Post-Operative Outcomes of Pre-Thoracic Surgery Respiratory Muscle Training Vs. Aerobic Exercise Training: A Systematic Review and Network Meta-analysis

Abstract

Objective:

To compare the postoperative outcomes of preoperative respiratory muscle training with a device (RMT) to preoperative aerobic exercise training (AET) in patients undergoing thoracic surgeries (cardiac and lung).

Data sources:

PubMed, EMBASE, Cochrane, and Web of Science were comprehensively searched upon inception to 9/2020.

Study Selection:

All randomized control studies, including preoperative RMT and preoperative AET compared to a non-training control group, were included.

Data Extraction:

The meta-analysis was performed for outcomes including postoperative pulmonary complications (PPC), pneumonia, postoperative respiratory failure (PRF), hospital length of stay (HLOS), and mortality. We performed a network meta-analysis based on Bayesian random-effects regression models.

Data Synthesis:

A total of 25 studies, 2070 patients were included in this meta-analysis. Pooled data for the patients who performed RMT with a device showed a reduction in PPCs, pneumonia, PRF with OR (Odds ratio) of 0.35 (P-value 0.006), 0.38 (P-value 0.002), and 0.22 (P-value 0.008), respectively. Pooled data for the patients who performed AET showed reduction in PPC, pneumonia with a OR of 0.33 (P-value <0.00001) and OR of 0.54 (P-value 0.01) respectively. HLOS was decreased by 1.69 days (P-value <0.00001) by performing RMT and 1.79 days (P-value 0.0008) by performing AET compared to the usual group. No significant difference in all-cause mortality compared to usual care in both RMT and AET intervention groups. No significant difference in the incidence of PRF compared to usual group in RMT + AET and AET alone intervention groups (OR 0.32; p= 0.21; OR 0.94; p= 0.87). Based on rank probability plots analysis, on network meta-analysis, RMT and AET ranked similarly on the primary outcome of PPC and secondary outcomes of pneumonia, PRF and HLOS.

Conclusion:

In thoracic surgeries, preoperative RMT is comparable to preoperative AET to prevent PPC, pneumonia, and PRF and reduce HLOS. It can be considered in patients in resource-limited settings.

Thoracic surgeries, including cardiac, lung resection, and aortic surgeries, are considered high-risk and are associated with significant postoperative pulmonary and cardiac complications. The definition of postoperative pulmonary complications (PPC) varies based on the classification used^1–3. PPCs are comprised of atelectasis, pneumonia, pleural effusion, pulmonary edema, bronchospasm, and respiratory failure⁴. PPCs are more common than postoperative cardiac complications and play a more significant role in health care costs, hospital length of stay (HLOS), readmission rates, and mortality^5–8. The incidence of PPC after elective cardiac or lung surgery ranges from 15%−43%^9–11. Various recognized risk factors predispose the person to develop PPCs. The procedure-related risk factors include the type of surgery, incision site, operative time, and type of anesthesia used. The patient-related risk factors include age > 60 years, ASA class 2 or above status, chronic obstructive pulmonary disease, congestive heart failure, and functional dependence. Patients at high risk of PPCs who are scheduled to undergo elective thoracic surgery should be considered for preoperative rehabilitation strategies as a standard of care.

Preoperative rehabilitation prior to thoracic surgeries, including respiratory muscle training with a device, aerobic exercises, and/or breathing exercises, has been shown to reduce the incidence of PPCs, pneumonia, and overall HLOS^12–14. A recent meta-analysis concluded that preoperative exercise training of any kind that improves physical fitness reduces the risk of PPCs in patients undergoing major surgery¹². A meta-analysis of 12 randomized control trials (RCT) by Katsura et al. preoperative respiratory muscle training (RMT) in participants undergoing cardiac and major abdominal surgery showed a reduction in postoperative atelectasis, pneumonia, and HLOS¹⁵. A meta-analysis of 17 RCTs by Kendall et al. for preoperative and postoperative RMT for all surgeries showed a reduction in PPCs and HLOS¹⁶. Similarly, for preoperative aerobic exercise training (AET), a meta-analysis of 5 RCTs by Cavalheri et al. in patients with non-small cell lung cancer awaiting lung resection surgery showed a reduction in PPCs and HLOS¹⁴. Patients with respiratory muscle weakness have a higher risk of PPCs due to an altered breathing pattern (more rapid and shallow), an increased tendency to develop micro atelectasis, and decreased cough strength. Preoperative RMT can significantly improve mean inspiratory muscle strength and endurance after thoracic surgery without significant side effects¹⁵. It can lead to deeper breathing after surgery, potentially resulting in a decreased incidence of PPCs. Preoperative AET includes aerobic and resistance training exercises in a supervised environment. Preoperative AET by improving exercise capacity can also improve PPCs.

