Distinct Trajectories of Physical Activity Among Patients with COPD During and After Pulmonary Rehabilitation

Abstract

Little is known about longitudinal trends in objectively measured physical activity (PA) during and after pulmonary rehabilitation (PR) for individuals with Chronic Obstructive Pulmonary Disease (COPD). The purpose of this study was to examine the PA trajectories of patients with COPD during and after PR and whether demographic, clinical, or program characteristics differed across these trajectories. The study was approved by Research Ethics Boards at all participating institutions, and written informed consent was obtained from each participant prior to study inclusion. COPD patients (N = 190) completed a questionnaire and wore a pedometer for 7 days at baseline, end of PR, and 3 and 9 months after completing PR. Latent class growth analyses showed that two distinct PA trajectories emerged. Active Maintainers averaged 9177 steps/day at baseline, and maintained this level throughout the assessment and post rehabilitation period. In contrast, Inactive Maintainers averaged 3133 steps/day at baseline, which also remained stable during and after PR. Follow-up analyses showed the Inactive Maintainers were more likely to be retired from work and have lower baseline scores for their stress tests and 6-minute walk tests compared to Active Maintainers (all p < 0.05). These results suggest that two distinct steps/day trajectories exist for COPD patients during and after completing PR that are partially explained by specific demographic and clinical characteristics.

Keywords:

• exercise
• latent class growth analysis
• pedometers
• steps per day

Introduction

Physical activity (PA) levels are lower among people with chronic obstructive pulmonary disease (COPD) than among their healthy peers (1–3). For instance, Pitta et al. (1) report that, in comparison to healthy age-matched peers, patients with COPD spend 50% less time walking, and 20% more time sitting down. These findings are of clinical relevance, given that low levels of PA are associated with increased COPD morbidity and mortality, independent of lung function (2–4). Thus, it is not surprising that increasing PA is recommended for individuals with COPD (2,3).

One intervention with the potential to increase PA in people living with COPD is pulmonary rehabilitation (PR). Indeed, research has shown that PR patients experience improvements in dyspnea, fatigue, health-related quality of life, emotional function, and exercise capacity (2, 5–8). However, the impact of PR on PA per se remains unclear (2,3,9–13). Using pedometers, Dallas et al. (10) observed no change in PA levels before and after 6–12 weeks of PR in 45 patients with COPD. In contrast, Sewell and colleagues (11) showed objectively measured PA increased by roughly 30% following a 7-week PR program, with similar findings reported by Mercken et al (12). Given the inconsistent findings across studies, it is important to clarify how physically active COPD patients are during PR using an objective PA measure (14).

As the vast majority of PA research has focused on PA during PR, little is known about the PA levels of patients after they complete these programs. Egan and colleagues (9) used an arm-worn activity monitor to examine changes in objectively measured moderate and vigorous physical activity (MVPA) in 47 patients with COPD following 7 weeks of PR. In comparison to baseline levels of MVPA (48 minutes/day), they observed little change 7 weeks after the start of PR (51 minutes/day) or 20 weeks after the start of PR (47.6 minutes/day).

On the other hand, a recent study by Soicher et al. (15) measured self-reported PA in 206 COPD patients 1, 3, 5, and 9 months after completing PR.

Using latent class growth analysis (LCGA), they found that 3 distinct groups (or classes) of patients emerged (i.e., 30% of patients were highly active during the entire follow-up period, 55% of patients were inactive the entire follow-up period, and 15% of patients were highly active 1 month after PR and became inactive by 9 months after PR). The discrepancies in the above studies are likely due to the analyses employed (i.e., traditional repeated measure ANOVA vs. LCGA), potential sample differences, and PA measurement (i.e., self-report vs. objective). Regardless of why differences emerged, it is clear that a study is needed that uses an objective PA measure and LCGA to describe the PA trajectories of COPD patients during and after completion of PR. Finally, it is also important to investigate factors which may help to differentiate among patients in different trajectories.

The first purpose of the present study was to employ an objective PA measure (i.e., pedometers) to examine the steps/day trajectories of COPD patients during (pre– post) and 3 and 9 months after completing PR using LCGA. The second purpose was to examine demographic, clinical, and program characteristics across the different classes to identify patients who may be in need of a PA intervention.

