Introduction

Pulmonary complications are among the most frequent preventable causes of death and disability after upper abdominal surgeries. The genesis of undesirable postoperative pulmonary adverse events is multifactorial. General anaesthesia is known to increase the alveolar collapse [1, 2]. Diminished ciliary clearance is demonstrated both during inhalational and total intravenous anaesthesia, although to a lesser degree with the later technique [3]. Anaesthesia-induced reduction in the ventilatory drive, along with decreased diaphragmatic activity can worsen the ventilation/perfusion mismatch. The cumulative effects of these intraoperative events might extend into the immediate postoperative period, contributing to higher rates of atelectasis, pneumonia and hypoxemia [4]. In spite of many modalities of treatment options available to address these unwanted postoperative pulmonary adverse events, the morbidity and mortality remains high. Intraoperative ventilatory strategies along with positive end expiratory pressure (PEEP) have shown to reduce these complications [5]. Conclusive evidence exists for the therapeutic benefit of postoperative CPAP in obese patients and in patients with obstructive sleep apnoea (OSA) [6]; however, for other population subsets definitive results do not exist. Many investigators have studied the role of immediate postoperative continuous positive pressure airway pressure (CPAP) in preventing postoperative pulmonary complications [7]. Trials vary significantly in the degree and duration of CPAP applied and as a result their direct effect has not been quantified. In this meta-regression analysis, we analyzed the ability of postoperative CPAP to pre-empt atelectasis, pneumonia and other pulmonary complications in patients undergoing upper abdominal surgeries under general anaesthesia.

Materials and Methods

The present study was conducted in accordance with preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines [8] (Figure 1). We used the population, intervention, control and outcome study (PICOS) design for identification of the relevant studies (Table 1). Studies fulfilling the above criterion for primary outcome goal (incidence of postoperative pulmonary complications, atelectasis or pneumonia) were included in the analysis. The salient features of these included trials which meet the above criterion are displayed in Table 2.

Fig. 1
figure 1

PRISMA flow diagram

Full size image
Table 1 Inclusion characteristics of the trial
Full size table
Table 2 Summary of all the trials included in the meta-analysis
Full size table

Literature Search

Two independent reviewers (A. B. and P. M. S.) searched the online literature available on Medline, Embase, Science Citation Index Expanded, Cochrane Central Register of Controlled Trials, clinical trials registry, Scopus and meta-register of controlled trials published until 4th November 2015. The medical subject heading terms Prophylactic postoperative CPAP, CPAP and postoperative pulmonary complications, CPAP and postoperative pneumonia/atelectasis, postoperative CPAP in upper abdominal surgeries, prophylactic CPAP in upper abdominal surgeries, CPAP prevention of pulmonary complications were searched on the aforementioned database. We used the following terms as exclusion during our search morbidly obese patients, postoperative CPAP in OSA patients, prophylactic non-invasive ventilation. Only randomized controlled trials comparing CPAP to control groups (oxygen therapy without positive pressure with or without chest physiotherapy) were further evaluated. Our search extended to articles published either as full manuscripts or meeting abstracts in peer-reviewed journals. References from comparable meta-analysis were also manually searched for relevant trials. Our search included both English and non-English language literature. The decision to include a potential study in the analysis was based upon the independent assessment of full text by the authors. Disagreements were harmonized by consensus and if necessary by arbitration by the third researcher. Quality assessment for bias in the included studies was carried out as per the other published meta-analysis and the guidelines laid by the Cochrane Collaboration by another independent researcher [9].

Data Extraction

Data were independently abstracted into a standardized format and entered into Microsoft Excel 2011 (Macintosh edition, Microsoft Corporation, USA). The following data were extracted from the individual studies: Study design, year and country of publication, nature of surgery, participant numbers primary/secondary outcome reported, use of CPAP (both value and duration- if available), number of patients reported to have atelectasis, pneumonia and overall pulmonary complications within 72 h of surgery. We also attempted to extract the data on preoperative and postoperative arterial oxygen saturations; however, due to inconsistency in reporting it was aborted. If the data were not complete or not comprehendible, authors were contacted. During the data extraction, minor variations were discovered among the studies with regards to definitions of relevant complications. We included the complication under the relevant headings based upon criterion listed below.

  1. (1)

    Atelectasis was defined as radiological evidence of lung volume loss (Computed tomography or Chest X-ray), with clinical signs of respiratory distress without signs of infection (fever or raised total leucocyte count) at the end of the third postoperative day.

