Physical activity is a key prognostic factor in chronic obstructive pulmonary disease (COPD), yet poorly understood. COPD patients are less active than healthy controls from the early stages of disease onwards1•3 and this reduced activity has been associated with impaired prognosis and accelerated progression of COPD.4,5 For this reason, several national and international COPD guidelines recommend encouraging patients to maintain a good physical activity level.6,7

Despite patients
tm) and health professionals
tm) efforts, physical activity has been shown to exhibit a spontaneous decline of an average of 400•500steps/day per year in COPD patients.8•14 Such observed decline has been related, although not consistently, to lower lung function levels,10,11 the presence of exacerbations15•17 or the seasonality of testing (e.g., decline observed in patients going from summer to winter).12,18 Given the heterogeneous nature and progression of COPD,19 it can be hypothesized that physical activity progression also displays different patterns, not captured by the mean physical activity values. Two previous studies support this hypothesis by showing distinct physical activity trajectories over 9 months after a pulmonary rehabilitation program.20,21 However, the reported progression in physical activity after rehabilitation programs probably does not reflect how physical activity evolves in the wider COPD population nor in an observational setting, where patients receive a variable combination of pharmacological and non-pharmacological treatments.

We aimed (1) to identify, using a hypothesis-free approach, distinct patterns of natural physical activity progression in COPD patients recruited from diverse settings (primary care, hospital and rehabilitation services) and followed during 12 months; and (2) to assess the baseline sociodemographic, interpersonal, environmental, clinical and psychological determinants for the identified patterns. Better understanding of the natural progression of physical activity, of potential distinct patterns and of their determinants could help to individualize strategies to increase (or prevent a decline in) physical activity.

From 412 patients at baseline, 291 (71%) completed the follow-up visit and were included in the current analyses (Fig. S1). These patients had a higher proportion of males, better functional status and were more active at baseline than those lost-to follow-up (Table S1). Included patients were 81% male and had a mean age of 68 years, FEV1 of 59% predicted, 6MWD of 477m, mMRC dyspnea score of 1.3, and walked 6720 steps/day (Table 1). Compliance with the activity monitor during waking hours was excellent: at baseline median (range) valid days of 7 (3•7) and mean±SD wearing hours of 14.6±0.5 in Urban Training, and 6 (3•7) days and 14.8±2.2 wearing hours in PROactive; and at follow-up, 7 (4•7) days and 14.6±0.6 wearing hours in Urban Training, and 6 (3•7) days and 14.4±2.2 wearing hours in PROactive.

At the group level, the step count did not change over 12 months. In the hypothesis-free approach, we identified three cluster groups (three distinct physical activity patterns) (Fig. 1, Table S2). A first cluster (n=173 [59%]), labeled Inactive pattern due to the low step count, walked at baseline mean±SD 4621±1757 steps/day and decreased their physical activity by 487±1201 steps/day over 12 months. A second cluster (n=49 [17%]), labeled ActiveImprovers, walked 7727±3275 steps/day at baseline and increased by 3378±2203 steps/day. The third cluster (n=69 [24%]), labeled ActiveDecliners, walked 11,267±3009 steps/day at baseline and decreased by 2217±2085 steps/day. Distribution of MVPA and walking intensity by physical activity pattern followed the same sequence as steps/day, except for walking intensity in Active Improvers that did not change. Sedentary time did not change for Inactive, decreased for Active Improvers and increased moderately for Active Decliners. The physical activity experience as expressed by C-PPAC scores did not change for the Inactive pattern; the Active Improvers increased the C-PPAC scores (i.e., increased amount and reduced difficulty); the Active Decliners decreased the C-PPAC amount and total scores while the C-PPAC difficulty score did not change (Fig. 2, Table S2).

Fig. 1.

Physical activity variables at baseline and at 12-month follow-up, overall and by PA progression pattern (Inactive, Active Improvers and Active Decliners). Notes: Data are presented as mean±SE (specific numbers are presented in Table S2). * P-value ≤.05. Abbreviations: MVPA: moderate-to-vigorous physical activity; MET: metabolic equivalent of task.

(0.43MB).

Fig. 2.

