Predictors of change in physical function among older adults in response to long-term, structured physical activity: The LIFE Study

Abstract

Objective

To evaluate the extent of variability in functional responses among participants in the LIFE study, and to identify the relative contributions of intervention adherence, physical activity, and demographic and health characteristics to this variability.

Design

Secondary analysis of the Lifestyle Interventions and Independence for Elders (LIFE) study.

Setting

Multicenter U.S. institutions participating in the LIFE study.

Participants

A volunteer sample of 1635 sedentary men and women aged 70 to 89 years who were able to walk 400 m, but had physical limitations, defined as a score on the Short Physical Performance Battery (SPPB) of ≤9.

Interventions

Moderate-intensity physical activity (PA, n=818) consisting of aerobic, resistance and flexibility exercises performed both center-based (twice/wk) and in or around the home environment (3-4 times/wk) or health education (HE, n=817) consisting of weekly to monthly workshops covering relevant health information.

Main Outcome Measures

Physical function: gait speed over 400-m and lower extremity function (SPPB) assessed at baseline, six, twelve, and 24 months.

Results

Greater baseline physical function (gait speed and SPPB score) was inversely associated with Δ gait speed (regression coefficient β=−0.185, p<0.001) and ΔSPPB score (β=−0.365, p<0.001), while greater number of steps per day measured by accelerometry was positively associated with Δ gait speed (β=0.035, p<0.001) and Δ SPPB score (β=0.525, p<0.001). Other baseline factors associated with positive Δ gait speed and/or SPPB score include younger age (p<0.001), lower body mass index (p<0.001), and higher self-reported physical activity (p=0.002).

Conclusions

Several demographic and physical activity-related factors were associated with the extent of Δ functional outcomes among participants in the LIFE study. These factors should be considered when designing interventions for improving physical function among older adults with limited mobility.

In older adults, limitations in physical abilities are associated with the onset of disability and loss of independence as well as increased rates of cardiovascular morbidity and mortality.^1-3 As the number of older men and women continues to rise in the U.S. and worldwide,^4-6 the maintenance of functional status and preservation of physical independence presents a major challenge with significant clinical implications. As a result, the development of scalable interventions to maintain the function and independence of older persons is an important public health goal.

To date, structured physical activity (PA; i.e. exercise) is the only intervention consistently demonstrated to attenuate functional decline among older adults, as numerous studies have reported that regular PA improves performance on a variety of functional tasks.^7-10 Despite these benefits, the change in function can be quite variable – even under well-controlled experimental conditions.¹¹ As a result, the benefits of regular PA on functional status are absent in many seniors despite strong adherence to training.^12-14 For instance, our group recently reported that – compared to health education – a long-term (median 2.7 years) PA intervention reduced the incidence of major mobility disability by 18% among older adults with functional limitations. Yet despite this overall benefit, approximately 30% of the individuals in the PA intervention still experienced major mobility disability.⁷

These results suggest that although PA is the best-known method for preserving physical function among older adults, exercise alone may be insufficient for preventing disability among at least some groups of older adults.¹⁵ Therefore, a need exists to explore the development of adjuvant therapies designed to enhance the beneficial effects of PA in improving the functional status of older adults. To successfully identify such therapies, additional information is first needed regarding the extent and sources of variability in functional responses of older adults to PA. The present investigation aimed to provide such information using data collected in the Lifestyle Interventions and Independence for Elders (LIFE) study. The LIFE study was a multi-site, phase three randomized controlled trial designed to determine if long-term PA—when compared to health education alone--reduces the incidence of major mobility disability among older adults with functional limitations.¹⁶ The objectives of this analysis were to evaluate the extent of variability in functional responses among participants in the LIFE study, and to identify interpersonal factors associated with the extent of this variability.

Methods

Study Population

The LIFE Study was a multicenter randomized controlled trial comparing a moderate-intensity PA program with a “successful aging” health education program in 1,635 nondisabled, community-dwelling persons aged 70 to 89 years.¹⁶ The assembly of this cohort has been described in detail elsewhere.¹⁷ In brief, eligibility criteria included low physical activity, defined as self-report of fewer than 20 min/wk of structured exercise over the previous month and fewer than 125 min/wk of any moderate-intensity physical activities according to the modified 18-item Community Healthy Activities Model Program for Seniors (CHAMPS) questionnaire,¹⁸ and lower extremity functional limitations, defined as a Short Physical Performance Battery (SPPB) score of ≤9^{19, 20} but able to complete a 400-m walk test in 15 minutes without sitting, leaning, or the help of another person. The institutional review boards of all participating centers approved all study procedures and all participants provided written documentation of informed consent prior to participating.

Interventions

The PA intervention involved walking, with a goal of 150 min/wk, strength, flexibility, and balance training. ¹⁶ The intervention included attendance at 2 center-based visits per week and home-based activity 3 to 4 times per week for the duration of the study. A protocol was in place to restart the intervention for the participants who suspended the physical activity for medical reasons. The physical activity sessions were individualized and progressed toward a goal of 30 minutes of walking daily at moderate intensity, 10 minutes of primarily lower extremity strength training by means of ankle weights (2 sets of 10 repetitions), 10 minutes of balance training, and large muscle group flexibility exercises. Intensity of PA was assessed using the Borg scale,²¹ a numerical rating of perceived exertion that ranges from six (minimal exertion) to twenty (maximal exertion). Intensity of PA was gradually increased over the first three weeks of the intervention. Participants were asked to walk at an intensity of 13 (somewhat hard) on the Borg scale and perform strength training at an intensity of 15-16 (hard).

The health education program (HE) consisted of weekly workshops held during the first 26 weeks of the intervention, followed by monthly sessions for the remainder of the intervention (bimonthly attendance was optional).⁷ Workshops covered topics relevant to older adults, such as nutrition, accessing reliable health information, recommended preventative services and screenings, etc. The workshops did not cover physical activity-related topics, but the program included a five- to ten-minute instructor-led program of gentle upper extremity stretching or flexibility exercises.

Measurements

Participants were assessed every 6 months at clinic visits. Home, telephone, and proxy assessments were attempted if the participants could not come to the clinic. The assessment staff was blinded to the intervention and remained separate from the intervention team. Participants were asked not to disclose their assigned group and not to talk about their interventions during the assessment. Self-reported physical activity was ascertained by a separate set of unblinded assessors.