  2. (2)

    Pneumonia was defined as abnormal chest X-ray finding and/or pathological clinical findings on auscultation along with fever (>38.5 °C) and a raised total leucocyte count after third postoperative day.

  3. (3)

    Postoperative pulmonary complication was defined in individual studies and included sum of significant atelectasis, pneumonia, significant hypoxia, tracheal reintubation, intensive care unit (ICU) admission.

In trials where a CPAP range rather than a fixed value was reported, we used the median value. However, if the study titrated CPAP to individual patients as per the patient’s response, the values for the study were excluded from the analysis. Most trials mentioned the daily duration of CPAP therapy individually; however, wherever intermittent CPAP was used (in 6 of the included trials), we calculated the cumulative daily duration of CPAP for the purposes of analysis.

Statistical Analysis

The statistical analysis of the pooled data was performed using Comprehensive Meta-analysis Version 3 (Biostat Inc, USA). This version allowed meta-regression to be performed. Meta-analysis was performed initially using fixed effect modelling and eventually with random-effect modelling if the heterogeneity with fixed modelling was high. I 2 statistics was used to quantify the heterogeneity between the trials. Values of I 2 < 40 % were considered non-significant, 40–60 % were considered to represent moderate heterogeneity, 60–90 % was reported as high heterogeneity. Results were expressed as Mantel–Haenszel pooled odds ratio with 95 % CI. A P < 0.05 was considered statistically significant. Number needed to treat (NNT) was also calculated for individual variables. Meta-regression was performed for CPAP values, and daily CPAP duration for each of individual parameters was reported (for studies reporting these values). Regression model was evaluated for goodness-of-fit, both for CPAP amount and duration. A log event rate curve along with 95 % CI was generated for all the desired endpoints. Potential publication bias was quantified using the Egger’s regression test and it was further evaluated by using the funnel plot.

Results

We found a total of 839 publications in the preliminary search using the above-mentioned database. Duplicates obtained from individual search by different reviewers were identified and were removed using Endnote (Thompson Reuters, USA). No unpublished or incomplete trial was identified. Additionally, none of the published abstracts met the inclusion criterion for the final analysis. Eventually, 11 trials were identified which reported the desired outcomes for inclusion in our analysis. Of these 9 were in English language and 2 were in German. These 2 German language trials were interpreted after consultation with a medical professional well versed with the German language. Two of these trials had either varying amounts of CPAP (Christensen et al. compared two different CPAP groups to control) [10] or variable daily CPAP therapy duration (Denehy et al.—compared long and short duration CPAP to control) [11]. Thus, we had groups for comparison of qualitative outcomes in the meta-analysis; however, these different values were used as individual values of the moderator variable used in the meta-regression.

The results were analyzed under the following groups.

Postoperative Pneumonia

Incidence of Postoperative Pneumonia

Pneumonia rates were reported in 8 subgroups, which included 310 and 316 patients in CPAP and control groups respectively. The incidence of postoperative pneumonia was 4.52 % (95 % CI being 2.71–7.44 %) in the CPAP group compared to 11.08 (95 % CI being 8.07–15.01 %) in the control group. The heterogeneity for the above meta-analysis was 0 % with a Q value of 4.56. The calculated number needed to treat, to prevent single episode of postoperative pneumonia was 15.24. The pooled odds ratio and final effect are shown in Fig. 2. Subgrouping based upon the nature of CPAP (continuous vs intermittent) was also performed. The individual odds ratio values failed to attain a statistical significance, thus, no valid conclusion could be made.

Fig. 2
figure 2

Pooled odds ratio for Pneumonia (CPAP vs Control)

Full size image

Meta-regression

Random effects modelling with “method of moments” was used for calculation. Values computed were 2-tailed P values.

CPAP Value

A total of 5 trials reported the use of constant CPAP during the trial. The R2 (variance) explained with this model was 7.87 %. The incidence of postoperative pneumonia displayed a positive correlation with increasing levels of CPAP. The regression coefficient for CPAP related to log odds ratio for postoperative pneumonia was 0.61 (95 % CI being 0.02–1.21) (SE = 0.30, P = 0.042). The heterogeneity for the above model was 8.14 %, τ 2 being 0.0787. The regression curve is shown in Fig. 3.

Fig. 3
figure 3

Regression on CPAP value for postoperative pneumonia

Full size image

CPAP Duration

The CPAP duration was negatively related to odds ratio of postoperative pneumonia. Overall regression coefficient derived from the included 5 trials was −0.124 (95 % CI being −0.37 to 0.12) (SE = 0.12, P = 0.311). Although a negative relationship between CPAP duration and log odds ratio of pneumonia was evident with intercept at 0.11, the numbers failed to achieve statistical significance. The regression curve with 95 % CI is shown in Fig. 4.