Physical activity experience variables at baseline and at 12-month follow-up, overall and by PA progression pattern (Inactive, Active Improvers and Active Decliners). Notes: Data are presented as mean±SE (specific numbers are presented in Table S2). * P-value ≤.05. Abbreviations: C-PPAC: Clinical visit•PROactive Physical Activity in COPD (higher numbers indicate a better score). C-PPAC variables have 87 missing values: 38 in Inactive, 21 in Active Improvers, and 28 in Active Decliners.

(0.33MB).

Patients in the Inactive physical activity pattern had a higher degree of education, a smaller proportion was living with a partner or grandparenting, and they presented with a worse general health status, lower lung function, poorer exercise capacity, worse quality of life and higher dyspnea and depression scores than those in the Active Improvers or Decliners patterns (Table 2). Active Improvers and Decliners were very similar in their baseline characteristics, except for their daily step count.

In the multivariable multinomial logistic regression model we used Active Decliners as the reference pattern to capture both the determinants of being Inactive vs Active and the determinants of being an Active Improver vs Decliner. A lower exercise capacity and a higher mMRC dyspnea score were independently related with being Inactive whereas no variable was identified as independently associated with being an Active Improver (Table 3). The final model showed good fit. Sensitivity analyses confirmed the results (Tables S3•S5).

This study identified, for the first time to our knowledge, the natural progression of physical activity in COPD patients. We used a hypothesis-free approach that allowed the identification of patterns without a priori assumptions about the physical activity changes over time. We found that (1) the natural change in physical activity over time was indeed heterogeneous; (2) the majority of patients (59%) was inactive at baseline and decreased their physical activity level subsequently; (3) among active individuals some increased and some decreased their physical activity level; and (4) although clinical COPD characteristics were related to the physical activity level at baseline they could not predict subsequent physical activity changes.

A first important finding is that physical activity progression in COPD is heterogeneous. In our 12-month study, mean changes in the full group were virtually zero; however when using hypothesis-free clustering methods, we identified one Inactive pattern which decreased and two Active patterns which increased or decreased physical activity. This observation is in line with previous reports of heterogeneous physical activity progression in patients with rheumatoid arthritis.30 The average lack of 12-month change in step count differs from previous studies that showed overall a decrease in physical activity.8,10,11,13 A potential explanation is that most of these studies recruited patients from outpatient or pulmonary clinics, who may have slightly more advanced disease and reduced variability in physical activity and COPD characteristics as compared to our sample including also primary care. Supporting this, the group of patients who started with a lower physical activity (59%) was similar in their baseline characteristics to previous studies and also had a comparable mean decrease of around 500 steps/day.10,11 Notably, the low overall dyspnea score may have positively influenced the physical activity level of our study population. A second explanation could lie in the high proportion of male subjects and regional differences in physical activity practice (a cohort of patients included in the Mediterranean region5 had a baseline physical activity comparable to the Urban Training sample). These characteristics of our sample could justify the two patterns with relatively high baseline physical activity and an average small physical activity change.

The second important finding is that there seems to exist a group of COPD patients (our Active Improvers) that spontaneously increase their physical activity over time. Of note, such observed increase of >3000steps/day is remarkably high given that the minimal important difference has been proposed between 600 and 1100steps/day.31 There are several possibilities that would explain this observed increase. First, some patients could have been inactive at baseline by chance; however, we tested this option against study records by screening for atypical events and it did not hold true. Second, regression to the mean could account for part of the increase, but in our data regression to the mean was estimated to account for maximal 25% of the effect. Third, changes in daily steps could be due to changes in wearing time, but this was not the case in our study (Table S4). Fourth, patients could have increased their physical activity after participation in rehabilitation programs, but this was dismissed in our analysis (Table S5). Finally, we considered that some patients in the usual care arm of Urban Training could have increased their physical activity due to being enrolled in a physical activity study. However, the proportion of patients from Urban Training was similar between Active Improvers and Active Decliners. Thus, we suggest that some patients do actually increase their physical activity.

