Respiratory Muscle Training in Chronic Airflow Limitation: A Meta-Analysis 1-3
KAREN SMITH, DEBORAH COOK, GORDON H. GUYATT, JOSHI MADHAVAN, and ANDREW D. OXMAN
Introduction
Respiratory muscle weakness appears to contribute to exerciselimitation in patients with chronic airflow limitation. It has therefore been suggestedthat strength or endurance training may, by strengthening the muscles of respiration, ameliorate dyspnea and improve exercise capacity (1-6). A number of randomized trials have been conducted to test this hypothesis, first by determining whether training can actually strengthen muscles, and second whether stronger respiratory musculature can improve exerciseperformance and day-to-day function. The results of these trials have been conflicting. Although a number of reviews of these studies have been published (7-10), the reviews all suffer from major methodologic limitations (11). These include incomplete identification of relevant studies, lack 0 f explicitcriteria regarding how studies were selected and appraised, and the absence of a quantitative analysis. Therefore, to clarify the issue of whether respiratory muscle training is of benefit to patients with chronic airflow limitation, we conducted a systematic quantitative overview of the randomized trials that have addressed this issue. Methods Does training the respiratory muscles improve spirometry, ventilatory muscle strength and endurance, exercise capacity and functional status in patients with chronic airflow limitation?
Study Identification A MEDLINE (National Library of Medicine, Bethesda, MD) search was conducted; it combined four concepts: (l) obstructive lung disease; (2) breathing exercises or exerciseor exertion or physical therapy or respiratory function tests; (3) random allocation or clinical trials or double-blind method; (4) human. Reference lists of all articles obtained were reviewed, and any additional possibly relevant citations were retrieved. Key references were searched forward in time using SCISEARCH (Institute for Scientific Informa-
SUMMARY The purpose of this study was to determine the effect of respiratory muscle training on muscle strength and endurance, exercise capacity, and functional status In patients with chronic airflow limitation. Computarlzed bibliographic data bases (MEDLINEAND SCISEARCH)were searched for published clinical trails, and an Independent revl_ of 73 articles by two of the Investi. gators Identified 17relevant randomized trials for Inclusion. Study quality wasassessed and descriptive Information concerning the study popUlations, Interventions, and ou1come measurements wes extracted. We combined effect sizes across studies (the difference between treatment and control groups divided by the pooled standard deviation of the outcome measure). Across all studies, the effect sizes and associated p-Yalues were as follows: maximal Inspiratory pressure 0.12, p .. 0.38; maximal voluntary ventilation 0.43, p = 0.02; respiratory muscle endurance 0.21, p = 0.14; laborato· ry exercise capacity -0.01, p = 0.43; functional exercise capacity 0.20, p = 0.15; functional status 0.06, p = 0.72. Secondary analyses suggested that endurance and function may be Improved if resistance training with control of breathing pattern is undertaken. Overall, there Is little evidence of clinically Important benefit of respiratory muscle training In patients with chronic airflow IImlta· tlon. The possibility that benefit may result If resistance training Is conducted in a fashion that ensures generation of adequate mouth pressures may be worthy of further stUdy. AM REV RESPIR DIS 1992; 145:533-539
tion, Philadelphia, PA). A comprehensive list of relevant articles was constructed and a letter was sent to the first author of each relevant article published in the last 5 yr asking for knowlege of any additional published or unpublished studies.
Study Selection The following criteria were used in selecting studies for inclusion in the overview. Studies included were those that had patients with chronic airflow limitation as a target population. Specific criteria included a clinical diagnosis of chronic airflow limitation and one of the following: (1) best recorded FEV,/FVC ratio of each subject < 0.7; (2) best recorded FEV, of each subject < 70070 of predicted; (3) mean FEV, of the group < 35% of predicted; (4) mean FEV,/mean FVC ratio of the group < 0.5. Only studies using respiratory muscle training, specifically resistive breathing exercises or isocapneic hyperventilation, and randomized control trials were considered. Studies were included if any of pulmonary function, respiratory muscle strength, respiratory muscle endurance, laboratory exercise capacity, functional exercise capacity, or functional status were measured. Initially, these criteria were applied by two of us (K.S. and J .M.) to the titles of all citations obtained. If an article's title suggested any possibility that it might be relevant, it was retrieved. The complete article was reviewed by three of us (K.S., J.M. and G.H.G.) and its relevance judged.
Study Evaluation and Data Extraction The methodologic quality of each study was evaluated by two reviewers(K.S. and G.H.G.) using the criteria listed in table 1. Based on this review, a total "methodologic quality" score was calculated. Agreement regarding each criterion was evaluated by weighted kappa. Relevant information regarding population, intervention, and outcome was abstracted from each relevant article by two reviewers (K.S.and D.J.C.).Disagreementsregarding relevance, methodologic quality, or information regarding population, intervention, or outcome were resolved by discussion between reviewers. We contacted the authors of all relevant papers and abstracts. Wesent the authors our summaries of the validity of their studies, and (Received in original form March 11, 1991 and in revised form September 27, 1991) i From the Departments of Medicine, Clinical Epidemiology and Biostatistics, and Family Medicine, McMaster University, Hamilton, Ontario, Canada, and the College of Medicine,'Irivandrum, Kerala State, India. 2 Dr. Cook is a career scientist of the Ontario Ministryof Health and a Scholar of the St. Joseph's Hospital Foundation; Dr. Guyatt is a career scientist of the Ontario Ministry of Health. 3 Correspondence and requests for reprints should be addressed to Karen Smith, M.D., MeMaster University, Department of Medicine, Chedoke Division, Holbrook Blvd., Room 90, Box 2000, Hamilton, Ontario L8N 3Z5, Canada
533
534
SMITH, COOK, GUYATT, MADHAVAN, AND OXMAN
TABLE 1 CRITERIA FOR METHODOLOGIC QUALITY' 1. Sample Random or consecutive sample (5) Arbitrary sample (or can't tell) (0) 2. Similarity of groups Number of following variable in which groups were comparable: age, gender forced expired volume, maximum inspiratory pressures, walk test distance (0-5) 3. Cointervention Number of following criteria met: comparable frequency of visits, comparable medication changes, comparable number of intercurrent illnesses (5 for 313; 4 for 213; 3 for 113; 0 for 0/3) 4. Sham or masking Sham or masking undertaken (5) No sham or masking (0) 5. Compliance Training was hospital supervised (5) Home program with reporting diary and patients periodically (4) Home program with either diary or periodic review (3) Compliance not measured or cannot tell (0) 6. Observer masking Observers masked as to treatment groups (5) Not masked or cannot tell (0) 7. Standardized testing Encouragement standardized (5) Not standardized or cannot tell (0) 8. Follow-up 90 to 1000Al follow-up (5) 80 to 89% follow-up (3) < 80% subjects accounted for (1) cannot tell (0) • Numbers in parentheses indicate number of points in summary score.
of the data wehad extracted, and asked them to ensure their accuracy. In addition, we asked the researchers to provide important additional information missing from their reports. Agreement between observers for study inclusion and assessment was evaluated using weighted kappa with quadratic weights (12).
Analysis The analysis was based on examination of the effect size for each individual study, where the effect was defined as the difference between treatment and control groups after treatment divided by the pooled standard deviation of the posttreatment outcome measure in the treatment and control groups. Effect sizes were weighted by the inverse of the standard deviation and combined to compute a weighted mean effect size. The methods used to test for statistical significance and for homogeneity were those described by Hedges and Olkin (13). For each outcome we reported the effect size and the confidence interval around the effect size in standard deviation units. We also reported the effect size in natural units. For outcomes in which different investigators chose different measures (for instance, maximal sustained ventilatory capacity [MSVC]versus breathing against resistance to test respiratory muscle endurance) to obtain a result in natural units, we chose the most commonly used measure (i.e., the one most likely to have intuitive meaning for the clinician) and converted the effect size back into natural units. When different measures related to laboratory exercise capacity were reported (for instance, peak workload, maximum ox-
ygen uptake, and submaximal exercise capacity all reported in the same study), a single effect size was calculated across these outcomes and that effect size used in the overall analysis. This effect size was derived in exactly the same manner as the effect sizesacross studies; that is the difference between treatment and control groups after treatment was divided by the pooled standard deviation of the posttreatment outcome measure in the treatment and control groups, and effect sizes were weighted by the inverse of the standard deviation and combined to compute a weighted mean effect size. In the sensitivity analysis of studies in which the breathing pattern was controlled versus studies in which it was not controlled, the likelihood of difference in effect size occurring as a result of the play of chance was calculated using a test of homogeneity in which the test statistic approximates the chi-square distribution. A priori Hypotheses Regarding Sources of Heterogeneity When conducting a meta-analysis, heterogeneity (that is, major differences in the apparent effect of the interventions across studies) is often found. When heterogeneity is present, it must be explained (11). Before analyzing the results, we developed a number of hypotheses concerning underlying differences in the studies which might explain heterogeneity. First, we speculated that increases in respiratory muscle strength or endurance might be restricted to studies in which training and testing occurred on the same apparatus. We therefore planned a separate analysis
for studies in which the training and testing apparatus were the same versus studies in which they were different. Second, we noted that there were two fundamentally different training strategies. One was based on volume (that is, having subjects breathe at a relatively high proportion of their maximum voluntary ventilation for appreciable periods of time) and the other was based on having subjects inspire against resistance. We speculated that these two strategies may have different effects on our outcomes. Third, we have been impressed by the recent findings reported by Belman and colleagues suggesting that ability to tolerate increasing respiratory resistance can be achieved by changes in breathing pattern, rather than by strengthening respiratory muscles (14). We therefore looked separately at studies in which the patient was required to target during training to achieve a specified flow rate. Fourth, if strengthening of the respiratory muscles has not been achieved, one cannot expect changes in exerciseperformance, functional capacity, or quality of life. We therefore conducted a separate analysis looking at clinically relevant endpoints restricted to studies in which improved respiratory muscle function had been achieved. We defined studies in which there was improved respiratory muscle function as those in which either strength or endurance improved by 0.4 or more standard deviation units. Finally, we hypothesized that results may be related to study quality. Wetherefore compared effect sizes in studies with a summary quality score of less than 20 from our assessment of methodologic quality to those with a score of 20 or greater.