Fig. 4
figure 4

Regression on CPAP duration for overall postoperative pneumonia

Full size image

Postoperative Atelectasis

Incidence of Postoperative Atelectasis

Ten subgroups reported atelectasis rates. The incidence of atelectasis in CPAP group was 25.90 % (95 % CI being 19.73–33.99 %) and in control group it was 35.27 % (95 % CI being 29.70–41.80 %). The heterogeneity for above comparison was 0 % with a Q value of 6.87. The number needed to treat to prevent single episode of atelectasis was 9.83. The overall effect size along with confidence interval is shown in Fig. 5. Effects of continuous versus intermittent CPAP were analyzed using subgroup analysis. Both CPAP groups showed protective effect for atelectasis. Continuous CPAP offered higher advantage with an odds ratio of 0.33 (95 % CI being 0.16–0.68, P = 0.003) compared to intermittent CPAP that showed an odds ratio of 0.67 (95 % CI being 0.37–1.19 P = 0.170—non-significant).

Fig. 5
figure 5

Pooled odds ratio for postoperative atelectasis (CPAP vs Control)

Full size image

Meta-regression

Random effects modelling with “method of moments” was used for calculation. Values computed were 2-tailed P values.

CPAP Value

CPAP values correlated negatively to the log odds ratio of atelectasis. The slope of regression curve was −0.012 (95 % CI being −0.21 to 0.19) with SE = 0.10, P value being 0.90. Such a small value of slope cannot predict strong correlation between the two variables. The values were reported for 9 subgroups, and the curve with 95 % CI is shown in Fig. 6.

Fig. 6
figure 6

Regression on CPAP value for postoperative atelectasis

Full size image

CPAP Duration

Duration was available for 9 subgroups and it correlated negatively with log of odds of postoperative atelectasis. The regression coefficient was found to be −0.047 (95 % CI being −0.18 to 0.08) with a P = 0.482 (Figure 7). The number of studies that used continuous CPAP were 5 and intermittent were 6. Comparison between these small subgroups in terms of odds ratio to control group showed statistically similar odds ratio.

Fig. 7
figure 7

Regression on CPAP duration for overall postoperative atelectasis

Full size image

Postoperative Pulmonary Complications

Incidence of Postoperative Pulmonary Complications

Thirteen subgroups reported overall incidence of pulmonary complications. The incidence was significantly higher in control group, at 31.66 % (95 % CI being 27.28–36.38 %) compared to CPAP group at 19.95 % (95 % CI being 16.31–24.17 %). The heterogeneity for above comparison was 0 % with Q value being 7.57. The number needed to treat to prevent a single episode of postoperative pulmonary complication was calculated to be 8.54. The pooled odds ratio for likelihood of complications in CPAP group in comparison to control group along with 95 % CI is shown in Fig. 8. We further performed a subgroup analysis based upon the nature of CPAP used. Continuous CPAP demonstrated a higher beneficial effect with an odds ratio of 0.30 (95 % CI being 0.16–0.53, P < 0.001) compared to intermittent CPAP with an odds ratio of 0.44 (95 % CI being 0.24–0.58, P = 0.006), both over the control group.

Fig. 8
figure 8

Pooled odds ratio for postoperative pulmonary complications (CPAP Vs Control)

Full size image

Meta-regression

Random effects modelling with “method of moments” was used for calculation. Values computed were 2-tailed P values.

CPAP Value

The curve relating CPAP strength to log odds ratio of pulmonary complications was nearly a horizontal line with regression coefficient being −0.015 (95 % CI being −0.206 to 0.176). Twelve subgroups reported the desired value, and the overall P value failed to achieve a statistical significance (P = 0.877). The regression curve with 95 % CI is shown in Fig. 9.

Fig. 9
figure 9

Regression on CPAP value for postoperative pulmonary complications

Full size image

CPAP Duration

Relevant values were found in 12 studies for the calculation. The regression coefficient was found to be -0.016 (95 % CI being −0.123 to 0.090) with a P = 0.675 (Figure 10).

Fig. 10
figure 10

Regression on CPAP duration for postoperative pulmonary complications

Full size image

Study Quality and Publication Bias

Quality assessment for bias in the included studies was carried out as per the other published meta-analysis and the guidelines laid by the Cochrane Collaboration, and the results are shown in Fig. 11.