There is no published meta-analysis for examining whether postoperative complications and hospital length of stay (HLOS) following major thoracic surgery is reduced by preoperative intervention of RMT or AET compared to usual care. Although individual studies demonstrated both interventions are beneficial, they have never been compared head-to-head to determine the optimal rehabilitation approach to reduce PPCs following major thoracic surgery. As RMT is logistically easier to implement in the preoperative period, we hypothesized that it would lead to improvements comparable to AET. As a result, we performed a network meta-analysis examining the benefits of preoperative RMT with a device [threshold loading device or a respiratory muscle endurance device] vs. preoperative AET on PPC, pneumonia, postoperative respiratory failure (PRF), HLOS, and mortality in patients undergoing thoracic surgeries (cardiac and lung procedures).

METHODS

Protocol and Registration

After developing appropriate PICO question and initial search criteria, and keywords, the protocol was registered with the International Prospective Register of Systematic Reviews (CRD 42021231315).

Eligibility Criteria

We included RCTs in adults undergoing elective thoracic surgery for coronary artery bypass surgery, lung resection, aortic aneurysm repair, open aortic valve or mitral valve replacement with the interventions of preoperative respiratory muscle training, preoperative aerobic exercises compared with usual care on postoperative outcomes including PPCs, pneumonia, HLOS, PRF and mortality. The studies were excluded if the surgery included an abdominal approach, the intervention was continued into the postoperative period, and published in any language other than English.

Information Sources and Search

We searched MEDLINE, Web of Science, Embase, Cochrane Library from inception to September 2020, including literature published in English using medical subject headings (MeSH) terms and free text related to thoracic surgery, inspiratory muscle training, preoperative aerobic exercises, and randomized controlled trials. The details of search criteria are provided in Appendix 1. We reviewed all the references of previously published meta-analysis associated with this analysis and added additional references to this study as deemed appropriate.

Study selection

Two authors (RK, AS) independently screened the titles and abstracts of potentially relevant records and selected the trials for meta-analysis based on full-text review using Covidence software. A third reviewer (MJM) resolved any disagreements between the two reviewers. A Prisma flow diagram representing the literature retrieval process is presented in Figure 1. A total of 10185 references were imported for screening. After removing all the duplicates there were 7359 references remaining. Following the initial screening of titles 7227 articles were excluded. From the 132 relevant studies identified for full text review, only twenty-five met the criteria for the current meta-analysis.

Figure 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram for the systematic review and meta-analysis.

Data Collection Process

From each study included in the analysis, we collected data on patient characteristics, study site, surgery performed, type of intervention provided, and its effects in postoperative period.

Data Items

The primary endpoint was the incidence of PPCs. Secondary endpoints were pneumonia, HLOS, PRF, and mortality.

Type of interventions

Eight studies included had the intervention of respiratory muscle strength training using a threshold inspiratory muscle trainer device^10,17–22 or respiratory muscle endurance with Spirotiger^® device²³(RMT vs. Usual Care). The workload for the threshold device ranged from a predetermined set of 30% to 40% of maximum inspiratory pressure (MIP)^10,17–21, except one study which used increments from 15% up to 60% of their MIP²². For respiratory endurance, the training started at 30% of maximal voluntary ventilation (MVV)²³. The duration of most of the studies ranged from 2–4 weeks, except one study of 5 days¹⁸, and the other study did not report the duration of the intervention²¹. The included studies had each session lasting from 20 to 40 minutes, 4 to 7 days per week. Fifteen studies had the intervention of preoperative aerobic exercises with and without breathing exercises^11,24–37(AET vs. Usual care). Most of the studies had intervention duration between 1–3 weeks^{11,25–32,34–37}, except two studies which had a duration of 8 weeks²⁴ to 16 weeks³³. The included studies had each session lasting from 30 to 60 minutes, 3 to 7 days per week. Although the type of intervention used was heterogenous, all included studies had some form of aerobic or high-intensity exercises as a treatment plan. Eight studies had breathing exercises without inspiratory muscle training^25–31,34, seven studies did not include breathing exercises in the treatment regimen^{11,24,32,33,35–37}. Two studies included had the intervention of both Inspiratory muscle training and preoperative aerobic exercise^38,39(RMT+AET vs. Usual care). One study had an intervention of 30 minutes of RMT and 40 minutes of AET per day for one week³⁸, and another study had an incremental increase of 10 minutes each week in the duration of RMT and AET to a duration of 30 minutes of RMT and 30 minutes AET five days a week over four weeks³⁹. The comparative intervention was usual care with routine preparation, including laboratory and radiologic examinations and preoperative education. In a few studies, abdominal and deep breathing maneuvers were included as part of usual care^10,18,23.

