Rehabilitation Interventions for Physical Capacity and Quality of Life in Adults With Post–COVID-19 Condition

1The World Health Organization defines post–COVID-19 condition (PCC) as, “the continuation or development of new symptoms 3 months after the initial SARS-CoV-2 infection, with these symptoms lasting for at least 2 months with no other explanation.” Although COVID-19 was initially recognized as a respiratory illness, PCC symptoms can range from mild impairment to severe systemic disease, with many patients experiencing dozens of symptoms across multiple organ systems.,, At least 65 million individuals worldwide are estimated to have PCC, but the number is likely much higher due to many undocumented cases. The incidence is estimated at 10% to 30% of nonhospitalized cases, 50% to 70% of hospitalized cases, and 10% to 12% of vaccinated cases. In Canada, 14.8% of those who reported a previous positive test or a suspected infection of SARS-CoV-2 experienced symptoms at least 3 months after their infection, which translates into approximately 1.4 million Canadian adults or 4.6% of the Canadian population aged 18 years and older. Similarly, in the US, as of February 23, 2023, 14% of adults with a previous positive test for COVID-19 reported having experienced PCC symptoms at some point, and 6.5% reported current symptoms.

2The 5 most frequently observed symptoms include fatigue (58%), headache (44%), attention disorder (27%), hair loss (25%), and dyspnea (24%). Emerging evidence suggests that female sex, socioeconomic determinants of health (eg, lower income), inability to adequately rest in the early weeks after developing COVID-19, and comorbid conditions appear to be independent factors associated with increased risk for the development of PCC.,

3According to a 2023 review, the existing research is insufficient to improve outcomes for people with PCC. There is an urgent need for evidence-based rehabilitation interventions to support people affected by PCC,,, because current guidelines are primarily based on expert opinion and observational data., The latest systematic review with meta-analysis on rehabilitation interventions for patients with PCC included a total of 3 trials,, (233 patients with PCC) and suggested that rehabilitation interventions may be associated with improvements in functional exercise capacity, but the association of rehabilitation interventions with respiratory function was inconsistent across studies. The accuracy of these early-evidence syntheses is likely to be compromised due to the small number of included studies. Since then, 11 additional trials,,,,,,,,,,, have emerged. The increased attention on PCC now allows us to conduct a more comprehensive, methodologically sound, and stable analysis. The purpose of this meta-analysis is to assess whether rehabilitation interventions are associated with improvements in physical capacity (functional exercise capacity, muscle function, dyspnea, and respiratory function) and quality of life in adults with PCC.

4Our protocol was registered in PROSPERO database and is available in the eAppendix in Supplement 1. We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) reporting guideline to report this systematic review.

5After consulting with a librarian, a systematic electronic search of the literature was performed from January 2020 until February 2023, in MEDLINE (via Ovid), Scopus, CINAHL, and the Clinical Trials Registry. Examples of the keywords that were used to identify potentially relevant studies were long-covid, post-covid, sequelae, exercise therapy, rehabilitation, physical activity, physical therapy, and randomized controlled trials. The full research strategy is summarized in eTable 1 in Supplement 1. Additionally, we conducted a manual search of the reference lists of the included studies to identify any additional studies not retrieved in the electronic search.

6We included randomized clinical trials that compared rehabilitation interventions such as respiratory training aerobic exercises and resistance exercises with either placebo, usual care, waitlist, or control in adults with PCC. We did not pose any restrictions with regard to comorbidities or medication use concurrently with the rehabilitation protocol. Trials that used medication-only treatments without a rehabilitation component and nonrandomized studies were excluded.

7As part of the study selection, 2 independent researchers (D.V.P. and E.S.) screened titles and abstracts of articles, reviewed the articles that met inclusion criteria in initial screening at the full text level, and extracted data from the eligible studies. We extracted authors, year of publication, participant and intervention characteristics, and outcome data.

8The primary outcome was functional exercise capacity, measured with the 6-minute walking test. Our secondary outcomes were fatigue; functional leg strength and endurance as measured by the 30-second sit-to-stand test; dyspnea; respiratory function; and quality of life. The time point of the primary and secondary outcomes was the shortest time point available upon completion of the rehabilitation program. Respiratory function was assessed through forced expiratory volume in the first second (FEV1) and forced vital capacity (FVC). If a study used more than 1 measure to assess dyspnea and quality of life, we planned to extract the measure reported as the primary outcome. We also assessed the safety of the rehabilitation interventions by extracting any treatment-emergent adverse events (including serious adverse events) across all studies. Safety outcomes were assessed at the last follow-up available.