Fig. 11
figure 11

Risk of bias-summary of all analysed studies

Full size image

Funnel plot was constructed, and distribution of studies was accessed for overall postoperative pulmonary complications. The publication bias was further estimated using the Egger’s test for all the parameters analyzed. The intercept at X axis was seen at 0.37 with a P value of 0.61. Thus, publication bias was unlikely in the present meta-analysis (Figure 12).

Fig. 12
figure 12

Funnel plot estimating the publication bias of the studies

Full size image

Adverse Events

The reporting of adverse events was not consistent across all the trials and therefore a statistical comparison could not be made. Table 3 shows the adverse events (wherever available) across individual trials along with frequency and P value as reported by the authors.

Table 3 Summary of complications reported in the included trials
Full size table

Discussion

Our meta-regression clearly demonstrates the ability of CPAP to reduce the incidence of atelectasis and other postoperative pulmonary complications, especially in patients undergoing major abdominal surgeries. This finding is in agreement with the results of many other studies and meta-analyses. However, by using meta-regression, we have demonstrated the futility of using increasing levels of PEEP for the purposes of reducing postoperative pulmonary complications. Apart from a lack of any demonstrable benefits, such an approach might result in escalation of other complications. We have also demonstrated that continuous CPAP is preferable to intermittent CPAP, especially in preventing postoperative atelectasis and reducing overall pulmonary complications.

Hypoxemia after major abdominal surgery is usually related to atelectasis and (or) pneumonia [12]. In addition to prolongation of the hospital stay, these complications can be a major cause of postoperative mortality [13] in about 25 % of patient population. The measures employed to reduce the incidence of postoperative pulmonary complications are application of CPAP in the postoperative period, use of chest physiotherapy including incentive spirometry, preoperative respiratory muscle training, limiting the percentage of inspired oxygen-both at induction and maintenance of anaesthesia, appropriate use of CPAP [13–31] and early mobilization. Among these measures, use of CPAP in the postoperative is most studied. Need for sedation, varying levels of tolerance and natural reluctance on the part of patients to accept such measures after an elective surgery are some of the factors that need to be addressed. As a result, the optimum CPAP and its place in the postoperative respiratory armamentarium need to be redefined.

The most important finding of our meta-analysis is the demonstration that increasing values of CPAP offered little additional benefit in preventing postoperative pneumonia. The positive correlation between increasing levels of CPAP and the likelihood of developing postoperative pneumonia is possibly multifactorial. Higher CPAP values are often associated with increasing aerophagia due to delivery of air under pressure. Aerophagia is known to be associated with gastroesophageal reflux (GERD) in patients with OSA treated with CPAP [18]. GERD is recognized to increase the risk of pneumonia (OR = 1.15) by acid reflux and micro-aspiration resulting in airway inflammation [3]. In fact, organizing pneumonia associated with GERD, often relapses and is more aggressive, [32]. All these factors increase the likelihood of possible loss of significant pneumonia protective effect associated with increasing degree of CPAP in the immediate postoperative period.

Evidence regarding the various sedation techniques employed to increase the tolerance of postoperative CPAP administration is limited. Letteri et al., demonstrated an improved patient compliance with increasing sedation, especially during titration phase of CPAP application [33]. It is reasonable to deduce that higher CPAP levels require deeper sedation to achieve similar patient acceptability. However, deeper sedation is likely to impair the cough reflex, resulting in a higher incidence of possible aspiration. Even without the use of sedation, the risk of aspiration with non-invasive ventilation has been found to be as high as 5 %. CPAP use can impair sputum expectoration and higher driving pressures would make the process even more difficult [34]. It is possible that benefits of CPAP in the postoperative setting are directly related to its ability to prevent airway collapse; however, the progression of these beneficial effects is not linear due to the factors discussed.

The conventional disadvantages of higher levels of CPAP are also well known. Patient discomfort from high flows and increased work of breathing due to exhalation against excessive expiratory pressure are associated with poor patient compliance, that might have contributed to our findings. Inadequate humidification is another shortcoming of using CPAP during the postoperative period. Prolonged use of CPAP in patients with OSA is known to be associated with increasing upper airway complications, possibly related to humidification, hygiene and skin pressure ulcers [35]. It is reported that, the use of CPAP with or without humidification is associated with higher pulmonary complications in patients with OSA [36]. Some or all of these factors could have contributed to our mathematical finding that higher degrees of CPAP do not withstand the scrutiny of cost-benefit analysis.

