Are incentive spirometry, intermittent positive pressure breathing, and deep breathing exercises effective in the prevention of postoperative pulmonary complications after upper abdominal surgery? A systematic overview and meta-analysis.

Pulmonary compfications are the most frequently occurring complications following upper abdominal surgery, with reported frequencies of up to 75% of all patients.[1] Some form of pulmonary physical therapy is often prescribed for prophylaxis or treatment of these complications.[1-4] Interventions such as deep breathing exercises (DBEX), incentive spirometry (IS), and intermittent positive pressure breathing (IPPB) and other strategies such as ambulation have been studied to determine their efficacy.[1-16]

The results of these studies have been conflicting, and this may contribute to a varied pattern of clinical practice and use of physical therapy for these patients. Although the treatment of postoperative pulmonary complications has been studied extensively, much of the work has methodological limitations. The absences of uniform criteria for establishing the presence of pulmonary complications and for describing postoperative therapheutic regimens, as well as the relative importance given to risk factors for pulmonary complications, may contribute to the discrepancy in results.[2] In some studies,[3,4] the numbers of subjects are small, and the studies therefore the lack adequate statistical power to demonstrate a treatment effect even though one may be present.

Of the many published reviews examining the efficacy of these treatments,[2,17-25] none have included a systematic overview and statistical synthesis of the results. This article will assess a body of literature concerning the efficacy of IS, IPPB, and DBEX in the prevention of postoperative pulmonary complications in patients undergoing upper abdominal surgery using a quantitative meta-analysis approach. Meta-analysis is a technique that is used to critically review of previous research to draw conclusions about therapuetic effectiveness or to plan new studies.[26] Meta-analysis is especially useful when results from studies disagree with regard to magnitude or direction of effect, or when sample sizes are individually too small to detect an effect and label it statistically significant.[26]

The analysis in this report attempts to answer two questions: (1) Is any treatment better than a control intervention (ie, no treatment)? and (2) Is one treatment modality better than another? Four separate meta-analyses were conducted: (1) any physical therapy intervention versus a control, (2) IS versus IPPB, (3) IS versus DBEX, and (4) IPPB versus DBEX.

Method

Study Identification

Both review and primary research articles were examined. Previously published reviews addressing the efficacy of any "chest physical therapy" modality for patients undergoing surgery were identified by searching the MEDLINE (National Library of Medicine, Bethesda, Md) database for the years 1966 through 1992. The year 1966 was chosen as a cutoff because we felt that prior to this date the differences in surgical techniques, physical therapy interventions, and study quality were great and that searching for this information would not yield any useful information. The medical subject headings used in the search for review articles were "chest," "lung diseases (rehabilitation)," "physical therapy (methods)," and "review." The identified reviews were evaluated for methodologic quality using the criteria developed by Oxman and Guyatt.[27]

Four different search strategies were used to locate primary research. First, the Cumulative Index to Nursing and Allied Health was searched for the years 1966 through 1992 using the term "chest physical therapy." Second, the MEDLINE database was searched for the same period. Search terms used were identical to those used in the search for reviews with the exclusion of the term "review." The MEDLINE database was also searched using the following expanded terms: "atelectasis (rehabilitation)," "respiratory therapy (methods)," "respiratory tract disease (prevention and control)," and "spirometry (methods)." Third, reference lists of the retrieved review articles were searched for relevant citations. Finally, reference lists and unpublished abstracts were collected from a Consensus Exercise on Physical Therapy for the Surgical Patient, 1989.[28] The authors of unpublished abstracts were asked for details of their study if they met the inclusion criteria. Only citations published in English were retrieved.

Study Selection

For criteria were used in selecting studies for inclusion in the overview. The study population consisted of adults undergoing any type of upper abdominal surgery. Studies of patients undergoing upper abdominal surgery were chosen because this population represents a uniform group of patients with similar probable mechanisms for the development of pulmonary complications. The interventions chosen were any combination of IS, IPPB, or DBEX. Studies that used the term "chest physical therapy" were included in the DBEX category only when deep breathing exercises constituted the majority of the treatment regimen. The outcomes chosen were any type of postoperative pulmonary complication. The study design chosen was randomized trials only. Studies had to meet all four criteria to be included in the overview.

Initial relevancy was assessed by a single reviewer (JAT) based on the title and abstract. The article was retrieved if the title or abstract suggested that it might be relevant. Study validity was assessed by two independent observers (JAT, JMM), and discrepancies were settled by discussion and consensus. Agreement between observers was measured by weighted kappa.[29] Disagreements, as indicated by a low kappa (Kw<.4), were then resolved by consensus.