Methods

The PR Programs

Seven PR programs from across Canada (i.e., 2 in Nova Scotia; 1 in New Brunswick; 2 in Quebec; 1 in Saskatchewan; 1 in British Columbia) provided medically supervised exercise training and patient education to reduce symptoms of COPD, promote behavior modification, and increase self-efficacy, exercise capacity and quality of life. The programs ranged in duration (i.e., 6, 10, or 12 weeks), frequency of exercise training sessions (i.e., 2× or 3×/week), total number of supervised exercise sessions (i.e., 18, 20, or 24 sessions), and location (i.e., community- or hospital-based). Patients exercised for 1–1.5 hours/session, with each session including aerobic and resistance training. Aerobic exercise training involved progression to achieve 60–80% of maximum power output for 30 minutes’ duration. The strength training included 2–3 sets of 8–10 repetitions at a moderate-to-high intensity (i.e., approximating 50-85% of max). Programs also included formalized patient education sessions on a number of topics, including respiratory anatomy and physiology, pathophysiology of COPD, airway clearance techniques, benefits of exercise, nutrition, medications and devices, smoking cessation, coping strategies, and advanced care planning.

The study was approved by Research Ethics Boards at all participating institutions, and written informed consent was obtained from all participants prior to participation in accordance with the principles expressed in the Declaration of Helsinki.

Participants

Patients were eligible to participate if they were ≥ 18 years of age, had a physician diagnosis of COPD according to Canadian Thoracic Society criteria (i.e., FEV₁/FVC < 0.70) (16), could read and write English, and provided informed consent. Patients were excluded if they did not consent to participate or failed to complete at least one pedometer assessment.

Measures

Demographic characteristics included self-reported age, sex, education, marital status, race, employment, and income. Baseline clinical characteristics were obtained via self-report questionnaires (i.e., the presence or absence of diabetes, stroke, lung disease, arthritis, cancer, circulation problems, back pain, hip pain, knee pain, and/or foot pain that were summed to form a co-morbidity score) and chart review (i.e., each site completed a standardized form to provide the FEV₁, FVC, stress test scores in METs, 6-minute walk test distance, and height and weight to calculate body mass index). Finally, the PR program characteristics were recorded (i.e., program region in Canada, program length, exercise training frequency, and% of exercise training sessions attended).

Steps/Day were assessed using the Yamax- DIGI-WALKER. The DIGI-WALKER has a sensitivity threshold of 0.35–0.50 Gs (17), and is a valid and reliable pedometer in both healthy populations (18,19) and in those with chronic respiratory disease (20). A research assistant at each site programmed the participant's stride length and weight into the pedometer prior to its use to ensure accuracy of the step counts obtained. Participants were asked to wear the pedometer on their right hip for all waking hours of the day for 7 consecutive days and record their daily steps in a log book at the end of each day. The total number of steps were aggregated and divided by the number of valid days to obtain a steps/day outcome variable.

Procedure

Following ethical approval at each site, patient charts were reviewed for eligibility upon program entry. PR staff asked eligible patients if they were interested in speaking to a research assistant regarding the study. Interested patients were approached by the research assistant in the 2^nd or 3^rd week of their program and those who agreed to participate were given the informed consent form and self-report social ecological questionnaire to complete that day. The decision to recruit in the 2^nd or 3^rd week was based on the fact that the social ecological questionnaire included intrapersonal-related questions (e.g., attitudes about physical activity, confidence to engage in physical activity, etc. ..). As such, it was deemed necessary for the patients to have some exposure to the activity itself and the PR program in order to accurately judge and respond to these types of questions (21), although this was not the focus of the current article.

Participants were then fitted with a pedometer and asked to record the date and daily steps in a log book for a 7-day period. Once all questions were addressed, an appointment was made to pick up the study materials at the following week's PR class. At the end of PR (i.e., within the last 2 weeks of program completion), patients were asked to complete the same questionnaire in addition to wearing the pedometer for a 7-day period and to return their study materials during the following week's class. Finally, patients were asked to complete the same questionnaire and wear the pedometer at 3 and 9 months after completing PR. Prior to each assessment, patients were contacted by a research assistant to provide them with the study materials in person at the PR site or via the mail (based on their preference). Participants completed the forms at home, and returned them in person at the PR site or via postage-paid envelope.

Analytical plan

Recruitment rates were calculated, after which the baseline demographic, clinical, and program characteristics were generated. Next, a series of LCGAs (22) were conducted in MPLUS 6.1 using full information maximum likelihood estimation in order to include all patients with at least one pedometer assessment. Unlike traditional repeated measure analyses (e.g., analysis of variance; multi-level modeling) that assume patients come from a single population and that a single trajectory can adequately approximate an entire population (23), LCGA attempts to group individuals who are similar to each other and different from individuals in other groups (24). Importantly, the results from the Soicher et al. (15) study supports the use of LCGA in COPD patients as they showed 3 distinct groups of patients exist (i.e., have their own unique trajectories) after completing PR. Therefore, the traditional single-trajectory approach may over-simplify changes in PA in patients during and after PR. LCGA can therefore be used to estimate the size and number of potential latent classes, as well as investigate which demographic and clinical characteristics are associated with class membership. Thus, LCGA allows researchers to identify key groups (i.e., those with especially low PA levels) for intervention, in order to more effectively target resources toward patients with the greatest need.