Results
A total of 1,085citations wereidentified; 775 from MEDLINE, 65 from SCISEARCH, and 245 from our personal files, reviewof reference lists or from direct communications with researchers.Of these 1,085 citations, reviewers agreed that 73 were potentially eligible. The weighted kappa on the agreement between the two reviewers in identifying potentially eligible citations was 0.77 (95070 confidence interval of 0.74, 0.80). Of the potentially eligible citations, 17 proved eligible on detailed review (1, 3, 15-29). Twoofthese wereobtained from direct communication with researchers, the data not having been published (18, 21). Reasons for exclusion 0 f potentially relevant articles included patients not meeting eligibility criteria, the intervention proving not directed at respiratory muscle training, and patients not randomized to treatment and control groups. In four reports (30-33) whether the study was actually randomized was unclear from the published description. In each
535
RESPIRATORY MUSCLE TRAINING IN CHRONIC AIRFLOW LIMITATION
case, extensive attempts were made to contact the investigators. True randomization wasnot undertaken in two of these studies; in the other two, we were unable to obtain the information. The weighted kappa on the agreement between two reviewers regarding the application of inclusion criteria to potentially relevant reports was 0.66 (95070 confidence interval 0.59,0.73). A kappa of 0.66 is conventionally considered "good agreement." Authors of 14 of the relevant reports responded to our requests for checking the accuracy of the information we abstracted and for providing additional data. Each reply provided helpful additional information. The 17relevant studies are summarized in table 2. Weighted kappa on agreement for the validity criteria varied from 0.69
to 1.0.Ofthe 17studies, information that could be used in the data analysis was available in 13. Other studies provided only p values or mean scores without associated standard deviations. Of the 13 studies included in the meta-analysis, 11 studied the effects of resistive training and two studied the effect of volume training. Of the 11 that studied the effects of resistance training, the flow rates (and thus the resistance) generated during training were controlled by the investigators in four. The results of the overall statistical analyses are summarized in table 3. The interpretation of the effect sizes is aided by a rule of thumb suggested by Cohen: an effect size of 0.2 standard deviation units represents a small effect size, 0.5 a medium effect size, and 0.8 a large effect size (34). The interpretation of the
results is further aided by noting the effect size in natural units. The analyses shows small nonsignificant trends in favor of respiratory muscle training for most outcomes. The best estimate of the effect size is in all cases small. A small to moderate effect size (0.43 standard deviation units) was found for respiratory muscle strength, as measured by maximum voluntary ventilation. This corresponds to 8.8 L difference in natural units (confidence interval 1.2 to 14.4 L), and reaches conventional levels of statistical significance (p = 0.02). Statistically significant heterogeneity was found in two analyses: laboratory exercise capacity and functional status. It could be argued that for the other outcomes, further exploration in terms of the a priori sensitivity analyses described above is not necessary (or even appro-
TABLE 2 SUMMARY OF SALIENT CHARACTERISTICS OF EACH STUDY
Reference
Type of Training
Sample Size
Masking Patients/Study Personnel
Duration of Follow-up
Outcome Measures Reported*t
Overall Methodology Score
Chen, 1985 (1)
Resistance uncontrolled
Training: 7 Control: 6
Yes/no
4 wk
FEV" FVC RMS: Plmax, MW RME: endurance time at 60% PIMMaxg LEC: maximal workload. peak vo.
24
Falk, 1985 (3)
Resistance uncontrolled
Training: 12 Control: 15
Yes/yes
2 months
Trend toward greater improvement in exercise capacity and more reduction in dyspnea in trained patients
30
Harver, 1989 (15)
Resistance controlled
Training: 10 Control: 9
No/no
8 wk
FEV" FVC. FRC RMS: Plmax• MW FC/OOL: transition dyspnea index
32
Noseda, 1987 (16)
Resistance uncontrolled
Training: 10 Control: 10
No/no
8 wk
FEV,. FRC RME: highest p-flex level tolerated for 10 minutes LEC: maximal workload. peak Vo. FEC: 12-minute walk
19
Belman. 1988 (17)
Resistance controlled
High intensity: 8 Low intensity: 9
No/no
6 wk
FEV" FVC. FRC RMS: Pl max, MW RME:MSVC
19
Guyatt, 1991 (18)
Resistance uncontrolled
Training: 43 Control: 39
Yes/yes
6 months
FEV" MSVC RMS: Pl max RME: progressive endurance test LEC: exercise duration FEC: 6-minute walk FS/OOL: CRO
38
Levine, 1986 (19)
Volume
Training: 17 Control: 15
No/no
6 wk
FEV" FVC, FRC RMS: MW RME:MSVC LEC: maximal workload, peak FEC: 12-minute walk test
38
Vo.
Asher, 1982 (20)
Resistance uncontrolled
Crossover 11 patients
No/yes
1 month each
RMS: Pl max LEC: maximal workload, peak endurance at 2/3 maximal workload
10
Mcintosh. 1987 (21)
Resistance uncontrolled
Training: 6 Control: 8
Yes/no
3 months
RMS: Pl max LEC: maximal workload, peak FEC: 6·minute walk test FS/OOL: CRO
36
vo,
vo,
(continued)
536
SMITH, COOK, GUYATT, MADHAVAN, AND OXMAN
TABLE 2 CONTINUED
Reference
Sample Size
Type of Training
Masking Patients/Study Personnel
Duration of Follow-up
Outcome Measures Reported*t
Overall Methodology Score
Larson, 1988 (22)
Resistance controlled
High intensity: 10 Low intensity: 11
Yes/yes
2 months
RMS: Plmax RME: Endurance time at 66% of PIM max FEC: 12-minute walk
34
McKeon, 1986 (23)
Resistance uncontrolled
Training: 11 Control: 11
Yes/no
6 wk
Trend toward greater increase in Plmax in control, stair climbing in training group; changes in laboratory and functional exercise capacity comparable
28
Reis, 1986 (24)
Volume
Training: 5 Walking: 7
No/no
6 wk
RMS: MW RME: MSVC LEC: peak Vo.; endurance at submaximal workload FEC: 12-minute walk
21
Goldstein, 1989 (25)
Resistance controlled
Training: 6 Control: 5
No/no
4 wk
RMS: Plmax, MW RME: endurance time at fixed load LEC: endurance time at submaximal load FEC: 6-minute walk
27
Kim, 1984 (26)
Resistance uncontrolled
Crossover eight patients; incentive spirometry with or without resistance
Yes/yes
4 wk
Plmax were higher after resistance device than control in group of 8 patients who trained with resistance first; not so for other eight; no difference in 12-minute walk distance or activities of daily living
Sobush, 1986
Resistance uncontrolled
Training: 10 Control: 9
No/no
16 wk
FEV" FVC RMS: Plmax, MW FS/OOL: physical function scale of Sickness Impact Profile
22
Bjerre.Jepson, 1981 (28)
Resistance uncontrolled
Training: 14 Control: 14
No/no
6 wk
No appreciable improvement in MW in either group; RME increased comparably in both groups as did stair-climbing ability
27
Dekhuijzen, 1991 (29)
Resistance controlled
Training: 20 Control: 20
No/no
6 wk
RMS: Plmax RME: endurance of static inspiratory maneuvers FEC: 12-minute walk FS/QOL: activities of daily living LEC: VO.; maximal workload
13
(27)
8
Definitionof abbreviations: RMS - raspiratory musclestrength; Plmax = maximalinspiratory pressure; MW = maximalvoluntary ventilation; RME - respiratory muscleendurance; PIMmax - maximalInspiratory mouthpressure; LEC = laboratory exercise capacity;Vo, = oxygen uptake; FSlQOL = functional capacity/quality of life; FEC = functional exercise capacity; MSVC _ maximalsustained ventilatory capacity; CRO - chronic respiratory questionnaire. * Outcomes listed include only those with data usablein meta-analysis. t For studiesIn which no date suitable for analysis was reported, outcomes are described.