Study Evaluation and Data

Extraction

Data were extracted by a sigle observer after pilot testing of the data extraction form. Validity was assessed by a weighted 20-point scale using the criteria listed in Table 1. Agreement for the validity of each study was measured by weighted kappa calculated for each question in the validity form. Disagreements were dealt with using the same process as for assessment of relevance.

Table 1. Criteria for Methodologic Quality(a)

1. Population
   Reproducible description of patient population, surgical
   procedure (1)
2. Preoperative risk
   Groups were comparable with respect to smoking history, past
   respiratory disease, and
   pulmonary function (2)
3. Intervention
   Experimental maneuver was well described (1)
   There was a control group (2)
4. Outcome
   Outcome measures well described and appropriate (1)
5. Design
   Random allocation to groups (3)
   Randomization process blinded (2)
   Withdrawals listed and why (2)
   Outcome measurement blinded (3)
   Prior estimate of power (1)
6. Compliance
   Test of compliance demonstrated (volume and number of
   repetitions) (2)

(a) Numbers in parentheses indicate number of points in summary
score.

Data Analysis

Summary statistics, common odds ratio (OCR), and test for heterogeneity were calculated only for outcome measures that were consistent across all studies. Chest radiographs were the most commonly used measure. Because definitions for a positive chest radiograph (ie, positive pulmonary complications) were varied, only atelectasis and pulmonary infiltrate were included as outcomes. Other outcomes, such as pleural effusion and pulmonary edema, were excluded when possible because there is no strong evidence or rationale to sugges that any physical therapy modality is effective in preventing these complications. All radiographic complications were included when it was not possible to delienate each specific radiographic complication.

Physical examination was used as a secondar outcome measure if it was not possible to separate chest radiograph findings. Definitions of a positive outcome usually included a combination of physical signs such as chest auscultation, temperature, and sputum production, which varied among studies.

The test for heterogeneity and the COR and its 95% confidence intervals (CIs) were calculated using the computer software program OR 2X2.[30] The COR calculation estimates the odds of an event or complication occurring in the control group versus the odds of an event or complication occurring in the treatment group for the combined group of studies. If the COR and its 95% confidence intervals are less than 1, there is a statistically significant difference between the treatment and control groups.[31] The 95% CIs were calculated using the Cornfield method.[32] A probability value of [greater than or equal to] 0.5 was considered significant. The test for heterogeneity was calculated using the Breslow-Day method.[33] The test for heterogeneity is a statistical test to determine whether the variation among studies was more than could be explained by chance. If significant heterogeneity exists, it is essential to try an locate the source of interstudy variation before continuing with the analysis. The approximate power of each meta-analysis was calculated, identifying a 10%, 30% and 50% risk reduction for pulmonary complications.[31]

Results

Ten review articles were identified using the search strategies described previously.[2,17-25] One hundred sixteen citations were found using the search strategies for primary research. Of these, 55 were considered potentially relevant and were retrieved. Fourteen of the 55 citations met the inclusion criteria and were included in the review.[1,3-15] Interobserver agreement for assessing relevance was good (Kw=.71). The most common reason for excluding studies was lack of randomization (ie, use of nonrandomized designs) (n=15). Other reasons for excluding studies includes the use of patient populations other than those undergoing upper abdominal surgery (n=18) and having a descriptive design (n=6). Characteristics of the studies are shown in Table 2. The mean agreement between observers for the validity criteria was good (Kw=.84).

[TABULAR DATA OMITTED]

The results of this meta-analysis answered four questions: (1) Is any treatment better than a control (ie, no treatment)? (This analysis included results of studies with any kind of treatment versus a control.[1,5-10) (2) Is IS different from IPPB?3,5,12,13 (3) Is Is different from DBEX?5,11,14,15 and (4) Is IPPB different from DBEX?4,5

Is Any Treatment Better Than

No Treatment?

The COR for studies examining any treatment versus a control was 0.85 (95% CI=0.59-1.2), indicating no statistically significant difference between the two approaches. The range of individual odds ratios was from 0.05 (very effective) to 1.45 (a negative effect). In only one study[5] was the difference between the treatment and control groups statistically significant. The test for heterogeneity was significant, indicating that the variation among study results was more than could be expected due to chance. It is therefore difficult to fully accept the results of this analysis without attempting to locate the source of heterogeneity. We hypothesizes that the significant heterogeneity of the data synthesis may have been due to the wide variation in treatment modalities. To eliminate the heterogeneity, the studies were separated into their specific modalities, and the COR was recalculated for each modality. It was only possible to calculate the COR for IS versus no physical therapy and for DBEX versus no physical therapy, because only one study5 examined IPPB versus no physical therapy with the appropriate outcome measure included in the overview.

