Prophylactic respiratory physiotherapy after cardiac surgery: systematic reviewBMJ 2003; 327 doi: http://dx.doi.org/10.1136/bmj.327.7428.1379 (Published 11 December 2003) Cite this as: BMJ 2003;327:1379
- 1Division of Surgical Intensive Care, Department of Anaesthesiology, Pharmacology and Surgical Intensive Care, Geneva University Hospitals, CH-1211 Geneva 14, Switzerland
- 2Division of Anaesthesiology, Department of Anaesthesiology, Pharmacology and Surgical Intensive Care, Geneva University Hospitals
- Correspondence to: P Pasquina
- Accepted 16 October 2003
Objective To assess whether respiratory physiotherapy prevents pulmonary complications after cardiac surgery.
Data sources Searches through Medline, Embase, Cinahl, the Cochrane library, and bibliographies, for randomised trials comparing any type of prophylactic respiratory physiotherapy with another type or no intervention after cardiac surgery, with a follow up of at least two days, and reporting on respiratory outcomes.
Review methods Investigators assessed trial validity independently. Information on study design, population, interventions, and end points was abstracted by one investigator and checked by the others.
Results 18 trials (1457 patients) were identified. Most were of low quality. They tested physical therapy (13 trials), incentive spirometry (eight), continuous positive airway pressure (five), and intermittent positive pressure breathing (three). The maximum follow up was six days. Four trials only had a no intervention control; none showed any significant benefit of physiotherapy. Across all trials and interventions, average values postoperatively were: incidence of atelectasis, 15-98%; incidence of pneumonia, 0-20%; partial pressure of arterial oxygen per inspired oxygen fraction, 212-329 mm Hg; vital capacity, 37-72% of preoperative values; and forced expiratory volume in one second, 34-72%. No intervention showed superiority for any end point. For the most labour intensive intervention, continuous positive airway pressure, the average cost of labour for each patient day was €27 (£19; $32).
Conclusions The usefulness of respiratory physiotherapy for the prevention of pulmonary complications after cardiac surgery remains unproved. Large randomised trials are needed with no intervention controls, clinically relevant end points, and reasonable follow up periods.
Pulmonary complications after cardiac surgery prolong hospital stay and increase healthcare costs.1 We performed a systematic review to determine to what extent respiratory physiotherapy prevents such complications, and the best type of physiotherapy intervention.
We chose the setting of cardiac surgery for three reasons. Firstly, patients are prone to pulmonary complications after surgery; up to 65% of patients may have an atelectasis, and 3% may develop pneumonia.2 3 Secondly, the prevalence of cardiac surgery is high; around 110 per 100 000 population annually in the Western world.4 Thirdly, the extra costs of pulmonary complications after cardiac surgery exceed €28 000 (£19 000; $32 000) for each patient.5
We carried out an extensive search, with no language restrictions, through Medline, Embase, CINAHL, and the Cochrane controlled trials register using these key words: physical therapy, respiratory therapy, breathing exercise, chest physiotherapy, continuous positive airway pressure, incentive spirometry, intermittent positive pressure breathing, noninvasive pressure support ventilation, noninvasive positive pressure ventilation, bilevel positive airway pressure ventilation, cardiac surgery, cardiac operation, coronary artery bypass grafting, and random. The last search was on 19 February 2003. We checked the bibliographies of retrieved reports and reviews.6–8 Not considered were data from abstracts, letters, and animal studies. All main authors of all included studies were contacted.
Inclusion criteria, end points, and definitions
We included full reports of randomised trials of adults or children who had undergone cardiac surgery. Inclusion criteria included any method of prophylactic respiratory physiotherapy compared with no intervention or with another method of respiratory physiotherapy, and an observation period of at least two days.
The trials also had to assess at least one of four end points: atelectasis, pneumonia, oxygenation (partial pressure of arterial oxygen, with the corresponding fractional inspired oxygen), and pulmonary function (vital capacity or forced expiratory volume in one second). If end points were reported at different time points after surgery, we considered the latest. For atelectasis, pneumonia, and adverse effects we extracted dichotomous data. We checked for cointerventions that may have influenced the efficacy of the physiotherapy9: analgesia, respiratory physical therapy other than the tested intervention, and mobilisation. One investigator (PP) abstracted the data, which were independently cross checked by the others. The investigators independently scored the methodological quality of the included studies.10 11
To establish the relative efficacy of physiotherapy in the absence of a gold standard intervention, we regarded as the most valid study design comparisons between an active intervention and a no intervention control. Active (head to head) comparisons were of secondary importance.