priate), However, before concluding definitively that respiratory muscle training was of little use, we felt that exploration of the between-study differences was warranted. The first hypothesis, that significant effects might only be seen when the training and testing devices were similar (and effects may thus be specific to the training modality) was generated to guard against a false positive result. It is therefore relevant only to maximum voluntary ventilation. Effect sizes were actually greater in studies in which resistance training was used (0.51 standard deviation units) than in studies using volume training (0.29 standard deviation units) in which similar training and testing
devices could be implicated. This analysis, therefore, raises no concerns about the validity of the finding of increased maximum voluntary ventilation. In addition to maximal voluntary ventilation, data are available for comparison of the effects of volume versus resistance training for respiratory muscle endurance, laboratory exercise capacity, and functional exercise capacity. For respiratory muscle endurance, the effect size was 0.09 standard deviation units for resistance training (p = 0.56) and 0.67 standard deviation units for volume training (p = 0.03). For laboratory exercise capacity, the results for both resistance and volume training, like the overall analysis, showed
negligible effects. For functional exercise capacity the effect size for volume training studies is 0.42 standard deviation units (p = 0.17); for resistance studies the effect is 0.14 standard deviation units (p = 0.64). The results of the third sensitivityanalysis are presented in table 4. These analysesshowed, for most outcomes, substantial differences in effect size between studies in which pressure generation was ensured and those in which it was not ensured. The moderate to large effect sizes seen in the former studies reach conventionallevels of statistical significance for respiratory muscle strength and functional status. For these two outcomes the difference in effect size between studies
537
RESPlRAlORY MUSCLE TRAINING IN CHRONIC AIRFLOW LIMITATION
TABLE 3 PRIMARY RESULTS OF ANALYSIS No. of Studies
Outcome Measure FEV, Vital capacity Respiratory muscle strength (maximum inspiratory pressures) Respiratory muscle strength (maximum voluntary ventilation) Respiratory muscle endurance Laboratory exercise capacity Functional exercise capacity Functional status (quality of life)
Effect Size (standard deviation units)
95% Confidence Interval (lower bound)
95% Confidence Interval (upper bound)
Effect Size (natural units)
p Value
Homogeneity p Value
8 7 11
0.12 0.12 0.15
-0.15 -0.16 -0.09
0.39 0.41 0.39
41 ml 82 ml 3.0 cm
0.38 0.40 0.23
0.18 0.18 0.21
7
0.43
0.07
0.80
8.8 L
0.02
0.28
9
0.22
-0.03
0.48
0.09
0.33
9
-0.04
-0.22
0.14
37.3 Llmin (maximum sustained ventilatory capacity) - 0.036 ml/kg/min
0.22
0.04
9
0.20
-0.06
0.45
40.7 m
0.13
0.25
5
0.12
-0.18
0.42
0.67 on dyspnea scale of CRQ in which 0.5 is minimal important difference
0.44
0.04
Definition of abbreviation: CRQ = chronic respiratory questionnaire.
TABLE 4 SENSITIVITY ANALYSIS OF RESISTANCE STUDIES WITH FLOW RATES CONTROLLED VERSUS THOSE WITH FLOW RATES NOT CONTROLLED
Variable/Condition Respiratory muscle strength (Plmax). controlled flow rate Respiratory muscle strength (Pl mox) . uncontrolled flow rate Respiratory muscle endurance. controlled flow rate Respiratory muscle endurance. uncontrolled flow rate Laboratory exercise capacity. controlled flow rate Laboratory exercise capacity. uncontrolled flow rate Functional exercise capacity. controlled flow rate Functional exercise capacity. uncontrolled flow rate Functional status. controlled flow rate Functional status. uncontrolled flow rate
No. of Studies
Effect Size (standard deviation units)
5
0.51
6
-0.09
4
0.41
Effect Size (natural units)
p Value
Homogeneity p Value
8.2cm
0.Q1
0.92
-1.8 cm
0.57
0.23
10.3 Llmin
0.06
0.93
p Value on Difference in Effect Size between Controlled and Uncontrolled Studies
0.02
0.09 3
-0.08
-14.6 L1min
0.67
0.23
2
-0.002
- 0.02 ml/kg/min
0.99
0.10
5
- 0.17
- 1.57 ml/kg/min
0.10
0.09
3
0.30
88.9 m*
0.22
0.75
4
0.07
18.4 m
0.71
0.05
2
0.65
1.81t
0.02
0.004
3
-0.13
-0.70
0.49
0.92
0.30
0.45
0.02
Definition of abbreviation: Pl max = maximal inspiratory pressure. * 12-minute walk test distance. t Dyspnea rate of Chronic Respiratory Questionnaire in which 0.5 is minimal important difference.
using uncontrolled flowrates versusthose using controlled flow rates reaches conventionallevels of statistical significance. The results of the fourth sensitivity
analysis, which divided studies according to whether or not improvements in strength or endurance were found, showed no effect on laboratory exercise
capacity (effect sizes of 0.0 and 0.03 for studies in which, respectively, strength or endurance did not or did improve). In three studies in which strength and en-
538
durance did not improve, functional exercise capacity was essentially unchanged (effect size 0.11,p = 0.55). In five studies in which strength or endurance did improve a moderate treatment effect on functional exercise capacity was found (effect size 0.57, p = 0.007, homogeneity p = 0.69). The single study in which neither strength nor endurance improved and functional status was measured showed no benefit of training (effect size -0.11). In three studies in which strength or endurance improved and functional status was measured, a small nonsignificant trend in favor of treatment was found (effect size 0.36, p = 0.22). In the final sensitivity analysis, studies receiving high validity scores were compared with studies receiving lower validity scores. In this analysis, some outcomes showed larger differences in the higher quality studies and other outcomes showed larger differences in the lower quality studies. We therefore concluded that study quality was not helpful in explaining differences between study results. Discussion This overview met most of the methodologic criteria that have been suggested for research overviews (35-39). Specifically, explicit inclusion and exclusion criteria were developed, the methodologic quality of the articles included was assessed, the reproducibility of the study selection and assessment criteria were demonstrated, a quantitative analysis was undertaken, and reasons for differences in results between studies were explored. The major limitation of this overview is that common to all overviews: differences in study methodology may render studies sufficiently noncomparable that attempts to combine results across studies become questionable. With respect to the current overview, the effect of muscle training depends on the mode of training and on its duration, frequency, and intensity. These features varied widely across studies. In addition, the optimal training regimen may differ in different disease states or in different stages of the same disease. Despite these limitations, we believe this overview provides important insights into the effects of respiratory muscle training in patients with chronic airflow limitation. When looking across all studies effects on strength and endurance, when present, are small and may be specific to the training regimen. Overall, there is little evidence of important ef-
SMITH, COOK, GUYATT, MADHAVAN, AND OXMAN
fects on outcomes of greater relevance to patients, including functional exercise capacity and quality of life. On the basis of these findings it seems appropriate to conclude that training the respiratory muscles is not, in general, of benefit to patients with chronic airflow limitation. This does not, however, exclude the possibility that a specific training regimen may be of benefit to all patients with chronic airflow limitation, or to a subgroup of these patients. We conducted a number of secondary analyses to explore this possibility. In the sensitivity analysis that separatelyexamined studies using resistance training versus studies using volume training, we found an effect size of 0.67 standard deviation units (p = 0.03) for respiratory muscle endurance in volume training studies. This finding appears to suggest that volume training may be much more effective for increasing respiratory muscle endurance than resistance training. The finding is, however, considerably less impressive when one considers that the endurance outcome, maximal sustained ventilatory capacity, was measured in a manner essentially identical to the training regimen. This strongly suggests an effect that may be specific to the training modality. Another sensitivity analysis suggested that resistance training may result in appreciable improvements in strength and endurance if the breathing pattern is controlled so that substantial pressures are generated during inspiration. The results of this secondary analysis further suggest that when breathing pattern is controlled, increases in respiratory muscle strength and endurance may translate into clinically important improvement in functional status. In two of the studies that we classified as controlling breathing pattern (22, 25), though inspiratory pressure generation was ensured, frequency of breathing was not controlled. In these studies, the intensity of training may have been less than would have been achieved had breathing frequency been controlled. If the intensity of training was suboptimal in these studies, it is possible that training effects would have been greater had breathing frequency been controlled. Criteria for judging the credibility of secondary subgroup analyses that explore possible differences in treatment effect due to differences in patient characteristics or (as in this case) differences in the intervention have been suggested (40). These criteria are relevant to deciding on
the credence that should be given to the hypothesis that clinically important benefit may be obtained by resistance training of respiratory muscles when breathing pattern is controlled. There are a number of factors that support the credibility of the hypothesis. The hypothesis preceded the analysis and was one of only a small number of secondary hypotheses tested. The finding appears consistent across studies, and the magnitude of the effect is moderate in size. In addition, the hypothesis is consistent with our current understanding of the determinants of respiratory muscle performance. The strength of the inference that the difference in effect in the subgroup of studies in which breathing pattern was controlled is different from the overall effect is weakened by the fact that it is supported by differences between (rather than within) studies. The statistical magnitude of the differences is modest; specifically, there are statistically significant differences between studies in which flow rate is controlled versus those in which it is uncontrolled for respiratory muscle strength and functional status, but not for respiratory muscle endurance, laboratory exercise capacity, and functional exercise capacity. In conclusion, this overview provides little support for respiratory muscle training as a treatment strategy for patients with chronic airflow limitation. It does suggest, however, that resistance training programs in which breathing pattern and flow rate are controlled may warrant further investigation.