In examining the difference between IS and a control, two studies,[5,10] with a total sample size of 212, were included in the analysis. The COR for IS versus a control was 0.44 (95% CI=0.18-0.99), demonstrating a statistically significant benefit of IS over a control (P=.034). The test for heterogeneity was not significant (P=.15) (Fig. 1)

In examining the difference between DBEX versus control, four studies,[1,5,8,9] with a total sample size of 564, were included in the analysis.

The COR for DBEX versus no physical therapy was 0.43 (95% CI=0.27-0.63), demonstrating a statistically significant benefit of DBEX over a control (P=.005) (Fig. 2). The test for heterogeneity was still significant (P=.001).

Is Incentive Spirometry Different

From Intermittent Positive

Pressure Breathing?

In comparing IS with IPPB, the COR was .76 in favor of IS, but this finding was not statistically significant. Odds ratios ranged from 0.3 (with IS being more effective than IPPB) to 1 (no difference between the two), and are presented in Table 3. The test for heterogeneity was not significant (P=.744). The study by Van de Water et al[3] was not included in this analysis because other treatment modalities confounded the results and separating outcomes was not possible.

Table 3. Meta-analysis Results of
Incentive Spirometry Versus Intermittent
Positive Pressure Breathing(a)

                                   95%
                            Odds   Confidence
Authors                     Ratio  interval (CI)

Jung et AI[13] (1980)       0.77   0.21-2.78
Dohi and Gold[12] (1978)    0.41   0.13-1.30
Celli et al[5] (1984)       1.10   0.40-3.05
(a) Common odds ratio = 0.73 (95% CI = 0.39-1.36:
P>.05); power for percentage of risk reduction
(RR): 10%RR = 13%, 30%RR = 8%, 50%RR = 88%.

Is Incentive Spirometry Different

From Deep Breathing Exercises?

In comparing IS and DBEX, the COR was 0.91 in favor of IS, but this finding was not statistically significant. The test for heterogeneity was not significant (P=.13) (Tab. 4).

Table 4. Meta-analysis Results qf
Incentive Spirometry, Versus Deep
Breathing Exercises(a)

                                    95%
                          Odds      Confidence
Authors                   Ratio     Interval (CI)

Celli et al[5] (1984)     0.78      0.28-2.13
Stock et al[15] (1982)    2.17      0.51-9.54
Lyager et al[14] (1979)   1.27      0.54-3.00
Craven et al[11] (1974)   0.39      0.13-1.15

(a) Common odds ratio = 0.91 (95% CI = 0.57-1.4;
P>.05); power for percentage of risk reduction
RR): 10%RR = 19%, 30%RR = 75%,
50%RR = 99%.

Is Intermittent Positive Pressure

Breathing Different From Deep

Breathing Exercises?

In comparing IPPB and DBEX, the COR was .94 in favor of IPPB, but this finding was not statistically significant (Tab. 5). The test for heterogeneity was again not significant (P=.25). Only two studies4,5 were included in this comparison, and neither study showed significant results. Both studies had very small sample sizes and low power, which may have accounted for the nonsignificant results.

Table 5  Meta-analysis Results of
Interminent Positive Pressure Breathing
Versus Deep Breathing Exercises(a)

                                    95%
                            Odds     Confidence
Authore                     Ratio    Interval (CI)

Celli et al[5] (1984)       1.02      0.22-4.56
Schupisser et al[4] (1980)  0.8       0.77-8.11

(a) Common odds ratio = 0.94 (95% CI = 0.28-3.17;
P>.05); power for percentage of risk reduction
(RR): 10%RR = 9%, 50%RR = 52%.