The main baseline assessments relevant to the present manuscript include self-reported demographic and contact information, medication inventory, physical examination, self-reported walking and resistance training assessed with the CHAMPS questionnaire²² steps per day assessed with accelerometry over 7-days^{23, 24} (Actigraph Inc), cognitive testing based on the Modified MiniMental State Examination (3MSE),²⁵ depression assessed with the Center for Epidemiological Studies Depression Scale (CESD),²⁶ sleep quality based on the Pittsburg Sleep Quality Index (PSQI),²⁷ 400-m walk test,²⁸ the SPPB,²⁹ body weight, blood pressure, and pulse rate. These measures were repeated during follow-up at varied intervals. Details of these measures and their frequency are described elsewhere.¹⁶ Race and ethnicity were reported by the participants and were collected according to National Institutes of Health requirements.

Outcome Assessment

The functional outcomes included usual-paced gait speed measured during a 400 m test and performance on the SPPB. These tests have high clinical relevance, as they have proven reliable and valid for predicting adverse health outcomes among seniors.^{1, 3, 19, 30} During the 400m walk test, time and distance walked were recorded, and gait speed was determined in meters/second. For participants who were unable to complete the 400m walk follow-up visits, gait speed was determined over 80m. The SPPB consisted of 4-m walk at usual pace, a timed repeated chair stand, and 3 increasingly difficult standing balance tests Each measure was assigned a categorical score ranging from 0 (inability to complete the test) to 4 (best performance). A summary score ranging from 0 (worst performers) to 12 (best performers) was calculated by summing the 3 component scores.

Major mobility disability (MMD) was defined as the inability to complete the 400m walk test in 15 minutes or less without sitting and without the help of another person or walker. Use of a cane was acceptable. Participants were asked to walk 400 m at their usual pace, without overexerting, on a 20-m course for 10 laps (40 m/lap). Participants were allowed to stop for up to 1 minute for fatigue or related symptoms. When major mobility disability could not be objectively measured because of the inability of the participant to come to the clinic and absence of a suitable walking course at the participant’s home, institution, or hospital, an alternative adjudication of the outcome was based on objective inability to walk 4 m in less than 10 seconds, or self-, proxy-, or medical record–reported inability to walk across a room. If participants met these alternative criteria, they would not be able to complete the 400-m walk within 15 minutes. Final determination of MMD onset was made by the Outcomes Committee. If the participant was unable to attend a follow-up session, a determination of MMD was made if four meter walking speed was slower than 0.4m/s at a home or clinical assessment, or if the participant was unable to walk across a room without assistance. The MMD outcome was defined as indeterminate if information was insufficient to adjudicate the outcome. Persistent MMD (PMMD) was defined as the inability to complete the 400m walk in 15 minutes or less at consecutive follow-up visits or the onset of MMD followed by death.

Data analysis and Statistics

The purpose of this study is to examine the associations between adherence (covariate of interest) and functional outcomes (gait speed and SPPB score), MMD and PMMD. Two primary sets of analyses were performed: 1) for the entire sample and 2) for the PA randomized arm only. For each set of analyses, three modeling approaches were used. First, repeated measures analysis of covariance (ANCOVA) was used to study the association between adherence measure and functional outcomes (gait speed and SPPB score) where the outcomes were treated as continuous measures. Second, ordinal logistic regression models incorporating generalized estimating equations were used when functional outcomes were treated as ordinal variables. Third, Cox regression models were used to study the association between adherence measure and MMD and PMMD. Variables selected for inclusion in both sets of analyses were similar, with the exception that center- and home-based walking minutes per week were included in the PA arm analysis only (as walking was not performed in the HE intervention). The statistical approaches were described in details below. The analyses were conducted using the SAS software version 9.4 (SAS Inc., Cary, NC).

Repeated measures ANCOVA models with unstructured covariance were used to evaluate the relative influence of demographic characteristics and intervention adherence on functional responses to the PA and HE interventions. The dependent variables were change of gait speed and change of SPPB score. The change was calculated using the follow-up measure minus baseline measure. Two models were fitted: 1) the basic model included sex and clinical site (stratifying variables) as well as randomization arm, baseline values of physical function, visit, time-dependent total steps from accelerometry, and the interaction between randomization arm and visit: and 2) the full model included all the covariates in the basic model and other factors not already included in the basic model selected based on variable selection results (Table 1). The candidate variables included age, race, married status, income, smoking status, waist circumference, BMI, weight, radial pulse rate, systolic BP, diastolic BP, HDL cholesterol, LDL cholesterol, triglycerides, total cholesterol, glucose, antihypertensive drugs, lipid-lowering drugs, co-morbidities, CESD score, PSQI score, and 3MSE score. Backward elimination was used to select the final model. The p-values for the factors needed to be smaller than 0.05 to remain in the final model. Continuous factors were standardized to assess their relative importance in the model. The partial coefficient of determination (partial R²) statistic was calculated by comparing the full model to the basic model. The model assumptions were checked. For PA only analysis, the randomization arm and interaction between randomization arm and visit were not included in the basic model.

Table 1.

Baseline characteristics of LIFE study participants for covariates included in full analysis model

	PA Arm n=818	HE Arm n=817
Physical Function
400m gait speed, m/s	0.83 ± 0.17	0.82 ± 0.17
SPPB Score	7.4 ± 1.6	7.3 ± 1.6
Physical Activity Engagement
Walking/ weight training activities (min/wk)*	75.1 ± 125.6	86.7 ± 134.5
Total steps/wk†	2667.5 ± 1385.2	2695.7 ± 1565.8
Demographics
Age	78.5 ± 5.2	79.1 ± 5.2
Female	547 (66.9%)	551 (67.4%)
Minority	214 (26.2 %)	182 (22.3%)
Married	523 (64.5%)	520 (63.8%)
Income (<$70,000/yr)	603 (83.3%)	602 (83.6%)
Smoking status (current/former)	407 (49.8%)	365 (44.7%)
Waist circumference (cm)	101.5 ± 15.4	101.9 ± 15.7
BMI (kg*m−2)	30.1 ± 5.9	30.3 ± 6.2
Weight (kg)	81.9 ± 18.4	82.0 ± 19.3
Radial Pulse Rate	66.7 ± 10.6	66.4 ± 10.0
Systolic BP	127.9 ± 18.1	127.0 ± 17.8
Diastolic BP	68.7 ± 10.3	67.7 ± 10.1
HDL Cholesterol (mg/dL)	61.1 ± 18.0	61.1 ± 17.6
LDL Cholesterol (mg/dL)	93.6 ± 32.2	93.1 ± 33.1
Triglycerides (mg/dL)	123.0 ± 56.6	121.8 ± 58.6
Total cholesterol (mg/dL)	179.3 ± 39.6	178.5 ± 39.9
Glucose	103.7 ± 23.4	104.7 ± 24.8
Medications
Antihypertensive drugs	637 (78.4%)	638 (78.2%)
Lipid-lowering drugs	515 (63.4%)	506 (62.0%)
Co-morbidities12	1.75 ± 1.14	1.82 ± 1.13
CESD score58	8.3 ± 7.7	8.8 ± 7.9
PSQI score59	5.9 ± 3.8	5.9 ± 3.8
3MSE score60	91.5 ± 5.5	91.6 ± 5.3

