Effects of Early Pulmonary Rehabilitation on Hospitalized Patients with Acute Exacerbation of Chronic Obstructive Pulmonary Disease: A Systematic Review and Meta-Analysis

Introduction

Chronic obstructive pulmonary disease (COPD), a clinical syndrome characterized by chronic respiratory symptoms and irreversible airflow limitation, is the third leading cause of death worldwide, with 3.2 million deaths in 2017 and a worldwide prevalence of 10.1%.^1,2 Acute exacerbation (AE) is a serious complication of COPD and has many adverse effects, including worsening of the quality of life, accelerating disease progression, and increased hospital admissions and mortality.^3–7 AE is an important poor prognostic factor for COPD and is also the most common cause of admission in COPD patients.⁸ It was reported that severe AE resulted in three times higher mortality in COPD patients than in those without severe AE.⁷

Pulmonary rehabilitation (PR) is considered a cornerstone for COPD management. According to current guidelines and literature, PR provides additional benefits on respiratory symptoms, exercise ability, quality of life (QoL), readmission rate, acute exacerbation, and survival in stable COPD patients than the standard medical therapy.^9,10 The recent American Thoracic Society (ATS) and European Respiratory Society (ERS) guideline also suggests that PR should be initiated within three weeks following discharge from acute exacerbations of chronic obstructive pulmonary disease (AECOPD).¹¹ PR is a comprehensive intervention and is designed to improve the physical and psychological condition of patients with the chronic respiratory disease.¹² Exercise training is a key component of PR and includes endurance training and resistance training. Neuromuscular electrical stimulation is an alternative rehabilitation method for patients too weak to perform exercise.¹² Other components of PR include respiratory training, breathing retraining, patient education, smoking cessation, nutrition, and psychological support.^10,13

The previous two meta-analyses, conducted in 2016 by Puhan et al¹⁴ and in 2018 by Ryrsø et al,¹⁵ showed that early pulmonary rehabilitation (PR) had significant benefits for patients with AECOPD. However, there were discrepancies in the inclusion of studies and the outcome of mortality. In fact, some of the studies included were not ideal, including some non-COPD patients or non-AECOPD patients, or some studies included control groups that actually received late rehabilitation exercise. In addition, in the 2018 study by Ryrsø et al,¹⁵ the subgroup analysis was designed to compare early PR between patients receiving PR from admission to within one week of discharge and those receiving PR from 1 to 4 weeks after discharge, which I personally think is not an ideal way to compare. The recent ATS and ERS guideline also suggested early PR initiated during hospitalization for AECOPD showed increased patient’s exercise capacity; however, increased mortality was also reported.¹¹ Owing to the limited evidence at that time, further evaluation is required to clarify the ideal PR timing after AECOPD with hospitalization. Therefore, the purpose of this meta-analysis is to collect and update the literature in a more pure and objective way to compare the outcomes of AECOPD patients receiving early rehabilitation and those who receive no rehabilitation and to analyze the differences in outcomes between early rehabilitation during admission and after discharge.

Materials and Methods

Search Strategy and Eligibility Criteria

The whole study followed the rule of Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) statement.¹⁶ Systematic search was performed to retrieve relevant RCTs from the PubMed, Embase, and Cochrane library from their inception to November 2021, without language restriction. Searches were updated for any new research that also fulfilled the eligibility criteria until November 2021. Articles were also searched from the reference lists of relevant review articles. The keywords, which include both the text word and MeSH term, used for searching included COPD, exacerbation, PR, and physical therapy. The full search strategy is presented in Figure 1. The inclusion criteria were (1) patients admitted to the hospital owing to AECOPD (2) RCTs compared the effect of early PR during their hospitalization or within four weeks after discharge (intervention group) with standard medical therapy without PR (control group). (3) RCTs reporting at least one of the outcomes of mortality, hospital readmission, exercise capacity, dyspnea and health-related QoL. The PR must have main component of exercise training and may also contain respiratory exercise, education, smoking cessation, and nutritional support. Studies using neuromuscular electrical stimulation as an alternative PR intervention were also included. Studies with control group having any exercise training were excluded.

Figure 1 Flow diagram of inclusion and exclusion of RCTs for meta- analyses.

