 Hypertension, present 900 (64.1%) 0.53 0.52 −2.06* −1.31 Diabetes, present 289 (20.6%) −0.30 −0.97 −1.38 −0.59 Heart failure, present 99 (7.1%) 2.28* −1.16 −1.83 0.04 TUG time, sec 12.75 (4.06) −0.01 0.04 −0.02 −0.23** PA, min/week 66.45 (115.12) −0.56** −0.33** −0.04 −0.04 Cycle 9 TUG time, sec 14.40 (4.31) 0.03 −0.02 0.24** 0.67** PA, min/week 33.61 (73.43) 0.12* 0.26** −0.07* −0.14** MMSE score 28.07 (3.86) 0.06 0.06 −0.04 −0.04 Worsening CDR score, present 120 (8.6%) 1.79 1.10 −3.76** −3.59** Participant characteristics Analytic samplea Early PA slopeb Late PA slopeb Early PF slopeb Late PF slopeb Cycle 1 n = 1,404 (Cycles 1–5) (Cycles 5–9) (Cycles 1–5) (Cycles 5–9) Age 76.96 (7.2) 0.01 0.05* 0.25** −0.07* Sex, female 876 (62.4%) −0.64 −1.41 −0.27 0.23 Race, white 1337 (95.2%) −1.52 0.52 −0.12 1.25 Education, >HS 609 (43.38%) 0.21 0.31 9.46** 1.72 MMSE score 27.84 (1.99) 0.05 0.01 0.01 0.02 CDR score, MCI (score = 0.5)# 338 (24.1%) 0.24 0.08 16.48** 3.44* Painful gait, present 158 (11.3%) −0.80 0.20 0.13 0.47 ≥1 Alcohol drinks/week 324 (23.1%) −0.30 0.26 1.26 0.34 Ever smoker 734 (52.3%) −0.64 −0.37 2.04* 0.32 BMI 28.06 (5.40) 0.05 0.06* 0.02 0.05 Stroke/TIA, present 153 (10.9%) 0.52 −0.66 −2.71* 0.75 Myocardial infarction, present 192 (13.7%) 0.52 0.65 −1.88 −0.61 Hypertension, present 900 (64.1%) 0.53 0.52 −2.06* −1.31 Diabetes, present 289 (20.6%) −0.30 −0.97 −1.38 −0.59 Heart failure, present 99 (7.1%) 2.28* −1.16 −1.83 0.04 TUG time, sec 12.75 (4.06) −0.01 0.04 −0.02 −0.23** PA, min/week 66.45 (115.12) −0.56** −0.33** −0.04 −0.04 Cycle 9 TUG time, sec 14.40 (4.31) 0.03 −0.02 0.24** 0.67** PA, min/week 33.61 (73.43) 0.12* 0.26** −0.07* −0.14** MMSE score 28.07 (3.86) 0.06 0.06 −0.04 −0.04 Worsening CDR score, present 120 (8.6%) 1.79 1.10 −3.76** −3.59** aMean (SD) presented for continuous and N(%) presented for categorical variables. bT-tests (t-statistics shown) or ANOVA (F-statistics shown) used for mean comparisons for two- and three-level categorical variables, respectively; Pearson’s correlations reported for continuous variables. *Indicates P value ≤0.05. **Indicates P value ≤0.001. #CDR score indicating cognitively normal n = 1,060 (75.5%) and CDR score indicating dementia n = 6 (0.04%). Table 1. Descriptive characteristics, and associations with early and late PA and PF slopes. Participant characteristics Analytic samplea Early PA slopeb Late PA slopeb Early PF slopeb Late PF slopeb Cycle 1 n = 1,404 (Cycles 1–5) (Cycles 5–9) (Cycles 1–5) (Cycles 5–9) Age 76.96 (7.2) 0.01 0.05* 0.25** −0.07* Sex, female 876 (62.4%) −0.64 −1.41 −0.27 0.23 Race, white 1337 (95.2%) −1.52 0.52 −0.12 1.25 Education, >HS 609 (43.38%) 0.21 0.31 9.46** 1.72 MMSE score 27.84 (1.99) 0.05 0.01 0.01 0.02 CDR score, MCI (score = 0.5)# 338 (24.1%) 0.24 0.08 16.48** 3.44* Painful gait, present 158 (11.3%) −0.80 0.20 0.13 0.47 ≥1 Alcohol drinks/week 324 (23.1%) −0.30 0.26 1.26 0.34 Ever smoker 734 (52.3%) −0.64 −0.37 2.04* 0.32 BMI 28.06 (5.40) 0.05 0.06* 0.02 0.05 Stroke/TIA, present 153 (10.9%) 0.52 −0.66 −2.71* 0.75 Myocardial infarction, present 192 (13.7%) 0.52 0.65 −1.88 −0.61 Hypertension, present 900 (64.1%) 0.53 0.52 −2.06* −1.31 Diabetes, present 289 (20.6%) −0.30 −0.97 −1.38 −0.59 Heart failure, present 99 (7.1%) 2.28* −1.16 −1.83 0.04 TUG time, sec 12.75 (4.06) −0.01 0.04 −0.02 −0.23** PA, min/week 66.45 (115.12) −0.56** −0.33** −0.04 −0.04 Cycle 9 TUG time, sec 14.40 (4.31) 0.03 −0.02 0.24** 0.67** PA, min/week 33.61 (73.43) 0.12* 0.26** −0.07* −0.14** MMSE score 28.07 (3.86) 0.06 0.06 −0.04 −0.04 Worsening CDR score, present 120 (8.6%) 1.79 1.10 −3.76** −3.59** Participant characteristics Analytic samplea Early PA slopeb Late PA slopeb Early PF slopeb Late PF slopeb Cycle 1 n = 1,404 (Cycles 1–5) (Cycles 5–9) (Cycles 1–5) (Cycles 5–9) Age 76.96 (7.2) 