The evolution of other physical activity variables provided complementary information. Time in MVPA and sedentary time (opposite direction) paralleled the progression of step count in all three patterns, supporting previous research that suggested that in COPD patients, physical activity and sedentary time provide information about the same concept.5 We also investigated the progression of physical activity from the perspective of patients. As expected, C-PPAC amount and total scores followed a trend similar to the objectively measured physical activity, as they include steps/day in their calculation. However, C-PPAC difficulty score remained unchanged in Inactive and Active Decliners and increased (i.e., less difficulty) in Active Improvers, suggesting that the observed increase in amount could be related to experiencing fewer difficulties (less dyspnea for instance32) while being active.

Our third main finding is the impossibility to predict the physical activity progression patterns, despite having included sociodemographic, interpersonal, environmental and psychological characteristics in addition to the typical clinical COPD variables. We found a large set of COPD-related, functional characteristics associated with the Inactive pattern, in accordance with previous, mainly cross-sectional, literature about the determinants of physical activity levels in COPD.2,4 Also higher education levels, lower social support (living alone, not taking care of grandchildren) and higher depression scores related to being in the Inactive pattern, although none of these factors remained in the multivariable model suggesting they were subject to confounding.

Most tellingly, we did not identify any factors that could predict among Active patients, the evolution to Improvers or Decliners. Surprisingly, the presence of severe exacerbations during follow-up did not play any role. It could be speculated that our harmonized exacerbation data was not detailed enough to distinguish the severity of exacerbations, the length of hospital stay or the time from the last exacerbation to physical activity assessment at follow-up. We also considered the role of incident comorbidities during follow-up, which could have influenced behavior, but they were not significantly different for the three patterns. Pharmacological treatment for COPD was not different across progression patterns discarding any potential role for treatment inappropriateness. Moreover, we did not find an association between the recruitment season and physical activity progression. This is in line with the hypothesis that the recruitment season, although possibly affecting the baseline levels of physical activity,12,18,33 would not affect the progression pattern during a follow-up of 12 months. Finally, we did not find an effect of accumulated rainfall on physical activity progression, as recently described cross-sectionally in the same PROactive population.14

Our study has several implications. It adds to the current knowledge that contrary to the general belief not all patients decline but some patients considerably improve their physical activity, which should be confirmed in future research and shows the importance of including a usual care group in intervention studies. The limitation of traditional clinical COPD characteristics to predict physical activity progression suggests that further research should broaden the view and give more attention to interpersonal and environmental factors potentially related to the individual's motivation. As the optimal timing and use of physical activity interventions to improve physical activity (especially in the long term) is still unclear,34 understanding the different COPD progression patterns may help to overcome a one-size-fits-all approach and customize physical activity promotion to reflect different physical activity practices and different treatment needs.35 Finally, our results highlight the limitation of using mean population values in phenomena that are heterogeneous in nature.

A major strength of our study is the inclusion of patients across a broad spectrum of disease severities and physical activity in several European cities. This makes our results applicable (i.e., more representative) to a larger COPD population than a single recruitment setting or severity group. In addition, the inclusion of patients from diverse geographic locations allowed us to indirectly control for residual confounding. Moreover, we included some variables beyond the conventional clinical COPD characteristics.36 The use of the hypothesis-free clustering approach allowed us to identify patterns of physical activity progression based on the distribution of the data without prior assumptions.

However, we acknowledge some shortcomings. We had a small sample size for some of the hypothesized determinants of physical activity progression patterns, such as dog walking, current pulmonary rehabilitation and knowledge of baseline physical activity, which precluded our ability to test their role. Similarly, we did not collect information on some physical activity barriers (e.g. costs or transportation difficulties), which precluded testing their role on physical activity progression. The drop-out was 29% which is comparable to previous studies37,38 but the excluded patients had worse functional parameters, and we cannot rule out that they would have presented with a fourth, potentially declining pattern. The two measurement points available for both studies allowed to investigate only linear patterns over time. Having more data points could provide more detailed information on the trajectories. In addition, a longer follow-up would have been desirable, but the 12-month span appears reasonably long to provide this first novel insight into physical activity patterns. Finally, one might argue that pooling of the two studies was not appropriate, although our sensitivity analyses showed similar cluster results and characteristics and it resulted in a broad spectrum of physical activity and COPD severity.