Discussion

Methodology

Studies included in the analysis were assessed for methodological rigor using a 20-point weighted scale. All studies scored low, with the mean score of 11.2/20 and a range of 7 to 16/20. Because of the small number of randomized controlled trials, it was decided a priori to include these studies in the meta-analysis. The potential impact of such a methodological limitation on the results needs to be recognized.[34] Only randomized trials were included in this analysis, thus ensuring a conform baseline for study validity. Despite this requirement, the method of randomization was not always clearly stated. Lack of a blinded randomization process can result in a systematic bias and would most likely favor the treatment group. The target population was fairly uniform across studies, and there were systematic attempts to ensure comparability of groups within studies. Factors compared included pulmonary function, smoking history, previous respiratory diseases, type and duration of anesthetic used, and type of surgery. Patient withdrawals were not large in any study (less than 5%) because most patients were not considered eligible for the study until they were admitted for surgery. Most patient withdrawals occurred because the patients did not undergo their operation, not because of the treatment interventions. Therefore, patient characteristics and withdrawals probably did not have a large effect on the outcome.

Few studies attempted to blind the outcome measures.[5,10,11,14,15] In all studies, only the chest radiograph outcome (and not the physical examination) was blinded. This limitation could also act to bias the results in favor of the treatment group.

All three therapeutic interventions being investigated require some effort on the part of the patient and the therapist, yet only two studies[14,15] measured patient compliance. Differences in study outcomes may be attributed to the dosage received as a function of ineffective efforts of the patient or therapist. With IPPB, the frequency and pressure were recorded, but no information was provided on the volume the patients attained during inspiration. Bartlett et[25] claimed that regular, sustained maximal inflations prevent atelectatic pulmonary complications, possibly inferring that a causative mechanism is a lack of deep breaths. Lyager et al[14] measured the volume achieved on the incentive spirometer and used that volume as a guideline for progression of treatment. They categorized their results into "good users" and "bad users" of the Bartlett Incentive Spirometer. Although there was no statistical significance between the "good user" and "bad user" groups, there was a trend for the "good user" group to have fewer complications. The lack of statistical significance may have been due to the lower power that resulted from dividing the IS group into two subgroups. The issue of patient compliance must be given more consideration in future studies.

Timing and Dosage of Treatment

Modalities

Optimal prescription and timing of the various treatment modalities is important, but no dosage studies were found. Deep breathing exercises and IPPB are prescribed four to five times daily during waking hours.[5] The prescription of IS is hourly, with 10 maximal breaths required per treatment session during waking hours. This prescription was fairly uniform across studies. The lack of strong positive evidence may be attributed to the fact that there was an insufficient dosage of treatment.

Outcome Measures

As previously mentioned, there are numerous definitions for pulmonary complications. Part of the variation in study findings can be attributed to the various definitions of pulmonary complications and their measurement. For optimal assessment of therapeutic needs and methods, it is essential to establish uniform criteria for identifying complications that have a true impact on the postoperative course compared with complications that are self-limiting and cause no clinical concern for both the patient and the clinician.[2]

Chest radiographs were the most commonly used measure across all studies. There were variations, however, in the choice of what constituted a positive pulmonary complication. Some authors[1,5,7,10-12] described chest radiograph findings as atelectasis, pulmonary infiltrate, or pleural effusion. Some authors[8,9] further stratified their findings into major and minor atelectasis or consolidation. For the purpose of this overview, all types of atelectasis and pulmonary infiltrate/consolidation were considered together. if it was possible, pleural effusion was excluded from the analysis.

Physical examination was also used as an outcome measure. This measure was more diverse in its scope of definitions for pulmonary complications than was chest radiography. Measures of temperature, tachypnea, dyspnea, and chest auscultation were used to define pulmonary complications, and most studies used some combination of these measures to define a complication. Celli et al,[5] for example, described a pulmonary complication as having any three of the following symptoms: coughing, increased sputum, dyspnea, temperature greater than 38 [degrees] C, or heart rate greater than 100 beats per minute. The many combinations of these different physical measures may explain the variation in the study findings. Physical examination data were used once[4] in the overview when radiographic data were unavailable.

The timing of administration of the outcome measure with respect to the underlying disease process is also important to consider. If the outcome is measured too early postoperatively, the rate of complications may be underestimated. If measured late, complications that may have been successfully treated may be missed. There is a significant decrease in functional residual capacity (FRC) following upper abdominal surgery.[35] If the FRC falls below the closing capacity of the lung, this may result in the postoperative atelectasis that is frequently reported in these patients.[35] It has been assumed in the past that FRC was decreased immediately postoperatively, secondary to the induction of anesthesia. Ali et al,[36] using measurements of FRC obtained 4, 10, and 16 hours following cholecystectomy, demonstrated that a significant decrease in FRC did not occur until 16 hours following cholectstectomy. Therefore, the timing of the outcome measurement may have a significant impact on the results of the study. For example, Celli et al[5] and Stock et al[15] only measured major outcomes at 24 hours after abdominal surgery and at this point may have been premature in evaluating the effect of treatment.

