Associations of total and type-specific physical activity with mortality in chronic obstructive pulmonary disease: a population-based cohort study

Compared to their healthy peers, people with chronic obstructive pulmonary disease (COPD) engage in a lower intensity of daily physical activity (PA), and spend significantly more time sitting and lying down [1, 2]. This has been attributed to disease-specific limitations such as dyspnoea, fatigue and reduced exercise tolerance, which reduce the capacity of people with COPD to engage in moderate-vigorous intensity physical activity (MVPA) [3]. Previous cohort studies have consistently shown that low levels of PA are associated with an increased risk of all-cause and respiratory mortality in people with COPD, as well as an increased risk of COPD exacerbations, irrespective of the degree of airflow obstruction [4]. Low levels of PA are also associated with an increased risk of hospitalisation, re-hospitalisation after an acute exacerbation, and longer hospital length of stay [4, 5].

Although regular PA is recommended for all patients with COPD [6], there is a lack of high-quality evidence on which to base specific PA recommendations for this population, in particular the dose of PA required to gain mortality benefit. A Spanish cohort study of 611 people with COPD found that those who “took walks regularly for >8 km, no less than 5 days a week, or practised sports” had a lower mortality risk compared to those who “don’t leave the house and life is limited to the bed or armchair, or to doing some domestic chores” (risk ratio (RR) 0.38, 95% confidence interval (CI) 0.11–1.29) [7]. Another cohort study of 173 people with moderate to very severe COPD, which used accelerometry to measure PA levels, found that every ten vector-magnitude unit (VMU) increase in daily PA was associated with a 14% reduction in mortality risk [8]. Such studies are limited by the use of small non-representative samples of people with COPD to examine the association between PA and mortality risk, and by arbitrary definitions of what constitutes “low” and “high” levels of PA. Many different questionnaires and activity monitors have been used to assess PA in people with COPD, and the optimal mode of reporting (e.g., energy expenditure, number of steps/day) is not yet known [4, 9].

The considerable heterogeneity in the way that PA has been classified and reported in previous studies has made it difficult to develop PA guidelines specific to people with COPD, and subsequently clinicians must rely on the PA guidelines for the general population to inform clinical recommendations [10]. It is unknown whether specific types of PA, such as walking and domestic PA, are associated with reductions in mortality risk. Exploring types of PA that are relevant to functionally limited populations such COPD, and the dose of PA required to gain mortality benefit, is necessary in order to guide the prescription of PA in clinical practice and to inform interventions which aim to increase PA.

The aim of this study was to examine the associations between total and type-specific PA and mortality risk in a series of representative cohorts of adults with COPD, specifically:

Of the 19,000 participants with available data on lung function and PA, 2398 participants met the eligibility criteria for COPD and were included in the analysis. The mean age of the sample was 62.6 (SD 11.5) years and 52% were men. There was a high prevalence of COPD of mild (42.7%) and moderate (37.5%) severity, and of never-smokers (27.9%). Baseline characteristics, grouped according to the degree of adherence to the general PA recommendations, are displayed in Table 1. Participants who reported no PA were older, had COPD of greater severity, had more medical comorbidities, and had a lower level of education (p < 0.001). Mortality data were available for 2155 participants. Over a mean 8.5 (3.9) years follow-up, there were 571 deaths (26%) of which 172 were CVD-related deaths and 105 were respiratory-related deaths, corresponding to a total of 18,222 person years.

The associations of MET-hours/week of total PA and minutes/week of MVPA with mortality risk are shown in Table 2. Dose-response associations with all-cause, CVD and respiratory mortality risk were demonstrated for total PA and MVPA. Participants in the ‘Insufficiently Active (High)’ group for total PA had a reduced risk of all-cause and CVD mortality compared to those who reported no PA, but no significant mortality risk reductions were observed in the ‘Insufficiently Active (Low)’ group for total PA (Fig. 1). Conversely, participants in the ‘Insufficiently Active (Low)’ group for MVPA had a reduced risk of all-cause mortality compared to those who reported no MVPA (Fig. 2). These dose-response associations were also observed in the sensitivity analyses (Additional file 1: Tables S1 and S2).