In conclusion, the natural change in physical activity over time in COPD patients is heterogeneous and three distinct patterns of physical activity progression have been identified: a predominant Inactive pattern, related to worse scores for clinical COPD characteristics, and two Active patterns, Improvers and Decliners, which cannot be predicted at baseline.

JGA had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. MK, HD and JGA drafted the first version of the manuscript; MK, HD, MB and JGA had full access to the data and were responsible for the statistical analysis; MK, HD, MB, AAE, EB, ABG, EGS, NH, CdJ, NK, ZL, MPo, MPu, RRa, RRo, PVC, IV, TT, JGA contributed to data collection and coordination. All authors (1) provided substantial contributions to the conception or design of the work; or the acquisition, analysis or interpretation of data for the work; (2) drafted or revised the manuscript for important intellectual content; (3) approved the final version; and (4) agreed to be accountable for all aspects of the work.

This study was conducted in accordance with the amended Declaration of Helsinki. Both studies were approved by all local institutional review boards and written informed consent, including re-use of data for COPD-related research, was obtained from all patients.

The Urban Training trial was approved by the ethics committees of all participating institutions (Comitè Ètic d‘Investigació Clínica Parc de Salut MAR 2011/4291/I, Comitè Ètic d‘Investigació Clínica de l
tm)IDIAP Jordi Gol i Gurina P11/116, Comitè Ètic d‘Investigació Clínica de l
tm)Hospital Universitari de Bellvitge PR197/11, Comitè Ètic d‘Investigació Clínica de l
tm)Hospital Universitari Germans Trias i Pujol AC-12-004, Comitè Ètic d‘Investigació Clínica de l
tm)Hospital Clínic de Barcelona 2011/7061, Comitè Ètic d‘Investigació Clínica de l
tm)Hospital de Mataró November 23rd, 2011).

The PROactive study was advised and approved by the PROactive ethics and patient advisory boards, and approved by the local ethics committee at each center (Commissie medische ethiek van de universitaire ziekenhuizen KU Leuven (Leuven, S-55919); Medische ethische toetsingscommissie universitair medisch centrum Groningen (Groningen, Metc 2013.362); RES Committee London•South East (London and Edinburgh, 13/LO/1660); Scientific Council of the ‘Sotiria
tm) General Hospital for Chest Diseases (Athens, 27852/7-10-13); Kantonale Ethikkommission Zñ/4rich and Ethikkommission Nordwest- und Zentralschweiz (Zurich, KEK-ZH-Nr. 2013-0469).

Prior abstract presentation/publication: This work was presented at the European Respiratory Society Congress 2019 in Madrid and the abstract was published in the European Respiratory Journal 2019; 54: Suppl. 63, OA5361.

The Urban Training study was funded by grants from Fondo de Investigación Sanitaria, Instituto de Salud Carlos III (ISCIII, PI11/01283 and PI14/0419), integrated into Plan Estatal I+D+I 2013•2016 and co-funded by ISCIII-Subdirección General de Evaluación y Fomento de la Investigación and Fondo Europeo de Desarrollo Regional (FEDER); Sociedad Española de Neumología y Cirugía Torácica (SEPAR, 147/2011 and 201/2011), Societat Catalana de Pneumologia (Ajuts al millor projecte en fisioteràpia respiratòria 2013). We acknowledge support from the Spanish Ministry of Science and Innovation through the “Centro de Excelencia Severo Ochoa 2019•2023” Program (CEX2018-000806-S), and support from the Generalitat de Catalunya through the CERCA Program.

The PROactive project was funded by the European Commission Innovative Medicines Initiative Joint Undertaking (IMI JU # 115011). HD is a post-doctoral research fellow of FWO Vlaanderen. ZL is a post-doctoral fellow of the FWO-Flanders (#12U5618N).

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

NK is employed by AstraZeneca. MP reports personal fees from Philips, during the conduct of the study. TT reports lecture fees to his institution from Boehringer Ingelheim, Chiesi Belgium and AstraZeneca outside the submitted work. JGA reports payments for consulting and lecture fees to her institution from AstraZeneca and lecture fees from Esteeve, Chiesi and Menarini outside the submitted work. The authors report no other conflict of interest in this work.