0.01 0.05* 0.25** −0.07* Sex, female 876 (62.4%) −0.64 −1.41 −0.27 0.23 Race, white 1337 (95.2%) −1.52 0.52 −0.12 1.25 Education, >HS 609 (43.38%) 0.21 0.31 9.46** 1.72 MMSE score 27.84 (1.99) 0.05 0.01 0.01 0.02 CDR score, MCI (score = 0.5)# 338 (24.1%) 0.24 0.08 16.48** 3.44* Painful gait, present 158 (11.3%) −0.80 0.20 0.13 0.47 ≥1 Alcohol drinks/week 324 (23.1%) −0.30 0.26 1.26 0.34 Ever smoker 734 (52.3%) −0.64 −0.37 2.04* 0.32 BMI 28.06 (5.40) 0.05 0.06* 0.02 0.05 Stroke/TIA, present 153 (10.9%) 0.52 −0.66 −2.71* 0.75 Myocardial infarction, present 192 (13.7%) 0.52 0.65 −1.88 −0.61 Hypertension, present 900 (64.1%) 0.53 0.52 −2.06* −1.31 Diabetes, present 289 (20.6%) −0.30 −0.97 −1.38 −0.59 Heart failure, present 99 (7.1%) 2.28* −1.16 −1.83 0.04 TUG time, sec 12.75 (4.06) −0.01 0.04 −0.02 −0.23** PA, min/week 66.45 (115.12) −0.56** −0.33** −0.04 −0.04 Cycle 9 TUG time, sec 14.40 (4.31) 0.03 −0.02 0.24** 0.67** PA, min/week 33.61 (73.43) 0.12* 0.26** −0.07* −0.14** MMSE score 28.07 (3.86) 0.06 0.06 −0.04 −0.04 Worsening CDR score, present 120 (8.6%) 1.79 1.10 −3.76** −3.59** aMean (SD) presented for continuous and N(%) presented for categorical variables. bT-tests (t-statistics shown) or ANOVA (F-statistics shown) used for mean comparisons for two- and three-level categorical variables, respectively; Pearson’s correlations reported for continuous variables. *Indicates P value ≤0.05. **Indicates P value ≤0.001. #CDR score indicating cognitively normal n = 1,060 (75.5%) and CDR score indicating dementia n = 6 (0.04%). Dementia status was strongly associated with PF but not with PA slopes (Table 1). Age was positively associated with early PF slope but negatively associated with late PF slope. There were no associations between age and PA slopes (Table 1). Smoking and cardiovascular risk factors (i.e. history of stroke/TIA, MI, hypertension and heart failure) were associated with early but not with later PF slope. Only BMI and heart failure were associated with PA slope (Table 1). Compared to early slopes, late PF slopes did not differ (Table 2); when the absolute change variables were computed, late PF slopes were significantly steeper (Table 2). Late PA slopes and change scores were not statistically different from early slopes or change scores (Table 2). Table 2. Mean (SD) PF and PA slope values and change score values for early (cycles 1–5) and later (cycles 5–9) portions of the study. Variable Mean (SD) Variance (SE) Comparison of early and late changes Early PF slopea 0.30 (0.64) 0.63 (0.04) Early (vs) late: Late PF slopea 0.37 (0.65) 0.90 (0.07) Chi-squared test (1) = 1.53, P = 0.22 Early PA slopea −4.55 (17.67) 491.62 (29.98) Early (vs) late: Late PA slopea −4.99 (14.53) 366.33 (23.37) Chi-squared test (1) = 0.14, P = 0.71 Early PF changeb 0.65 (3.44) 14.70 (0.65) Early (vs) late: Late PF changeb 1.73 (2.73) 16.52 (1.08) Chi-squared test (1) = 22.4, P < 0.001 Early PA changeb −14.29 (98.19) 11175.24 (473.61) Early (vs) late: Late PA changeb −21.32 (71.07) 7701.41 (391.53) Chi-squared test (1) = 1.91, P = 0.17 Variable Mean (SD) Variance (SE) Comparison of early and late changes Early PF slopea 0.30 (0.64) 0.63 (0.04) Early (vs) late: Late PF slopea 0.37 (0.65) 0.90 (0.07) Chi-squared test (1) = 1.53, P = 0.22 Early PA slopea −4.55 (17.67) 491.62 (29.98) Early (vs) late: Late PA slopea −4.99 (14.53) 366.33 (23.37) Chi-squared test (1) = 0.14, P = 0.71 Early PF changeb 0.65 (3.44) 14.70 (0.65) Early (vs) late: Late PF changeb 1.73 (2.73) 16.52 (1.08) Chi-squared test (1) = 22.4, P < 0.001 Early PA changeb −14.29 (98.19) 11175.24 (473.61) Early (vs) late: Late PA changeb −21.32 (71.07) 7701.41 (391.53) Chi-squared test (1) = 1.91, P = 0.17 aValues reflect estimated annual change, in minutes per week for PA and seconds for PF. bValues reflect