Abbreviations: PA , physical activity; CHAMPS, Community Healthy Activities Model Program for Seniors questionnaire ; BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; HDL, high density lipoprotein; LDL, low density lipoprotein; CESD, Center for Epidemiological Studies Depression Scale ; PSQI, Pittsburg Sleep Quality Index; 3MSE, Modified Mini-Mental State Examination.

^*Based on self-report (CHAMPS survey¹⁸)

^†Determined by accelerometry

In addition to modeling functional measures as continuous outcomes,³¹ we also modeled them using three response categories (i.e. strong, modest, poor). Participants were grouped based on the magnitude of change in functional outcomes from baseline to the 24-month visit using clinically meaningful cut-points (for Δ gait speed, poor response: ≤−0.05 m/s decline, modest response: 0.049 m/s decline to 0.049 m/s increase, and strong response: ≥ 0.05 m/s increase; for Δ SPPB score, poor response: ≥1 point decline, modest response: 0 change to 1 point improvement, and strong response: >1 point increase). The poor, modest, and strong functional response groups were further divided according to level of adherence to the intervention. As an exploratory analysis, the response subgroups were divided according to level of adherence to the intervention in PA only. Participants in the PA arm were dichotomized according to the percentage of center-based sessions attended (<60% or ≥60%) and center- and home-based walking minutes per week (≥120 minutes per week or <120 minutes per week).

Ordinal logistic regression models incorporating generalized estimating equations were used to identify the factors related to the response sub-groups. Proportional odds assumptions were checked before model fitting. Sex, clinical sites, randomization arm, visit, and the interaction between randomization arm and visit were always adjusted in the model. Other baseline characteristics (listed in Table 1) were included in the model based on the result of variable selection. For PA only analysis, the randomization arm and interaction between randomization arm and visit were not adjusted in the model. Cox regression modeling was used to identify baseline demographic and intervention adherence factors associated with risk of incident MMD and PMMD. Sex and clinical site were treated as stratified variables to allow different baseline hazards for different strata. Because of the expected small number of persistent MMD events, the Cox regression model was stratified by sex alone to study the effect of PMMD. Failure time was measured from the time of randomization. Follow-up was censored at 24 months. Adherence measures and baseline characteristics (listed in Table 1) were included in the model where adherence measures were treated as time-dependent variables. Randomization arm was always included in the model Backward elimination was used to select the other factors that were associated with incident MMD and PMMD, respectively. The proportional hazards assumptions were checked for each covariate included in the final model. For analysis using PA participants only, randomization arm was not adjusted in the model.

Sensitivity analyses were performed to investigate the effect of loss to follow-up on the outcomes using the inverse probability weight approach.³² We assigned a weight for each participant equivalent to the inverse probability of remaining in the study at each visit based on baseline age, sex, race, SPPB group (SPPB<8 vs. ≥8), number of co-morbidities, married status, and clinical sites. The analyses mentioned above were repeated, but weighted by the inverse probability weight.

Results

Demographics

Baseline characteristics of the study sample were reported previously.⁷ Briefly, the mean age of the study sample was 78.9 ± 5.2 years, 67.2% of the participants were women, and the average BMI was 30.2 ± 6.0 kg*m⁻². The mean baseline SPPB score was 7.4 ±1.6, with 35.4% of the sample having moderate functional impairment (SPPB score <8). Mean baseline gait speed during the 400m walk was 0.83 m/s, reflecting the age and functional level of the sample.^{1, 33} Participants averaged 80.9 ± 130.2 (quartile1=0, median=30, quartile 3=105) minutes per week of walking and resistance exercise at baseline according to the CHAMPS questionnaire. .

Distributions of functional responses

The distributions of Δ gait speed and Δ SPPB after 24 months of the PA and HE interventions are shown in Figure 1. A total of 23.8% of participants in the PA intervention improved gait speed by at least 0.05 m/s at 24 months, compared to 19.2% of participants in the HE intervention. For SPPB, 52.1% of participants in the PA intervention improved from baseline by at least one point, compared to 50.4% in the HE intervention. A total of 36.7% of participants in the PA intervention maintained their baseline gait speed at 24 months compared to 35.8% of participants in the HE group. Additionally, 30.7% of participants in the PA intervention maintained their baseline SPPB score at 24 months compared to 30.3% for the HE intervention.

Figure 1.

24-month Δ gait speed (top) and short physical performance battery (SPPB) performance (bottom) among participants in the physical activity (left) and health education (right) intervention group. Data for gait speed indicate individual responses to the interventions, while SPPB data reflect the proportion of individuals with a respective Δ SPPB score following treatment.

Predictors of change in continuous functional outcomes for the entire sample

Greater baseline physical function was negatively associated with Δ gait speed (β=−0.185, p<0.001) and ΔSPPB score (β=−0.365, p<0.001) from baseline to 24 months (Table 2a). Each standard deviation increase in steps per day (± 1611.17 steps) measured by accelerometry was positively associated with Δ gait speed (β=0.035, p<0.001) and Δ SjPPB score (β=0.525, p<0.001). In addition, older age (β=−0.013, p<0.001), higher BMI (β=−0.013, p=<0.001) and greater baseline CHAMPS score (β=−0.008, p=0.002) were significant factors associated with poorer Δ gait speed. Older age (β=−0.204, p<0.001) and higher baseline serum glucose (β=−0.086, p=0.046) were associated with poorer Δ SPPB score, while black race (β=0.323, p=0.012), and greater baseline 3MSE score (β=0.105, p=0.026) were associated with a better Δ SPPB score. Overall, the factors included in the model accounted for 28.3% of the variability in gait speed response and 22.8% for SPPB response.