The associations of MET-hours/week of walking, domestic PA, and sport/exercise with mortality risk are shown in Table 3. Dose-response associations with all-cause, CVD and respiratory mortality risk were demonstrated for walking, with no significant mortality risk reductions observed in the ‘low’ walking group. There was a dose-response association between sport/exercise and all-cause mortality risk, but not CVD or respiratory mortality risk. The few number of CVD and respiratory deaths in the ‘low’ and ‘high’ categories of sport/exercise have likely contributed to this result. Domestic PA of any level was not associated with reductions in mortality risk. Similar results were demonstrated in the sensitivity analyses (Additional file 1: Tables S3 and S4).

This study of 2398 participants who met the GOLD spirometric criteria for COPD provides a detailed analysis of the dose-response associations between PA and mortality risk within the context of current public health recommendations, and for specific types of PA. The use of established thresholds to examine PA in both MET-hours/week and minutes/week provides clinicians with a meaningful reference standard to guide prescription of PA, and enables future studies to make direct comparisons to the magnitude of mortality risk described in this sample. This is the first population-based cohort study of people with COPD to examine whether specific types of PA confer mortality benefit, which has high clinical relevance when developing interventions to increase PA in people with COPD.

Dose-response associations with mortality risk were demonstrated for total PA, MVPA, walking and sport/exercise, but not domestic PA. A previous Danish population-based cohort study of people with COPD, which compared subjects who reported low, moderate, or high levels of PA to subjects who reported a very low level of PA, showed that some level of regular PA was associated with a lower risk of all-cause and respiratory mortality [18]. In our study, reductions in mortality risk of the greatest magnitude were observed in participants who adhered to the PA guidelines. Compared to those who reported no PA, participants who were ‘Sufficiently Active’ showed a 45% reduction in all-cause mortality risk, a 50–55% reduction in CVD mortality risk, and a 55–60% reduction in respiratory mortality risk.

Although there is added benefit in adhering to the PA guidelines, meeting such recommendations is often difficult for people with symptomatic COPD. This study showed that participants who did not meet the guidelines but reported at least 3.75 MET-hours/week of total PA, or reported some level of MVPA, still had significant reductions in mortality risk. After taking the sensitivity analyses into consideration, which yielded similar mortality risk reductions in participants without existing CVD, cancer and diabetes, these results suggest that the dose of PA required to gain mortality benefit in COPD may be lower than is currently estimated. This is consistent with previous observations that relatively low levels of PA have beneficial effects on mortality risk in people with COPD. It has been suggested that as little as 2 h/week of walking or cycling is associated with a 30–40% risk reduction in respiratory mortality and COPD hospitalisation [18].

In both the main and sensitivity analyses, participants in the ‘Insufficiently Active (Low)’ group for MVPA had a significant reduction in all-cause mortality risk (HR 0.75, 95% CI 0.61–0.94). However, no significant reductions in all-cause mortality risk were observed in participants who reported < 3.75 MET-hours/week of total PA (HR 0.86, 95% CI 0.67–1.10). This subtle difference between the two equivalent measures of PA is likely due to error in the calculation of MET-hours/week of total PA, in which the intensity of PA is approximated according to the nature of the activity as indexed in the MET compendium. Whether higher intensity PA has beneficial effects on mortality risk independent of the volume of activity is not yet known in people with COPD. A Spanish cohort study demonstrated a 20% reduction in the risk of COPD hospitalisation for every additional 1000 daily steps at low average intensity, but no change in hospitalisation risk with additional steps at high average intensity [19]. Conversely, higher intensity PA was associated with a lower prevalence of metabolic syndrome in 223 subjects with COPD drawn from the National Health and Nutrition Examination Survey 2003–2006 dataset [20]. Considering the innate challenges of increasing MVPA in people with COPD, encouraging more light-intensity PA in this population may be a first and more realistic step than targeting MVPA alone.

Although there was a dose-response association between sport/exercise and mortality risk, sport is not feasible for most people with COPD due to the older age of the population and the high prevalence of comorbidities such as poly-pharmacy and mobility problems [21], which are associated with poorer physical performance [22]. Walking, however, is an achievable and accessible type of PA. It requires no specialised exercise equipment and, in the form of ground walking or treadmill walking, is commonly undertaken as part of pulmonary rehabilitation programs with good effect [23]. This study showed that participants who reported at least 5.25 MET-hours/week of walking (approximately equivalent to 105 min/week of brisk walking) had a 30% reduction in all-cause mortality risk, a 44% reduction in CVD mortality risk, and a 53% reduction in respiratory mortality risk compared to those who reported no walking. Future studies should explore strategies for incorporating more walking into daily life for people with COPD. Nordic walking is one such strategy, where improvements in PA were maintained to 3 months in the group of participants [24].