Acknowledgment The writers acknowledge the authors of the original articles who replied to inquiries and provided additional data beyond that included in the published manuscripts. They also acknowledge William Mcilroy and thank Jim Julian for his assistance with the statistical analysis.
References 1. Chen H, Dukes R, Martin BJ. Inspiratory muscle training in patients with chronic obstructive pulmonary disease. Am RevRespir Dis 1985; 131:251-5. 2. McKeon JL, Dent A, Turner J, et al. The effect of inspiratory resistive training on exercise capacity in optimally treated patients with severe chronic airflow limitation. Aust N Z J Med 1986; 16:648-52. 3. Falk P, Eriksen A, Kolliker K, Andersen J. Relieving dyspnea with the inexpensive and simple method in patients with severechronic airflow limitation. Eur J Respir Dis 1985; 66:181-6. 4. Pardy RL, Rivington RN, Despas PJ, Macklem PT. Inspiratory muscle training compared with physiotherapy in patients with chronic airflow limitation. Am Rev Respir Dis 1981; 123:421-5. 5. Andersen JB, Falk P. Clinical experience with
RESPIRATORY MUSCLE TRAINING IN CHRONIC AIRFLOW LIMITATION
inspiratory resistive breathing training. Int Rehab Med 1984; 6:183-5. 6. Sonne LJ, Davis JA. Increased exercise performance in patients with severe COPD following inspiratory resistive training. Chest 1982; 81:436-8. 7. Braun NMT, Faulkner J, Hughes RL, et af. When should respiratory muscles be exercised? Chest 1983; 84:76-84. 8. Sharp JT. Therapeutic considerations in respiratory muscle function. Chest 1985; 88:118S-23S. 9. Aldrich TK. The application of muscle endurance training techniques to the respiratory muscles in COPD. Lung 1985; 163:15-22. 10. Pardy RL, Reid WD, Belman MJ. Respiratory muscle training. Clin Chest Med 1988;9:287-95. 11. Oxman AD, Guyatt GH. Guidelines for reading literature reviews. Can Med Assoc J 1988; 138:697-703. 12. Cicchetti DV, Fleiss JL. A comparison of the null distributions of weighted kappa and the c ordinal statistic. Appl Psychol Measurement 1977; 1:195-201. 13. Hedges LV, Olkin I. Statistical methods for meta-analysis. New York: Academic Press, Inc., 1985; 108-38. 14. Belman MJ, Thomas SG, Lewis MI. Resistive breathing training in patients with chronic obstructive pulmonary disease. Chest 1986; 90:662-8. 15. Harver A, Mahler DA, Daubenspeck A. Targeted inspiratory muscle training improves respiratory muscle function and reduces dyspnea in patients with chronic obstructive pulmonary disease. Ann Intern Med 1989; 111:117-24. 16. Noseda A, Carpiaux JP, Vandeput W, et af. Resistive inspiratory muscle training and exercise performance in COPD patients. A comparative study with conventional breathing retraining. Bull Eur Physiopathol Respir 1987; 23:457-63. 17. Belman MJ, Shadmehr R. Targeted resistive ventilatory muscle training in chronic obstructive pulmonary disease. J Appl Physiol 1988; 65: 2726-35. 18. Guyatt GH, Keller J, Singer J, et at. A con-
trolled trial of respiratory muscle training. Thorax 1992 (In Press). 19. Levine S, Weiser P, Gillen J. Evaluation of a ventilatory muscle endurance training program in the rehabilitation of patients with chronic obstructive pulmonary disease. Am Rev Respir Dis 1986; 133:400-6. 20. Asher MI, Pardy RL, Coates AL, et af. The effects of inspiratory muscle training in patients with cystic fibrosis. Am Rev Respir Dis 1982; 126:855-9. 21. Mcintosh JM, Berman LB, Goldsmith CH, Jones NL. Voluntary exercisein chronic airflow obstruction. Final Report to the Ministry of Health, DM 545, Toronto, Ontario, Canada 1987. 22. Larson JL, Kim MJ, Sharp JT, Larson DA. Inspiratory muscle training with a pressure threshold breathing device in patients with chronic obstructive pulmonary disease. Am Rev Respir Dis 1988; 138:689-96. 23. McKeon JL, Dent A, Turner J, et af. The effect of inspiratory resistive training on exercise capacity in optimally treated patients with severe chronic airflow limitation. Aust N Z J Med 1986; 16:648-52. 24. Ries AL, Moser KM. Comparison of isocapneic hyperventilation and walking exercise training at home in pulmonary rehabilitation. Chest 1986; 90:285-9. 25. Goldstein R, DeRosie J, Long S, et af. Applicability of a threshold loading device for inspiratory muscle testing and training in patients with COPD. Chest 1989; 96:564-71. 26. Kim MJ, Larson M, Sachs P, Sharp JT. Respiratory muscle training in patients with chronic obstructive pulmonary disease. Am Rev Respir Dis 1984; 129:S129. 27. Sobush DC, Kutty K, Fehring R, et al. The physiologic comparison between singing and resistive inspiratory exercise in persons having advanced chronic airway obstruction. Cardiol Pulmonol Records 1986; 1:19-20. 28. Bjerre-jepsen K, Secher NH, Kok-Jensen A.
539 Inspiratory resistance training in severe chronic obstructive pulmonary disease. Eur J Respir Dis 1981; 62:405-11. 29. Dekhuijzen PNR, Folgering HTM, van Herwaarden CLA. Target-flow inspiratory muscle training during pulmonary rehabilitation in patients with COPD. Chest 1991; 99:128-33. 30. Pardy RL, Rivington RN, Despas PJ, Macklem PT. Inspiratory muscle training compared with physiotherapy in patients with chronic airflow limitation. Am Rev Respir Dis 1981; 123:421-5. 31. Altose MD, Kendis C, Connors AF, Dinacor AF. Comparison of the effects of inspiratory muscle training and physical reconditioning on exercise capacity and dyspnea in chronic obstructive lung disease (abstract). Am Rev Respir Dis 1986; 33:AI02. 32. Sonne LJ. Increased exercise performance in patients with severe chronic obstructive pulmonary disease following inspiratory resistivetraining. Chest 1982; 81:436-9. 33. Martin L. Respiratory msucle function: a clinical study. Heart Lung 1984; 13:346-8. 34. Cohen J. Statistical power analysis for the behavioural sciences. New York: Academic Press. 1977; 24-7. 35. Mulrow CD. The medical review article: state of the science. Ann Intern Med 1987; 106:485-8. 36. Sacks H8, Berrier J, Reitman D, Ancona-Berk VA,Chalmers C. Meta-analyses ofrandornized controlled trials. N Eng J Med 1987; 316:450-5. 37. L'Abbe KA, Detsky AS, Orourke K. Metaanalysis in clinical research. Ann Intern Med 1987; 107:224-33. 38. Gerbarg ZB, Horwitz RI. Resolving conflicting clinical trials: guidelines for meta-analysis. J Clin Epidemiol 1988; 41:503-9. 39. Squires BP. Biomedical review articles: what editors want from authors and peer reviewers. Can Med Assoc J 1989; 141:195-7. 40. Oxman AD, Guyatt GH. A consumer's guide to subgroup analyses. Ann Intern Med (In Press).