Table 2a.

Demographic and/or physical activity-related factors significantly associated with changes in gait speed and SPPB score during 24 months of follow-up modeled as continuous outcomes through mixed effects modeling.

Variable	Regression Coefficient	SE of Regression Coefficient	SD	p-value	95% CI	Partial R2*	Model R2
ΔGait speed							0.283
Baseline gait speed	−0.185	0.017		<0.001	(−0.218,−0.152)	0.070
Steps/day	0.035	0.002	1611.17	<0.001	(0.031,0.040)	0.128
Age	−0.013	0.003	5.23	<0.001	(−0.019,−0.008)	0.015
BMI	−0.013	0.003	5.98	<0.001	(−0.018,−0.008)	0.015
CHAMPS score	−0.008	0.002	130.24	0.002	(−0.013,−0.003)	0.006
ΔSPPB score							0.228
Baseline SPPB score	−0.365	0.028		<0.001	(−0.420,−0.310)	0.094
Steps/day	0.525	0.038	1611.17	<0.001	(0.450,0.599)	0.105
Age	−0.204	0.046	5.23	<0.001	(−0.295,−0.114)	0.012
Race†				0.022 ‡		0.005
Black	0.323	0.127		0.012	(0.072,0.572)
Other	0.322	0.210		0.146	(−0.106,0.716)
Glucose	−0.086	0.043	24.14	0.046	(−0.169,−0.001)	0.002
3MSE score	0.105	0.047	5.42	0.026	(0.012,0.196)	0.003

Data reflect the relative influence of the corresponding factors to Δ gait speed and/or SPPB score over the course of the intervention according to mixed-effects modeling. Model was adjusted for clinical site, sex, randomization arm, visit, and interaction between randomization arm and visit (results not shown). Regression coefficients based on per standard deviation increase of the variable. SE indicates standard error; SD, standard deviation (for continuous variables only); BP, blood pressure; BMI, body mass index; SPPB, Short Physical Performance Battery; CHAMPS, Community Health Activities Model Program for Seniors; 3MSE, Modified Mini-Mental State Examination

^*Compare the full model with the model without this variable

^†White is the reference group

^‡2df test

Predictors of change in ordinal functional outcomes for the entire sample

When participants were categorized as strong, modest, or poor responders based on Δ gait speed and Δ SPPB score at 24 months, each standard deviation increase in steps per day increased the likelihood of being a strong responder by 81% (OR 1.81, 95% CI 1.61-2.03) for Δ gait speed and 85% (OR 1.85, 95% CI 1.66-2.07) for Δ SPPB score (Table 2b). Each standard deviation increase in age (± 5.23 years) was associated with a 22% decreased likelihood (OR 0.78, 95% CI 0.71-0.86) of being a strong responder for Δ gait speed and a 22% decreased likelihood (OR 0.78, 95% CI 0.71-0.85) for Δ SPPB score. Additionally, higher BMI at baseline was associated with a decreased likelihood (OR 0.81, 95% CI 0.73-0.89) for achieving a strong response for Δ gait speed, and increased baseline total serum cholesterol was associated with an increased likelihood (OR 1.11, 95% CI 1.01-1.21) of achieving a strong response for Δ SPPB score.

Table 2b.

Demographic and/or physical activity-related factors significantly associated with gait speed and SPPB score response categories for PA and HE groups over time identified through ordinal logistic regression.

Variable (SD)	OR (95% CI) of strong response vs. moderate and poor responses*
Gait speed
Steps/day (1611.17)	1.81 (1.61-2.03)
Age (5.23)	0.78 (0.71-0.86)
BMI (5.98)	0.81 (0.73-0.89)
SPPB score
Steps/day (1611.17)	1.85 (1.66-2.07)
Age (5.23)	0.78 (0.71-0.85)
Cholesterol (39.76)	1.11 (1.01-1.21)

Values represent the odds ratio (OR) and 95% confidence interval (CI) for each predictive factor indicating likelihood of experiencing strong response to the intervention

^*Relative to moderate and poor response categories according to ordinal logistic regression modeling with generalized estimating equations adjusting for sex, clinical site and other selected covariates. The sex and site effects are not shown. Response categories were chosen based on established clinically-meaningful cut-points⁶⁷,⁶⁸ and defined as strong (≥0.05 m/s Δ gait speed, >1 point increase in SPPB score), modest (−0.49-0.49 m/s, 0.999-1 points) or poor (≤−0.05 m/s, ≤−1 points). BMI indicates body mass index; SPPB, Short Physical Performance Battery.

Predictors of major mobility disability and persistent major mobility disability for the entire sample

Results of Cox regression analysis of factors associated with the development of MMD and PMMD in our study cohort are presented in Figure 2. A greater number of steps per day at baseline (HR 0.25, 95% CI 0.18-0.34) and being married at baseline (HR 0.67, 95% CI 0.46-0.97) were associated with a decreased risk of experiencing MMD, while low baseline SPPB score (HR 1.67, 95% CI 1.21-2.29), high baseline BMI (HR 1.37, 95% CI 1.20-1.57) and elevated heart rate at baseline (HR 1.19, 95% CI 1.03-1.38) were associated with an increased risk of developing MMD. For PMMD, a greater number of steps per day at baseline (HR 0.14, 95% CI 0.09-0.23), taking blood pressure medication at baseline (HR 0.50, 95% CI 0.27-0.94) and being married at baseline (HR 0.57, 95% CI 0.35-0.90) were associated with a decreased risk of developing PMMD, while low baseline SPPB score (HR 2.18, 95% CI 1.42-3.35) and higher baseline BMI (HR 1.42, 95% CI 1.20-1.67) were associated with an increased risk of developing PMMD. Overall, limited mobility at baseline as indicated by an SPPB score <8 was the strongest independent predictor of MMD and PMMD.

Figure 2.

Forest plot of the hazard ratio (HR) and 95% confidence interval (CI) of each risk factor for development of major mobility disability (MMD) according to Cox regression modeling stratified by sex. MMD was defined as inability to complete a 400m walk test within 15 minutes without sitting or help of another person or walker. Persistent MMD (PMMD) was defined by two consecutive major mobility disability assessments or major mobility disability followed by death. SPPB=Short Physical Performance Battery; BMI= body mass index; BP=blood pressure. *0=No and 1=Yes.