Reported outcomes

The impact of the preoperative intervention on the incidence of post pulmonary complications (PPCs) and pneumonia were reported by 14 studies and 17 studies. The PPCs in these studies were defined using Calvien-Dindo Classification Score^{11,23,26,27,30,31}, Kroenke’s grading system^10,18, Melbourne Group scale³⁴, or other specific criteria^{22,28,29,38,39}. The impact of preoperative intervention on HLOS was reported by 16 studies^{10,17,18,20,23,25–33,38,39}, PRF by 11 studies^{10,11,19,22,23,26–28,30,38,39} and mortality by 13 studies^{10,11,18,19,24–32}. PRF was defined as respiratory failure requiring ventilatory support six hours after completion of surgery.

Risk of Bias in Individual Studies

The risk of bias of all studies was assessed using the Cochrane collaboration’s tool - RoB 2⁴⁰. The tool addresses five domains: randomization process, deviations from the intended interventions, missing outcome data, measurement of the outcome and selection of the reported result. Risk of bias assessment was performed for all the included studies independently by two authors (RK and AS), and a third author (MJM) was available to resolve any disagreements (Figure 2). Eleven studies were deemed to be low risk for bias^{10,11,18,20,23,26,28,29,31,32}, eight were rated to have high-risk bias^{17,21,25,33–35,37,39}, and six were rated to have some concerns^{19,22,24,27,36,38}.The study designs, participant’s characteristics, search strategy, inclusion and exclusion criteria, reasons for full article exclusion are elaborated in the supplemental material (Appendix 1).

Figure 2.

Risk of Bias for all included Randomized Controlled Trials.

Statistical analysis:

Meta-analysis

Meta-analyses were performed if data from at least two trials could be combined. Data was pooled into Cochrane Collaboration’s Review Manager (RevMan) version 5.4. RevMan was used for statistical analysis and generation of forest plots. For the dichotomous variables, Odds ratios (OR) with 95% confidence intervals (CIs) were computed to compare intervention and control groups at the study level. The effect estimates of individual studies were combined into a pooled weighted estimate using Mantel-Haenszel weights. For continuous outcomes, the mean differences in effects between the intervention and control groups were computed at the study level and pooled into mean differences (MDs) using the inverse-variances method. The continuous variables reported as median and IQR were tested for the skewness of the data from the sample size and the five-number summary⁴¹. The sample mean and standard deviation were estimated from the sample size, median, and IQR if not skewed^42,43. We expected a priori that the included RCTs would be heterogeneous and used a random-effects model throughout. Data heterogeneity was defined as I² >50%. Subgroup analyses was performed if data heterogeneity was reported more than 50%.

Network Meta-analysis

We performed a study-level network meta-analysis (NMA) based on Bayesian random-effects regression models. We used a logit link function for the binary outcomes (PPC, mortality, atelectasis, etc.) and an identity link function for the HLOS. Analyses were performed using Markov chain Monte-Carlo methods. A vague prior uniform distribution was chosen for the between-trial variances, which we assumed to be equal across comparisons. To ensure that the prior distribution was sufficiently vague, uniform distribution was chosen as U(0, S), Specifically S was selected heuristically based on the outcome scale following the approach by Valkenhoef et al., in which S was the maximum of the effect sizes reported in selected studies for the outcomes of interest⁴⁴. Convergence was assessed by checking plots of the Gelman-Rubin statistics, which indicated that the width of pooled runs and individual runs stabilized around the same value, and their ratio was approximately 1.