Our results also favour the use of immediate postoperative continuous CPAP over delayed intermittent CPAP for achieving higher protective benefits. Considering that inflammation is a cascade reaction, it is better prevented than contained. It is possible that, patients who received continuous CPAP (all studies within this group used CPAP in the immediate postoperative period) did not develop any atelectatic changes, and the cascade of inflammation was averted. However, in the intermittent CPAP group, as a result of CPAP free intervals, inflammatory cascade that was initiated could not be contained. As a result, the patients did realize the full benefits of protective/preventive CPAP intervention. However, we do accept that intermittent CPAP may be more acceptable for some of the patients. The use of newer modes of CPAP titration in OSA patients is evolving. Auto-adjustable CPAP devices (APAP) are increasingly being used and studied with the aim of delivering lower pressures and thus improving patient compliance [37, 38]. It would be interesting to evaluate the effects of APAP or titrated CPAP on the frequency and severity of pneumonia in patients without OSA.

Although prolonged CPAP use was associated with falling incidence of pneumonia, the regression was not statistically significant. Evidence from OSA patients suggests, adherence to CPAP results in improved cardiovascular and metabolic outcomes [39, 40]. Aaron et al. in his meta-analysis noted a decrease in levels of inflammatory markers in the OSA patient group compliant with CPAP [41]. Paschalis et al. found similar results with the use of CPAP, even with after 6 h of use [42]. No significant downward trend in IL-6 levels was observed in these studies with the use CPAP. Inflammatory markers such as peak IL-6 and CRP levels were found to be reliably associated with the magnitude of injury after elective surgery [43]. The interplay between inflammatory mediators and duration of CPAP application in non-OSA patients possibly exists, however, it is difficult to quantify currently.

We were also able to demonstrate qualitatively that prophylactic use of CPAP lowers the overall incidence of both atelectasis and pneumonia postoperatively. A physiological explanation of such clinical benefits is a possible migration of the “equal pressure point” towards more proximal airways [44]. The resulting “splinting” of the smaller airways improves the gas exchange, while the more patent conducting airways allow better pulmonary toileting. The direct benefits of CPAP on the postoperative pulmonary complications are beyond doubt, as heterogeneity is negligible in our analysis, which incidentally is a major strength of our study. By including the studies that involved only upper abdominal surgeries and excluding surgeries involving sternotomy or thoracotomy, we could reduce the heterogeneity to negligible levels (Table 2). Considering that CPAP requirements are likely to be higher in thoracic surgeries, it was necessary to exclude them.

Limitations

Heterogeneity is often the major limitation of any meta-analysis [45, 46]. In contrast to many meta-analyses, our results had negligible heterogeneity. However, we could not quantitatively relate the duration of CPAP to desired outcomes, because of significant variability among the included trials. Thus, a statistically valid conclusion in terms of CPAP duration could not be established. Although the use of prophylactic CPAP was a common denominator in all the included studies, the respective aims and methods were varied and resultantly, studies reporting multiple varying values of CPAP were excluded. Even though we have demonstrated that increasing degree of CPAP is of limited benefit and may be even detrimental, the factors responsible for such a finding are not clear in the studies analyzed. Furthermore, we were only able to derive a correlation between CPAP degree and incidence of pneumonia. We however could not define the exact values for optimal CPAP for individual set of patients. These CPAP values should be adjusted by treating anaesthesiologist on individual case basis keeping in mind that increasing CPAP does not necessarily increase protective effects towards pneumonia. The reporting of long-term outcomes like duration of ICU/hospital stay and cost benefit was very inconsistent among the included trials, and as a result, a comparative analysis for these factors could not be performed. Another limitation arose from the nature of intervention being evaluated. CPAP is a treatment strategy that cannot be subjected to true blinding (although trials report the use of sham CPAP masks) and the possibility of reporting bias cannot be eliminated. Furthermore, it must be borne in mind that a small number of studies qualified for a regression analysis. This could have affected our results and lowered the strength our conclusions.

Conclusions

We conclude that higher levels of CPAP are of questionable benefit and present a variety of challenges especially in preventing postoperative pneumonias. Additionally, the findings of our meta-regression demonstrate significant reduction in atelectasis, pneumonia and postoperative pulmonary complications with the use of prophylactic CPAP without heterogeneity. We however suggest that the clinician should use his/her judgement on a case to case basis, before escalating the CPAP level for the purposes of reducing postoperative pulmonary complications. A large multicentre study using different levels of CPAP might conclusively answer this question.