Trajectories in muscular strength and physical function among men with and without prostate cancer in the health aging and body composition study

Objectives To examine and compare changes in strength and physical function from pre- to post-diagnosis among men with prostate cancer (PC, [cases]) and matched non-cancer controls identified from the Health, Aging and Body Composition (Health ABC) study. Materials and methods We conducted a longitudinal analysis of 2 strength and 3 physical function-based measures among both cases and controls, identified from a large cohort of community living older adults enrolled in the Health ABC study. We plotted trajectories for each measure and compared cases vs. controls from the point of diagnosis onwards using mixed-effects regression models. For cases only, we examined predictors of poor strength or physical function. Results We identified 117 PC cases and 453 matched non-cancer controls (50% African Americans). At baseline, there were no differences between cases and controls in demographic factors, comorbidities or self-reported physical function; however, cases had slightly better grip strength (44.6 kg vs. 41.0 kg, p<0.01), quadriceps strength (360.5 Nm vs. 338.7 Nm, p = 0.02) and Health ABC physical performance battery scores (2.4 vs. 2.3, p = 0.01). All men experienced similar declines in strength and physical function over an equivalent amount of time. The loss of quad strength was most notable, with losses of nearly two-thirds of baseline strength over approximately 7 years of follow up. Conclusions Among both cases and controls, strength and physical function decline with increasing age. The largest declines were seen in lower body strength. Regular assessments should guide lifestyle interventions that can offset age- and treatment-related declines among men with PC.


Materials and methods
We conducted a longitudinal analysis of 2 strength and 3 physical function-based measures among both cases and controls, identified from a large cohort of community living older adults enrolled in the Health ABC study. We plotted trajectories for each measure and compared cases vs. controls from the point of diagnosis onwards using mixed-effects regression models. For cases only, we examined predictors of poor strength or physical function.

Results
We identified 117 PC cases and 453 matched non-cancer controls (50% African Americans). At baseline, there were no differences between cases and controls in demographic factors, comorbidities or self-reported physical function; however, cases had slightly better grip strength (44. 6  clinically meaningful changes in strength and/or physical function among patients with PC. We hypothesized that patients with PC would have similar trajectories of strength and physical function to non-cancer controls pre-diagnosis, but that following diagnosis, men with PC would experience greater declines than non-cancer controls. Further we hypothesized that among cases, treatment with ADT, more comorbidities, lower levels of physical activity and a greater severity of depressive symptoms would be associated with worse trajectories in strength and physical function than non-cancer controls.

Materials and methods
The Health ABC Study enrolled 3075 black and white community dwelling older adults. Participants were recruited from a random sample of white

Study sample
For the current analyses, we identified 117 adjudicated PC diagnoses that were confirmed by pathology reports or other medical record information over 10 years of follow-up (study visits 1,2,4,6,8,10). An analytic dataset was created to include socio-demographics, strength and physical function outcomes data for all 117 PC cases and 468 matched non-cancer controls (Fig 1). Each participant was assigned an index visit with the participant having had to attempt the 400m walk during the visit and have attempted a 400m walk at a later visit. The Health ABC study collected data on the 400m walk at years 1 (baseline), 2, 4, 6, 8, and 10. For each PC case with an incident diagnosis since Health ABC baseline, we used the last pre-diagnosis visit as an index visit. Frequency matching, weighted by race, was used to randomly assign an index visit for the non-cancer controls at a ratio of 4:1 for a total analytic sample of 585 men.

