Acute exacerbations of chronic obstructive pulmonary disease (COPD) represent a major burden for patients and health care systems. For patients, acute exacerbations are a common reason for hospital admissions and severely affect health-related quality of life (HRQL) [1] and prognosis[2]. Mortality rates during hospitalisations are around 10% [3, 4] and during the year following a hospitalisation may be as high as 40% [3, 5].

From the health care provider's perspective, COPD is resource consuming [6]. A small proportion of COPD patients of around 10% suffering from acute exacerbations accounts for over 70 percent of costs caused by COPD because of emergency visits and hospitalisations [6–8]. The readmission rate is typically high in these high-risk patients. A recent large study found a readmission rate of 63% during a mean follow-up of 1.1 year with physical inactivity amongst the significant predictors for readmissions[9].

Recent position papers of the American College of Physicians and American College of Chest Physicians provided recommendations on the management of acute exacerbations [10, 11]. However, a weakness of these papers was that they did not provide recommendations how future exacerbations and hospitalisations could be prevented despite being one of the main goals of COPD management [11, 12]. One solution that has been adopted in clinical practice is to provide rehabilitative care after treatment of acute exacerbation including physical exercise, patient education focusing on self-management strategies and psychosocial support. The rationale to offer rehabilitation in patients recently treated for acute exacerbation is to enhance HRQL as in stable COPD patients [13], but also to modify factors associated with increased risk for post-exacerbation morbidity and mortality. A recent study showed that exacerbations results in acute muscle deconditioning and weakness[14]. Hence patients with frequent exacerbations have more pronounced skeletal muscle weakness and a more limited six minute walking distance [15], which is in turn a risk factor for exacerbations and mortality [3, 16].

Thus respiratory rehabilitation may have the potential to reduce hospital admissions by improving exercise capacity. It is hence surprising, and in contrast to the large body of evidence supporting respiratory rehabilitation in stable patients [13, 17], that the effects of respiratory rehabilitation in patients after acute exacerbation has never been studied systematically. Therefore, our aim was to conduct a systematic review of all randomized controlled trials that compared respiratory rehabilitation after acute exacerbation and usual care.

Initial agreement of reviewers on quality assessment was 85% for all items (chance corrected kappa = 0.70). All disagreement could be resolved by consensus. The quality of trials was moderate (Table 2). Three trials provided details about the randomisation procedures and three trials about concealment of random allocation, while in none of the trials the outcome assessors were blinded.

Effect on hospital admissions

Figure 2 shows the effect of respiratory rehabilitation on unplanned hospital admissions for each study [20, 21, 24] and the pooled relative risk ratio of 0.26 (0.12–0.54). The other trials included either only inpatients [22, 23] or did not record hospital admissions during the follow-up [25].

Effect on HRQL

Three trials assessed HRQL using the Chronic Respiratory Questionnaire (CRQ) [20, 24] and the St Georges Respiratory Questionnaire (SGRQ) [21, 24] (Figure 3). With both instruments, the trials found large effects exceeding the minimal important difference of 0.5 on the CRQ and of 4 on the SGRQ. Weighted mean differences (expressed as points change on a scale from 1 to 7) on the CRQ were 1.37 (95% CI 1.13–1.61) for the fatigue domain, 1.36 (0.94–1.77) for emotional function and 1.88 (1.67–2.09) for mastery. Weighted mean differences for the SGRQ total score, impacts and activities domains were -11.1 (95% CI -17.1 to -5.2), -17.1 (95% CI -23.6 to -10.7) and -9.9 (95% CI -18.0 to -1.7). For the CRQ dyspnea and SGRQ symptoms domain, results were too heterogeneous to be pooled (Q = 6.44, p = 0.01 for CRQ dyspnea domain and Q = 3.50, p = 0.06 for SGRQ symptoms domain), but all studies showed a consistent effect in favor of the rehabilitation intervention.

Man and Murphy also used generic HRQL instruments and found improvements by respiratory rehabilitation of 10.6 (-0.3 to 21.6) and 20.1 (3.3 to 36.8) on the physical composite and mental composite score of the Short-Form Survey 36 [24] and of 0.18 (95% CI 0.04 to 0.32) with the EuroQol score [21].

Effect on exercise capacity

All trials showed a significant benefit of respiratory rehabilitation on the six-minute walking distance (Figure 4). We did not pool the results of the six-minute walking tests because of statistically significant heterogeneity (Q = 28.33, p < 0.001), which could not be explained by our a priori defined sources for heterogeneity. The trials reported by Behnke [19] and Kirsten[22] were conducted in the same institution and showed much larger effects (mean effects of 215 and 158 meters on the six minute walking test) compared to the trials of Nava [23] (68 meters) and Troosters [25] (64 meters). All studies showed a consistent benefit in favor of the rehabilitation group, which exceeded the minimal clinically important difference of 53 meters. The meta-analysis of the shuttle walk tests results showed a weighted mean difference of 81 meters (95% CI 48 to 115) between the rehabilitation and usual care groups.

Effect on mortality

The individual study relative risks for mortality ranged from 0.40 (0.18–0.86) to 1.00 (0.07–15.04, Figure 5). The pooled risk ratio was 0.45 (0.22–0.91). Although no significant heterogeneity was present, it should be noted that the length of follow-up differed substantially between these studies. We did not include one trial [23] in the primary meta-analysis because severity of disease of included patients differed considerably from those of the other studies. For this trial a mortality of 20% for patients of either group (12/60 in rehabilitation group and 4/20 in control group) was observed while staying in the respiratory intensive care unit with a mean survival of 18.1 days (SD 7.2) for patients with and 12.4 days (SD 11.1) for patients without rehabilitation (p > 0.05). Of the 12 patients of the rehabilitation group who died, only five started a walking training (stage 2, Table 1). If this trial is included in the meta-analysis the pooled risk ratio is 0.59 (0.34–1.05) favoring the rehabilitation group.

