Short and long-term effects of pulmonary rehabilitation in interstitial lung diseases: a randomised controlled trial

Patients with chronic ILD referred to the ILD clinic in the University Hospital of Leuven were considered for enrolment. According to eligibility criteria, potential candidates were informed about the protocol and written informed consent was obtained, between March 2009 and September 2011. The study was approved by the local ethics committee (B32220095560) of this hospital.

Inclusion criteria were a diagnosis of ILD according to internationally established criteria with a formal workup and multidisciplinary discussion [1, 2], dyspnoea on exertion and having a stable medical therapy with no infection/exacerbation in the previous 4 weeks.

Exclusion criteria were comorbidities (unstable angina, recent myocardial infarction or cerebrovascular accident, active cancer, severe orthopedic disorders) or systemic manifestations (active myopathy, arthralgia, synovitis) that do not allow training and a life expectancy below 3 months.

Patients were randomly assigned to the rehabilitation or control group using sealed envelopes prepared and shuffled before the start of the study by an independent person unrelated to the study protocol. The envelope was opened by the allocator sequentially, only after the participant’s name was written on it.

Patients assigned to the rehabilitation group were invited to attend a 6-month outpatient rehabilitation program with a total of 60 sessions, 3 times per week for the first 3 months and thereafter twice weekly. The program was performed in accordance with the guidelines [16] and is detailed in the Additional file 1 (online data supplement). Patients assigned to the control group were treated with maximal medical care and with an identical medical follow-up, similarly to the active intervention arm.

All measurements were made at enrolment and at 3, 6 and 12 months after the start of the study.

The primary outcome of the study, functional exercise capacity, was measured as the distance walked during the best of two 6-minute walking tests (6MWD) [17] and also expressed as percentage of the predicted values [18]. Use of oxygen was standardized to a minimal flow rate of 2 l-per-minute and subsequent follow-up tests were conducted in the same conditions as at baseline. The difference in 6MWD between treatment and control group at 6 months was considered as the primary endpoint.

The secondary outcomes were forced vital capacity (FVC); slow vital capacity (SVC); diffusing capacity for carbon monoxide (DL_CO); arterial partial pressure of oxygen (PaO₂); maximal work rate (W_max); 6MWD; quadriceps (QF) and hand grip (HF) muscle force; quality of life evaluated by the the St George’s respiratory (SGRQ) and the chronic respiratory disease questionnaires (CRQ) and dyspnoea (MRC 1 to 5). More details on the assessments are provided in the Additional file 1 (online data supplement).

Patients were provided with an activity monitor SenseWear Armband (Bodymedia, Pittsburgh, USA), in order to assess objectively their physical activity. They were instructed to wear the monitor for at least 7 consecutive days, from waking up until going to bed. Days with less than 8 h of wearing time during waking (defined as 7 AM-8 PM) were excluded. Only week days were used to minimize the variability. A valid measurement was defined as having at least 2 valid weekdays. The influence of the duration of the daylight time was included as a covariate [19].

Mean step count and mean time in at least moderate intense physical activity (MPA) were chosen as the physical activity outcomes. MPA was defined as any activity with an energy expenditure above 3 METs [20]. The SenseWear has been validated in COPD patients [21], a respiratory population with a comparable inactive lifestyle.

Physical activity at 1 year was not included in the analysis due to the large number of missing data at this time point, caused by incompliance of patients and technical problems. For this reason the data were not judged as being representative of the entire sample and therefore not included in the analyses.

All adverse events during the rehabilitation program were registered.

All variables were expressed as means ± standard deviation (SD) or 95% confidence interval (95% CI). A minimal 6MWD benefit of 38 ± 43 m might be expected after 3 months [10]. To anticipate a similar difference of 40 ± 45 m, a sample size of 20 patients in each arm is needed to show a statistically significant difference at the 0.05 p-level with 80% power. Anticipating on 40% of dropouts, a total number of 60 patients were randomized. Statistical analysis was performed using mixed models including the outcomes as measured in the 4 visits (baseline, after 3, 6 and 12 months). ‘Visit’ was included as a class variable. P values below 0.05 were considered as significant in all analyses (R version 3.2.1) [22]. Group, time and group*time interactions were retrieved. This interaction term assesses the pure effect of rehabilitation since it is the difference, at each time-point, between changes from baseline in the rehabilitation group and in the control group. Duration of daylight was including as a (time varying) covariate when investigating the physical activity outcomes [19]. The number needed to treat (NNT) to achieve an increase in functional exercise capacity was defined as a clinically significant improvement (> 30 m) in the 6MWD.