KAREN SMITH, DEBORAH COOK, GORDON H. GUYATT, JOSHI MADHAVAN, and ANDREW D. OXMAN
Introduction
Respiratory muscle weakness appears to contribute to exerciselimitation in patients with chronic airflow limitation. It has therefore been suggestedthat strength or endurance training may, by strengthening the muscles of respiration, ameliorate dyspnea and improve exercise capacity (1-6). A number of randomized trials have been conducted to test this hypothesis, first by determining whether training can actually strengthen muscles, and second whether stronger respiratory musculature can improve exerciseperformance and day-to-day function. The results of these trials have been conflicting. Although a number of reviews of these studies have been published (7-10), the reviews all suffer from major methodologic limitations (11). These include incomplete identification of relevant studies, lack 0 f explicitcriteria regarding how studies were selected and appraised, and the absence of a quantitative analysis. Therefore, to clarify the issue of whether respiratory muscle training is of benefit to patients with chronic airflow limitation, we conducted a systematic quantitative overview of the randomized trials that have addressed this issue. Methods Does training the respiratory muscles improve spirometry, ventilatory muscle strength and endurance, exercise capacity and functional status in patients with chronic airflow limitation?
Study Identification A MEDLINE (National Library of Medicine, Bethesda, MD) search was conducted; it combined four concepts: (l) obstructive lung disease; (2) breathing exercises or exerciseor exertion or physical therapy or respiratory function tests; (3) random allocation or clinical trials or double-blind method; (4) human. Reference lists of all articles obtained were reviewed, and any additional possibly relevant citations were retrieved. Key references were searched forward in time using SCISEARCH (Institute for Scientific Informa-
SUMMARY The purpose of this study was to determine the effect of respiratory muscle training on muscle strength and endurance, exercise capacity, and functional status In patients with chronic airflow limitation. Computarlzed bibliographic data bases (MEDLINEAND SCISEARCH)were searched for published clinical trails, and an Independent revl_ of 73 articles by two of the Investi. gators Identified 17relevant randomized trials for Inclusion. Study quality wasassessed and descriptive Information concerning the study popUlations, Interventions, and ou1come measurements wes extracted. We combined effect sizes across studies (the difference between treatment and control groups divided by the pooled standard deviation of the outcome measure). Across all studies, the effect sizes and associated p-Yalues were as follows: maximal Inspiratory pressure 0.12, p .. 0.38; maximal voluntary ventilation 0.43, p = 0.02; respiratory muscle endurance 0.21, p = 0.14; laborato· ry exercise capacity -0.01, p = 0.43; functional exercise capacity 0.20, p = 0.15; functional status 0.06, p = 0.72. Secondary analyses suggested that endurance and function may be Improved if resistance training with control of breathing pattern is undertaken. Overall, there Is little evidence of clinically Important benefit of respiratory muscle training In patients with chronic airflow IImlta· tlon. The possibility that benefit may result If resistance training Is conducted in a fashion that ensures generation of adequate mouth pressures may be worthy of further stUdy. AM REV RESPIR DIS 1992; 145:533-539
tion, Philadelphia, PA). A comprehensive list of relevant articles was constructed and a letter was sent to the first author of each relevant article published in the last 5 yr asking for knowlege of any additional published or unpublished studies.
Study Selection The following criteria were used in selecting studies for inclusion in the overview. Studies included were those that had patients with chronic airflow limitation as a target population. Specific criteria included a clinical diagnosis of chronic airflow limitation and one of the following: (1) best recorded FEV,/FVC ratio of each subject < 0.7; (2) best recorded FEV, of each subject < 70070 of predicted; (3) mean FEV, of the group < 35% of predicted; (4) mean FEV,/mean FVC ratio of the group < 0.5. Only studies using respiratory muscle training, specifically resistive breathing exercises or isocapneic hyperventilation, and randomized control trials were considered. Studies were included if any of pulmonary function, respiratory muscle strength, respiratory muscle endurance, laboratory exercise capacity, functional exercise capacity, or functional status were measured. Initially, these criteria were applied by two of us (K.S. and J .M.) to the titles of all citations obtained. If an article's title suggested any possibility that it might be relevant, it was retrieved. The complete article was reviewed by three of us (K.S., J.M. and G.H.G.) and its relevance judged.
Study Evaluation and Data Extraction The methodologic quality of each study was evaluated by two reviewers(K.S. and G.H.G.) using the criteria listed in table 1. Based on this review, a total "methodologic quality" score was calculated. Agreement regarding each criterion was evaluated by weighted kappa. Relevant information regarding population, intervention, and outcome was abstracted from each relevant article by two reviewers (K.S.and D.J.C.).Disagreementsregarding relevance, methodologic quality, or information regarding population, intervention, or outcome were resolved by discussion between reviewers. We contacted the authors of all relevant papers and abstracts. Wesent the authors our summaries of the validity of their studies, and (Received in original form March 11, 1991 and in revised form September 27, 1991) i From the Departments of Medicine, Clinical Epidemiology and Biostatistics, and Family Medicine, McMaster University, Hamilton, Ontario, Canada, and the College of Medicine,'Irivandrum, Kerala State, India. 2 Dr. Cook is a career scientist of the Ontario Ministryof Health and a Scholar of the St. Joseph's Hospital Foundation; Dr. Guyatt is a career scientist of the Ontario Ministry of Health. 3 Correspondence and requests for reprints should be addressed to Karen Smith, M.D., MeMaster University, Department of Medicine, Chedoke Division, Holbrook Blvd., Room 90, Box 2000, Hamilton, Ontario L8N 3Z5, Canada
533
534
SMITH, COOK, GUYATT, MADHAVAN, AND OXMAN
TABLE 1 CRITERIA FOR METHODOLOGIC QUALITY' 1. Sample Random or consecutive sample (5) Arbitrary sample (or can't tell) (0) 2. Similarity of groups Number of following variable in which groups were comparable: age, gender forced expired volume, maximum inspiratory pressures, walk test distance (0-5) 3. Cointervention Number of following criteria met: comparable frequency of visits, comparable medication changes, comparable number of intercurrent illnesses (5 for 313; 4 for 213; 3 for 113; 0 for 0/3) 4. Sham or masking Sham or masking undertaken (5) No sham or masking (0) 5. Compliance Training was hospital supervised (5) Home program with reporting diary and patients periodically (4) Home program with either diary or periodic review (3) Compliance not measured or cannot tell (0) 6. Observer masking Observers masked as to treatment groups (5) Not masked or cannot tell (0) 7. Standardized testing Encouragement standardized (5) Not standardized or cannot tell (0) 8. Follow-up 90 to 1000Al follow-up (5) 80 to 89% follow-up (3) < 80% subjects accounted for (1) cannot tell (0) • Numbers in parentheses indicate number of points in summary score.
of the data wehad extracted, and asked them to ensure their accuracy. In addition, we asked the researchers to provide important additional information missing from their reports. Agreement between observers for study inclusion and assessment was evaluated using weighted kappa with quadratic weights (12).
Analysis The analysis was based on examination of the effect size for each individual study, where the effect was defined as the difference between treatment and control groups after treatment divided by the pooled standard deviation of the posttreatment outcome measure in the treatment and control groups. Effect sizes were weighted by the inverse of the standard deviation and combined to compute a weighted mean effect size. The methods used to test for statistical significance and for homogeneity were those described by Hedges and Olkin (13). For each outcome we reported the effect size and the confidence interval around the effect size in standard deviation units. We also reported the effect size in natural units. For outcomes in which different investigators chose different measures (for instance, maximal sustained ventilatory capacity [MSVC]versus breathing against resistance to test respiratory muscle endurance) to obtain a result in natural units, we chose the most commonly used measure (i.e., the one most likely to have intuitive meaning for the clinician) and converted the effect size back into natural units. When different measures related to laboratory exercise capacity were reported (for instance, peak workload, maximum ox-
ygen uptake, and submaximal exercise capacity all reported in the same study), a single effect size was calculated across these outcomes and that effect size used in the overall analysis. This effect size was derived in exactly the same manner as the effect sizesacross studies; that is the difference between treatment and control groups after treatment was divided by the pooled standard deviation of the posttreatment outcome measure in the treatment and control groups, and effect sizes were weighted by the inverse of the standard deviation and combined to compute a weighted mean effect size. In the sensitivity analysis of studies in which the breathing pattern was controlled versus studies in which it was not controlled, the likelihood of difference in effect size occurring as a result of the play of chance was calculated using a test of homogeneity in which the test statistic approximates the chi-square distribution. A priori Hypotheses Regarding Sources of Heterogeneity When conducting a meta-analysis, heterogeneity (that is, major differences in the apparent effect of the interventions across studies) is often found. When heterogeneity is present, it must be explained (11). Before analyzing the results, we developed a number of hypotheses concerning underlying differences in the studies which might explain heterogeneity. First, we speculated that increases in respiratory muscle strength or endurance might be restricted to studies in which training and testing occurred on the same apparatus. We therefore planned a separate analysis
for studies in which the training and testing apparatus were the same versus studies in which they were different. Second, we noted that there were two fundamentally different training strategies. One was based on volume (that is, having subjects breathe at a relatively high proportion of their maximum voluntary ventilation for appreciable periods of time) and the other was based on having subjects inspire against resistance. We speculated that these two strategies may have different effects on our outcomes. Third, we have been impressed by the recent findings reported by Belman and colleagues suggesting that ability to tolerate increasing respiratory resistance can be achieved by changes in breathing pattern, rather than by strengthening respiratory muscles (14). We therefore looked separately at studies in which the patient was required to target during training to achieve a specified flow rate. Fourth, if strengthening of the respiratory muscles has not been achieved, one cannot expect changes in exerciseperformance, functional capacity, or quality of life. We therefore conducted a separate analysis looking at clinically relevant endpoints restricted to studies in which improved respiratory muscle function had been achieved. We defined studies in which there was improved respiratory muscle function as those in which either strength or endurance improved by 0.4 or more standard deviation units. Finally, we hypothesized that results may be related to study quality. Wetherefore compared effect sizes in studies with a summary quality score of less than 20 from our assessment of methodologic quality to those with a score of 20 or greater.