Study Selection

Two authors independently reviewed all articles for identifying qualified studies (HY Lu and CF Chen). Titles and abstract were screened for relevant studies. The search of RCTs was not limited to any language. Duplicate studies were removed from systematic review searches. If there was unclear information, concerned authors were contacted through email.

Data Extraction

Data were extracted and arranged in excel sheets, including author’s name, publish year, country, study design, patients’ characteristics, interventions and outcomes. All the data were independently checked by two authors (HY Lu and CF Chen). If there were disagreements of extract data, it would be discussed by a third author (PC Lin) to reach a consensus.

Outcome Measures

The primary outcome was readmission rate. The secondary outcomes were as follows: mortality, 6-minute walking distance (6MWD) for exercise capacity, Modified Medical Research Council (mMRC) scale and Borg Scale for dyspnea evaluation, and St. George’s Respiratory Questionnaire (SGRQ) for QoL score.

Risk of Bias in Individual Studies

Cochrane risk of bias (RoB) tool¹⁷ was used for assessing the quality of included studies. The RoB tool evaluates six domains, including selection bias, performance bias, detection bias, attrition bias, reporting bias, and other bias. The RoBs were independently judged by two authors (HY Lu and CF Chen). If there were disagreements on RoBs judgement, a third author (PC Lin) will join the discussion to reach a consensus. The reviewer’s assessment can be of a high risk of bias, some concerns of bias, and a low risk of bias. If a study has at least one domain judged as high risk of bias, or there are multiple domains judged as some concerns of bias, the overall result would be a high risk of bias. If all domains were judged as low risk of bias, the overall result would be a low risk of bias. If a study was judged to have some concerns or was unclear, the overall result of the risk of bias would be an unclear risk of bias.

Statistical Analyses

Review Manager 5.3 software (Cochrane Library Software, Oxford, UK) was used for the meta-analysis. Weighted mean difference (WMD) was used to pool outcome for continuous outcomes. For dichotomous data such as mortality and readmission rates, relative risk and 95% confidence interval (CI) were used to pool the effect size. Random-effect model was chosen due to the considerable difference of PR method of the included RCTs. Heterogeneity was tested using I² statistics:¹⁷ P < 0.1 or I²> 50% were considered as significant heterogeneity. Subgroup analysis was performed to compare the outcome difference between PR during admission and PR after discharge. Funnel plots were used to check for the presence of publication bias. Egger’s test was used to evaluate the significance of publication bias. Both of them were analyzed by The Comprehensive Meta-Analysis Version 3 software (Borenstein et al 2013).

Results

Study Selection

The flowchart of studies selection is shown in Figure 1. Until November 2021, a total of 1400 studies were searched from the electronic database, and 102 duplicates were removed. Non-RCT studies were excluded from the remaining 1298 studies, leaving 211 studies. We continued to select eligible studies based on the article’s title, abstract, and full text among these 211 studies. Among these, 10 studies included stable COPD patients, two studies included patients with other chronic respiratory diseases in additional to COPD, and 179 studies had pulmonary rehabilitation intervention 1 month after AECOPD, and therefore, these studies were excluded. Finally, 21 articles were included in the qualitative synthesis, and 20 were included in the quantitative synthesis. There were 1274 patients in total (Figure 1).

Characteristic of Included Studies

The summary of included 20 studies is shown in Tables 1 and S1. In the 20 RCTs included in the analysis, all patients were hospitalized because of AECOPD,^18–37 and 3 studies included patients with mean FEV1 predicted ≧ 50%.^34–36 Most studies included patients with mean/median age of ≧ 60 years (9 studies with mean/median age of 60–69 years, and 9 studies with age of ≧ 70 years), and only 2 studies included patients with mean age between 50 and 59 years.^22,35 Most of the articles were male predominant, and only two studies were female predominant,^21,29 and the other two studies did not report the gender of their patients.^22,34 In the included studies, the baseline characteristics between the intervention group and the control group were all similar. There were two kinds of early PR studies. Ten studies performed PR during hospitalization.^{18–21,23,24,28,31,33,37} Among these, 2 studies continued the PR program after discharge.^19,21 The other 10 studies initiated the PR program after discharge.^{22,25–27,29,30,32,34–36} The duration of PR program of enrolled studies ranged from 4 days to 3 months. The frequency of the PR program ranged from 2 times per week to 2 times per day. In the intervention group (PR group) of each study, training methods included endurance exercise, resistance exercise, or neuromuscular electrostimulation. The control group (usual care group) received standard hospital care without exercise training (Table 1).