9To assess the risk of bias, 2 independent reviewers (D.V.P. and E.S.) used the Cochrane risk of bias tool version 2. We assessed selection bias, performance bias, detection bias, attrition bias, and reporting bias. We used the GRADE approach,,,,,, to evaluate the certainty of evidence with regard to risk of bias, inconsistency, indirectness, imprecision, and publication bias. Discrepancies were resolved with a senior research team member (P.B.).

10We described variables with a normal distribution as mean (SD) and those with a skewed distribution as median (IQR). Categorical variables were presented as numbers (percentages).

11The different rehabilitation interventions were compared against a common comparison group, termed control (ie, placebo, sham, waiting list, or usual care). We used bayesian pairwise, random-effects, meta-analysis models to summarize results across trials. We chose random-effects models to account for the anticipated clinical and methodological diversity between studies.

12For continuous outcomes, we used the normal likelihood and an identity link. For binary outcomes, we used the binomial likelihood and the logit link. All models were fitted with noninformative prior distributions. We used normal prior distributions with a mean of 0 and a large variance for treatment effects. For the main analysis, we used uniform, vague prior distributions for the between-trial variance. In sensitivity analyses, we used empirical prior distributions for the between-trial variance. We quantified the between-trial heterogeneity using the between-study variance, τ2, and 95% credible intervals (95% CrIs). In the presence of substantial heterogeneity, we performed sensitivity analysis on the basis of allocation concealment that was defined a priori as potential moderator.

13For continuous outcomes, summary results were presented as standardized mean differences (SMDs) with 95% CrIs. We coined all outcomes such that negative treatment effects (SMD <0) indicate better outcomes in the intervention group compared with the comparison. We calculated the probability of rehabilitation interventions being superior to usual care (ie, probability of the true SMD <0) and the probability of rehabilitation interventions providing a treatment effect more pronounced than the between-group minimum important difference (MID) of 0.30 SD units (ie, the probability that the true SMD is ≤−0.30).

14For binary outcomes, summary results were presented as odds ratios (ORs) with 95% CrIs, with an OR greater than 1 representing a higher risk of the event among patients who received the intervention than those who received the comparison.

15Parameters were estimated via Markov chain Monte Carlo methods. We used a burn-in period of 50 000 iterations and 3 chains with 100 000 simulations each (300 000 in total). Summary estimates were obtained via posterior medians (the 2.5 percentile and the 97.5 percentile). We monitored model convergence via the Brooks-Gelman-Rubin R statistic and trace plots. Model diagnostics also included visual inspection of autocorrelation plots and the posterior densities.

16We used the Egger test for continuous outcomes and Peters test for binary outcomes to assess for publication bias. In the presence of large between-study heterogeneity, we used allocation concealment as a moderator. A 2-sided P <.10 was considered evidence of publication bias. All analyses were conducted in OpenBUGS statistical software version 3.2.3 (University of Cambridge) and Stata statistical software version 16 (StataCorp). Details on the prior distributions and models are provided in the eMethods in Supplement 1.

17Our search identified 1834 records. After removal of duplicates, we carried out title and abstract screening of 1193 references, leaving 67 articles selected for full text review. Of these studies, 14 trials (15 records) were deemed eligible (Figure 1). Excluded references from the full text screening and the reasons for exclusion are provided in eTable 2 in Supplement 1. The characteristics of the included studies are summarized in Table 1.

18The analysis included 1244 participants (median [IQR] age, 50 [47-56] years; 45% female participants). Six trials,,,,, included patients who had previously been hospitalized due to a COVID-19 infection (range of intensive care unit admission, 6.2%-100%), and 3 trials,,, (4 records) included patients who had not been hospitalized following SARS-CoV-2 infection. Five trials,,,, included a mixed population of both patients who had been hospitalized following initial COVID-19 infection and those who had not been hospitalized following initial COVID-19 infection. The level of dyspnea of the participants at the baseline ranged from mild to severe.