Figure 1 shows the Consort flow diagram. From 271 patients referred to the ILD clinic, 209 were eligible for inclusion; 60 (22%; 37 M/23F) were randomized.

Sixty-seven percent of patients completed the rehabilitation program and 60% of the randomized patients were followed up at 1 year.

Table 1 shows the baseline characteristics of the participants. At baseline, 6MWD (expressed as percentage of the predicted value) was slightly lower (p = 0.045) and hypersensitivity pneumonitis was more represented in the rehabilitation group. The groups were matched for other parameters (Table 1, Additional file 1: Table S1). Patients who dropped out did not differ from those who completed the study in the rehabilitation group; in the control group, drop-outs appeared to be younger and to have worse diffusing capacity (Additional file 1: Table S2).

At 6 months, the 6MWD of the intervention group was mean, 95% CI [ll to ul] + 72, [36 to 108]m better as compared to the usual care group (Table 2, Fig. 2). The benefit was largely maintained between groups at 1-year follow-up (73, [28 to 118] m). Taking a conservative intention to treat approach where none of the dropouts in either group is counted as a responder, 11 patients (37% of the initially selected 30 patients) had a clinically significant improvement in their 6MWD at the 1-year time point, whereas only 5 (17%) of the controls had such an improvement. This results in a NNT of 5.

Similarly, the intervention yielded significant increases in maximal exercise capacity (W_max), health status (in all the domains) and quadriceps force at the end of PR (Table 2, Fig. 2, Additional file 1: Tables S3-S4), as compared to usual care. Most of these benefits were maintained at 1 year. A significant improvement was already shown after 3 months of PR for exercise tolerance, muscle force, total scores of health status, SGRQ activity domain and CRQ dyspnoea, emotion and mastery (Table 2, Additional file 1: Tables S3-S4).

The rehabilitation program did not change MRC dyspnoea and physical activity significantly as compared to the control group at any time point (Table 3, Fig. 2, Additional file 1: Table S4).

PR did not influence lung function and arterial blood gases with the exception of forced vital capacity which was slightly better in the rehabilitation group as compared to usual care (Table 2, Additional file 1: Table S5).

The patients participated in a mean of 50 ± 13 sessions. Aerobic training intensity for cycling started at 65 ± 8% of the initial W_max. Walking was initiated at 72 ± 8% of the baseline mean 6MWD speed. These intensities were progressed to 95 ± 17% and 106 ± 15%, respectively. The resistance training load also increased significantly over time (Fig. 3).

No adverse events related to the exercise training were recorded during the PR program.

To our knowledge, there are only 5 randomized controlled trials [9‐13] and a meta-analysis [14] demonstrating improvements in exercise performance and symptoms after exercise training in ILD. Only one non-randomised study showed a benefit on muscle force after PR [28]. Overall, the magnitude of the effect in our trial was more pronounced than in previous shorter studies [10, 11, 13, 28‐33].

Unfortunately, the large benefits after PR in ILD were not translated in objective gains in PA. This is in line with research in COPD [34]: some trials showed an increase in PA after PR and others failed to do so. In the negative studies, the lack of improvement was also observed despite gains in physiological function and health status. This is the first trial measuring the PR-effect on PA in ILD objectively. Previously, 2 studies [6, 35] have evaluated IPF patients with PA questionnaires and have suggested small benefits. However, questionnaires may not provide a reliable estimation of true PA behaviour. In our study, no specific interventions were introduced to coach patients towards more PA. Future studies could consider for instance the provision of feedback on PA using step-counters providing promising results in COPD [36].