Analyses of Physical activity intervention effects

Participants in the PA intervention arm were categorized into three response groups based upon Δ SPPB score and gait speed at 24 months. The poor, modest, and strong functional response groups were further divided according to level of adherence to the intervention (Figure 3). A total of 36.0% of participants with ≥60% center-based attendance increased SPPB by more than 1 point, compared to only 24.2% of those with <60% attendance (p<0.01). Similarly, 40.9% of participants who walked at least 120 minutes per week increased SPPB score by one or more points, versus 27.2% of those who walked less than 120 minutes per week (p<0.01).

Figure 3.

Percentage of individuals obtaining poor (≤ −0.05 m/s or ≤ −1 point), modest (−0.049 to 0.049 m/s or 0-1 point) or strong (≥ 0.05 m/s or >1 point) responses to the PA intervention, dichotomized by adherence, for gait speed and SPPB score at 24 months. Minutes walked per week includes total center- and home-based walking over the course of the intervention.

Predictors of change in continuous functional outcomes for the PA intervention

Baseline physical function was inversely associated with Δ gait speed (β=−0.105, p<0.001) and ΔSPPB score (β=−0.343, p<0.001) in response to the PA intervention (Table 3). A higher percentage of center-based sessions attended (β=0.019, p<0.001 for gait speed and β=0.154, p=0.018 for SPPB score) and greater minutes walked at baseline (β=0.010, p=0.004 for gait speed and β=0.230, p<0.001 for SPPB score) were associated with a better Δ both gait speed and SPPB score. In addition, older age (β=−0.017, p<0.001) and higher BMI (β=−0.013, p=0.001) at baseline were associated with a poorer Δ gait speed, while higher diastolic BP (β=0.008, p=0.011) associated with a better Δ gait speed. Older age (β=−0.385, p<0.001) and higher CESD score (β=−0.205, p=0.002) at baseline associated with a poorer Δ SPPB score. Overall, the covariates included in the model accounted for 25.2% of the variability in gait speed response to the PA intervention and 19.8% of the variability in SPPB score response.

Table 3.

Demographic and/or adherence factors significantly associated with changes in gait speed and SPPB among participants in the PA arm during 24 months of follow-up.

Variable	Standardized Regression Coefficient	SE of Regression Coefficient	SD	p-value	95% CI	Partial R2*	Model R2
Δ Gait speed							0.252
Baseline  gait speed	−0.105	0.024		<0.001	(−0.153,− 0.058)	0.023
Attendance	0.019	0.004	21.72	<0.001	(0.011,0.027)	0.029
Age	−0.017	0.004	5.22	<0.001	(−0.025,− 0.009)	0.021
Minutes  Walked	0.010	0.003	72.27	0.004	(0.003,0.017)	0.010
BMI	−0.013	0.004	5.72	0.001	(−0.021,− 0.005)	0.013
Diastolic  BP	0.008	0.003	9.79	0.011	(0.002,0.014)	0.008
Δ SPPB score							0.198
Baseline  SPPB score	−0.343	0.041		<0.001	(−0.425,− 0.262)	0.078
Attendance	0.154	0.065	21.72	0.018	(0.027,0.280)	0.007
Age	−0.385	0.066	5.22	<0.001	(−0.514,− 0.256)	0.040
Minutes  Walked	0.230	0.059	72.27	<0.001	(0.114,0.347)	0.018
CESD  score	−0.205	0.066	7.70	0.002	(−0.334,− 0.075)	0.012

Data reflect the relative influence of the corresponding factors to Δ SPPB score and/or gait speed over the course of the intervention according to mixed-effects modeling Model was adjusted for clinical site, sex, randomization arm, visit, and interaction between randomization arm and visit (results not shown). Regression coefficients based on per standard deviation increase of the variable. SE indicates standard error; SD, standard deviation (for continuous variables only); BP, blood pressure; BMI, body mass index; SPPB, Short Physical Performance Battery; CESD, Center for Epidemiological Studies Depression Scale.*Includes total center- and home-based walking minutes over course of the intervention. *Compare the full model with the model without this variable

Predictors of change in ordinal functional outcomes for the PA intervention

Each standard deviation increase in attendance to the center-based PA sessions (± 21.72%) was associated with a 26% increased probability of achieving a strong response for both Δ gait speed (OR 1.26, 95% CI 1.10-1.45) and Δ SPPB score (OR 1.26, 95% CI 1.12-1.41, Table 4). Each standard deviation increase in age at baseline (±5.22 years) was associated with a decreased probability of achieving a strong response in both Δ gait speed (OR 0.80, 95% CI 0.69-0.91) and Δ SPPB score (OR 0.65, 95% CI 0.58-0.74). Additionally, each standard deviation increase in baseline waist circumference (±15.41 cm) was associated with a decreased probability of achieving a strong response for Δ gait speed (OR 0.83, 95% CI 0.72-0.96), while each standard deviation increase in baseline diastolic BP (±9.79 mmHg) was associated with an increased probability of achieving a strong response for Δ gait speed (OR 1.16, 95% CI 1.02-1.31). Finally, each standard deviation increase in baseline CESD score (±7.70 points) was associated with a decreased probability of achieving a strong response for Δ SPPB score (OR 0.82, 95% CI 0.73-0.93).

Table 4.

Demographic and/or adherence factors significantly associated with gait speed and SPPB score response categories for PA participants over time identified through ordinal logistic regression .

Variable (SD)	OR (95% CI) of strong response vs. moderate and poor responses*
Gait speed
Attendance (21.72)	1.26 (1.10-1.45)
Age (5.22)	0.80 (0.69-0.91)
Waist Circumference (15.41)	0.83 (0.72-0.96)
Diastolic BP (9.79)	1.16 (1.02-1.31)
SPPB score
Attendance (21.72)	1.26 (1.12-1.41)
Age (5.22)	0.65 (0.58-0.74)
CESD score (7.70)	0.82 (0.73-0.93)

Values represent the odds ratio (OR) and 95% confidence interval (CI) for each predictive factor indicating likelihood of experiencing a strong response to the intervention

^*Relative to poor and moderate response categories according to ordinal logistic regression modeling adjusting for sex and clinical site and other selected covariates. The site effect is not shown. Response categories were chosen based on established clinically-meaningful cut-points⁶⁷,⁶⁸ and defined as strong (≥0.05 m/s Δ gait speed, >1 point increase in SPPB score), modest (−0.49-0.49 m/s, 0.999-1 points) or poor (≤−0.05 m/s, ≤−1 points). BP indicates blood pressure; SPPB, Short Physical Performance Battery; CESD, Center for Epidemiological Studies Depression Scale.