We estimated the cost of physiotherapy, assuming that one physiotherapist was treating one patient at a time. The cost for purchase or maintenance of equipment was not considered, but we estimated the cost of labour from reported labour time. If no such data were given, we made three assumptions. Firstly, incentive spirometry comprised 10 inspirations, each session lasting five minutes, and a physiotherapist supervised two sessions a day—a total of 10 minutes for each patient day. Secondly, for continuous and intermittent positive pressure breathing, 10 minutes were needed for installation, 10 minutes for adjustments for each hour of therapy, and five minutes for disconnection—a total of 25 minutes for each patient day. Thirdly, for physical therapy, the physiotherapist needed to be present during the entire treatment period except for breathing exercises, when the same assumptions were made as for incentive spirometry. The average salary of a physiotherapist in Europe was estimated at €13/h (Switzerland €19/h, Belgium €13/h, and France €9/h; data from personal communication in 2003 with physiotherapists working in public hospitals in these countries).
Of 107 papers screened, 27 randomised controlled trials were eligible for inclusion; nine were subsequently excluded (figure).12–20 We analysed data from 18 trials (1457 patients) from nine countries, published between 1978 and 2001 (table 1).21–38 Three authors responded to our inquiries23 24 28: all provided supplementary information, which resulted in one additional trial being identified.22 The average group size was 32 patients (range 12-95 patients). Four trials described an adequate randomisation method, two reported on concealment of treatment allocation, and 14 reported on blinding of observers. Three trials used an intention to treat analysis.
Thirteen trials tested 11 different physical therapy regimens; incentive spirometry (n = 8), continuous positive airway pressure (n = 5), intermittent positive pressure breathing (n = 3), and blow bottles (n = 2). Cointerventions were used in most trials but adequately described in only four. One trial studied children, one trial studied children and adults, and 16 trials studied adults. Average length of stay in the intensive care unit was 2 to 2.8 days and in the hospital was 7.5 to 13 days.
Active intervention versus no intervention control
Four trials had a no intervention control.22 28 33 35 They tested three physical therapy regimens; deep breathing, deep breathing and cough, and deep breathing and costal expansion exercises. Two also tested incentive spirometry.33 35 We found no evidence of superiority of any active intervention for the end points.
Head to head comparisons
Overall, 14 trials (1266 patients) reported on the incidence of atelectasis (table 2). One study (44 children) found a significantly lower incidence when less intensive physical therapy was compared with more intensive physical therapy.36
Nine trials (942 patients) reported on the incidence of pneumonia (table 2). No statistically significant differences were evident.
Ten trials (752 patients) reported on partial pressure of arterial oxygen per inspired oxygen fraction (table 3). One trial (58 patients) found a significant increase with continuous positive airway pressure compared with physical therapy.32
Overall, 11 trials (921 patients) reported on vital capacity and eight trials (748 patients) reported on forced expiratory volume in one second (table 3). One trial (96 patients) found a significant increase in both with both continuous positive airway pressure and non-invasive ventilation compared with incentive spirometry.23
Four trials provided dichotomous data on adverse effects; gastric distension in 2-10% of patients and nausea in 0-12% of patients.25 29 32 38 Inconvenience of the mask was reported in 43% of patients receiving continuous positive airway pressure.29 32 During physical therapy, 4% of patients had a percutaneous capillary oxygen saturation of less than 90% and 1% of patients had tachycardia.25 Eleven trials did not mention any adverse effects, and none were observed in two trials.24 28
The median time patients spent receiving physiotherapy was 80 minutes (range 20-120 minutes) for incentive spirometry, 480 minutes (70 to 720 minutes) for continuous positive airway pressure, 80 minutes (80 to 120 minutes) for intermittent positive pressure breathing, and 120 minutes (data from one trial only) for physical therapy.23 29–32 34 37 38 Physiotherapy lasted on average 0.3 to 5 days.21 22 28 29 31–34 37 The average daily cost of labour for each patient was €6 for incentive spirometry, €10 for physical therapy, €20 for intermittent positive pressure breathing, and €27 for continuous positive airway pressure (table 4).