For each pairwise comparison of the binary outcomes, we used odds ratios (ORs) with 95% credible intervals (95% CrIs) as a measure of the association between the treatment used and its efficacy. Mean difference with 95% CIs were used for HLOS. Within the Bayesian framework, the NMA estimated the overall rankings of treatments by calculating their posterior probabilities. Given the lack of loops in the networks, assessments of inconsistency (consistency describes the agreement between estimates of various studies for a specific comparison) and coherence (coherence describes the agreement between direct and indirect estimates for a specific comparison) was irrelevant in this study. Network meta-analysis was performed using R 4.1.2, R gemtc package and JAGS 4.3.0.

Synthesis of Results

Primary Outcome:

• Post-operative pulmonary complications

RMT vs. Usual care group

Four studies reported PPC outcomes^10,18,22,23(Figure 3). Three studies used a threshold inspiratory muscle trainer device^10,18,22, and one study used respiratory muscle endurance with Spirotiger^® device²³. The incidence of PPCs was 32.4% and 15.4% in a usual care group and RMT intervention group, respectively. The pooled analysis of four studies with preoperative exercise using RMT was associated with a reduced risk of PPCs compared to the usual care group (OR 0.35; 95% CI, 0.16 – 0.74; I²= 61%, P-value 0.006).

Figure 3.

Forrest plot of the effect of Respiratory Muscle Training on postoperative pulmonary complications, including all randomized controlled trials. CI = confidence interval; df = degree of freedom; I² = percentage of variation across studies due to heterogeneity; M-H = Mantel-Haenszel.

AET vs. Usual care group

Seven studies reported PPC outcomes^{11,26–30,34} (Figure 4). Six studies used aerobic exercises, breathing exercises, and incentive spirometry^26–30,34. One study used aerobic exercises alone¹¹. The incidence of PPCs was 37.0% and 16.6% in a usual care group and AET intervention group, respectively. The pooled analysis of seven studies with preoperative AET was associated with a reduced risk of PPCs compared to the usual care group (OR 0.33; 95% CI, 0.22 – 0.51; I²= 0%, P-value < 0.00001).

Figure 4.

Forrest plot of the effect of Aerobic Exercise Training on postoperative pulmonary complications, including all randomized controlled trials. CI = confidence interval; df = degree of freedom; I² = percentage of variation across studies due to heterogeneity; M-H = Mantel-Haenszel.

RMT + AET vs. Usual care group

Two studies reported PPC outcomes^38,39 (Figure 5). One study used aerobic exercises, a threshold inspiratory muscle trainer device, and breathing exercises³⁸, other study used aerobic exercises and a threshold inspiratory muscle trainer device³⁹. The incidence of PPCs was 60.0% and 23.8% in a usual care group and RMT + AET intervention group, respectively. The pooled analysis of two studies with preoperative RMT + AET was associated with a reduced risk of PPCs compared to the usual care group (OR 0.20; 95% CI, 0.05 – 0.81; I²= 0%, P-value 0.02).

Figure 5.

Forrest plot of the effect of Respiratory Muscle Training and Aerobic Exercise Training on postoperative pulmonary complications, including all randomized controlled trials. CI = confidence interval; df = degree of freedom; I² = percentage of variation across studies due to heterogeneity; M-H = Mantel-Haenszel.

Secondary outcomes and Subgroup analysis:

The secondary outcomes for postoperative pneumonia, PRF, HLOS, all-cause mortality at 30 days, subgroup analysis for RMT for PPC, and AET for HLOS outcomes with supplemental figures are elaborated in the supplementary material (Appendix 2).

Network Meta-analysis:

Similar to the analyses above, for the PPC, all interventions are superior to usual care (Figure 6). The differences among the intervention groups were not significant, although RMT + AET showed the highest efficacy. A similar pattern can be observed for the secondary outcomes, for which RMT + AET is generally most beneficial, but no significant difference was found among the three types of interventions (Table 1). The Bayesian ranking results suggest RMT + AET has the highest probability of being ranked first for PPC and most secondary outcomes except for PRF. For PRF, RMT has the highest probability of being ranked first (Figure 7). The ranking probabilities of interventions were further summarized by the surface under the cumulative ranking curve (SUCRA) values⁴⁵. High SUCRA values are expected for the best treatments, and low SUCRA values are expected for the worst treatments (Table 2). Similarly, RMT + AET has the highest score for all outcomes except for PRF, with an averaged SUCRA of 0.8409. No clear difference was observed for RMT and AET alone (Table 2). On average, the RMT has a higher score than AET alone (0.6201 vs. 0.4975).