Predictors of major mobility disability and persistent major mobility disability for the PA intervention

Results of Cox regression analysis of factors associated with the development of MMD and PMMD among participants in the PA intervention arm are presented in Figure 4. Higher attendance to the center-based intervention sessions (HR 0.72, 95% CI 0.58-0.89), greater minutes walked (HR 0.68, 95% CI 0.51-0.92), and higher baseline diastolic blood pressure (HR 0.62, 95% CI 0.48-0.80) were associated with a decreased risk of experiencing MMD. Low baseline SPPB score (HR 1.83, 95% CI 1.20-2.79), greater baseline waist circumference (HR 1.30, 95% CI 1.02-1.65), higher baseline systolic blood pressure (HR 1.49, 95% CI 1.17-1.90), and elevated resting heart rate at baseline (HR 1.33, 95% CI 1.07-1.64) were associated with an increased risk of developing MMD. For PMMD, greater minutes walked (HR 0.66, 95% CI 0.46-0.94), higher baseline diastolic BP (HR 0.75, 95% CI 0.56-0.99) and higher baseline 3MSE score (HR 0.76, 95% CI 0.60-0.98) were associated with a decreased risk of developing PMMD, while low baseline SPPB score (HR 2.52, 95% CI 1.41-4.51) higher systolic BP at baseline (HR 1.39, 95% CI 1.05-1.85), and higher CHAMPS score at baseline (HR 1.30, 95% CI 1.03-1.65) were associated with an increased risk of developing PMMD.

Figure 4.

Forest plot of the hazard ratio (HR) and 95% confidence interval (CI) of each risk factor for development of major mobility disability (MMD) among participants in the PA intervention alone according to Cox regression modeling stratified by sex. MMD was defined as inability to complete a 400m walk test within 15 minutes without sitting or help of another person or walker. Persistent MMD (PMMD) was defined by two consecutive major mobility disability assessments or major mobility disability followed by death. SPPB=Short Physical Performance Battery; BP=blood pressure; 3MSE=modified mini-mental state examination; CHAMPS=community health activities model program for seniors. *0=No and 1=Yes.

Sensitivity analyses

When the analyses were weighted by inverse probability weight, the effects were all similar (usually different at the 3^rd decimal point) to what we have presented (Supplementary Tables 1-5).

Discussion

The primary objective of the present analysis was to identify demographic and physical activity-related factors associated with variability in functional responses among older adults participating in the LIFE study. Secondarily, we sought to identify factors predictive of variability in functional responses to long-term PA. The primary results from the present analysis indicate that greater steps per day at baseline were associated with better longitudinal Δ functional outcomes, while older age, greater physical function and higher BMI at baseline were associated with poorer Δ functional outcome responses among all participants (PA and HE) in the LIFE study. Similarly, better intervention adherence (in-center session attendance and total minutes walked) was associated with a more positive Δ in functional outcomes among participants in the PA arm of the LIFE study, while higher baseline physical function and older age at baseline were associated with a poorer Δ in functional outcomes.

Our objective was to identify factors that may provide predict responsiveness to the PA and HE interventions as opposed to randomization arm alone. Perhaps unsurprisingly, more steps per day were positively associated with Δ SPPB score and gait speed. This finding is in agreement with recent publications indicating that sedentary behavior is associated with lower physical function^34-37 and greater losses in function over time,³⁸ even independent of time spent engaging in moderate or vigorous physical activity.³⁹ Thus, decreasing sedentary time through increased daily walking time and frequency as advocated by the recent Surgeon General’s 2015 Call to Action⁴⁰ may improve physical function and promote independence among older adults, regardless of whether they participate in moderate or vigorous physical activity. Interestingly, we found that participants who accumulated at least 120 walking minutes per week were significantly more likely to achieve a strong response to the PA intervention compared to those accumulating less than 120 walking minutes per week. Current PA guidelines suggest at least 150 minutes per week of moderate to vigorous physical activity are needed to improve major health outcomes.⁴¹ However, our findings indicate that this threshold may be lower for improving physical function among older adults.

The negative association between baseline function and longitudinal Δ physical function in response to the interventions is consistent with recent findings from Chmelo et al.¹⁴ indicating that changes in 400m walk time, usual paced gait speed, chair rise time and SPPB score were negatively associated with baseline function among 95 overweight and obese older adults. As suggested by Chmelo et al., this could be due to a greater capacity of participants with lower baseline function to improve, or due to a ceiling effect for those with higher baseline function. It should be noted that, despite this association with greater absolute improvements in function, lower baseline function was still associated with an increased risk of experiencing MMD or PMMD. This is an important distinction to be considered in evaluating the present data and considering the overall utility of the exercise intervention. For instance, a participant who improves their SPPB score from 5 to 7 obtains “more benefit” from exercise but still remains at higher risk of MMD than a participant with a baseline score of 9 that remains unchanged throughout follow-up.

We also found that older age at baseline was negatively associated with Δ functional outcomes in response to the interventions It is well known that aging is associated with declines in physical function due in large part to age-related losses in skeletal muscle mass, power, and aerobic capacity.^42-44 However, data concerning the relationship between age and responsiveness to PA are limited, and reports are equivocal. For example, training studies indicate that the relative strength gain in response to resistance training can be similar^{45, 46} or lower⁴⁷ in older adults compared to younger adults. Additional reports from the HERITAGE family study⁴⁸ and Kohrt et al.⁴⁹ indicate that there is no relationship between age and responsiveness to endurance training using VO2 max as the primary outcome in some populations. This discrepancy is likely due to methodological differences between the studies, particularly in regard to the outcomes measured and the age ranges of the participants. For example, participants in the HERITAGE study were 17-65 years compared to 70-89 years for the LIFE study.