Although domestic PA is feasible for most people with COPD, this study showed no association between domestic PA of any level and mortality risk. There is conflicting evidence about the protective effect of domestic PA on mortality. The domain is often poorly defined; for example, housework and gardening are both treated as domestic PA despite most housework activities being classified as light intensity, since they do not require locomotion or the use of large muscles groups [25]. An English population-based study of 15,000 adults found that PA “around the home”, such as lawn mowing and digging, was inversely related to all-cause mortality risk [26]. However, a recent study of the SHS cohort found no association between intense domestic PA and CVD mortality risk after adjusting for relevant confounders [17]. The inclusion of more light intensity activities in our definition of domestic PA may explain the absence of a dose-response association between domestic PA and mortality risk.

Our study has several limitations. A fixed FEV₁/FVC ratio was used to identify participants with COPD due to a lack of other information that may have helped to confirm a diagnosis. Although COPD can be identified with high diagnostic accuracy using spirometry, a fixed ratio of obstruction would not take into account any reversible factors that may have influenced FEV₁, such as asthma or a respiratory infection on the day of testing. Spirometry performed on 219 adults in a general practice setting showed a sensitivity of 0.92 (95% CI 0.80–0.97) for diagnosing COPD, with eight “false positives” identified from further laboratory testing [27]. To account for misclassification of COPD in our study, we conducted sensitivity analyses which excluded participants with asthma and/or a recent respiratory infection, and these yielded similar results to the main analysis.

Misclassification of the PA variables is also possible, since self-reported methods were used to quantify PA and any changes in PA levels during follow-up were not accounted for. This would have most likely reduced the magnitude of the effect of PA on mortality risk [28]. It is well-established that self-report PA questionnaires tend to overestimate the time spent physically active [29] as they are reliant on respondent recall and are often subject to social desirability bias [30], and thus should be interpreted with caution. Nevertheless, self-report questionnaires are frequently used to estimate PA levels in national populations for reasons of feasibility, cost-effectiveness, and to obtain information on specific types and domains of PA, which are not captured by objective measures. In comparison with accelerometry, the PASBAQ has demonstrated criterion validity similar to that of other questionnaires used in national surveys [15]. While there is no universally acceptable level for the magnitude of criterion validity coefficients for self-report PA questionnaires versus accelerometry, multiple reviews have indicated that correlations rarely exceed 0.40 [31‐33].

Smoking history was not used as an eligibility criterion to identify participants with COPD due to a lack of available data to quantify smoking history in pack-years, and to preserve sample size. While the high prevalence of never-smokers in our sample (27.9%) is surprising, it is supported by a recent case-finding study of 74,818 patients with COPD showing that 20–26% of patients were never-smokers [34]. Exposure to second-hand smoke, pollutants, and occupational exposure to dusts and chemicals are other risk factors for COPD [3], and may explain the high proportion of never-smokers in the sample. Furthermore, sensitivity analyses excluding never-smokers yielded similar results to the main analysis. It is therefore unlikely that the inclusion of never-smokers in the analysis would have biased the magnitude of the effect of PA on mortality risk.

Finally, given that the HSE and SHS are general population-based health examinations and are not specifically designed to investigate people with COPD, some potential confounders may not have been captured and this may limit the generalisability of the results. These include history of previous exacerbations, COPD hospitalisations, exercise capacity at baseline, and severity of symptoms, which are known predictors of mortality [35‐37]. The high proportion of mild and moderate COPD in the sample may also limit the generalisability of the results to people with COPD of greater severity.

We observed a linear dose-response association between PA and all-cause, CVD and respiratory mortality risk in people with COPD. There is added mortality benefit in adhering to the PA guidelines, although engaging in PA above a threshold of 3.75 MET-hours/week (at least 75 min/week of MPA, 32.5 min of VPA, or an equivalent combination of MVPA) is still associated with significant reductions in mortality risk. Walking and structured exercise should be encouraged in people with COPD, but the effects of domestic PA on mortality remain unclear. Our results reinforce the importance of promoting regular PA as part of routine management of COPD, and the need for effective strategies to increase PA in this population.