overall cumulative change during time period of 5 years. Table 2. Mean (SD) PF and PA slope values and change score values for early (cycles 1–5) and later (cycles 5–9) portions of the study. Variable Mean (SD) Variance (SE) Comparison of early and late changes Early PF slopea 0.30 (0.64) 0.63 (0.04) Early (vs) late: Late PF slopea 0.37 (0.65) 0.90 (0.07) Chi-squared test (1) = 1.53, P = 0.22 Early PA slopea −4.55 (17.67) 491.62 (29.98) Early (vs) late: Late PA slopea −4.99 (14.53) 366.33 (23.37) Chi-squared test (1) = 0.14, P = 0.71 Early PF changeb 0.65 (3.44) 14.70 (0.65) Early (vs) late: Late PF changeb 1.73 (2.73) 16.52 (1.08) Chi-squared test (1) = 22.4, P < 0.001 Early PA changeb −14.29 (98.19) 11175.24 (473.61) Early (vs) late: Late PA changeb −21.32 (71.07) 7701.41 (391.53) Chi-squared test (1) = 1.91, P = 0.17 Variable Mean (SD) Variance (SE) Comparison of early and late changes Early PF slopea 0.30 (0.64) 0.63 (0.04) Early (vs) late: Late PF slopea 0.37 (0.65) 0.90 (0.07) Chi-squared test (1) = 1.53, P = 0.22 Early PA slopea −4.55 (17.67) 491.62 (29.98) Early (vs) late: Late PA slopea −4.99 (14.53) 366.33 (23.37) Chi-squared test (1) = 0.14, P = 0.71 Early PF changeb 0.65 (3.44) 14.70 (0.65) Early (vs) late: Late PF changeb 1.73 (2.73) 16.52 (1.08) Chi-squared test (1) = 22.4, P < 0.001 Early PA changeb −14.29 (98.19) 11175.24 (473.61) Early (vs) late: Late PA changeb −21.32 (71.07) 7701.41 (391.53) Chi-squared test (1) = 1.91, P = 0.17 aValues reflect estimated annual change, in minutes per week for PA and seconds for PF. bValues reflect overall cumulative change during time period of 5 years. Initial PF significantly predicted early and later PA slopes and these associations remained significant after adjustment for baseline age, MMSE and CDR scores (Table 3 and see Appendix 7 in the supplementary data, available at Age and Ageing online). Early PF slope did not significantly predict later PA slope (Table 3). Initial PA significantly predicted later, but not early PF slope (Table 3). Early PA slope was not a significant predictor of later PF slope (Table 3). The root mean square error or approximation and the comparative fit index (Table 3) indicate that the unadjusted and adjusted models had acceptable fit to the data. Table 3. Associations of initial and early slopes with later slopes for PF and PA. Unadjusted Adjusteda Standardised β (SE) P value Standardised β (SE) P value Predictive associations Initial PF value → Early PA slope −0.10 (0.03) <0.001 −0.09 (0.03) 0.003 Initial PA value → Early PF slope −0.03 (0.04) 0.47 −0.02 (0.03) 0.66 Initial PF value → Later PA slope −0.09 (0.03) 0.01 −0.07 (0.04) 0.05 Initial PA value → Later PF slope −0.09 (0.04) 0.04 −0.09 (0.03) 0.003 Early PF slope → Later PA slope −0.03 (0.04) 0.44 −0.02 (0.04) 0.69 Early PA slope → Later PF Slope −0.04 (0.04) 0.39 −0.03 (0.04) 0.45 Residual correlations Initial PF value ←→ Initial PA value −0.16 (0.03) <0.001 −0.13 (0.03) <0.001 Early PF slope ←→ Early PA slope −0.12 (0.03) <0.001 −0.11 (0.03) <0.001 Late PF slope ←→ Late PA slope −0.13 (0.04) <0.001 −0.12 (0.04) 0.002 Model fit Root mean square error of approximation (90% confidence Interval) 0.066 (0.062, 0.070) 0.058 (0.055, 0.061) Comparative fit index 0.95 0.94 Unadjusted Adjusteda Standardised β (SE) P value Standardised β (SE) P value Predictive associations Initial PF value → Early PA slope −0.10 (0.03) <0.001 −0.09 (0.03) 0.003 Initial PA value → Early PF slope −0.03 (0.04) 0.47 −0.02 (0.03) 0.66 Initial PF value → Later PA slope −0.09 (0.03) 0.01 −0.07 (0.04) 0.05 Initial PA value → Later PF slope −0.09 (0.04) 0.04 −0.09 (0.03) 0.003 Early PF slope → Later PA slope −0.03 (0.04) 0.44 −0.02 (0.04) 0.69 Early PA slope → Later PF Slope −0.04 (0.04) 0.39 −0.03 (0.04) 0.45 Residual correlations Initial