Obesity is another factor known to contribute to declines in physical function and the development of mobility disability among older adults.⁵⁰ In the present study, we found that BMI was negatively associated with Δ gait speed over time. These findings are in agreement with data from the LIFE-Pilot study indicating that obese individuals (BMI ≥ 30 kg*m⁻²) experienced a decline in 400-m walk gait speed following 12 months of PA (-3.1%), whereas non-obese participants improved gait speed (+1.5%) in response to the PA intervention.¹³ However, Δ SPPB score was similar between obese and non-obese participants in the PA intervention arm of the LIFE-Pilot study (+11.3% for obese participants vs +13.5% for non-obese participants). Likewise, BMI was not significantly associated with Δ SPPB score in the present analysis. Excessive body mass may limit performance of endurance-based outcome measures, while having less of an impact on short-duration tests such as the SPPB.¹³ Nevertheless, these findings and results from other trials^51-54 suggest that diet-induced weight loss may be a necessary adjuvant therapy for improving responsiveness to PA and preventing mobility disability among obese older adults.

Factors associated with experiencing MMD and PMMD included fewer baseline steps per day, low baseline SPPB score (<8), and higher baseline BMI. Additionally, being married was associated with a decreased risk of experiencing MMD and/or PMMD, which is consistent with other studies that suggest marriage is associated with longevity⁵⁵ and greater physical function among older adults. Higher resting heart rate at baseline was associated with increased risk of experiencing MMD, possibly due to poor cardiovascular conditioning⁵⁶ or underlying morbidity.^57-59 Finally, antihypertensive medication use was associated with decreased risk of incident PMMD. This finding is in agreement with prior suggestions that antihypertensive medications may confer protective benefits later in life, including preservation of muscle function and improved body composition.^60-62

The secondary objective of this analysis was to identify factors associated with Δ functional outcomes in response to long-term PA. Similar to the whole sample, baseline physical function, age, and minutes walked were significantly associated with Δ functional outcomes among participants in the PA intervention. In addition, diastolic blood pressure at baseline was positively associated with longitudinal Δ gait speed and probability of achieving a high response in gait speed following the PA intervention. These results add to a relatively small literature related to factors that predict physiologic responses to long-term PA in older age. Presently, a full understanding of the factors contributing to inter-individual differences in physiologic responses to PA is limited by this scarcity of findings, as only a few studies to our knowledge have been explicitly designed for this purpose.^63-65 Well-designed, prospective studies are needed to further characterize the factors which underlie these individual differences in PA responsiveness.

Study Limitations

The strengths of the present study include a clinically relevant population and outcomes, a large sample size drawn from multiple sites in diverse settings (i.e., urban/rural, different regions of the country), excellent retention, and a relatively long period of intervention and follow up assessments. However, the LIFE study was not explicitly designed to evaluate individual differences in responsiveness to PA. As such, there are likely a number of potential factors which contribute to response variability which we were unable to evaluate. For instance, the trial did not collect certain types of biologic data (i.e., changes in skeletal muscle proteins and mRNA, DNA genotyping, circulating hormones) which may have explained additional variability in the functional outcomes. Furthermore, these results may not be generalizable to older adults with very high or very low levels of physical function. In addition, it is known that different variable selection approaches may result in different final models and have their own pros and cons. Due to limitations of the analytic software, we manually conducted backward elimination to select the variables in the final model for repeated measures ANCOVA and ordinal logistic regression. As such, the results must be interpreted with caution. For Cox models, the built-in automatic variable selection is available. The backward, forward, and stepwise selections resulted in the same model, which alleviates some of this concern. However, a cross-validation study in the future might be helpful to validate the present findings. Finally, using gait speed over 80m for participants unable to complete the 400m walk test likely overestimated gait speed for those participants.

Conclusions

We identified several phenotypic factors associated with variability in response to interventions intended to improve physical function among older adults. In particular, higher PA participation (steps per day, walking minutes per week, attendance) was associated with better Δ functional outcomes, while older age, BMI and higher physical function at baseline were associated with poorer Δ functional outcomes. These results add to the ongoing efforts to identify sources of the heterogeneity of functional responses among individuals and populations. Future studies should focus not only on identifying such factors, but also on identifying and implementing adjuvant therapies to improve responsiveness to PA.⁶⁶ Such therapies will be critical for improving quality of life and preventing future disability among at-risk older adults.

Abbreviations

BMI

Body mass index

CHAMPSy

Community Healthy Activities Model Program for Seniors

HR

Hazard ratio

HE

Health education

MMD

Major mobility disability

3MSE

Modified mini-mental state examination

OR

Odds ratio

PMMD

Persistent MMD

PA

Physical activity

SPPB

Short Physical Performance Battery

Appendix

Research Investigators for the LIFE Study

LIFE investigators are also partially supported by the following:

Dr. Thomas Gill (Yale University) is the recipient of an Academic Leadership Award (K07AG3587) from the National Institute on Aging.

Dr. Carlos Vaz Fragoso (Spirometry Reading Center, Yale University) is the recipient of a Career Development Award from the Department of Veterans Affairs.

Dr. Roger Fielding (Tufts University) is partially supported by the U.S. Department of Agriculture, under agreement No. 58-1950-0-014. Any opinions, findings, conclusion, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the view of the U.S. Dept of Agriculture.

Administrative Coordinating Center, University of Florida, Gainesville, FL

Marco Pahor, MD – Principal Investigator of the LIFE Study

Jack M. Guralnik, MD, PhD – Co-Investigator of the LIFE Study (University of Maryland School of Medicine, Baltimore, MD)

Christiaan Leeuwenburgh, PhD Connie Caudle

Lauren Crump, MPH

Latonia Holmes

Jocelyn Lee, PhD

Ching-ju Lu, MPH

Data Management, Analysis and Quality Control Center, Wake Forest University, Winston Salem, NC

Michael E. Miller, PhD – DMAQC Principal Investigator

Mark A. Espeland, PhD – DMAQC Co-Investigator

Walter T. Ambrosius, PhD

William Applegate, MD

Daniel P. Beavers, PhD, MS

Robert P. Byington, PhD, MPH, FAHA

Delilah Cook, CCRP

Curt D. Furberg, MD, PhD

Lea N. Harvin, BS

Leora Henkin, MPH,

Med John Hepler, MA

Fang-Chi Hsu, PhD

Laura Lovato, MS

Wesley Roberson, BSBA

Julia Rushing, BSPH, MStat

Scott Rushing, BS

Cynthia L. Stowe, MPM

Michael P. Walkup, MS

Don Hire, BS

W. Jack Rejeski, PhD

Jeffrey A. Katula, PhD, MA

Peter H. Brubaker, PhD

Shannon L. Mihalko, PhD

Janine M. Jennings, PhD

National Institutes of Health, Bethesda, MD

Evan C. Hadley, MD (National Institute on Aging)