Low-Risk Patient Populations

Most studies included in this analysis involved patients of all risk groups admitted for surgery. In all studies, some attempt was made to ensure comparability of groups for the factors that are known to have some influence on the incidence of pulmonary complications such as age, smoking history, history of previous respiratory disease, pulmonary function, and obesity. Numerous researchers[9,10,37] have attempted to study patients who are at low risk of developing pulmonary complications with respect to these factors.

Roukema et al[9] and Roy et al[37] studied a regimen of DBEX versus a control regimen in a low-risk population. Roukema and colleagues found that there was a greater total number of complications in the treatment group compared with the control group. Roy et al found an overall low rate of complications and no difference between the groups. Schweiger et al[10] compared a regimen of IS with a control regimen and found no difference between the interventions. The two trials with no diference observed between the groups[10,37] had a low overall complication rate (20%), whereas the trial that demonstrated a beneficial treatment effect[9] had an overall complication rate of 60%. This difference may be due to differing definitions of pulmonary complications. The first two studies relied solely on radiologic data, whereas the third study used a combination of radiologic and physical assessment findings to determine complication rates. It is important to note that most "control groups" did receive some form of treatment. This control intervention, referred to as a "stir-up" regimen by Pontoppidan,[2] included early ambulation, coughing, and turning. It might therefore be concluded that patients who are at low risk of developing postoperative pulmonary complications do not require anything above the "routine nursing care" provided in many hospitals. This sub-group of patients could not be assessed separately in the meta-analysis due to differing outcome measures and treatment modalities, although results of the individual trials were included in the overall analysis.

Conclusions

The following conclusions can be made on the basis of this metaanalysis. Deep breathing exercises (chest physical therapy) are more effective than no physical therapy. There was significant heterogeneity within these studies that may be attributable to poor methodology or to the difference in treatment regimens. Incentive spirometry is also more effective than no physical therapy. There was no significant heterogeneity within these studies. It is difficult to make any conclusions regarding the effectiveness of IPPB versus a control because there were not enough data to warrant a calculation. Comparisons of the different modalities (ie, IS, DBEX, and IPPB) revealed no statistically significant differences among them. All these comparisons demonstrated low power and showed no significant heterogeneity. The lack of statistical significance may be attributed to a lack of clinically important differences, poor study design, suboptimal treatment dosage or timing, poor patient compliance, or poor outcome measures and definitions. Incentive spirometry and DBEX should be recommended as treatment modalities in the prevention of atelectasis and pneumonia in patients undergoing upper abdominal surgery.

Future research should consider that the frequency and optimal dosage of treatment must be standardized before effectiveness can be determined. More regular treatments for shorter periods of time may be of benefit. A standardized definition for "clinically important pulmonary complications" should be delineated, and larger samples should be used. The "low-risk" subgroup of patients warrants future consideration, as resources may be more effectively allocated to "high-risk" patients.