PF value ←→ Initial PA value −0.16 (0.03) <0.001 −0.13 (0.03) <0.001 Early PF slope ←→ Early PA slope −0.12 (0.03) <0.001 −0.11 (0.03) <0.001 Late PF slope ←→ Late PA slope −0.13 (0.04) <0.001 −0.12 (0.04) 0.002 Model fit Root mean square error of approximation (90% confidence Interval) 0.066 (0.062, 0.070) 0.058 (0.055, 0.061) Comparative fit index 0.95 0.94 Notes: Initial = estimated intercept at year 1; Early slope = linear slope from year 1 to year 5 (using time points 1–5); Late slope = linear slope from year 5 to year 9 (using time points 5–9). aAll paths adjusted for baseline age, baseline MMSE, baseline CDR and baseline hypertension. Late slopes are additionally adjusted for year 9 CDR. Table 3. Associations of initial and early slopes with later slopes for PF and PA. Unadjusted Adjusteda Standardised β (SE) P value Standardised β (SE) P value Predictive associations Initial PF value → Early PA slope −0.10 (0.03) <0.001 −0.09 (0.03) 0.003 Initial PA value → Early PF slope −0.03 (0.04) 0.47 −0.02 (0.03) 0.66 Initial PF value → Later PA slope −0.09 (0.03) 0.01 −0.07 (0.04) 0.05 Initial PA value → Later PF slope −0.09 (0.04) 0.04 −0.09 (0.03) 0.003 Early PF slope → Later PA slope −0.03 (0.04) 0.44 −0.02 (0.04) 0.69 Early PA slope → Later PF Slope −0.04 (0.04) 0.39 −0.03 (0.04) 0.45 Residual correlations Initial PF value ←→ Initial PA value −0.16 (0.03) <0.001 −0.13 (0.03) <0.001 Early PF slope ←→ Early PA slope −0.12 (0.03) <0.001 −0.11 (0.03) <0.001 Late PF slope ←→ Late PA slope −0.13 (0.04) <0.001 −0.12 (0.04) 0.002 Model fit Root mean square error of approximation (90% confidence Interval) 0.066 (0.062, 0.070) 0.058 (0.055, 0.061) Comparative fit index 0.95 0.94 Unadjusted Adjusteda Standardised β (SE) P value Standardised β (SE) P value Predictive associations Initial PF value → Early PA slope −0.10 (0.03) <0.001 −0.09 (0.03) 0.003 Initial PA value → Early PF slope −0.03 (0.04) 0.47 −0.02 (0.03) 0.66 Initial PF value → Later PA slope −0.09 (0.03) 0.01 −0.07 (0.04) 0.05 Initial PA value → Later PF slope −0.09 (0.04) 0.04 −0.09 (0.03) 0.003 Early PF slope → Later PA slope −0.03 (0.04) 0.44 −0.02 (0.04) 0.69 Early PA slope → Later PF Slope −0.04 (0.04) 0.39 −0.03 (0.04) 0.45 Residual correlations Initial PF value ←→ Initial PA value −0.16 (0.03) <0.001 −0.13 (0.03) <0.001 Early PF slope ←→ Early PA slope −0.12 (0.03) <0.001 −0.11 (0.03) <0.001 Late PF slope ←→ Late PA slope −0.13 (0.04) <0.001 −0.12 (0.04) 0.002 Model fit Root mean square error of approximation (90% confidence Interval) 0.066 (0.062, 0.070) 0.058 (0.055, 0.061) Comparative fit index 0.95 0.94 Notes: Initial = estimated intercept at year 1; Early slope = linear slope from year 1 to year 5 (using time points 1–5); Late slope = linear slope from year 5 to year 9 (using time points 5–9). aAll paths adjusted for baseline age, baseline MMSE, baseline CDR and baseline hypertension. Late slopes are additionally adjusted for year 9 CDR. Secondary analyses yielded similar results. Baseline PF significantly predicted early and later PA changes (see Appendix 5 in the supplementary data, available at Age and Ageing online). These associations remained significant after adjustment for baseline age, MMSE score, CDR score and history of hypertension (see Appendix 5 in the supplementary data, available at Age and Ageing online). Baseline PA also significantly predicted early and later PF changes, and adjustment for covariates did not attenuate these associations (see Appendix 5 in the supplementary data, available at Age and Ageing online). Early PF change did not significantly predict later PA change; similarly, early PA change did not predict later PF change (see Appendix 5 in the supplementary data, available at Age and Ageing online). Sensitivity analyses with varying degrees of missing