19The most common interventions in the treatment group were breathing exercises, either alone,,,,, (6 trials [815 participants]), or in combination with resistance and/or aerobic training,,,, (5 trials [298 participants]). In 2 trials,, (3 records [71 participants]), the intervention included strengthening and aerobic exercises without a breathing exercise component, and 1 trial (60 participants) included only aerobic exercises. According to the TiDiER checklist, 12 trials (86%) gave appropriate descriptions of the intervention components, 11 trials (79%) stated the dose and duration of the intervention, and 9 trials (64%) described the tailoring process. A detailed analysis of the reporting of each component for each study is summarized in eTable 3 in Supplement 1. The most common comparator was usual care in the form of respiratory training and exercise-based, self-management education,,,,,,,,,,,,, (13 studies [14 records; 1156 participants]). A single study (88 participants) used sham or placebo comparisons in the form of a respiratory training with no resistance. Minimum exercise dose was not well defined for our population. The most common limitations of unsupervised respiratory training (ie, usual care) were wrong use of equipment and ineffective execution of the breathing techniques. Hence, all comparators will be referred to as usual care.

20Figure 2 summarizes the risk of bias judgment for each individual trial. Overall risk of bias was deemed low in 1 study (7%), deemed as having some concerns in 6 trials,,,,,, (7 records [43%]), and deemed high in 7 trials,,,,,, (50%). Risk of bias arising from the randomization process and the selection of the reported outcomes was deemed low in 9 trials,,,,,,,,, (10 records [64%]). Bias due to deviations of the included intervention was deemed low in 2 trials, (14%). The most common sources of bias due to deviations of the included intervention were nonblinding of the patients and/or the clinicians. Bias due to missing outcome data was deemed low in 4 trials,,, (29%). Bias arising from the measurement of the outcome was deemed low in 6 trials,,,,, (43%).

21A total of 7 trials,,,,,, involving 389 participants with PCC reported treatment outcomes on functional exercise capacity measured by the 6-minute walking test. The median (IQR) follow-up time for evaluating the primary outcome after randomization was 6 (5.5-7.0) weeks. Rehabilitation interventions were associated with a greater improvement in functional exercise capacity compared with usual care (SMD, −0.56; 95% CrI, −0.87 to −0.22) (Figure 3 and eFigure 1 in Supplement 1). Rehabilitation interventions demonstrated a 99.6% posterior probability of superiority and were associated with 95% probability of reaching the MID threshold when compared with controls. Specifically, patients in the intervention group were able to cover a mean (SD) of 35.84 (6.55) m more than patients in the usual care group during the 6-minute walking test (95% CI, 34.97 m to 36.71 m). The value of τ2 (0.04; 95% CrI, 0.00 to 0.60) indicated low heterogeneity for exercise capacity. The confidence in the certainty of evidence was moderate and was rated down due to high risk of bias (Table 2).

22The available evidence on fatigue outcomes was limited and could not be synthesized. A single study measured the presence or absence of fatigue during everyday activities as a binary outcome without quantifying it. A total of 2 studies, assessed only exercise-induced fatigue levels, and 3 studies (4 records),,, did not adequately report the circumstances under which fatigue was measured. Additionally, 2 studies, recorded the presence of exercise-induced fatigue as an adverse event.

23Four trials,,, reported treatment outcomes on lower limb muscle function among 371 participants with PCC. The summary point estimate suggested that rehabilitation interventions could be associated with improvements in lower limb muscle function when compared with usual care. However, given that the CrI included both positive and negative values, the magnitude and direction of the association remain uncertain (SMD, −0.77; 95% CrI, −1.94 to 0.44) (Figure 3and eFigure 2 in Supplement 1). Rehabilitation interventions demonstrated a 93% posterior probability of superiority and were associated with 84% probability of reaching the MID threshold when compared with usual care. The value of τ2 (0.8; 95% CrI, 0.15 to 8.70) indicated substantial statistical heterogeneity for the 30-second sit to stand test. The confidence in the certainty of evidence of the treatment outcomes was very low and was rated down due to high risk of bias, imprecision, and indirectness (Table 2).

24Eight trials (9 records),,,,,,,, involving 573 participants with PCC reported treatment outcomes on dyspnea. Rehabilitation interventions were associated with a greater improvement in functional exercise capacity compared with usual care (SMD, −1.00; 95% CrI, −1.94 to −0.10) (Figure 3 and eFigure 3 in Supplement 1). Rehabilitation interventions were associated with a 98% posterior probability of superiority when compared with usual care and were associated with a 95% probability to reach the MID threshold. The magnitude of τ2 indicated substantial statistical heterogeneity (τ2, 1.18; 95% CrI, 0.36 to 5.50). Heterogeneity was partially explained when adjusting for allocation concealment (eFigure 4 in Supplement 1). The confidence in the certainty of evidence was low and rated down due to high risk of bias and inconsistency (Table 2).