A single class model was specified first with a latent intercept growth factor (i.e., steps/day at the beginning of PR) and a latent slope growth factor (0 = beginning of PR; 3 = end of PR; 6 = 3 months after PR; 12 = 9 months after PR) to reflect the monthly change in steps/day. Next, a quadratic term was added to test its necessity in the model. The same procedure was then used to examine two, three, and 4-class models. To identify the number of classes (objective #1), the Bayesian Information Criterion (BIC), the Vuong-Lo-Mendell-Rubin likelihood ratio (LRT), and entropy indices were used. When comparing a 2-class versus single class model, for example, a change in the BIC > 10, higher entropy value (near 1.0), and a LRT P-value < .05 would be considered as evidence favoring the 2-class over the single class model (25). Once the final number of classes was generated, a series of χ² and between-subject ANOVAs were conducted in SPSS 20.0 to determine potential demographic, clinical, and/or program differences across the classes. Prior to doing so, however, missing values for the baseline demographic, clinical, and program (i.e.,% of classes attended) variables were imputed using the multiple imputation (26) procedure in SPSS 20.0.

Results

A total of 294 patients were eligible for the study. Ninety of these patients declined participation, primarily due to lack of interest (41%) or health reasons (12.2%), and 14 consented to participate but did not provide at least 1 pedometer assessment, leaving a final sample of 190 patients who agreed to participate and completed at least one pedometer assessment. Of the 190 patients who participated, 159 completed an assessment at the end of PR, which on average, was 49.73 days (SD = 16.53) after the baseline assessment. A total of 144 participants completed the assessment three months after PR, and 116 completed the assessment 9 months after PR. Patients who completed the end of PR assessment were more likely to be retired (91.6%) compared to those who were employed, unemployed, homemakers, or on disability (74.5%) (p < 0.01) and also walked a further distance during the baseline 6-minute walk test (r = 0.20, p < 0.01). Patients who completed the 3-month post PR assessment had a higher baseline FVC compared to those who did not (r = 0.21, p < 0.01), whereas older (69.4% ≥ 65 years; 50% < 65 years, p < 0.01) and retired (i.e., 69.7% versus the 47.3% of those who were employed, unemployed, homemakers or on disability, p < 0.01) patients were more likely to complete the 9-mo post PR assessment.

Results from the LCGAs are presented in Table 1. Results from the single class model show that patients averaged 4183.23 steps/day at the beginning of PR. The linear trend was non-significant and indicates that the patients’ steps/day slightly declined each month by 16.03, which translates into the patients engaging in 4155.16 steps/day at the end of PR. In examining the fit indices, however, one can see that a 2-class model is a better fit to the data than a single class model as the change in the BIC was > 10, the LRT was significant, and the entropy value was very high (.95). However, when comparing the 2- vs. 3-class model, although the BIC improved, the LRT was non-significant for the 3-class model, and the entropy value decreased. Two distinct patient groups emerged: Class 1 (Active Maintainers) comprised 16% of the sample and averaged 9177 steps/day at baseline, which remained stable throughout PR and the subsequent follow-up period (end of PR = 9418 steps/day; 3 months after = 8619 steps/day; 9 months after PR = 9330 steps/day) (Figure 1). Class 2 (Inactive Maintainers) comprised 84% of the sample. Patients in this group averaged 3133 steps/day at baseline, which also remained stable during and after PR (end of PR = 3117 steps/day; 3 months after = 3008 steps/day; 9 months after PR = 2863 steps/day).

Figure 1. Physical activity trajectories identified in the LCGA analysis.