Measures
Muscular strength. Muscular strength was assessed with 2 objective measures to capture upper and lower body strength: 1) Isometric grip strength (kg) using a hand-held dynamometer (JAMAR Technologies, Inc., Hatfield, PA); and 2) Quadriceps (quad) strength (Nm) using an isokinetic dynamometer.
Physical function. Physical function was assessed with a further 3 objective measures: 1) a 400m walk test recorded as the proportion of men completing the test at each visit (failure to complete 400m in 15 minutes is indicative of major mobility disability (MMD) [19,20]); 2) Health ABC physical performance battery (HABCPPB), a combination test (5x timed chair stands, double and single-leg balance tests held for 30 seconds, and narrow walk test for balance and 6m gait speed) designed to measure a wide range of function in the Health ABC cohort of well-functioning older adults. Scores for the HABCPPB are on a scale of 0-4 with a scoring system designed to limit floor and ceiling effects [21]; and 3) 20m gait speed at usual pace over a 20m course assessed on a continuous scale in meters per second (m/s). The correlations between all measures of muscular strength and physical function ranged from 0.24 and 0.37.
Covariates. We examined several covariates (as potential confounders of the relationship between PC status and our primary outcomes) and their relationship with muscular strength and physical function among both cases and controls at baseline, index visit and a follow-up visit 2-3 years post-diagnosis. Demographic variables included age at index visit, race, education and marital status. We also considered the presence of multiple comorbidities including diabetes, previous heart attack, hypertension, previous stroke, chronic heart failure (CHF), arthritis, other cancers and total number of comorbidities. Depressive symptoms were measured with the Center for Epidemiologic Studies Depression Scale (CESD) [22]. Body mass index (BMI; kg/m 2 ) was used as an index of obesity status. Self-reported functional status was further assessed by the number of falls in the previous 12 months, perceived capacity to easily walk a quarter mile, lift or carry 10lbs, and having engaged in high intensity exercise over the past 12 months, or past 7 days.

Statistical analysis
Summary statistics were calculated for both the PC cases and controls at the Health ABC baseline (Table 1), index visit, and at a short-term follow-up visit (2 or 3 years depending on the index year). T-tests were used to compare continuous measures and Chi-square tests were used for categorical measures. To visualize longitudinal data, we plotted the means of each outcome by visit year, centered at the index visit. Nonparametric loess regression was used to plot a smooth curve enabling a better visualization of the effect of time (actual time from diagnosis/ index visit) on outcomes. Repeated measures mixed-effects regression analyses were used to test for differences in the effect of time between the two groups and to test whether the slope of the function among cases was different pre-to post-diagnosis. Cubic, quadratic, and linear terms for time were investigated to see which provided the best fit to the data and to verify what was seen in the loess plots. Residual plots were created to verify the assumptions of the model (not shown). To assess the association of relevant covariates with trajectories of upper body strength (grip strength), lower body strength (quad strength) and gait (20m usual pace gait speed), we placed PC cases into 1 of 4 groups for each outcome as follows: group 1decreasing; group 2 -consistently low; group 3 -consistently high; group 4-increasing. Cut offs for determining a meaningful change (±) and the groups men were placed into (decreasing, consistently low, consistently high or increasing) are shown in Table 3 and were based on published literature, except in the case of quad strength (where a 1.0 standard deviation (SD) was used). Bivariate association of all covariates was by group (1)(2)(3)(4). Loess plots were produced with the ggplot2 package in R 3.4.1 and all other analyses were done using SAS version 9.4 (SAS institute Inc. Cary, NC).