data and excluding extreme values of PA showed results were highly consistent, with the exception of a sample limited to only n = 450 with complete data at all nine cycles where all results were non-significant (see Appendix 6 in the supplementary data, available at Age and Ageing online). Discussion In this study, PF and PA both changed over time and were related in complex ways. We found initial PF strongly predicted both early and late PA slopes. Initial PA also predicted late but not early PF slope. There were no significant associations between early PF predicting later PA or early PA predicting later PF suggesting the starting values of PA and PF were more robust indicators than early slopes of subsequent PA and PF changes. Results were independent of factors related to PF or PA slope, including baseline age, MMSE score or CDR score, hypertension and converting to worse CDR scores, suggesting while covariates were related to PF and PA, they did not fully attenuate the observed associations. These results were also supported in sensitivity analyses allowing varying degrees of missing data, and when placing restrictions on upper limits of PA. Together, these results suggest that the associations are not driven solely by those at the highest or lowest levels of function. Change score analyses largely supported our primary findings (see Appendix 5 in the supplementary data, available at Age and Ageing online) except baseline PA significantly predicted early PF change. Most likely this discrepancy arose because slope values reflected an estimated annual change in seconds for PF, whereas change scores reflected overall cumulative change over 5 years (Table 2). This difference resulted in much larger values in change score than in primary analyses. These findings validate previous results [23], by demonstrating the relationship persists in a less highly selected, more representative sample of older adults, and when PF and PA measures are collected annually as opposed to every 4–5 years. Other studies examining the relationships between PA and PF, in older adults, have yielded mixed results. A meta-analysis concluded high-intensity but not moderate- or low-intensity, exercise promoted faster-walking speed [22]. The LIFE study, a large-scale PA intervention found moderate-intensity, multi-component PA significantly reduced physical disability but did not promote PF [12]. Better understanding the nature of the relationship between PF and PA will highlight potential mechanisms for improving these measures. It may also identify pathways to reduce other health-related outcomes, such as major mobility disability, falls, hospitalisations, poor cardiovascular health and declining cognitive function. An important consideration when interpreting these results is the complex nature of both PF and PA, especially in older adults. For example, PF may be influenced by a large number of factors, including disorders and diseases that contribute to disuse, muscle loss and weight gain in older adults, which adds a complexity in understanding the true nature of this relationship. Some diseases of ageing may affect PF progress slowly without overt clinical symptoms until years after underlying pathology has begun, which makes measuring these contributing factors difficult. An example of this is the accumulation of white matter lesions with age, which are common among older adults [31] and are associated with slower gait [31, 32]. Similarly, PA is affected by the motivation to be physically active, depressive symptoms, cognitive health and socioeconomic factors (i.e. safe place to walk, money to buy shoes or exercise equipment, etc.). PA is also likely conditional on maintaining some sort of baseline PF level. Thus, it is important to consider as many of these complex