Results
A total of 1,085citations wereidentified; 775 from MEDLINE, 65 from SCISEARCH, and 245 from our personal files, reviewof reference lists or from direct communications with researchers.Of these 1,085 citations, reviewers agreed that 73 were potentially eligible. The weighted kappa on the agreement between the two reviewers in identifying potentially eligible citations was 0.77 (95070 confidence interval of 0.74, 0.80). Of the potentially eligible citations, 17 proved eligible on detailed review (1, 3, 15-29). Twoofthese wereobtained from direct communication with researchers, the data not having been published (18, 21). Reasons for exclusion 0 f potentially relevant articles included patients not meeting eligibility criteria, the intervention proving not directed at respiratory muscle training, and patients not randomized to treatment and control groups. In four reports (30-33) whether the study was actually randomized was unclear from the published description. In each
535
RESPIRATORY MUSCLE TRAINING IN CHRONIC AIRFLOW LIMITATION
case, extensive attempts were made to contact the investigators. True randomization wasnot undertaken in two of these studies; in the other two, we were unable to obtain the information. The weighted kappa on the agreement between two reviewers regarding the application of inclusion criteria to potentially relevant reports was 0.66 (95070 confidence interval 0.59,0.73). A kappa of 0.66 is conventionally considered "good agreement." Authors of 14 of the relevant reports responded to our requests for checking the accuracy of the information we abstracted and for providing additional data. Each reply provided helpful additional information. The 17relevant studies are summarized in table 2. Weighted kappa on agreement for the validity criteria varied from 0.69
to 1.0.Ofthe 17studies, information that could be used in the data analysis was available in 13. Other studies provided only p values or mean scores without associated standard deviations. Of the 13 studies included in the meta-analysis, 11 studied the effects of resistive training and two studied the effect of volume training. Of the 11 that studied the effects of resistance training, the flow rates (and thus the resistance) generated during training were controlled by the investigators in four. The results of the overall statistical analyses are summarized in table 3. The interpretation of the effect sizes is aided by a rule of thumb suggested by Cohen: an effect size of 0.2 standard deviation units represents a small effect size, 0.5 a medium effect size, and 0.8 a large effect size (34). The interpretation of the
results is further aided by noting the effect size in natural units. The analyses shows small nonsignificant trends in favor of respiratory muscle training for most outcomes. The best estimate of the effect size is in all cases small. A small to moderate effect size (0.43 standard deviation units) was found for respiratory muscle strength, as measured by maximum voluntary ventilation. This corresponds to 8.8 L difference in natural units (confidence interval 1.2 to 14.4 L), and reaches conventional levels of statistical significance (p = 0.02). Statistically significant heterogeneity was found in two analyses: laboratory exercise capacity and functional status. It could be argued that for the other outcomes, further exploration in terms of the a priori sensitivity analyses described above is not necessary (or even appro-
TABLE 2 SUMMARY OF SALIENT CHARACTERISTICS OF EACH STUDY
Reference
Type of Training
Sample Size
Masking Patients/Study Personnel
Duration of Follow-up
Outcome Measures Reported*t
Overall Methodology Score
Chen, 1985 (1)
Resistance uncontrolled
Training: 7 Control: 6
Yes/no
4 wk
FEV" FVC RMS: Plmax, MW RME: endurance time at 60% PIMMaxg LEC: maximal workload. peak vo.
24
Falk, 1985 (3)
Resistance uncontrolled
Training: 12 Control: 15
Yes/yes
2 months
Trend toward greater improvement in exercise capacity and more reduction in dyspnea in trained patients
30
Harver, 1989 (15)
Resistance controlled
Training: 10 Control: 9
No/no
8 wk
FEV" FVC. FRC RMS: Plmax• MW FC/OOL: transition dyspnea index
32
Noseda, 1987 (16)
Resistance uncontrolled
Training: 10 Control: 10
No/no
8 wk
FEV,. FRC RME: highest p-flex level tolerated for 10 minutes LEC: maximal workload. peak Vo. FEC: 12-minute walk
19
Belman. 1988 (17)
Resistance controlled
High intensity: 8 Low intensity: 9
No/no
6 wk
FEV" FVC. FRC RMS: Pl max, MW RME:MSVC
19
Guyatt, 1991 (18)
Resistance uncontrolled
Training: 43 Control: 39
Yes/yes
6 months
FEV" MSVC RMS: Pl max RME: progressive endurance test LEC: exercise duration FEC: 6-minute walk FS/OOL: CRO
38
Levine, 1986 (19)
Volume
Training: 17 Control: 15
No/no
6 wk
FEV" FVC, FRC RMS: MW RME:MSVC LEC: maximal workload, peak FEC: 12-minute walk test
38
Vo.
Asher, 1982 (20)
Resistance uncontrolled
Crossover 11 patients
No/yes
1 month each
RMS: Pl max LEC: maximal workload, peak endurance at 2/3 maximal workload
10
Mcintosh. 1987 (21)
Resistance uncontrolled
Training: 6 Control: 8
Yes/no
3 months
RMS: Pl max LEC: maximal workload, peak FEC: 6·minute walk test FS/OOL: CRO
36
vo,
vo,
(continued)
536
SMITH, COOK, GUYATT, MADHAVAN, AND OXMAN
TABLE 2 CONTINUED
Reference
Sample Size
Type of Training
Masking Patients/Study Personnel
Duration of Follow-up
Outcome Measures Reported*t
Overall Methodology Score
Larson, 1988 (22)
Resistance controlled
High intensity: 10 Low intensity: 11
Yes/yes
2 months
RMS: Plmax RME: Endurance time at 66% of PIM max FEC: 12-minute walk
34
McKeon, 1986 (23)
Resistance uncontrolled
Training: 11 Control: 11
Yes/no
6 wk
Trend toward greater increase in Plmax in control, stair climbing in training group; changes in laboratory and functional exercise capacity comparable
28
Reis, 1986 (24)
Volume
Training: 5 Walking: 7
No/no
6 wk
RMS: MW RME: MSVC LEC: peak Vo.; endurance at submaximal workload FEC: 12-minute walk
21
Goldstein, 1989 (25)
Resistance controlled
Training: 6 Control: 5
No/no
4 wk
RMS: Plmax, MW RME: endurance time at fixed load LEC: endurance time at submaximal load FEC: 6-minute walk
27
Kim, 1984 (26)
Resistance uncontrolled
Crossover eight patients; incentive spirometry with or without resistance
Yes/yes
4 wk
Plmax were higher after resistance device than control in group of 8 patients who trained with resistance first; not so for other eight; no difference in 12-minute walk distance or activities of daily living
Sobush, 1986
Resistance uncontrolled
Training: 10 Control: 9
No/no
16 wk
FEV" FVC RMS: Plmax, MW FS/OOL: physical function scale of Sickness Impact Profile
22
Bjerre.Jepson, 1981 (28)
Resistance uncontrolled
Training: 14 Control: 14
No/no
6 wk
No appreciable improvement in MW in either group; RME increased comparably in both groups as did stair-climbing ability
27
Dekhuijzen, 1991 (29)
Resistance controlled
Training: 20 Control: 20
No/no
6 wk
RMS: Plmax RME: endurance of static inspiratory maneuvers FEC: 12-minute walk FS/QOL: activities of daily living LEC: VO.; maximal workload
13
(27)
8
Definitionof abbreviations: RMS - raspiratory musclestrength; Plmax = maximalinspiratory pressure; MW = maximalvoluntary ventilation; RME - respiratory muscleendurance; PIMmax - maximalInspiratory mouthpressure; LEC = laboratory exercise capacity;Vo, = oxygen uptake; FSlQOL = functional capacity/quality of life; FEC = functional exercise capacity; MSVC _ maximalsustained ventilatory capacity; CRO - chronic respiratory questionnaire. * Outcomes listed include only those with data usablein meta-analysis. t For studiesIn which no date suitable for analysis was reported, outcomes are described.