Figure 6.

Forest plot for PPC, pneumonia, PRF, HLOS and all-cause mortality with interventions versus usual care (control). An odds ratio <1 or mean difference <0 favors the interventions. 95% CrI indicates the 95% credible interval. I², the percentage of variance attributable to study heterogeneity.

Table 1.

Pooled estimates (logORs [95% CrIs] or mean difference [95% CrIs] for HLOS) for each outcome of the network meta-analysis. PPC - Postoperative pulmonary complications, PRF - Postoperative respiratory failure, HLOS – Hospital length of stay, RMT - Respiratory muscle training, AET - Aerobic exercise training.

		Control	RMT	RMT-AET
PPC	AET	1.14 (0.61, 1.7)	0.09 (−0.76, 0.87)	−0.6 (−2.34, 0.95)
	Control	-	−1.05 (−1.71, −0.5)	−1.74 (−3.35, −0.27)
	RMT	-	-	−0.68 (−2.37, 0.91)
Pneumonia
	AET	0.66 (0.1, 1.2)	−0.33 (−1.24, 0.54)	−1.34 (−4.86, 1.09)
	Control	-	−0.98 (−1.70, −0.34)	−1.97 (−5.54, 0.32)
	RMT	-	-	−0.98 (−4.62, 1.38)
PRF
	AET	0.38 (−0.66, 1.92)	−1.29 (−3.29, 1.23)	−0.9 (−3.75, 1.91)
	Control	-	−1.67 (−3.47, 0.11)	−1.31 (−4.04, 0.99)
	RMT	-	-	0.39 (−2.86, 3.28)
HLOS
	AET	1.74 (0.73, 2.88)	0.29 (−1.49, 2.32)	−2.75 (−6.26, 0.76)
	Control	-	−1.45 (−2.97, 0.16)	−4.51 (−7.96, −1.16)
	RMT	-	-	−3.07 (−6.88, 0.56)
All-cause mortality
	AET	1.72 (−0.65, 4.81)	0.86 (−3.03, 5.08)	-
	Control	-	−0.84 (−4.01, 2.07)	-
		-	-	-

Figure 7.

Bayesian ranking profiles of interventions. Profiles indicate the probability of each intervention being ranked from first to last with regard to the each of the primary and secondary outcomes.

Table 2.

SUCRA scores for the interventions with respect to each outcome. PPC - Postoperative pulmonary complications, PRF - Postoperative respiratory failure, HLOS - Hospital length of stay, RMT - Respiratory muscle training, AET - Aerobic exercise training.

	PPC	Pneumonia	HLOS	PRF	All-cause mortality*	Average
Control	0.0039	0.0182	0.0153	0.1293	0.1713	0.0417
AET	0.6089	0.4441	0.5673	0.3695	0.8026	0.4975
RMT	0.5325	0.6697	0.4603	0.8178	0.5261	0.6201
RMT_AET	0.8547	0.8680	0.9572	0.6835	-	0.8409

^*All-cause mortality was not included for calculating the average SUCRA score.

Discussion:

This systematic review and meta-analysis demonstrated that in patients who undergo elective thoracic surgeries, RMT or AET, either alone or in combination, can reduce PPCs compared to usual care. The preoperative RMT with a device or preoperative AET intervention reduces the incidence of pneumonia and is associated with reduced HLOS compared to usual care. Preoperative RMT also reduces incidence of PRF compared to usual care, which was not seen with preoperative AET. There was no significant difference noted in mortality in both interventions. However, mortality rates were low in the usual care group, so more subjects are likely to be required to determine if there are any mortality differences.