Table 1.  Results from the latent class growth analyses for steps/day

	Model	Coefficient	BIC	LRT	Entropy
Analysis 1
Class 1 (n = 190)	Intercept	4183.23^a
	Linear Trend	−16.04	11623.79	–	–
Analysis 2
Class 1 (n = 31)	Intercept	3133.41^a
	Linear Trend	−21.01
Class 2 (n = 159)	Intercept	9177.09^a		1 vs. 2 classes
	Linear Trend	−4.73	11069.44	748.15^b	.95
Analysis 3
Class 1 (n = 25)	Intercept	9819.65^b
	Linear Trend	14.16
Class 2 (n = 46)	Intercept	5176.43^b
	Linear Trend	−29.24
Class 3 (n = 119)	Intercept	2435.00^b		2 vs. 3 classes
	Linear Trend	−24.83	10960.24	183.94	.84

Note. ^ap < .001; ^bp < .05; BIC = Bayesian Information Criteria; LRT = Vuong-Lo-Mendell-Rubin likelihood ratio test; All models were tested with a quadratic trend and in all cases, the quadratic trends were non-significant and the fit indices did not improve.

Descriptive statistics for the demographic, clinical, and program characteristics of the group as a whole, as well as the two PA classes are displayed in Table 2. As can be seen, Inactive Maintainers were more likely to be retired and have lower baseline scores for their stress tests and 6-minute walk tests compared to Active Maintainers (all p < 0.05). A borderline significant sex difference (p = 0.05) was present, with males more likely being Inactive Maintainers.

Table 2.  Descriptive statistics for all participants, and by class

Patient Characteristics	All Participants	Active Maintainers (Class 1)	Inactive Maintainers (Class 2)	Comparison of Classes 1 and 2
Demographics
Age (% ≥ 65 years)	60.5	51.6	62.3	1.23 (p = 0.27)
Education (% > Grade 12)	35.3	38.7	34.6	.19 (p = 0.66)
Female (%)	45.3	61.3	42.1	3.84 (p = 0.05)
Married/Common-Law (%)	66.3	74.2	64.8	1.03 (p = 0.31)
Retired (%)	66.8	48.4	70.4	5.69 (p = 0.02)
Income (% > = $32,500)	51.1	51.6	50.9	.01 (p = 0.95)
Clinical
BMI	28.8 (6.9)	27.7 (5.0)	29.0 (7.2)	.91 (p = 0.34)
# Co-morbidities	3.28 (2.51)	2.65 (1.92)	3.41 (2.60)	2.42 (p = 0.12)
FEV1 (L/s)	1.33 (.59)	1.41 (.62)	1.31 (.59)	.79 (p = 0.37)
FVC (L)	2.68 (1.02)	2.77 (.95)	2.66 (1.04)	.31 (p = 0.58)
FEV1 (% Predicted)	48.96 (26.40)	50.31 (23.69)	48.69 (26.95)	.10 (p = 0.76)
FVC (% Predicted)	72.03 (31.06)	81.09 (27.14)	70.27 (31.54)	3.18 (p = 0.08)
FEV/FVC	.55 (.42)	.53 (.18)	.55 (.45)	.10 (p = 0.76)
Stress Test (METs)	4.15 (1.99)	5.03 (2.18)	3.97 (1.91)	7.65 (p = 0.01)
6-Minute Walk Test (m)	348.59 (103.85)	427.7 (76.0)	333.2 (101.7)	24.12 (p < 0.01)
Current Smoker (%)	18.9	12.9	20.1	.88 (p = 0.35)
Program
Length
6 weeks	24.7	29.0	23.9
10 weeks	10	12.9	9.4
12 weeks	65.3	58.1	66.7	.88 (p = 0.64)
Frequency
2x week	47.9	48.4	47.8
3x week	52.1	51.6	52.2	.004 (p = 0.95)
Region of Canada
Western	29.5	32.3	28.9
Central	23.2	19.4	23.9
Atlantic	47.4	48.4	47.2	.34 (p = 0.85)
% of Classes Attended to PR	76.42	76.12	76.48	.01 (p = 0.92)

Note: METS = metabolic equivalents. Categorical data are presented as percentages, and compared using chi square. Continuous variables are presented as mean (SD), and compared using a one-way ANOVA.

Discussion

The single class model found that participants averaged 4183 steps/day during and after PR, a level which has previously been defined as “severely inactive” for individuals with COPD (27). This level of PA is also in line with previous research reporting that patients with COPD average between 2000 and 7000 steps/day (9, 28–33). However, our results suggest that not all patients follow the same PA trajectory during and after PR, which is consistent with previous research in the post-PR phase (15). Instead, we have identified two distinct trajectories; those who maintained a relatively high level of PA throughout the 12-mo period, and those who maintained a relatively low level of PA.