priate), However, before concluding definitively that respiratory muscle training was of little use, we felt that exploration of the between-study differences was warranted. The first hypothesis, that significant effects might only be seen when the training and testing devices were similar (and effects may thus be specific to the training modality) was generated to guard against a false positive result. It is therefore relevant only to maximum voluntary ventilation. Effect sizes were actually greater in studies in which resistance training was used (0.51 standard deviation units) than in studies using volume training (0.29 standard deviation units) in which similar training and testing
devices could be implicated. This analysis, therefore, raises no concerns about the validity of the finding of increased maximum voluntary ventilation. In addition to maximal voluntary ventilation, data are available for comparison of the effects of volume versus resistance training for respiratory muscle endurance, laboratory exercise capacity, and functional exercise capacity. For respiratory muscle endurance, the effect size was 0.09 standard deviation units for resistance training (p = 0.56) and 0.67 standard deviation units for volume training (p = 0.03). For laboratory exercise capacity, the results for both resistance and volume training, like the overall analysis, showed
negligible effects. For functional exercise capacity the effect size for volume training studies is 0.42 standard deviation units (p = 0.17); for resistance studies the effect is 0.14 standard deviation units (p = 0.64). The results of the third sensitivityanalysis are presented in table 4. These analysesshowed, for most outcomes, substantial differences in effect size between studies in which pressure generation was ensured and those in which it was not ensured. The moderate to large effect sizes seen in the former studies reach conventionallevels of statistical significance for respiratory muscle strength and functional status. For these two outcomes the difference in effect size between studies
537
RESPlRAlORY MUSCLE TRAINING IN CHRONIC AIRFLOW LIMITATION
TABLE 3 PRIMARY RESULTS OF ANALYSIS No. of Studies
Outcome Measure FEV, Vital capacity Respiratory muscle strength (maximum inspiratory pressures) Respiratory muscle strength (maximum voluntary ventilation) Respiratory muscle endurance Laboratory exercise capacity Functional exercise capacity Functional status (quality of life)
Effect Size (standard deviation units)
95% Confidence Interval (lower bound)
95% Confidence Interval (upper bound)
Effect Size (natural units)
p Value
Homogeneity p Value
8 7 11
0.12 0.12 0.15
-0.15 -0.16 -0.09
0.39 0.41 0.39
41 ml 82 ml 3.0 cm
0.38 0.40 0.23
0.18 0.18 0.21
7
0.43
0.07
0.80
8.8 L
0.02
0.28
9
0.22
-0.03
0.48
0.09
0.33
9
-0.04
-0.22
0.14
37.3 Llmin (maximum sustained ventilatory capacity) - 0.036 ml/kg/min
0.22
0.04
9
0.20
-0.06
0.45
40.7 m
0.13
0.25
5
0.12
-0.18
0.42
0.67 on dyspnea scale of CRQ in which 0.5 is minimal important difference
0.44
0.04
Definition of abbreviation: CRQ = chronic respiratory questionnaire.
TABLE 4 SENSITIVITY ANALYSIS OF RESISTANCE STUDIES WITH FLOW RATES CONTROLLED VERSUS THOSE WITH FLOW RATES NOT CONTROLLED
Variable/Condition Respiratory muscle strength (Plmax). controlled flow rate Respiratory muscle strength (Pl mox) . uncontrolled flow rate Respiratory muscle endurance. controlled flow rate Respiratory muscle endurance. uncontrolled flow rate Laboratory exercise capacity. controlled flow rate Laboratory exercise capacity. uncontrolled flow rate Functional exercise capacity. controlled flow rate Functional exercise capacity. uncontrolled flow rate Functional status. controlled flow rate Functional status. uncontrolled flow rate
No. of Studies
Effect Size (standard deviation units)
5
0.51
6
-0.09
4
0.41
Effect Size (natural units)
p Value
Homogeneity p Value
8.2cm
0.Q1
0.92
-1.8 cm
0.57
0.23
10.3 Llmin
0.06
0.93
p Value on Difference in Effect Size between Controlled and Uncontrolled Studies
0.02
0.09 3
-0.08
-14.6 L1min
0.67
0.23
2
-0.002
- 0.02 ml/kg/min
0.99
0.10
5
- 0.17
- 1.57 ml/kg/min
0.10
0.09
3
0.30
88.9 m*
0.22
0.75
4
0.07
18.4 m
0.71
0.05
2
0.65
1.81t
0.02
0.004
3
-0.13
-0.70
0.49
0.92
0.30
0.45
0.02
Definition of abbreviation: Pl max = maximal inspiratory pressure. * 12-minute walk test distance. t Dyspnea rate of Chronic Respiratory Questionnaire in which 0.5 is minimal important difference.
using uncontrolled flowrates versusthose using controlled flow rates reaches conventionallevels of statistical significance. The results of the fourth sensitivity
analysis, which divided studies according to whether or not improvements in strength or endurance were found, showed no effect on laboratory exercise
capacity (effect sizes of 0.0 and 0.03 for studies in which, respectively, strength or endurance did not or did improve). In three studies in which strength and en-
538
durance did not improve, functional exercise capacity was essentially unchanged (effect size 0.11,p = 0.55). In five studies in which strength or endurance did improve a moderate treatment effect on functional exercise capacity was found (effect size 0.57, p = 0.007, homogeneity p = 0.69). The single study in which neither strength nor endurance improved and functional status was measured showed no benefit of training (effect size -0.11). In three studies in which strength or endurance improved and functional status was measured, a small nonsignificant trend in favor of treatment was found (effect size 0.36, p = 0.22). In the final sensitivity analysis, studies receiving high validity scores were compared with studies receiving lower validity scores. In this analysis, some outcomes showed larger differences in the higher quality studies and other outcomes showed larger differences in the lower quality studies. We therefore concluded that study quality was not helpful in explaining differences between study results. Discussion This overview met most of the methodologic criteria that have been suggested for research overviews (35-39). Specifically, explicit inclusion and exclusion criteria were developed, the methodologic quality of the articles included was assessed, the reproducibility of the study selection and assessment criteria were demonstrated, a quantitative analysis was undertaken, and reasons for differences in results between studies were explored. The major limitation of this overview is that common to all overviews: differences in study methodology may render studies sufficiently noncomparable that attempts to combine results across studies become questionable. With respect to the current overview, the effect of muscle training depends on the mode of training and on its duration, frequency, and intensity. These features varied widely across studies. In addition, the optimal training regimen may differ in different disease states or in different stages of the same disease. Despite these limitations, we believe this overview provides important insights into the effects of respiratory muscle training in patients with chronic airflow limitation. When looking across all studies effects on strength and endurance, when present, are small and may be specific to the training regimen. Overall, there is little evidence of important ef-
SMITH, COOK, GUYATT, MADHAVAN, AND OXMAN
fects on outcomes of greater relevance to patients, including functional exercise capacity and quality of life. On the basis of these findings it seems appropriate to conclude that training the respiratory muscles is not, in general, of benefit to patients with chronic airflow limitation. This does not, however, exclude the possibility that a specific training regimen may be of benefit to all patients with chronic airflow limitation, or to a subgroup of these patients. We conducted a number of secondary analyses to explore this possibility. In the sensitivity analysis that separatelyexamined studies using resistance training versus studies using volume training, we found an effect size of 0.67 standard deviation units (p = 0.03) for respiratory muscle endurance in volume training studies. This finding appears to suggest that volume training may be much more effective for increasing respiratory muscle endurance than resistance training. The finding is, however, considerably less impressive when one considers that the endurance outcome, maximal sustained ventilatory capacity, was measured in a manner essentially identical to the training regimen. This strongly suggests an effect that may be specific to the training modality. Another sensitivity analysis suggested that resistance training may result in appreciable improvements in strength and endurance if the breathing pattern is controlled so that substantial pressures are generated during inspiration. The results of this secondary analysis further suggest that when breathing pattern is controlled, increases in respiratory muscle strength and endurance may translate into clinically important improvement in functional status. In two of the studies that we classified as controlling breathing pattern (22, 25), though inspiratory pressure generation was ensured, frequency of breathing was not controlled. In these studies, the intensity of training may have been less than would have been achieved had breathing frequency been controlled. If the intensity of training was suboptimal in these studies, it is possible that training effects would have been greater had breathing frequency been controlled. Criteria for judging the credibility of secondary subgroup analyses that explore possible differences in treatment effect due to differences in patient characteristics or (as in this case) differences in the intervention have been suggested (40). These criteria are relevant to deciding on
the credence that should be given to the hypothesis that clinically important benefit may be obtained by resistance training of respiratory muscles when breathing pattern is controlled. There are a number of factors that support the credibility of the hypothesis. The hypothesis preceded the analysis and was one of only a small number of secondary hypotheses tested. The finding appears consistent across studies, and the magnitude of the effect is moderate in size. In addition, the hypothesis is consistent with our current understanding of the determinants of respiratory muscle performance. The strength of the inference that the difference in effect in the subgroup of studies in which breathing pattern was controlled is different from the overall effect is weakened by the fact that it is supported by differences between (rather than within) studies. The statistical magnitude of the differences is modest; specifically, there are statistically significant differences between studies in which flow rate is controlled versus those in which it is uncontrolled for respiratory muscle strength and functional status, but not for respiratory muscle endurance, laboratory exercise capacity, and functional exercise capacity. In conclusion, this overview provides little support for respiratory muscle training as a treatment strategy for patients with chronic airflow limitation. It does suggest, however, that resistance training programs in which breathing pattern and flow rate are controlled may warrant further investigation.