25A total of 6 trials (7 records),,,,,, reported treatment outcomes on FEV1 (estimated percentage) and FVC (estimated percentage) among 363 participants with PCC. We found no difference between rehabilitation interventions and usual care in either FEV1 (SMD, −0.16; 95% CrI, −0.42 to 0.11) or FVC (SMD, −0.13; 95% CrI, −0.38 to 0.12) (Figure 3 and eFigure 5 and eFigure 6 in Supplement 1). Rehabilitation interventions demonstrated 89% and 85% posterior probability of superiority when compared with usual care for FEV1 and FVC respectively. The magnitude of τ2 indicated very low statistical heterogeneity for both FEV1 (τ2, 0.02; 95% CrI, 0.00 to 0.28) and FVC (τ2, 0.01; 95% CrI, 0.00 to 0.23). The confidence in the certainty of evidence was low for both outcomes and downgraded for high risk of bias and indirectness (Table 2).

26Five trials,,,, reported treatment outcomes on overall quality of life among 366 participants with PCC. Rehabilitation interventions were associated with larger improvement in quality of life compared with the comparison group (SMD, −0.41; 95% CrI, −0.73 to −0.06) (Figure 3 and eFigure 7 in Supplement 1). Rehabilitation interventions were associated with a 98.4% posterior probability of being superior to the comparison group and were associated with a 76.9% probability to reach the MID threshold. The magnitude of τ2 indicated very low statistical heterogeneity (τ2, 0.04; 95% CrI, 0.00 to 0.63). The confidence in the certainty of evidence was moderate and rated down for high risk of bias (Table 2). Figure 3 and eFigure 8 and eFigure 9 in Supplement 1 display the association of rehabilitation interventions with physical and mental health separately.

27Five trials,,,, did not provide any information on adverse events, and 4 trials (5 records),,,, reported no adverse events related to the intervention. Adverse events were reported in 5 trials.,,,, However, we did not find compelling evidence for a difference in the odds of adverse events occurring. The wide 95% CrI indicates a high level of uncertainty and imprecision in this estimate (OR, 1.68; 95% CrI, 0.32 to 9.94) (eTable 4 and eTable 5 in Supplement 1).

28We found no indication of publication bias (eTable 6 in Supplement 1). No major concerns of autocorrelation or nonconvergence were identified. The autocorrelation and the trace plots for the main outcome are presented in eFigure 10 and eFigure 11 in Supplement 1.

29Analyses considering empirical prior distribution for the between-trial variability rendered virtually identical results (eTable 7 in Supplement 1). Analyses incorporating a thinning of 10 are presented in eFigure 12 and eFigure 13 in Supplement 1.

30Our meta-analysis of 14 randomized clinical trials examining different rehabilitation programs for people with PCC found that the patients undergoing rehabilitation experienced larger improvement in functional exercise capacity, dyspnea, and quality of life outcomes than patients receiving usual care. The analysis consistently showed that rehabilitation interventions had a greater probability of being superior to usual care across all outcomes, with probabilities ranging between 85% and 99%. Additionally, rehabilitation interventions were associated with higher probability of reaching the predefined between-group MID threshold for functional aerobic capacity, functional lower limb strength and endurance, dyspnea, and quality of life, with probabilities ranging between 84% and 95%.

31Substantial advances in vaccines and prevention strategies of acute SARS-CoV-2 infection may help reduce the burden of PCC. However, as pharmacological advances improve the prognosis of people living with comorbidities who develop an acute SARS-CoV-2 infection, the number of patients living with PCC is expected to grow. Thus, it is of high importance to develop a safe and effective strategy that will be based on high-quality evidence and will be applicable to a broad population. Given the small number of randomized clinical trials and the recent emergence of PCC, it is of no surprise that both the clinical practice and the evidence is rapidly evolving. Current treatment guidelines based on expert opinion suggest a supervised, individualized, symptom-based approach with close monitoring for adverse events such as orthostatic intolerance and postexertional symptom exacerbation., Yet, current standard care is still provided in the form of self-management recommendations in an ambulatory or home setting. The results of this study indicate that respiratory training and exercise-based rehabilitation interventions may have a greater benefit than current standard care. Inspiratory muscle training requires strict control and high adherence to achieve effective results., Research on people with chronic obstructive pulmonary disease has shown that education-based, nonsupervised rehabilitation programs are likely to be less effective than clinician-guided ones, mainly due to the wrong use of equipment, ineffective execution of the breathing techniques, and lower adherence., Additionally, we found a high level of uncertainty and imprecision in the probability of experiencing adverse events. These results further highlight the importance of supervised interventions with continuous monitoring and tailoring to ensure fidelity and patient safety until indicated otherwise.

