Effectiveness of Respiratory Muscle Training for Pulmonary Function and Walking Ability in Patients with Stroke: A Systematic Review with Meta-Analysis

Background: Neurological dysfunction due to stroke affects not only the extremities and trunk muscles but also the respiratory muscles. Aim: to synthesise the evidence available about the effectiveness of respiratory muscle training (RMT) to improve respiratory function parameters and functional capacity in poststroke patients. Methods: a systematic electronic search was performed in the MEDLINE, EMBASE, SPORTDiscus, PEDro and Web of Science databases, from inception to May 2020. Study selection and data extraction: randomised controlled trials (RCTs) that examined the effects of RMT versus non-RMT or sham RMT in poststroke patients. We extracted data about respiratory function, respiratory muscle strength and functional capacity (walking ability, dyspnea, balance, activities of daily life), characteristics of studies and features of RMT interventions (a type of RMT exercise, frequency, intensity and duration). Two reviewers performed study selection and data extraction independently. Results: nineteen RCTs met the study criteria. RMT improved the first second forced expiratory volume (FEV1), forced vital capacity (FVC), peak expiratory flow (PEF), maximal expiratory pressure (MEP), maximal inspiratory pressure (MIP) and walking ability (6 min walking test), but not Barthel index, Berg balance scale, and dyspnea. Conclusions: RMT interventions are effective to improve respiratory function and walking ability in poststroke patients.


Introduction
Stroke is a loss of focal neurological function due to infarction or haemorrhage in an essential part of the brain [1]. This neurological damage affects not only peripheral but also respiratory muscles, causing respiratory weakness, changes in the respiratory pattern and decreases in respiratory volumes and flows [2,3]. These breathing alterations are related to a decrease in physical activity and therefore, to a reduction in the ability to carry out the activities of daily life [4].

Study Selection
Reviewers were not blinded to authors, journals, or institutions. Included articles were RCTs studies that analysed the effectiveness of RMT (IMT and/or EMT) on improving pulmonary function, respiratory muscle strength and functional capacity parameters in patients with stroke.

Data Extraction
For data extraction, two authors (DPP-C and AIC-C), independently, used a standardised data collection form. From each selected study, the following data were collected: (1) first author's name and year of publication; (2) country; (3) period of data collection; (4) characteristics of study sample: sample size, age mean, gender; (5) data concerning inclusion/exclusion criteria; (6) main study outcomes. We emailed the corresponding author requesting the data when some information was lacking.

Risk of Bias Assessment
While in the Prospero register we stated that the intent to use the Cochrane Collaboration tool to assess the risk of bias (RoB2) [13], we observed that all previous systematic reviews and meta-analysis on this topic had used the PEDro Scale to assess the quality of included studies. Thus, with the aim of increase the homogeneity and the comparability, we decided to use this scale instead of RoB2.
PEDro Scale is a useful tool for assessing the quality of physical therapy trials [14]. This scale has 11 items designed for rating the methodological quality of RCTs assessing the internal validity (e.g., random and concealed allocation, baseline similarity, blinding, and intention-to-treat analysis) and statistical reporting. The total PEDro score ranges from 0 to 10 points (the first item, eligibility criteria, is not included in the total score); higher score means better study's quality [15].
Data extraction and risk of bias assessment were independently completed by two reviewers (DPP-C and AIC-C) if appeared inconsistencies were solved by consensus or involving a third researcher (JAL-A).