Evidence is lacking as to whether prophylactic respiratory physiotherapy prevents pulmonary complications after cardiac surgery. Two published systematic reviews examined the relation between respiratory physiotherapy and outcome after different operations, but they obtained conflicting results. One found benefits from incentive spirometry and deep breathing exercises after upper abdominal surgery, but pooled data came from different end points such as atelectasis and pulmonary infiltrates or consolidation.39 The other review found incentive spirometry to be of no benefit after cardiac and upper abdominal surgery.8 Again, data were combined from trials with a variety of different end points. Our conclusions reflect more uncertainty, showing several limitations in the original trials. These limitations are the main weakness of our systematic review.
Eighteen trials tested eight regimens of prophylactic respiratory physiotherapy. This variety, which is not dissimilar to other settings, may be due to the lack of a gold standard method for respiratory physiotherapy.40 If ethically acceptable, the best comparator is then a placebo or, as in the physiotherapy setting, a no intervention control.41 Four trials only had a no intervention control group, and each tested a different method of physiotherapy.22 28 33 35 Based on these trials, it was therefore difficult for us to determine the efficacy of different methods of respiratory physiotherapy.
On average the quality of the trials was low. Only a minority reported on an appropriate method of randomisation or on concealment of allocation, although bad reporting may not mean bad practice. In only a few trials was the follow up of patients adequately reported and data analysed according to intention to treat. One inherent problem of trials in this setting is that at best the observer can be blinded. Over two thirds of the trials attempted to blind the observers. We do not know if trials of better quality would have reached different conclusions.
Practical management of physiotherapy was inconsistent. For example, the reported duration of daily continuous positive airway pressure varied by a factor of 10. Inconsistency suggests that there is uncertainty about how each method should be applied and how frequently.
For most end points there was variability in event rates. The average incidence of pneumonia was 0-20%.22 23 26 38 Two reasons may explain this variability. Firstly, there were no uniform definitions of pneumonia; one study used established criteria only.26 Secondly, most trials were of limited size. Only two studies included groups of more than 50 patients.24 37 In small trials anything can happen by random chance.42
The longest observation period was six days. This may be too short in which to identify all respiratory complications. Nosocomial pneumonia, for instance, occurs on average eight days after cardiac surgery.43
In large randomised controlled trials, cointerventions are usually balanced between the groups. In small trials, however, we cannot exclude bias related to an imbalance of cointerventions. Sixteen of the 18 trials had less than 50 patients in each group, thus cointerventions may have affected the efficacy of physiotherapy. Method and intensity of postoperative analgesia may have an impact on pulmonary function.44 Early mobilisation may also have an effect on outcome. Only three trials adequately controlled for concomitant analgesia or mobilisation.25 26 33
Because there was no evidence of any benefit from respiratory physiotherapy, we were unable to determine the cost incurred to generate one patient who would profit from an intervention compared with doing nothing. If there is no benefit, there are only costs, and these are not negligible in this context.
What is already known on this topic
Prophylactic respiratory physiotherapy after cardiac surgery is widely used
It is thought to reduce the risk of pulmonary complications such as pneumonia or atelectasis
What this study adds
Evidence is lacking on benefit from any method of prophylactic respiratory physiotherapy after cardiac surgery
It is likely that there are adverse effects and costs only
We thank Daniel Haake (medical libraries, Centre Medical Universitaire, Geneva University) for his help in searching electronic databases, and Kathy Stiller, Jean Crowe, and Pascal Matte who provided additional information.
Contributors PP initiated, designed, and organised the study, and extracted and analysed the data. He will act as guarantor for the paper. MRT designed the study, cross checked the studies, and analysed the data. BW initiated and designed the study, cross checked the studies, and analysed the data.
Funding MRT is a beneficiary of a PROSPER grant from the Swiss National Science Foundation (No 3233-051939.97/2)
Competing interests None declared
Ethical approval None required