Specifically, participants in the Active Maintainers group entered PR with an average PA level above 9000 steps/day at baseline, which remained stable throughout PR and the 9-mo follow-up period. This level of PA meets current recommendations for moderate-intensity physical activity, and is thought to be associated with positive health benefits (34–37). Given that they are physically active during and after PR, a focus on relapse prevention may be all that is needed for individuals in this class. However, Active Maintainers comprised just 16% of the study sample. In other words, 84% of the PR patients were in the Inactive Maintainers group, and would therefore be considered “severely inactive” (27) or “sedentary” (34,35,38). Though previous studies found positive health benefits from PR in the absence of increases in PA (9,10), our results nonetheless suggest the majority of PR patients are not experiencing the full health benefits of adopting a physically active lifestyle. Future studies should therefore investigate ways to increase PA levels and decrease sedentary behaviour among patients in this particular class.

The second purpose of this study was to examine differences in demographic, clinical and program characteristics across the PA trajectory classes. Non-retirees and individuals with relatively high values on the pre-PR stress test and 6-minute walk test were more likely to be active at baseline, and to remain active throughout PR (i.e., be Active Maintainers). It may therefore be useful to focus resources on PR patients who are retired and have low levels of PA at baseline, as opposed to younger patients who are already sufficiently active. Interestingly, though retired individuals were over-represented in the Inactive Maintainers group, they were less likely to drop-out as the study progressed. However, this finding is likely due mostly to time constraints among those who were still employed, and of little clinical significance.

The average 6-minute walk distance for participants in the Active Maintainers group was 427.7 m; slightly below the 10^th percentile for healthy populations (39). In contrast, participants in the Inactive Maintainers group walked an average of just 333.2 m—almost 100m less than the Active Maintainers—which falls well below the 10^th percentile for both sexes (39). The 95% confidence intervals for the stress test, 6-minute walk test and steps/day were roughly 3.67–4.27 METs, 317–349 m and 2850–3434 steps/day, respectively in the Inactive Maintainers group. These findings suggest that clinicians may be best served by targeting PA interventions on patients that fall in or below these ranges at the beginning of PR, in addition to those who are retired. The stability of PA throughout the assessment period in both groups highlights the importance of objectively measuring PA at the beginning of PR, and suggests that this alone may be sufficient to identify those with the greatest need for intervention. Given that the present study found few differences in terms of demographic or clinical differences between Active and Inactive trajectories, future studies should examine whether other behavioural or psychological factors (e.g., self-efficacy, motivation, anxiety) can more readily distinguish between the two groups.

Strengths and weaknesses of the current study warrant mention. Strengths include a relatively large sample size representing PR programs from across Canada, as well as an objective measure of PA (e.g., pedometers) that was assessed during and after PR. This study also employed the use of LCGA, allowing us to identify distinct PA trajectories, rather than assuming that all participants followed a similar trajectory. However, participation in this study was self-selected, which is a potential source of bias. Daily step counts and co-morbidities were self-recorded, which may have introduced bias into these measures. Future studies should employ accelerometers or other tools which permit the quantification of MVPA, as well as sedentary behaviour (e.g., sitting (40)). This would allow researchers to investigate the impact of PR on PA intensity (or walking cadence), as well as the total volume of PA. It should be noted that pedometers may undercount steps at low walking speeds (20), which may have artificially lowered step values in this population. The use of accelerometers may therefore detect PR-related changes in the volume of PA which were not seen when assessing steps/day via pedometers in the current study.

Because of the sole focus on Canadian PR programs, future studies should examine whether similar trajectories are observed in other regions. Baseline measures of PA were taken 2–3 weeks after the beginning of PR. It is therefore possible that PR participation may have influenced baseline PA measures for some individuals. Finally, clinical outcomes were examined at the beginning of PR only, and thus we were not able to assess whether these variables changed over time. Future studies should therefore investigate whether health outcomes also differ longitudinally across PA trajectory classes.

Conclusion

In conclusion, we identified two PA trajectories for participants in PR programs: one group with relatively high levels of PA during and after PR, and another with low levels of PA during and after PR. Our findings suggest that women, non-retirees, and individuals with high values on the pre-PR stress test and 6-minute walk test are more likely to be active at baseline, and to remain active throughout PR. Future research should examine whether health outcomes differ across PA trajectory classes following PR, and whether different interventions may be more effective for the demonstrated PA trajectory classes.

Declaration of Interest Statement

The present study was supported by a grant from the Social Sciences and Humanities Research Council of Canada. TJS was supported by a Post-Doctoral Fellowship from the Heart and Stroke Foundation of Canada. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The authors report no conflicts of interest.

Acknowledgments

The authors wish to thank the study participants, as well as the staff at each of the participating PR programs, for their contribution to this research project.