Table 1 Characteristics of Randomized Controlled Trials Enrolled for Meta-Analysis

Risk of Bias Within Studies

In the 20 RCT studies, there was only one low risk of bias study,³² sixteen high risks of bias studies,^{18–23,25–31,33,36,37} and three unclear risks of bias studies^24,34,35 (Figure 2). Overall, there was high risk of biases in performance bias, detection biases, attrition bias, reporting bias and other bias. For selection bias, including random sequence generation and allocation concealment, there were some concerns and was determined as unclear risk of bias.

Figure 2 Summary of risk of bias.

Outcome Assessments

Primary Outcome: Readmission Rate

Hospital readmission rate was reported in 10 RCTs (n = 733). The readmission rate was 34.9% (127/364) in early PR group, and 52.6% (194/369) in usual care group. There was a significant improvement in the readmission rate in early PR when compared with usual care (random-effect, Risk ratio (RR) 0.68, 95% CI 0.50–0.92, I² = 56%). However, there was substantial heterogeneity (Figure 3).

Figure 3 Hospital readmission forest plot: early PR versus usual care.19,21,25–27,29,30,32,36,37

Secondary Outcomes

Compared with usual care, early PR was associated with a significant improvement of 6MWD (13 RCTs, n = 797, random-effect, Mean difference (MD) 59.73 meter, 95% confidence interval (CI) 36.34–83.12; I² = 88%; Figure 4), a better SGRQ total score (9 RCTs, n = 639, random-effect, MD −10.65; 95% CI −14.78 to −6.52; I² = 51%; Figure 5), a lower Borg score (8 RCTs, n = 390, random-effect, MD −0.79, 95% CI: −1.26 to −0.32; I² = 56%; Figure 6), a lower mMRC score (8 RCTs, n = 568, random-effect, MD −0.38, 95% CI −0.50 to −0.25; I² = 0%; Figure 7), and a similar mortality (6 RCTs, n = 521, random-effect, RR 0.72, 95% CI 0.39–1.34; I² = 0%; Figure 8). Six of the 20 RCTs evaluated adverse outcomes,^{18,21,23,29,32,37} and all reported no adverse event during the study period.

Figure 4 6MWD forest plot: early PR versus usual care.18,20–28,34,35,37

Figure 5 SGRQ total score forest plot: early PR versus usual care.20,25–27,29,30,34–36

Figure 6 Borg score forest plot: early PR versus usual care.19,21,24,25,28,30,31,33

Figure 7 mMRC forest plot: early PR versus usual care.21,23,25–27,30,34,35

Figure 8 Mortality forest plot: early PR versus usual care.19,25–27,29,32

Subgroup Analysis

A subgroup analysis was conducted for comparison of PR during admission, PR after discharge, and both (mixed inpatient and outpatient PR) (Table 2 and Figures S1–6), and showed no significant difference in mortality (RR 0.14 vs 0.77, P = 0.54), readmission rate (RR 1.26 vs 0.65, P = 0.2), 6MWD (MD 81.82 vs 39.42 meters, P = 0.18), SGRQ score change (MD −2.8 vs −11.12, P = 0.32), Borg score change (MD −0.87 vs −0.30, P = 0.48), and mMRC change (MD −0.7 vs −0.35, P = 0.26). However, heterogeneity did not reduce in most subgroups (except for a partial reduction in heterogeneity in the Borg score).

Table 2 Summary of Subgroup Analysis

Publication Bias

No publication bias was observed by visual examination of funnel plots and tests using Egger’s test (P = 0.37) (Figure 9).

Figure 9 Funnel plot for evaluation of publication bias.

Discussion

The result of this meta-analysis showed that early PR during admission or within four weeks after discharge for severe AECOPD had statistically significant beneficial effect on readmission rate, 6MWD, SGRQ, Borg score, and mMRC than usual care without PR. However, the trend of mortality benefit was not significant. Significant heterogeneity was found in readmission rate, 6MWD, SGRQ and Borg score. Heterogenicity was low in mortality and mMRC. When considering the minimal clinically important difference (MCID), both 6MWD and SGRQ had achieved clinically significant benefit (MCID for 6MWD = 26 meters, SGRQ = 4 units, Borg score = 1 unit).³⁸ In the subgroup analysis, there was no significant outcome difference between PR during admission and PR after discharge. Adverse event after PR was either reported as none or not been reported in enrolled PR studies.