Statistical Analysis
For each of the main outcomes (FEV1, MIP, MEP, PEF, FVC, 6-MWT, dyspnea, MBI and BBS), a separate pooled estimate of effect size (ES) and their respective 95% confidence intervals (95% CI) was calculated. The ES parameters from preintervention to postintervention between groups (exercise intervention vs control) [16] in each study were calculated assuming a correlation coefficient of 0.5 and using a random-effects model based on the DerSimonian and Laird method [17]. Statistical heterogeneity was analysed by the I2 statistic. Heterogeneity was considered as not important (I2: 0% to 40%), moderate (I2: 30% to 60%), substantial (I2: 50% to 90%) or considerable (I2: 75% to 100%); the corresponding p-values were considered [12]. When a study reported results separately for IMT and EMT, we calculated a combined estimate.
According to the Cochrane Handbook recommendations, when mean and standard deviation (SD) were lacking, and available statistics were median and interquartile range (IQR), the IQR was divided by 1.35 to transform these estimates on CI [18]. Finally, when studies were scaled inversely (i.e., lower values indicated worse outcomes), the mean in each group was multiplied by −1.
Subgroup analyses were conducted when it was possible due to the number of studies, to assess whether the intervention of RMT (IMT/EMT) or only IMT allowed better results on outcome variables. Sensitivity analyses were performed by removing studies one by one to assess the robustness of the summary estimates and to detect whether any study accounted for a large proportion of heterogeneity among RMT and ES pooled estimations. Meaningful results of the sensitivity analyses were considered when the resulting estimates were modified beyond the CIs of the original summary estimate.
Finally, Egger regression asymmetry method and visual inspection of funnel plots were used to assess publication bias. Statistical analyses were performed using Stata Statistical software, version 16.0 (Stata, College Station, TX, USA).
The nineteen studies included involved 643 patients with stroke, with the sample size varying between 12 and 82 participants. The mean age of participants was 61.3 years.
The interventions were very diverse; the number of sets per session varied between two and ten sets; the repetitions in each set ranged between five and 30. Other studies listed time rather than the number of repetitions; the load of RMT in the majority of studies was at 30% of MIP/MEP at the beginning, increasing with the intervention. However, some studies used higher loads at 40% or 50% at the beginning of the intervention.
The nineteen studies included involved 643 patients with stroke, with the sample size varying between 12 and 82 participants. The mean age of participants was 61.3 years.
The interventions were very diverse; the number of sets per session varied between two and ten sets; the repetitions in each set ranged between five and 30. Other studies listed time rather than the number of repetitions; the load of RMT in the majority of studies was at 30% of MIP/MEP at the beginning, increasing with the intervention. However, some studies used higher loads at 40% or 50% at the beginning of the intervention.
The length of the interventions ranged between three and ten weeks. In the control group, most studies conducted the conventional stroke rehabilitation program, however, four out of twenty studies performed sham respiratory training without resistance and/or progression [21,30,31,35]. Sham respiratory training without any resistance and progression.

Risk of Bias Assessment
The risk of bias of included studies was assessed using the PEDro Scale. All RCTs included complied with the following items: random allocation, to report between-group differences and in to report point estimate and variability ( Table 2). The highest score was eight points for two studies [21,30]. Four studies scored seven [19,31,35,37], mainly due to the non-blinding of therapist and assessors and the non-presentation of an intention-to-treat analysis. Five studies scored six and four studies obtained five points. The lowest score was three points for the study of the NJ Jung et al. [25]. The three resting studies obtained four points [24,33,34]. Thus, fifteen studies obtained higher or equal to five points, a moderate quality.
Equally, MEP was analysed in seven studies [19][20][21]28,30,31,37], and only one of these performed exclusively IMT exercises [37]; thus, the overall ES was 0.55 (95% CI: 0.12 to 0.98; I 2 = 71.9%, p = 0.002) ( Figure 5).  The MIP was examined in nine studies [19][20][21][22]28,30,31,35,37], and obtained a pooled ES= 0.84 (95% CI: 0.44 to 1.24; I 2 = 69.2%, p = 0.001) ( Figure 6); three of these studies performed only IMT exercises [22,35,37], and six studies carried out both IMT and EMT exercises, showing statistically significant improvements for both groups of interventions. Sensitivity analyses for all pulmonary function variables showed that the pooled ES estimates were not significantly modified in magnitude or direction when individual study data were removed from the analysis one at a time. The exception was the FVC analysis, which lost the statistical significance when the studies of KM Jung et al. [25] and K Kim et al. [36] were removed from the analysis.
There was no evidence of publication bias by funnel plot asymmetry and Egger's test in all pulmonary function variables studied.
Publication bias was not assessed by funnel plot and Egger test, due to the reduced number of studies that evaluated these variables of functional capacity [38].