factors in older adults as we continue to disentangle these complex relationships. This study has several strengths, including the representative nature of the cohort, the duration of time over which participants were followed and the large number of visits within that time. Several potential limitations should also be considered. PA was self-reported, not objectively measured and there may be a discrepancy between what participants believe they usually do, and what they actually perform. If most participants exaggerated their activity, our results may be inflated; if the reverse is the case, our results would be underestimations. Future studies should examine these associations with more objective measures like accelerometers. Furthermore, the only type of PA measured is self-reported time spent walking at a moderate-intensity. However, a large proportion of older adults reports low- or moderate-intensity walking as their only PA, even after interventions to increase PA [7, 10]. Participant survivor bias may have underestimated the associations meaning participants included in these analyses were healthier and likely at a lower risk for PF impairment because they were healthy enough to make it to every visit over nine cycles and a natural variation in PF is lost. We see this supported in sensitivity analyses limited to n = 450 with complete data, and as we noted, those excluded from our sample were less healthy than those included. This is also supported by older age being associated with a greater change in early PF slope but a slower change in late PF slope, indicating less variation in PF in later cycles where those surviving are likely more similar than those lost to follow-up. This study found a bidirectional relationship between initial PF and PA predicting later PA and PF. Thus, a person with difficulties in PF may decrease PA, and in turn, worsen in PF. Conversely, a person who declines in PA, perhaps due to inclement weather or poor social surroundings, may decline in PF, resulting in a struggle to return to previous PA levels, even when or if the barrier is removed. Of course, this is an over-simplification of all the physical, social and health-related challenges older adults face that may affect PF and PA, but the idea suggests helping older adults maintain PF and PA will work at keeping them both healthier and free of mobility impairments. Interventions to improve early PF may support increased PA, and vice versa, thus altering a vicious cycle of lower PF leading to lower PA, to a virtuous cycle of higher PF conveying higher PA. Such a shift could reduce associated risk factors, including impaired cardiovascular health [9], cognitive function [8], mood and increased social isolation [16]. More work is needed to determine if targeted interventions can effectively reduce the risk for PA decline and associated co-morbidities in older adults, and to determine if one factor more strongly predicts the other. Key points: Physical function predicts decline in physical activity, in the short and long-term. Physical activity predicted long-term decline in physical function. Interventions to improve Physical Function could support increased Physical Activity, and vice versa, thus altering a vicious cycle to a virtuous cycle. Supplementary data Supplementary data mentioned in the text are available to subscribers in Age and Ageing online. Acknowledgements The authors would like to acknowledge K.M. in MYHAT project coordination, and E.J. for compiling and providing the analytic datasets. All the authors meet the criteria for authorship stated in the Uniform Requirements for Manuscripts Submitted to Biomedical Journals. Conflicts of interest None. Funding This work was supported in part by the National Institute on Aging (Grant nos. R01 AG023651 and K24 AG022035). The sponsor had no role in the study or manuscript writing. References 1 Werner CA . The Older Population: 2010 Census Briefs . Washington, DC : U.S. Department of Commerce , 2012 . 