Acknowledgment The writers acknowledge the authors of the original articles who replied to inquiries and provided additional data beyond that included in the published manuscripts. They also acknowledge William Mcilroy and thank Jim Julian for his assistance with the statistical analysis.
References 1. Chen H, Dukes R, Martin BJ. Inspiratory muscle training in patients with chronic obstructive pulmonary disease. Am RevRespir Dis 1985; 131:251-5. 2. McKeon JL, Dent A, Turner J, et al. The effect of inspiratory resistive training on exercise capacity in optimally treated patients with severe chronic airflow limitation. Aust N Z J Med 1986; 16:648-52. 3. Falk P, Eriksen A, Kolliker K, Andersen J. Relieving dyspnea with the inexpensive and simple method in patients with severechronic airflow limitation. Eur J Respir Dis 1985; 66:181-6. 4. Pardy RL, Rivington RN, Despas PJ, Macklem PT. Inspiratory muscle training compared with physiotherapy in patients with chronic airflow limitation. Am Rev Respir Dis 1981; 123:421-5. 5. Andersen JB, Falk P. Clinical experience with
RESPIRATORY MUSCLE TRAINING IN CHRONIC AIRFLOW LIMITATION
inspiratory resistive breathing training. Int Rehab Med 1984; 6:183-5. 6. Sonne LJ, Davis JA. Increased exercise performance in patients with severe COPD following inspiratory resistive training. Chest 1982; 81:436-8. 7. Braun NMT, Faulkner J, Hughes RL, et af. When should respiratory muscles be exercised? Chest 1983; 84:76-84. 8. Sharp JT. Therapeutic considerations in respiratory muscle function. Chest 1985; 88:118S-23S. 9. Aldrich TK. The application of muscle endurance training techniques to the respiratory muscles in COPD. Lung 1985; 163:15-22. 10. Pardy RL, Reid WD, Belman MJ. Respiratory muscle training. Clin Chest Med 1988;9:287-95. 11. Oxman AD, Guyatt GH. Guidelines for reading literature reviews. Can Med Assoc J 1988; 138:697-703. 12. Cicchetti DV, Fleiss JL. A comparison of the null distributions of weighted kappa and the c ordinal statistic. Appl Psychol Measurement 1977; 1:195-201. 13. Hedges LV, Olkin I. Statistical methods for meta-analysis. New York: Academic Press, Inc., 1985; 108-38. 14. Belman MJ, Thomas SG, Lewis MI. Resistive breathing training in patients with chronic obstructive pulmonary disease. Chest 1986; 90:662-8. 15. Harver A, Mahler DA, Daubenspeck A. Targeted inspiratory muscle training improves respiratory muscle function and reduces dyspnea in patients with chronic obstructive pulmonary disease. Ann Intern Med 1989; 111:117-24. 16. Noseda A, Carpiaux JP, Vandeput W, et af. Resistive inspiratory muscle training and exercise performance in COPD patients. A comparative study with conventional breathing retraining. Bull Eur Physiopathol Respir 1987; 23:457-63. 17. Belman MJ, Shadmehr R. Targeted resistive ventilatory muscle training in chronic obstructive pulmonary disease. J Appl Physiol 1988; 65: 2726-35. 18. Guyatt GH, Keller J, Singer J, et at. A con-
trolled trial of respiratory muscle training. Thorax 1992 (In Press). 19. Levine S, Weiser P, Gillen J. Evaluation of a ventilatory muscle endurance training program in the rehabilitation of patients with chronic obstructive pulmonary disease. Am Rev Respir Dis 1986; 133:400-6. 20. Asher MI, Pardy RL, Coates AL, et af. The effects of inspiratory muscle training in patients with cystic fibrosis. Am Rev Respir Dis 1982; 126:855-9. 21. Mcintosh JM, Berman LB, Goldsmith CH, Jones NL. Voluntary exercisein chronic airflow obstruction. Final Report to the Ministry of Health, DM 545, Toronto, Ontario, Canada 1987. 22. Larson JL, Kim MJ, Sharp JT, Larson DA. Inspiratory muscle training with a pressure threshold breathing device in patients with chronic obstructive pulmonary disease. Am Rev Respir Dis 1988; 138:689-96. 23. McKeon JL, Dent A, Turner J, et af. The effect of inspiratory resistive training on exercise capacity in optimally treated patients with severe chronic airflow limitation. Aust N Z J Med 1986; 16:648-52. 24. Ries AL, Moser KM. Comparison of isocapneic hyperventilation and walking exercise training at home in pulmonary rehabilitation. Chest 1986; 90:285-9. 25. Goldstein R, DeRosie J, Long S, et af. Applicability of a threshold loading device for inspiratory muscle testing and training in patients with COPD. Chest 1989; 96:564-71. 26. Kim MJ, Larson M, Sachs P, Sharp JT. Respiratory muscle training in patients with chronic obstructive pulmonary disease. Am Rev Respir Dis 1984; 129:S129. 27. Sobush DC, Kutty K, Fehring R, et al. The physiologic comparison between singing and resistive inspiratory exercise in persons having advanced chronic airway obstruction. Cardiol Pulmonol Records 1986; 1:19-20. 28. Bjerre-jepsen K, Secher NH, Kok-Jensen A.
539 Inspiratory resistance training in severe chronic obstructive pulmonary disease. Eur J Respir Dis 1981; 62:405-11. 29. Dekhuijzen PNR, Folgering HTM, van Herwaarden CLA. Target-flow inspiratory muscle training during pulmonary rehabilitation in patients with COPD. Chest 1991; 99:128-33. 30. Pardy RL, Rivington RN, Despas PJ, Macklem PT. Inspiratory muscle training compared with physiotherapy in patients with chronic airflow limitation. Am Rev Respir Dis 1981; 123:421-5. 31. Altose MD, Kendis C, Connors AF, Dinacor AF. Comparison of the effects of inspiratory muscle training and physical reconditioning on exercise capacity and dyspnea in chronic obstructive lung disease (abstract). Am Rev Respir Dis 1986; 33:AI02. 32. Sonne LJ. Increased exercise performance in patients with severe chronic obstructive pulmonary disease following inspiratory resistivetraining. Chest 1982; 81:436-9. 33. Martin L. Respiratory msucle function: a clinical study. Heart Lung 1984; 13:346-8. 34. Cohen J. Statistical power analysis for the behavioural sciences. New York: Academic Press. 1977; 24-7. 35. Mulrow CD. The medical review article: state of the science. Ann Intern Med 1987; 106:485-8. 36. Sacks H8, Berrier J, Reitman D, Ancona-Berk VA,Chalmers C. Meta-analyses ofrandornized controlled trials. N Eng J Med 1987; 316:450-5. 37. L'Abbe KA, Detsky AS, Orourke K. Metaanalysis in clinical research. Ann Intern Med 1987; 107:224-33. 38. Gerbarg ZB, Horwitz RI. Resolving conflicting clinical trials: guidelines for meta-analysis. J Clin Epidemiol 1988; 41:503-9. 39. Squires BP. Biomedical review articles: what editors want from authors and peer reviewers. Can Med Assoc J 1989; 141:195-7. 40. Oxman AD, Guyatt GH. A consumer's guide to subgroup analyses. Ann Intern Med (In Press).