Recent meta-analysis on 2018 by Ryrsø et al concluded that early PR after AECOPD significantly improved mortality, hospital duration, readmission, SGRQ, and 6MWD.¹⁵ However, in this meta-analysis, a study evaluating COPD patients without AE within four weeks was included,³⁹ and two studies comparing early PR and late PR were also included (not a true control group).^40,41 Another previous Cochrane systematic review on 2016 by Puhan et al also showed that PR after AECOPD significantly improved QoL, exercise capacity, and readmission rate, but the mortality benefit was not significant.¹⁴ However, in their meta-analysis some studies with control groups performing exercise training (not true control group) were also included.^42,43 In contrast, our meta-analysis made a more focused comparison, with all hospitalized AECOPD patients, and all control groups received standard medical therapy without exercise training. In our study, five additional studies were included that were not analyzed by previous meta-analyses.^{18,22,27,31,32} The comparison of differences in the enrolled studies between the above two meta-analyses^14,15 and our study is presented in Table S2.

Previous ATS/ERS guidelines on 2017 concluded that early PR during hospitalization increases mortality but according to limited evidence.¹¹ This recommendation was based on a study published in 2014 by Greening, which showed early PR within 48 h of admission had higher 12-month mortality than the usual care group.⁴² However, in that study, the usual care group also received outpatient PR three months after discharge, and therefore, was not a true control group. Besides, the study participants included patients with COPD (approximately 80%), asthma, interstitial lung disease, and bronchiectasis. The author also reported a significantly lower rate of outpatient PR (three months after discharge) in the early PR group than in the usual care group.⁴² However, in our meta-analysis, the subgroup analysis of the mortality outcome showed that only one study included both inpatient PR and outpatient PR, while all the others included only outpatient PR. Therefore, there is still inadequate evidence about the effectiveness of inpatient PR on mortality, and it is not possible to draw a conclusion currently.

The issue that whether early PR should be initiated during admission of AECOPD is still controversial. In the meta-analysis on 2018 by Ryrsø et al, subgroup analysis was also done to compare different timing of PR (during admission or within 1 week after discharge vs between 1 and 4 weeks after discharge).¹⁵ And their result showed no significant difference in mortality, readmission rate, SGRQ, and 6MWD between the 2 subgroups.¹⁵ However, we suggest this subgroup analysis was not ideal, because PR during admission and those PR within 1 week after discharge were combined in a same group. Therefore, it cannot clearly answer the problem of whether early PR should be initiated during admission or not. In our study, the subgroup analysis compared early PR during admission vs those PR within 4 weeks after discharge. All the enrolled studies had true control groups (without any exercise training). However, our subgroup analysis result still showed no significant outcome difference between 2 groups.

There are some limitations to this study. The first is clinical heterogeneity. There was considerable variance in exercise training programs between each PR study (Table 1), which includes endurance training (walking and cycling), resistance training (upper and lower limbs), or both. In one study, neuromuscular electrostimulation was used instead of exercise training. Despite the difference in exercise training programs, the goal of exercise training was similar. In addition, the start timing of PR (from second admission day to within 2~3 weeks after discharge), frequency of PR (from 2 times/day to 2 times/week), and duration of PR were also variable (from four days to six months). These variations should be considered carefully when interpreting the results. Second, blinding of participants and personnel was not possible in most studies, and the performance bias and detection bias were inevitable. Third, an insufficient sample size-related imprecision was observed in the mortality outcome. Beside, statistic heterogeneity was observed in the readmission, 6MWD, SGRO total score, Borg score outcomes in this study.

Conclusion

In conclusion, this meta-analysis showed that early PR for AECOPD with hospitalization had a significant beneficial effect on hospital readmission rate, exercise capacity, QoL, and dyspnea. The benefit of early PR on mortality is still controversial. The subgroup analysis showed that the beneficial effects are similar between PR initiated during admission and those initiated within four weeks after discharge. However, heterogeneity was present, and more studies with larger sample size are still needed to verify this conclusion. Early PR timing should also be decided according to the stability of AECOPD patient and the PR program facility of the hospital.