2 Guralnik J , Ferrucci L , Pieper CF et al. . Lower extremity function and subsequent disability: consistency across studies, predictive models, and value of gait speed alone compared with the Short Physical Performance Battery . J Gerontol 2000 ; 55A : M221 – 31 . Google Scholar CrossRef Search ADS 3 Guralnik JM , Ferrucci L , Simonsick EM , Salive ME , Wallace RB . Lower-extremity function in persons over the age of 70 years as a predictor of subsequent disability . N Engl J Med 1995 ; 332 : 556 – 61 . Google Scholar CrossRef Search ADS PubMed 4 Hardy SE , Perera S , Roumani YF , Chandler JM , Studenski SA . Improvement in usual gait speed predicts better survival in older adults . J Am Geriatr Soc 2007 ; 55 : 1727 – 34 . Google Scholar CrossRef Search ADS PubMed 5 Studenski S , Perera S , Patel K et al. . Gait speed and survival in older adults . JAMA 2011 ; 305 : 50 – 8 . Google Scholar CrossRef Search ADS PubMed 6 Baker G , Gray SR , Wright A et al. . The effect of a pedometer-based community walking intervention ‘Walking for Wellbeing in the West’ on physical activity levels and health outcomes: a 12-week randomized controlled trial . Int J Behav Nutr Phys Act 2008 ; 5 : 44 . Google Scholar CrossRef Search ADS PubMed 7 Berkemeyer K , Wijndaele K , White T et al. . The descriptive epidemiology of accelerometer-measured physical activity in older adults . Int J Behav Nutr Phys Act 2016 ; 13 : 2 . Google Scholar CrossRef Search ADS PubMed 8 Falck RS , Davis JC , Liu-Ambrose T . What is the association between sedentary behaviour and cognitive function? A systematic review . Br J Sports Med 2017 ; 51 : 800 – 11 . Google Scholar CrossRef Search ADS PubMed 9 Wilmot EG , Edwardson CL , Achana FA et al. . Sedentary time in adults and the association with diabetes, cardiovascular disease and death: systematic review and meta-analysis . Diabetologia 2012 ; 55 : 2895 – 2905 . Google Scholar CrossRef Search ADS PubMed 10 Kimura M , Moriyasu A , Kumagai S et al. . Community-based intervention to improve dietary habits and promote physical activity among older adults: a cluster randomized trial . BMC Geriatr 2013 ; 13 : 8 . Google Scholar CrossRef Search ADS PubMed 11 Mutrie M , Doolin O , Fitzsimons C et al. . Increasing older adults’ walking through primary care: results of a pilot randomized controlled trial . Fam Pract 2012 ; 29 : 633 – 42 . Google Scholar CrossRef Search ADS PubMed 12 Pahor M , Guralnik JM , Ambrosius WT et al. . Effect of structured physical activity on prevention of major mobility disability in older adults: the LIFE study randomized clinical trial . JAMA 2014 ; 311 : 2387 – 96 . Google Scholar CrossRef Search ADS PubMed 13 Sink KM , Espeland MA , Castro CM et al. . Effect of a 24-month physical activity intervention vs health education on cognitive outcomes in sedentary older adults: the LIFE randomized trial . JAMA 2015 ; 314 : 781 – 90 . Google Scholar CrossRef Search ADS PubMed 14 Buracchio T , Dodge HH , Howieson D , Wasserman D , Kaye J . The trajectory of gait speed preceding mild cognitive impairment . Arch Neurol 2010 ; 67 : 980 – 6 . Google Scholar CrossRef Search ADS PubMed 15 Quan M , Xun P , Chen C et al. . Walking pace and the risk of cognitive decline and dementia in elderly populations: a meta-analysis of prospective cohort studies . J Gerontol A Biol Sci Med Sci 2017 ; 72 : 266 – 70 . Google Scholar CrossRef Search ADS PubMed 16 Bauman A , Merom D , Bull FC , Buchner DM , Fiatarone Singh MA . Updating the evidence for physical activity: summative reviews of the epidemiological evidence, prevalence, and interventions to