Sample characteristics
Briefly, we identified 117 men who reported a diagnosis of PC at a Health ABC follow-up visit and 468 men without cancer, who were randomly chosen from remaining HABC cohort, weighted by race (Fig 1). S1 Table shows sample characteristics at index visit (last visit before diagnosis, including a breakdown of the proportion of patients and matched controls who had their index visit in years 1 (baseline), 2, 4, 6 and 8, which ranged from 9% in year 8 to 33.35% in year 2.The mean age of the PC cases and controls at baseline was similar (74.0 vs. 74.1 years, respectively). Approximately half of cases and controls were black (50.6% vs. 50.4%) with the majority of men having a post-secondary education and being married or partnered. There were no statistically significant differences (all p's >0.05) between cases and controls for demographic factors, comorbidities, self-reported functional capacity or levels of physical activity. There were, however, significant differences in grip strength (mean ± SD: 44.6 ± 8.6kg vs. 41.0 ± 8.7kg; p <.001), quad strength (mean ± SD: 360.5 ± 81.5Nm vs. 338.7 ± 88.5Nm; p <.001), and in HABCPPB scores (mean ± SD: 2.4 ± 0.4 vs. 2.3 ± 0.5 p = 0.01), at baseline. There were no statistically significant differences between the proportion of cases vs. controls completing 400m walk or for 20m gait speed. Four people included in our dataset (all cases) were taking ADT drugs identified in the study medications list (Leuprolide, Goserelin, Buserelin, Nafarelin). Due the small numbers we did not conduct any further analyses using ADT as a sub-population. At index visit, a significantly higher number of controls had a previous heart attack than cases (23.3% vs. 13.7% respectively) while quad strength and HABCPPB scores were no longer significantly different between cases and controls.      Table 2 shows the results of repeated measures models for each outcome. All models include a linear effect for time. For quad strength, HABCPPB, and gait speed there was a significant quadratic effect for time, a comparison of functional trajectory in cases vs controls, and effect of time since diagnosis in PC cases. Grip strength declined significantly over time in both groups (p<0.01), though less quickly in cancer cases compared to controls (p = 0.02). Among cases, from the point of diagnosis onwards grip strength declined more rapidly than before diagnosis (p = 0.02). Quad strength declined significantly over time among both cases and controls (p<0.01) and at the same rate (p = 0.52). For the proportion of cases and controls completing the 400m walk test, the effects of both time (p<0.01) and time since diagnosis (p<0.04) were significant, though in opposite directions. In general, fewer cases and controls completed Table 3 the 400m walk test over time but from the point of diagnosis onwards, an increasing proportion of cases remaining in the study completed the test. In terms of the HABCPPB scores, performance declined significantly over time (p<0.01), though this was equivalent in cases and in controls and the rate of decline did not change after diagnosis. For 20m gait speed, both cases and controls experienced declines, with no differences in trajectory (p = 0.89), though cases declined at a slightly slower rate from the point of diagnosis onwards (p<0.02).  symptoms (p = 0.01) were significantly more likely to maintain a high level of grip strength. Men with PC who were older at index visit (p <.01), who had more follow up years in the study (p�.01), and who were taking PC drugs (p<0.01) were significantly more likely to have decreased quad strength. PC cases who had arthritis (p = 0.02) and who had more years of follow up in the study (p = 0.01) were significantly more likely to have experienced a decrease in their 20m gait speed.