The high number of attended sessions and the high exercise intensity reached during the training sessions indicate that PR is feasible to be implemented in ILD in a similar way as in COPD [8]. In the present study, ILD patients trained alongside patients with COPD and other chronic respiratory conditions. However, it remains a point of attention that many patients refuse the offer for rehabilitation and 33% dropped out during the program although the reasons were not related to the provided intervention.

All ILD patients used oxygen supplements per protocol in order to avoid desaturation [37] and all the exercise tests were performed with oxygen in order to better show the improvements related to the training [38].

The novel finding of this trial is the maintenance of the benefits in exercise capacity, health status and muscle force at 1 year in a subgroup of patients. So far, 5 studies [10, 13, 15, 28, 39] have evaluated long-term effects. Only 1 randomized trial [15] did find a discrete gain in quality of life at 1 year and one non-randomized study [39] showed an improvement on 6MWD and health status at 6 months.

The larger and long-term benefits in our study could potentially be explained by the differences in the rehabilitation program, the differences in disease aetiology and disease severity.

First, our regimen proposed a longer training than reported previously [9‐13, 28‐30, 32, 33, 35, 39]. In COPD, longer programs showed to have more effects than shorter programs [8, 40]. In addition, in line with recommendations for PR [16], a multidisciplinary approach consisting of occupational therapy, psychosocial and nutritional support was offered in our program.

Second, the lower proportion of IPF patients (23% in our trial compared to 100% [11‐13] or around 50% of the patients [9, 10] in the other 5 randomised trials) could explain the more pronounced treatment effects. Indeed, the magnitude of improvement following PR in IPF was less than other ILD [10, 41, 42] or COPD [28]. However, a mixed population reflects more the clinical reality of most of ILD units.

Third, greater PR-benefits were described in patients with less functional impairment [9, 10, 28]. Our patients had better baseline 6MWD and muscle force [10, 11, 13, 28‐33], whereas, the impairment in health status was comparable [43]. In fact, one trial with a similar baseline 6MWD did find similar PR-gains [12]. However, this remains controversial since other studies found greater improvements in the most disabled patients [29, 39].

Finally, it is important to note that, among the immunosuppressive medications proposed in the majority of non-IPF [2] and the recent recommended antifibrotic agents in IPF [3], none of them did increase exercise capacity, only pirfenidone has demonstrated to reduce the decline in 6MWD [44‐47]; and nintedanib [48, 49] showed a discrete, not clinically significant, improvement in quality of life in 1 of the 2 phase three trials [49]. The fact that PR showed a large and maintained benefit on health status and exercise tolerance and without any clinically meaningful side effects is of utmost importance to ILD patients.

In contrast to previous trials, our study did not show a lung function decline over time in the whole group [9, 10]. This finding might be due to our larger amount of non-IPF patients.

Curiously, our study showed an improvement over time of muscle force in the entire group. Steroid therapy contributes to muscle dysfunction in ILD [50], a treatment often proposed to non-IPF patients [1, 2, 51]. In our study, 47% were under prednisolone at baseline or during the preceding 6 months (Table 2) and only 28% at 1 year. We suggest that muscle dysfunction of our ILD patients might be partly explained by the steroid therapy and the improvement over time by the gradual steroid tapering.

First, an a priori expected limitation of the study was the dropout rate (33%) over the course of the 6-month PR program. This number was slightly higher than observed in ILD following some shorter programs [10‐13, 30, 33, 39], but similar as the Australian trial [9] and long programs in COPD [8, 40]. However, since the patients who dropped out from the rehabilitation group were very similar to those who completed the PR program, this has likely had minimal impact on the main results of the study. In addition, it is important to see that long-term benefits were obtained in a significant fraction of patients in the rehabilitation group resulting in a very acceptable NNT of 5 for the outcome 6MWD. Finally, a treatment of missing data penalizing the PR-effect did not affect the significance at any time-point.

Second, assessors were not blinded for 6MWD (main outcome) which may have led to an overestimation of the benefit, however this probably will not have affected the results of quality of life.

Third, only 20% from the eligible candidates were randomized in the present study (Fig. 1). This may compromise the external validity of the study. Whereas it shows the general difficulty of performing randomised controlled trials in rare diseases like ILD and patients may have considerable travel distances to attend our centre of reference. This has likely not impacted on the observed differences between groups.