promote ‘active aging’ . Gerontologist 2016 ; 56 ( Suppl 2 ): S268 – 280 . Google Scholar CrossRef Search ADS PubMed 17 Fox KR , Ku PW , Hillsdon M et al. . Objectively assessed physical activity and lower limb function and prospective associations with mortality and newly diagnosed disease in UK older adults: an OPAL four-year follow-up study . Age Ageing 2015 ; 44 : 261 – 8 . Google Scholar CrossRef Search ADS PubMed 18 Macmillan F , Fitzsimons C , Black K et al. . West End Walkers 65+: a randomised controlled trial of a primary care-based walking intervention for older adults: study rationale and design . BMC Public Health 2011 ; 11 : 120 . Google Scholar CrossRef Search ADS PubMed 19 Perera S , Patel KV , Rosano C et al. . Gait speed predicts incident disability: a pooled analysis . J Gerontol A Biol Sci Med Sci 2016 ; 71 : 63 – 71 . Google Scholar CrossRef Search ADS PubMed 20 Brach JS , Lowry K , Perera S et al. . Improving motor control in walking: a randomized clinical trial in older adults with subclinical walking difficulty . Arch Phys Med Rehabil 2015 ; 96 : 388 – 94 . Google Scholar CrossRef Search ADS PubMed 21 Brach JS , Van Swearingen JM , Perera S , Wert DM , Studenski S . Motor learning versus standard walking exercise in older adults with subclinical gait dysfunction: a randomized clinical trial . J Am Geriatr Soc 2013 ; 61 : 1879 – 86 . Google Scholar CrossRef Search ADS PubMed 22 Lopopolo RB , Greco M , Sullivan D , Craik RL , Mangione KK . Effect of therapeutic exercise on gait speed in community-dwelling elderly people: a meta-analysis . Phys Ther 2006 ; 86 : 520 – 40 . Google Scholar PubMed 23 Best JR , Liu-Ambrose T , Metti AL et al. . Longitudinal associations between walking speed and amount of time spent walking over a 9-year period in older women and men . J Gerontol A Biol Sci Med Sci 2017 Jun 22 [Epub ahead of print]. 24 Ganguli M , Snitz B , Vander Bilt J , Chang CC . How much do depressive symptoms affect cognition at the population level? The Monongahela–Youghiogheny Healthy Aging Team (MYHAT) study . Int J Geriatr Psychiatry 2009 ; 24 : 1277 – 84 . Google Scholar CrossRef Search ADS PubMed 25 Mathias S , Nayak US , Isaacs B . Balance in elderly patients: the ‘get-up and go’ test . Arch Phys Med Rehabil 1986 ; 67 : 387 – 9 . Google Scholar PubMed 26 Folstein MF , Folstein SE , McHugh PR . ‘Mini-mental state’. A practical method for grading the cognitive state of patients for the clinician . J Psychiatr Res 1975 ; 12 : 189 – 98 . Google Scholar CrossRef Search ADS PubMed 27 Hughes CP , Berg L , Danziger WL , Coben LA , Martin RL . A new clinical scale for the staging of dementia . Br J Psychiatry 1982 ; 140 : 566 – 72 . Google Scholar CrossRef Search ADS PubMed 28 Morris JC . The Clinical Dementia Rating (CDR): current version and scoring rules . Neurology 1993 ; 43 : 2412 – 4 . Google Scholar CrossRef Search ADS PubMed 29 Selig JP , Little TD . Autoregressive and cross-lagged panel analysis for longitudinal data . New York : The Guilford Press , 2012 . 30 Muthén LK , Muthén B . Mplus: Statistical analysis with latent variables , 7th edition . Los Angeles, CA : Muthén & Muthén , 2012 . 31 Smith EE , O’Donnell M , Dagenais G et al. . Early cerebral small vessel disease and brain volume, cognition, and gait . Ann Neurol 2015 ; 77 : 251 – 61 . Google Scholar CrossRef Search ADS PubMed 32 Bolandzadeh N , Liu-Ambrose T , Aizenstein H et al. . Pathways linking regional hyperintensities in the brain and slower gait . NeuroImage 2014 ; 99C : 7 – 13 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of the British Geriatrics Society. All rights reserved. For permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Age and AgeingOxford University Press

Published: Mar 13, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off