Sergei Romashkan, MD, PhD (National Institute on Aging)

Kushang V. Patel, PhD (National Institute on Aging)

National Heart, Lung and Blood Institute, Bethesda, MD

Denise Bonds, MD, MPH

Field Centers

Northwestern University, Chicago, IL

Mary M. McDermott, MD – Field Center Principal Investigator

Bonnie Spring, PhD – Field Center Co-Investigator

Joshua Hauser, MD – Field Center Co-Investigator

Diana Kerwin, MD – Field Center Co-Investigator

Kathryn Domanchuk, BS

Rex Graff, MS Alvito Rego, MA

Pennington Biomedical Research Center, Baton Rouge, LA

Timothy S. Church, MD, PhD, MPH – Field Center Principal Investigator

Steven N. Blair, PED (University of South Carolina)

Valerie H. Myers, PhD

Ron Monce, PA-C

Nathan E. Britt, NP

Melissa Nauta Harris, BS

Ami Parks McGucken, MPA, BS

Ruben Rodarte, MBA, MS, BS

Heidi K. Millet, MPA, BS

Catrine Tudor-Locke, PhD, FACSM

Ben P. Butitta, BS

Sheletta G. Donatto, MS, RD, LDN, CDE

Shannon H. Cocreham, BS

Stanford University, Palo Alto, CA

Abby C. King, PhD – Field Center Principal Investigator

Cynthia M. Castro, PhD

William L. Haskell, PhD

Randall S. Stafford, MD, PhD Leslie A. Pruitt, PhD

Kathy Berra, MSN, NP-C, FAAN

Veronica Yank, MD

Tufts University, Boston, MA

Roger A. Fielding, PhD – Field Center Principal Investigator

Miriam E. Nelson, PhD – Field Center Co-Investigator

Sara C. Folta, PhD – Field Center Co-Investigator

Edward M. Phillips, MD

Christine K. Liu, MD

Erica C. McDavitt, MS

Kieran F. Reid, PhD,

MPH Won S. Kim, BS

Vince E. Beard, BS

University of Florida, Gainesville, FL

Todd M. Manini, PhD – Field Center Principal Investigator

Marco Pahor, MD – Field Center Co-Investigator

Stephen D. Anton, PhD

Susan Nayfield, MD

Thomas W. Buford, PhD

Michael Marsiske, PhD

Bhanuprasad D. Sandesara, MD

Jeffrey D. Knaggs, BS

Megan S. Lorow, BS

William C. Marena, MT, CCRC

Irina Korytov, MD

Holly L. Morris, MSN, RN, CCRC (Brooks Rehabilitation Clinical Research Center, Jacksonville, FL)

Margo Fitch, PT (Brooks Rehabilitation Clinical Research Center, Jacksonville, FL)

Floris F. Singletary, MS, CCC-SLP (Brooks Rehabilitation Clinical Research Center, Jacksonville, FL)

Jackie Causer, BSH, RN (Brooks Rehabilitation Clinical Research Center, Jacksonville, FL)

Katie A. Radcliff, MA (Brooks Rehabilitation Clinical Research Center, Jacksonville, FL)

University of Pittsburgh, Pittsburgh, PA

Anne B. Newman, MD, MPH – Field Center Principal Investigator

Stephanie A. Studenski, MD, MPH – Field Center Co-Investigator

Bret H. Goodpaster, PhD

Nancy W. Glynn, PhD

Oscar Lopez, MD

Neelesh K. Nadkarni, MD, PhD

Kathy Williams, RN, BSEd, MHSA

Mark A. Newman, PhD

George Grove, MS

Janet T. Bonk, MPH, RN

Jennifer Rush, MPH

Piera Kost, BA (deceased)

Diane G. Ives, MPH

Wake Forest University, Winston Salem, NC

Stephen B. Kritchevsky, Ph.D. – Field Center Principal Investigator

Anthony P. Marsh, PhD – Field Center Co-Investigator

Tina E. Brinkley, PhD

Jamehl S. Demons, MD

Kaycee M. Sink, MD, MAS

Kimberly Kennedy, BA, CCRC

Rachel Shertzer-Skinner, MA, CCRC

Abbie Wrights, MS

Rose Fries, RN, CCRC

Deborah Barr, MA, RHEd, CHES

Yale University, New Haven, CT

Thomas M. Gill, MD – Field Center Principal Investigator

Robert S. Axtell, PhD, FACSM – Field Center Co-Investigator (Southern Connecticut State University, Exercise Science Department)

Susan S. Kashaf, MD, MPH (VA Connecticut Healthcare System)

Nathalie de Rekeneire, MD, MS

Joanne M. McGloin, MDiv, MS, MBA

Karen C. Wu, RN

Denise M. Shepard, RN, MBA

Barbara Fennelly, MA, RN

Lynne P. Iannone, MS, CCRP

Raeleen Mautner, PhD

Theresa Sweeney Barnett, MS, APRN

Sean N. Halpin, MA

Matthew J. Brennan, MA

Julie A. Bugaj, MS

Maria A. Zenoni, MS

Bridget M. Mignosa, AS

Cognition Coordinating Center, Wake Forest University, Winston Salem, NC

Jeff Williamson, MD, MHS – Center Principal Investigator

Kaycee M Sink, MD, MAS – Center Co-Investigator

Hugh C. Hendrie, MB, ChB, DSc (Indiana University)

Stephen R. Rapp, PhD

Joe Verghese, MB, BS (Albert Einstein College of Medicine of Yeshiva University)

Nancy Woolard

Mark Espeland, PhD

Janine Jennings, PhD

Electrocardiogram Reading Center, University of Florida, Gainesville, FL

Carl J. Pepine MD, MACC

Mario Ariet, PhD

Eileen Handberg, PhD, ARNP

Daniel Deluca, BS

James Hill, MD, MS, FACC

Anita Szady, MD

Spirometry Reading Center, Yale University, New Haven, CT

Geoffrey L. Chupp, MD

Gail M. Flynn, RCP, CRFT

Thomas M. Gill, MD

John L. Hankinson, PhD (Hankinson Consulting, Inc.)

Carlos A. Vaz Fragoso, MD

Cost Effectiveness Analysis Center

Erik J. Groessl, PhD (University of California, San Diego and VA San Diego Healthcare System) Robert M. Kaplan, PhD (Office of Behavioral and Social Sciences Research, National Institutes of Health)