The subgroup analysis in RMT shows a reduction in PPCs in patients who receive RMT using threshold devices or endurance devices. The dramatic effect of endurance device intervention can be attributed to data collection from only one study. We believe more studies are required to assess any potential difference in outcomes between respiratory muscle strength and endurance training. The diaphragm works continuously, and pulmonary disease states tend to increase diaphragmatic oxidative capacity, suggesting that efforts to increase respiratory muscle strength might be more fruitful than efforts to increase endurance. The limited evidence currently available does not support this conjecture. The subgroup analysis in AET shows a reduction of HLOS significantly in those who receive short intensive AET intervention close to surgery rather than over a few weeks. This is contrary to the usual goals of exercise training, where six weeks of training is felt to be the minimum required to induce structural changes within the exercising muscle. Thus, there appear to be clear benefits from a short perioperative training period. The mechanism by which these benefits are achieved requires further study. It is unclear why more extended periods of preoperative exercise showed lesser effects on HLOS. Unfortunately, all four studies receiving intervention for more than a week did not report the outcomes of PPCs^25,31–33.

To answer which intervention is better, a network meta-analysis based on Bayesian ranking results and SUCRA values showed that RMT is ranked similarly to AET in reducing PPCs in thoracic surgery patients. As one might expect, patients who can perform both RMT and AET had somewhat better results in lowering the incidence of PPCs compared to either intervention alone. However, the number of participants who received both interventions was very small, so this finding should be interpreted cautiously. These findings confirm that both interventions can improve postoperative outcomes with potentially addictive benefits by independently improving respiratory muscle strength and exercise capacity. RMT is ranked on par with AET on Bayesian ranking results in reducing the incidence of pneumonia and HLOS and PRF. We believe in patients undergoing elective thoracic surgeries who are at high risk for PPCs, at the very least RMT with a device can be considered, when the resources for AET before surgery are not feasible or if feasible the patient is unable to perform them due to musculoskeletal or other factors. It is easier to implement a home-based RMT program than an AET program, so this modality may allow more high-risk patients to receive preoperative training.

Strengths:

The strengths of this meta-analysis are, we selected studies with only preoperative intervention and excluded studies that carried intervention from the preoperative period into the postoperative period to allow accurate outcome assessments for the preoperative intervention. We were able to assess postoperative pneumonia as a separate variable rather than including it in assessing PPCs. All our included studies were randomized controlled trials providing more validation to our outcomes analysis results. The studies analyzed for outcome analysis have a low percentage of high-risk of bias studies for either RMT or AET.

Limitations:

There are limitations in our study; the first limitation is a potential for overestimation of treatment effect due to lack of adequate blinding, small-study effects, and publication bias which needs to be considered when interpreting the present findings. The second limitation is that AET had non-RMT breathing exercises as part of their intervention. Some of the observed effects in AET could indeed have been due to breathing exercises. We performed subgroup analysis and had one study¹¹ without breathing exercises behaving similarly to AET with breathing exercises, suggesting that non-RMT effects were relatively modest. It further substantiates our finding that RMT results were at least similar to AET since AET results might have been slightly lower if breathing exercises were not included in the AET group. The third limitation of this study is the majority of participants who received RMT were cardiac surgery patients, and the majority of patients who received AET were lung surgery patients. However, the incidence of postoperative pulmonary complications in the control group was similar for both groups. The fourth limitation is that although esophageal surgeries were included in the initial screening process, these were subsequently excluded as they were performed using thoracic, abdominal, or combined approaches with no separation of outcomes in the individual studies for the different surgical approaches. So, these findings cannot be generalized to patients receiving esophageal surgeries. None of the included studies had a patient population who underwent aortic surgeries; therefore, these findings cannot be generalized to these patients.

Conclusions:

Our network meta-analysis suggests that in thoracic surgeries, preoperative RMT is comparable to preoperative AET to prevent postoperative pulmonary complications, pneumonia and reduce hospital length of stay. RMT can be considered in patients in resource-limited settings or who are unable to perform AET due to associated co-morbidities. A short period of preoperative AET close to surgery surprisingly showed better postoperative outcomes than longer periods of AET. More studies are required to confirm whether there is added benefit of combining short-term AET with RMT in thoracic surgery patients.