The meta-analyses showed that respiratory rehabilitation after acute exacerbation of COPD reduced the risk for hospital admissions and mortality and led to large improvements of HRQL and exercise capacity.

Strengths of this systematic review include the extensive literature search, rigorous adherence to a predefined protocol and contacts to authors of the included trials who all provided additional information about their data. A limitation is the small number of patients included in the trials and methodological shortcomings that limit conclusions.

The effect of respiratory rehabilitation after acute exacerbation appears to be large. For HRQL and exercise capacity, the effects were well above the threshold for the minimal important difference for the CRQ (0.5 point difference [27]), St. Georges Respiratory Questionnaire (4 points [28]), SF-36 (5 points[29]) and Six-minute walking distance (around 53 meters [30]). In addition, the number of unplanned hospital admissions and mortality was reduced substantially. When one assumes that respiratory rehabilitation improves activity level in patients with COPD, it seems plausible that rehabilitation reduces readmission rate as inactivity has been shown to be a predictor of readmissions[9].

Compared to respiratory rehabilitation in stable COPD patients [13], its effects tend to be even larger after acute exacerbation. Several factors may contribute to this. First, as mentioned above, exacerbations lead to significant reductions in muscle function[14] and quality of life [1]. This initial deterioration may render patients more likely to improve from respiratory rehabilitation. Second, since patients were hospitalized, there may be a deficiency in self-management, or education. This may be partially targeted with the rehabilitation intervention, and patient education, as an additional part of multidisciplinary rehabilitation programs, may be of particular benefit to modify behavior. Indeed, a recent study showed impressive results of a patient management program including home exercises for COPD patients after acute exacerbation [31]. The mean number of hospital admissions per patient was reduced from 1.6 to 0.9 in the year following a hospital admission due to acute exacerbation. It is well known from earlier studies that the recovery period is long even in patients who have no further exacerbations and that another exacerbation within 6 months limits recovery markedly [32]. Our meta-analyses showed that respiratory rehabilitation during the recovery period is superior compared with usual care to improve prognosis and HRQL.

A word of caution is needed when interpreting the current analysis. A clear limitation of the trials is their relatively small sample size. All trials, in particular the trials reported by Behnke [20] and Kirsten[22] showed large effects of respiratory rehabilitation on HRQL and exercise capacity. Small trials tend to overestimate the effect of an intervention compared to large trials [33–36]. This phenomenon can partly be attributed to a publication bias, that is, the fact that small trials are more likely to be published if they show statistically significant treatment effects [37]. On the other hand, methodological shortcomings of small trials such as inadequate generation of the randomisation code, insufficient concealment of random allocation and lack of blinding contribute to discrepancies between the results of single large trials and pooled estimates based on small trials[35]. In our systematic review, the trials had methodological limitations and it cannot be excluded that the estimates provided by the meta-analyses represent overestimations of the effect of respiratory rehabilitation after acute exacerbation.

Larger trials seem justified to challenge the data presented in this article. Such trials should assess the effect of respiratory rehabilitation on unplanned out- and inpatient care but also include data on patient-important outcomes such as HRQL. Conducting trials on respiratory rehabilitation after acute exacerbation is, however, challenging. First, recruitment of patients is difficult because not all of them may want to be randomly allocated to respiratory rehabilitation or usual care in a situation of poor health status. Second, one needs to take into consideration that exercise capacity is particularly low after acute exacerbations[14] so that the exercise program should be designed carefully. Strength exercise and tolerable whole body exercise modalities such as interval exercise may be particularly suitable for these patients [38, 39]. Third, the definition of usual care raises a number of difficulties. Patients willing to participate in the trial are likely to have a preference for respiratory rehabilitation. If they are randomized to the control group, they might ask for respiratory rehabilitation at any time during the follow-up. Given the clear benefits of this intervention in stable patients, confirmed in meta-analyses [13], patients should not be refrained from rehabilitative strategies. It would perhaps be ethically justifiable to conduct a large rehabilitation trial in places where respiratory rehabilitation is currently not readily available to the general patient. This appears to be the case in many countries including Switzerland [40], the UK [41] and Canada [42]. These countries are just few examples of countries where the lack of access to rehabilitation has been pointed out as an important caveat in health care. In these places patients could be randomized to additional respiratory rehabilitation or standard treatment by general practitioners and respirologists because respiratory rehabilitation can be offered to a small proportion of COPD patients only. Alternatively relatively short term studies (3–6 months follow-up) could be conducted with re-admission as a primary end point. It has been shown that re-admission occurs often soon after discharge [43, 44]. Obviously, such studies could never address mortality as a primary end point, due to a lack of events. Whatever design investigators choose, a careful discussion of ethical and methodological issues is necessary before conducting large trials.

The present data show that respiratory rehabilitation has the potential to reduce the large COPD-related costs due to hospital admissions. It may not only reduce the number of acute exacerbations but also their severity. Patients may learn to notice imminent exacerbations and seek medical attention earlier leading to a shift from inpatient to the less costly outpatient treatment of acute exacerbations. The significant reduction in hospital readmissions is suggestive of a beneficial cost-benefit balance. However, larger trials should provide the final evidence base for formal cost analyses to test the hypothesis that respiratory rehabilitation after acute exacerbation is cost effective.

The data presented in this review are the first to show a survival benefit of respiratory rehabilitation in patients at risk. Although the results should be interpreted with caution, as mentioned above, this study provides the most solid evidence currently available that mortality is reduced. In summary, current evidence suggests that respiratory rehabilitation reduces unplanned hospital admissions and mortality and improves HRQL and exercise capacity when initiated immediately after acute exacerbations.