Discussion
In the current study, the trajectories of strength and physical function among both older men diagnosed with PC and among race-matched men without cancer declined at similar rates over an equivalent amount of time, with the greatest declines seen in lower leg strength as indexed by quad strength testing. Starting from approximately 80 years of age, both cases and controls lost around 70% of their quad strength over the following 6 years compared with having lost only 7% over the previous decade. Given the low rates of ADT use among this sample of men diagnosed with prostate cancer, it appears the majority of the decline in lower body strength was due to aging related changes. Unexpectedly, the overall rate of decline for grip strength was slower in men with PC compared to men without cancer, which may be a result of the fact that PC cases had significantly greater grip strength than controls at study baseline and at index visit unlike for other measures of strength and function. Previous studies examining upper and lower body strength changes among PC patients (both receiving and not receiving ADT) have found mixed results. This may relate to the specific tests being used or the time frame across which patients were tested. For example, studies using maximal leg press have found no differences in strength while those examining leg extension have found differences (Cheung). Among men with PC, mobility (proportion completing 400m walk and 20m gait speed) remained relatively intact; however, changes in muscle strength were apparent. The majority of men with PC either had significant declines in physical performance or maintained a low level of physical performance on upper and lower body strength tests (63% and 69% respectively) and gait (78%). Our study differed from previously reported data in noteworthy ways. Alibhai and colleagues [11], examined differences in physical function between men with PC (some receiving ADT) and non-cancer controls over time and found that grip strength declined only among PC patients receiving ADT. In our study, grip strength decreased in both cases and controls, but was the most dissimilar outcome between the two groups. These estimates may have also been affected by a smaller number of participants at the beginning and end of the follow-up periods. The Alibhai study found no changes over time among any group for the timed up and go test (TUG), used as a measure of lower body strength, in contrast to our finding of a large decline in quad strength over time. The discrepancy may be due to their shorter follow up (12 months) and the slightly different testing components [23]. The most significant difference in our study was that physical function was examined before men were diagnosed with PC, illustrating the acute effects of diagnosis and first line treatments on physical function over time.
To our knowledge, no other studies have used objective measures of functioning to evaluate change across this period. Moreover, we did not discriminate between men who were on/off ADT. A relatively small proportion of men in this study had ADT exposure, highlighting the significance of the change in quad strength with advancing age.
Our sample contained 50% African Americans, vs only 6% African Americans in the SEER cohort [17], a critical contribution considering that the majority of PC literature underrepresents African Americans. Reeve et al reported significant differences in physical health-related quality of life (HRQL) between cases and controls over 6 months following diagnosis, including worse depressive symptoms. Except for grip strength, we did not find significant differences between cases and controls in objective measures of physical performance over time. Furthermore, Chambers and colleagues [24] examined trajectories of self-reported HRQL in 2250 men recently diagnosed with PC over a 6-year period, reporting poorer physical HRQL was predicted by older age, lower education, lower income, comorbidities and receiving hormonal therapy. We found no associations between level of education, number of comorbidities and any physical performance measure. Overall, among men with PC, increasing age was the factor most consistently associated with meaningful declines in function over time.
Previous studies have reported that up to 50% of men diagnosed with PC will at some point receive ADT [7] . Though use may be declining for men with low-risk early stage disease, due to its adverse effects [25], ADT remains the cornerstone of treatment for advanced PC. Preparing for possible disease progression in conjunction with inevitable aging-related functional declines may promote maintenance of physical health-related QOL among men diagnosed with early-stage disease. Largely in contrast to our hypothesis, we found men diagnosed with PC who were healthy at baseline (eligibility criteria for Health ABC study) did not experience accelerated declines in long-term physical function when compared to men without cancer. Similar self-reported changes in long-term physical function have been described among women with breast cancer [26] but other findings suggest a cancer history is associated with marked declines in short-term function [27]. In our study, all older men experienced decreases in their function over time, with the largest declines seen in lower body strength. Helping all older men, particularly those diagnosed with PC, to maintain function as they age is an important clinical goal. Previous studies show that physical function declines less rapidly during exercise intervention [28]. Therefore, it is critical to identify strategies for maintaining effective and sustainable lifestyle behaviors that support optimal physical function [29]. In addition, aerobic training to support cardiovascular health is also critical among men who are at risk of developing CVD [30,31]. Our finding that older men are, in general, not engaging in regular exercise provides an opportunity for such interventions.
Maintaining muscle strength and quality [32] through the adoption of exercise training may be particularly important for promoting healthy survivorship among patients with PC. In the RENEW study [28], cancer survivors (breast, prostate and colorectal), in a wait-listed control group experienced significant declines in self-reported physical function over 12 months when compared with those receiving a home-based exercise intervention. In a separate study, Klepin et al. [14] found that lower extremity physical performance predicted both 2-year progression to disability and overall survival among cancer survivors from the Health ABC cohort. Therefore, early intervention to offset functional declines in aging patients with PC may be critical, with diagnosis providing an important opportunity (teachable moment) [33] to engage with these men.
We believe this study had a number of important strengths including: 1) the diverse sample (50% African Americans); with 2) multiple years of follow-up; which allowed us to 3) characterize change in physical function from pre-diagnosis; using 4) objective measures of function on multiple domains; and 5) a matched non-cancer cohort. We also recognize the following limitations: We did not have reliable information on the disease stage at diagnosis, making comparison by stage within and across other studies impossible. Relatively few men were on ADT, suggesting the majority had early stage disease at diagnosis [34] and making it unlikely that we could detect significant differences between groups. A further limitation of the study is the lack of other treatment information such as surgery or radiation as these may also have impacted aspects of strength or physical function, especially in the early post diagnosis period. Men with high risk disease or more advanced cancer at diagnosis would also likely have different treatments and/or psychosocial reactions to diagnosis. Additionally, while we did not find specific associations between BMI and trajectories of physical function over time, the use of more precise measures of body composition may have provided better information. Finally, we did not evaluate diet or nutrition, a critical factor contributing to changes in body composition in men who begin hormone therapy [35].
In conclusion, our findings reinforce the importance of health status at diagnosis. Furthermore, as part of multidisciplinary care, all aging men should undergo assessments of strength and physical function to identify those with deficits. Routine assessments will provide a basis for the prescription of exercise programs and for assessing the risk of disablement and mortality as these men age. Older patients with PC should be prescribed resistance exercise training to promote the maintenance of muscular strength [36,37].
Supporting information S1 Table. Characteristics of the study sample at last pre-diagnosis (index) visit.