Supplementary Material

e-Figure 1

e-Figure 1. Forrest plot of the effect of Respiratory Muscle Training on postoperative pneumonia, including all randomized controlled trials. CI = confidence interval; df = degree of freedom; I² = percentage of variation across studies due to heterogeneity; M-H = Mantel-Haenszel.

e-Figure 2

e-Figure 2. Forrest plot of the effect of Aerobic Exercise Training on postoperative pneumonia, including all randomized controlled trials. CI = confidence interval; df = degree of freedom; I² = percentage of variation across studies due to heterogeneity; M-H = Mantel-Haenszel.

e-Figure 3

e-Figure 3. Forrest plot of the effect of Respiratory Muscle Training and Aerobic Exercise Training on postoperative pneumonia, including all randomized controlled trials. CI = confidence interval; df = degree of freedom; I² = percentage of variation across studies due to heterogeneity; M-H = Mantel-Haenszel.

e-Figure 4

e-Figure 4. Forrest plot of the effect of Respiratory Muscle Training on postoperative respiratory failure, including all randomized controlled trials. CI = confidence interval; df = degree of freedom; I² = percentage of variation across studies due to heterogeneity; M-H = Mantel-Haenszel.

e-Figure 5

e-Figure 5. Forrest plot of the effect of Aerobic Exercise Training on postoperative respiratory failure, including all randomized controlled trials. CI = confidence interval; df = degree of freedom; I² = percentage of variation across studies due to heterogeneity; M-H = Mantel-Haenszel.

e-Figure 6

e-Figure 6. Forrest plot of the effect of Respiratory Muscle Training and Aerobic Exercise Training on postoperative respiratory failure, including all randomized controlled trials. CI = confidence interval; df = degree of freedom; I² = percentage of variation across studies due to heterogeneity; M-H = Mantel-Haenszel.

e-Figure 7

e-Figure 7. Forrest plot of the effect of Respiratory Muscle Training on hospital length of stay, including all randomized controlled trials. CI = confidence interval; df = degree of freedom; I² = percentage of variation across studies due to heterogeneity.

e-Figure 8

e-Figure 8. Forrest plot of the effect of Aerobic Exercise Training on hospital length of stay, including all randomized controlled trials. CI = confidence interval; df = degree of freedom; I² = percentage of variation across studies due to heterogeneity.

e-Figure 9

e-Figure 9. Forrest plot of the effect of Respiratory Muscle Training and Aerobic Exercise Training on hospital length of stay, including all randomized controlled trials. CI = confidence interval; df = degree of freedom; I² = percentage of variation across studies due to heterogeneity.

e-Figure 10

e-Figure 10. Forrest plot of the effect of Respiratory Muscle Training on all cause post operative 30 days mortality, including all randomized controlled trials. CI = confidence interval; df = degree of freedom; I² = percentage of variation across studies due to heterogeneity; M-H = Mantel-Haenszel.

e-Figure 11

e-Figure 11. Forrest plot of the effect of Aerobic Exercise Training on all cause post operative 30 days mortality, including all randomized controlled trials. CI = confidence interval; df = degree of freedom; I² = percentage of variation across studies due to heterogeneity; M-H = Mantel-Haenszel.

e-Figure 12

e-Figure 12. Forrest plot of Respiratory Muscle Training subgroup analysis based the device used for intervention on postoperative pulmonary complications, including all randomized controlled trials. CI = confidence interval; df = degree of freedom; I² = percentage of variation across studies due to heterogeneity; M-H = Mantel-Haenszel.

e-Figure 13

e-Figure 13. Forrest plot of Aerobic Exercise Training subgroup analysis based on duration of intervention on hospital length of stay, including all randomized controlled trials. CI = confidence interval; df = degree of freedom; I² = percentage of variation across studies due to heterogeneity.

e-Table 1

e-Table 1. Main characteristics of all included Randomized Controlled Trials. AET – Aerobic exercises treatment, RMT – Respiratory muscle training, CABG – Coronary artery bypass surgery, VATS - Video-assisted thoracoscopic surgery.

Abbreviations:

PPC

Postoperative pulmonary complications

HLOS

Hospital length of stay

ASA

American Society of Anesthesiologists

RMT

Respiratory muscle training

AET

Aerobic exercise training

PRF

Postoperative respiratory failure

MIP

Maximum inspiratory pressure

MVV

Maximal voluntary ventilation

CI

Confidence interval

MD

Mean difference

IQR

Interquartile range

I²

Heterogeneity

OR

Odds ratio

NMA

Network meta-analysis

Crls

Credible intervals

SUCRA

Surface under the cumulative ranking curve

SD

Standard deviation

PICO

Population, Intervention, Comparison and Outcomes