Discussion
In the present study, the results show a significant positive relationship between the RMT interventions on the improvement of pulmonary function parameters (FEV1, FVC, PEF, MEP, MIP) and the functional capacity such as walking ability (6-MWT) but not in balance (BBS), MBI and dyspnea, in patients with stroke.
The subgroup analyses by type of RMT exercises did not show differences between the exclusive IMT interventions and the IMT/EMT interventions, except in the PEF, in which the combined IMT and EMT exercises showed a greater effect than only IMT exercises on the improvements of this pulmonary parameter.
A previously published meta-analysis that included eleven trials also obtained similar results [7]. Nevertheless, the main difference was found in dyspnea. Contrary to this study, our meta-analysis did not find a positive effect of RMT on dyspnea in patients with stroke. These differences could be due to several reasons: first, the previous study only included two moderate methodological quality RCTs, a reduced number of studies to provide consistent conclusions about the effectiveness of RMT on dyspnea; second, patients with stroke usually have a low perception of dyspnea, due to their dissociation between respiratory effort and dyspnea [39].
Regarding the pulmonary function parameters, our results showed positive effectiveness of RMT interventions to improve FEV1, FVC, PEF, MEP, and MIP in patients with stroke. These results are in line with previous systematic reviews and meta-analyses results [7,8]. Patients who have suffered a stroke present abdominal and diaphragm dysfunction, causing a decrease of respiratory muscle strength [2,40,41]. This weakness of respiratory muscles is usually associated with reduced lung volumes, flows and restrictive ventilatory patterns [42]. Thus, it seems obvious, and the results of this meta-analysis corroborate this, that the training of the respiratory muscle will improve the lung volumes and flows, noting an increase of FEV1, FVC, PEF, MEP and MIP.
Aiming to facilitate the clinical interpretation of these results, we expressed the pooled effect estimate of each pulmonary function parameter on clinical measurement improvements, using methods recommended by the Cochrane Collaboration [43]. Thus, our data show that RMT interventions increase baseline FEV1 values in 12.2% of predicted-FEV1 by age and sex, FVC baseline values improve in 6.75% of predicted FVC, PEF baseline measure increases in 46.97 litres per second. Finally, MEP and MIP improved their baseline values in 10.05 and 22.40 cm H 2 0, respectively. These results align with previously published meta-analysis [7][8][9][10].
The improvement of inspiratory and expiratory muscle strength had significant implications for the functional capacity of poststroke patients. The weakness of respiratory muscles, secondary to stroke, can lead to a reduced tolerance to exercise [8,44]. Thus, the benefits gained by the increase of respiratory muscle strength and lung function could improve the tolerance to exercise [8].
Our results showed that RMT achieved an improvement of walking ability measured by 6-MWT, with an ES= 0.39, which when translated in terms of clinical measure improvement, using methods endorsed by the Cochrane Collaboration [43], is equal to an increase of 25 m travelled in 6 min. Although it could be seen as little progress, walking ability is a significant predictor of physical activity and social and community participation after stroke [45,46], so even small improvements could imply immense benefits for the physical and social health of stroke patients. They can help these patients to carry out daily activities more efficiently.
In line with this, three studies reported results about the Barthel index, a scale used to measure disability or dependence in activities of daily living [47]. The pooled estimate showed a positive but not statistically significant trend. The same occurred with the BBS [48]. Nevertheless, the limited number of studies that reported these variables do not allow for robust conclusions about it.
Despite the wide variety of characteristics of RMT interventions, the previous meta-analysis concluded that 30 min of RMT, five times/week, for five weeks could be sufficient to increase respiratory muscle strength in very weak poststroke patients [3].
This study has several limitations that must be considered, some of them, such as the heterogeneity, are inherent to meta-analyses. First, although a higher number of studies than the previous meta-analysis was included, not all of them report the same variables, so some variables such as MBI or BBS only were informed in three studies. Second, it is known that patients with more significant respiratory muscle weakness usually respond better to RMT [8], and in the present work around the half studies [19][20][21][22][28][29][30][31]37] included poststroke patients with respiratory muscle weakness (MIP < 50 cm H 2 O) which could have influenced the results. Third, the inclusion/exclusion criteria were different in the studies included, equally to the RMT characteristics of interventions. Nevertheless, the subgroup analysis by type of RMT exercises allowed for the comparison of the effect of IMT alone with the IMT and EMT exercises. Fourth, despite the number of studies included, the total sample of participants included was scarce.

Conclusions
The RMT interventions are effective in improving pulmonary function parameters (FEV1, FVC, PEF), strength of expiratory and inspiratory muscles (MEP and MIP), and walking ability in poststroke patients. More well-designed RCTs with larger sample sizes are needed to examine the most appropriate features of interventions: IMT, EMT or both, duration, frequency, and intensity to establish the highest clinical efficacy.