References

[1] Hallbook T, Lindblad B, Lindroth B, Wolff T. Prophylaxis against pulmonary complications in patients undergoing gall-bladder surgery. Ann Chir Gynawol. 1984;73:55-58. [2] Pontoppidan H. Mechanical aids to lung expansion in non-intubated surgical patients. Am Rev Respir Dis. 1980; 122:109-119. [3] Van de Water JM, Watring WG, Linton LA, et al. Prevention of postoperative pulmonary complications. Surg Gynecol Obstet. 1972; 135: 1-5. [4] Schupisser JP, Brandli O, Meili U. Postoperative intermittent positive presssure breathing versus physiotherapy. Am J Surg. 1980; 140: 682-686. [5] Celli B, Rodriguez G, Snider G. A controlled trial of intermittent positive pressure breathing, incentive spirometry, and deep breathing exercises in preventing pulmonary complications after abdominal surgery. Am Rev Respir Dis. 1984;130:12-15. [6] Crawford BL, Blunnie WP, Elliot AGP. The value of self-administered peri-operative physiotherapy. Int J Med Sci. February 1990:51-52. [7] Baxter WD, Levine RS. An evaluation of intermittent positive pressure breathing in the prevention of postoperative pulmonary complications. Arch Surg. 1969;98:795-798. [8] Morran C, Findlay I, Mathieson M, et al. Randomized controlled trial of physiotherapy for postoperative pulmonary complications. Br J Anaesth. 1983;55:1113-1116. [9] Roukema J, Carol E, Prins J. The prevention of pulmonary complications after upper abdominal surgery in patients with noncompromised pulmonary status. Arch Surg. 1988;123: 30-34. [10] Schwieger I, Gamulin Z, Forster A, et al. Absence of benefit of incentive spirometry in low-risk patients undergoing elective cholecystectomy. Chest. 1986;89:652-656. [11] Craven JL, Evans GA, Davenport JL, Williams RHP. The evaluation of the incentive spirometer in the management of postoperative pulmonary complications. Br J Surg. 1974;61: 793. [12] Dohi S, Gold MI. Comparison of two methods of postoperative respiratory care. Chest 1978;73:592-595. [13] Jung R, Wight J, Nusser R, Rosoff L. Comparison of three methods of respiratory care following upper abdominal surgery. Chest. 1980;78:31-35. [14] Lyager S, Wernberg M, Rajani N. Can post-operative pulmonary complications be improved by treatment with the Bartlett-edwards incentive spirometer? Acta Anaesthesiol Scand. 1979;23:31. [15] Stock MC, Downs JB, Gauer PK, Cooper RB. Prevention of atelectasis after upper abdominal operations Anesthesiology. 1982; 57(3A):A457. [16] Dull JL, Dull WL. Are maximal inspiratory breathing exercises or incentive spirometry better than early mobilization after cardiopulmonary bypass? Phys Ther. 1983;63:655-659. [17] Selsby D, Jones JG. Some physiological and clinical aspects of chest physiotherapy. Br J Anaesth. 1990;64:621-631. [18] Sutton PP, Pavia D, Bateman JRM, Clarke SW. Chest physiotherapy: a review. Eur J Respir Dis, 1982;63:188-201. [19] Sutton PP. Chest physiotherapy: time for reappraisal. Br J Dis Chest. 1988;82:127-137. [20] Kirilloff LH, Owens GR, Rogers RM, Mazzocco MC. Does chest physical therapy work? Chest. 1985;88:436-444. [21] Kigin CM. Chest physical therapy for the postoperative or traumatic injury patient. Phys Ther. 1981;61:1724 1735. [22] Bartlett RH. Respiratory therapy to prevent pulmonary complications of surgery. Resp Care. 1984;29:667 677. [23] Grimby G. Aspects of lung expansion in relation to pulmonary physiotherapy. Am Rev Respir Dis. 1974;110:145-153. [24] Orlandi O, Perino B, Testi R. Old and new in chest physiotherapy. Eur Respir J Suppl. 1989;7:595s-598s. [25] Bartlett RH, Gazzaniga AB, Geraghty TR. Respiratory maneuvers to prevent postoperative pulmonary complications: a critical review. JAMA. 1973;224:1017-1021. [26] L'Abbe KA, Detsky AS, O'Rourke K Meta-analysis in clinical research. Ann Intern Med. 1987;107:224-233. [27] Oxman AD, Guyatt GH. Guidelines for reading literature reviews. Can Med Assoc J. 1988; 138:697-703. [28] Kelsey CJK, McIntosh J. Consensus statement on perioperative physiotherapy for the surgical patient. Contact (Newsletter for the Canadian Physiotherapy Association). April 1990. [29] Cicchetti DV, Fleiss JL. A comparison of the null distribusions of weighted kappa and the c ordinal statistic, Appl Psychol Measurement. 1977;1:195-201. [30] Julian JA. Summary Odds Ratio Analysis for 2X2xk Tables, Version 1.0. Hamilton, Ontario, Canada, McMaster University. [31] Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natt Cancer Inst. 1959;22:719-748. [32] Cornfield J. A statistical problem arising from retrospective studies. In: Neyman J, ed. Proceedings of the Third Berkeley Symposium IV. Berkeley, Calif University of California Press; 1956:135-148. [33] Breslow NE, Day NE. Statistical Methods in Cancer Research, Vol 1: The Analysis of Case-Control Studies. Lyon, France: LARC; 1980. [34] Chalmers TC, Smith H, Blackburn B, et al. A method for assessing the quality of a randomized control trial. Controlled Clin Trials. 1981;2:31-49. [35] Craig D. Postoperative recovery of pulmonary function. Anesth Analg. 1981;60:46-52. [36] Ali J, Weisel R, Layug A, et al. Consequences of postoperative alterations in respiratory mechanics. Am J Surg. 1974; 128:376-382. [37] Roy PD, Macneil AR, Dechman G. Is routine perioperative chest physiotherapy beneficial? Chest. 1989;96(suppl). Abstract.