32Female sex and inability to adequately rest in the early weeks after developing COVID-19 are associated with increased risk of development of PCC. Because women often take on societal roles such as caring and household duties, which may not allow for adequate rest in the early weeks after infection, women may face increased risk for PCC due to both biological and social factors. Yet, only 45% of the population in the included studies were women. This finding highlights an important research gap that should be addressed in future studies.

33Despite research indicating that fatigue is the most frequently observed symptom among individuals with PCC, we found limited evidence quantifying the association of rehabilitation interventions with fatigue during everyday activities. This scarcity of high-quality evidence for managing fatigue has been previously underscored by de Sire et al in a systematic review that explored the association of rehabilitation with fatigue in patients with PCC. The review identified only 1 observational study and 5 intervention-based studies that lacked a control group. Evidence-based practice hinges on high-quality research, and this considerable gap in the current evidence should be addressed in future studies.

34Although several systematic reviews on PCC have emerged,,,,,,,,,,,,, our findings add to the literature on PCC recovery in several ways. Some of the limitations of the currently published evidence are the lack of a meta-analysis component,,,,,,,,, and the synthesis of data from both acute care patients and PCC patients in the same analysis. This study is the first, to our knowledge, to conduct a systematic review with a bayesian meta-analysis on rehabilitation interventions for patients with PCC. Our findings are in agreement with a previous systematic review with meta-analysis on PCC. Chen et al included a total of 3 trials (233 patients with PCC) and found that rehabilitation interventions were associated with a clinical important difference in favor of pulmonary rehabilitation for exercise capacity, but the outcomes on respiratory function were inconsistent across studies. Our review identified 11 additional trials (ie, a total of 14 trials with 1244 participants) and examined a full range of outcomes. To increase the robustness of our results we calculated the probability to indicate superiority of the rehabilitation interventions when compared with usual care, and we prespecified all of our analyses. The robustness and accuracy of our results are further supported by the low between-trial heterogeneity and absence of publication bias. In addition, we used an extended and comprehensive search strategy and searched all relevant sources to retrieve all the potentially eligible randomized clinical trials. We therefore believe that it is unlikely that we missed any relevant trials.

35The quality of our analysis is limited by the quality of the underlying data. Although we only included randomized clinical trials, these studies were not without flaws, and the overall evidence grade was deemed to be moderate. Bias in the included studies in terms of allocation concealment, blinding, and missing data could have led to an overestimation of the treatment outcomes. Currently, there is no criterion standard measurement tool for functional aerobic capacity, fatigue, dyspnea, and quality of life in patients with PCC. Therefore, it is possible that certain outcomes did improve, but the measurement tools that were chosen by the researchers did not have the ability to measure this change, which could have led to an underestimation of the treatment outcomes. Some trials included estimates from a per-protocol analysis only, which may have also led to overestimation of the treatment outcomes. Even though female sex is associated with higher risk of PCC, only 45% of the participants in the included studies were women, which could have potentially reduced the external validity of our results. We did not pose any restriction on concurrent medication use with the rehabilitation protocol. Including studies that allowed for concurrent medications increased the chances of overestimating the treatment outcomes due to potential synergy effects. However, not allowing for concurrent medications would have made our results inapplicable to a large proportion of people living with comorbidities who depend on several medication-based treatments. Because comorbidities are associated with higher risk of development of PCC, we believe that not allowing for concurrent medication would have been unfeasible, impractical, and unethical. Our analysis was focused on the shortest time point available after the completion of the interventions, which allowed us to measure the effects of the interventions more accurately and minimize any potential impact from external trial factors that could influence treatment outcomes but prevented us from assessing any associations in the long-term follow-ups.

36Rehabilitation interventions demonstrated an association with improved outcomes in functional exercise capacity, dyspnea, and quality of life, with a high probability of improvement compared with the current standard of care. The certainty of evidence was moderate for functional exercise capacity and quality of life and low for other outcomes. Given the uncertainty surrounding our safety outcomes, additional trials with enhanced monitoring of adverse events are necessary.