High inspired oxygen versus low inspired oxygen for reducing surgical site infection: a meta‐analysis

Abstract

To perform a meta‐analysis of published literature to assess the role of high‐concentration inspired oxygen in reducing the incidence of surgical site infections (SSIs) following all types of surgery, a comprehensive search for published randomized controlled trials (RCTs) comparing high‐ with low‐concentration inspired oxygen for SSIs was performed. The related data were extracted by two independent authors. The fixed and random effects methods were used to combine data. Twelve RCTs involving 6750 patients were included. Our pooled result found that no significant difference in the incidence of SSIs was observed between the two groups, but there was high statistic heterogeneity across the studies [risk ratio (RR): 0·91; 95% confidence interval (CI): 0·72–1·14; P = 0·40; I ² = 54%]. The sensitivity analysis revealed the superiority of high‐concentration oxygen in decreasing the SSI rate (RR: 0·86; 95% CI: 0·75–0·98; P = 0·02). Moreover, a subgroup analysis of studies with intestinal tract surgery showed that patients experienced less SSI when high‐concentration inspired oxygen was administrated (RR: 0·53; 95% CI: 0·37–0·74; P = 0·0003). Our study provided no direct support for high‐concentration inspired oxygen in reducing the incidence of SSIs in patients undergoing all types of surgery.

Introduction

Surgical site infections (SSIs) remain one of the most common nosocomial infections in surgical patients 1. The overall incidence of SSIs is estimated to be 2–5%, accounting for about 20% of all health care‐associated infections 2, 3. The development of SSIs is associated with significant increased postoperative morbidity, prolonged hospitalisation and expensive costs of medical care or even with higher risk of mortality in patients undergoing surgical procedures 4, 5. The causes of SSIs can be attributed to the site and complexity of surgery, the patient's perioperative physiological status and the administration of prophylactic antibiotics.

Although the factors are often multifactorial, it is well known that tissue oxygen tension, which mediates the oxidative killing of bacteria by neutrophils, is an important risk factor for the development of SSI. In the last decade, many investigators have paid great attention to increasing the inspired oxygen concentration to evaluate its effect on the incidence of SSIs in patients undergoing surgery 6, 7, 8, 9, 10, 11, 12, 13, 14. However, the conclusions have reached no consensus. As a result, systematic reviews were continuously performed to statistically analyse the study data collected from individual trials 15, 16, 17, 18, 19, 20, 21, 22. A recent meta‐analysis focusing on the rate of SSIs after caesarean sections included four studies 23. One study by Gardella et al. 11 has already been analysed as an eligible trial in previous reviews, while the other three studies 24, 25, 26 were never included in those meta‐analyses that were designed without the limitation of a specific surgical type.

Therefore, the aim of the present study was to revaluate the overall effect of high‐concentration inspired oxygen on the incidence of SSIs following all types of surgery.

Materials and methods

We carried out this study in accordance with the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) statement 27.

Literature search

We utilised the recommended methods of the Cochrane Collaboration to search literature 28. The electronic databases searched included MEDLINE, EMBASE, Cochrane Central Register of Controlled Trials and Web of Science. We used the following Medical Subject Headings (MeSH) terms and text words without language restrictions: ‘oxygen’, ‘hyperoxia’, ‘infection’ and ‘surgical wound infection’. Two of us reviewed the titles and abstracts (when available) identified in the search independently. All references cited in the articles were also searched by hand to identify additional publications. The latest search was conducted on 1 August 2015.

Study selection

We evaluated the eligible studies for all the following conditions: (i) the publication was an RCT; (ii) it compared perioperative high‐concentration (≥50%) with low‐concentration (<50%) oxygen in patients undergoing surgery; and (iii) it reported SSIs as an outcome. If two or more studies from the same institution were identified, the most recent or the most informative was selected unless they were reports from different time periods or if the data of overlapping patients could be subtracted. Non‐randomized or non‐comparative studies, reviews and case reports were excluded. If it was impossible to calculate SSIs from the published results, then the assessed study was excluded.

Data extraction

Two authors independently extracted data from the final full‐text articles. Any disagreements about data extraction were resolved by consensus or arbitration by a third reviewer. The following data were extracted from each study using standardised data extraction forms: first author, year of publication, country, study interval, patient characteristics (female sex, mean age and cases), type of surgery, use of prophylactic antibiotics and duration of follow‐up. The primary outcome in the systematic analyses was the incidence of SSI. The methodological quality of the included RCTs was assessed and calculated using the Jadad scale, which evaluates a study in terms of the description of randomisation, double‐blinding and withdrawals or dropouts 29.

Statistical analysis

The extracted outcomes were pooled as an estimate of the overall intervention effect for the meta‐analysis performed using Review Manage, version 5.1.0 (The Cochrane Collaboration, London, United Kingdom). We pooled data by using the RR with the corresponding 95 % CIs in the Mantel–Haenszel (M–H) method. The statistical heterogeneity among studies was assessed with the Q‐test and I ² statistics 30. As previously described 31, if there was no significant heterogeneity, the fixed effects model would be used; otherwise, the random effects model would be applied. Sensitivity analysis was performed by excluding studies to remove heterogeneity when it was significant. In addition, subgroup analysis was planned for studies with different durations of follow‐up and types of surgery. An estimate of potential publication biases was evaluated by the funnel plot 32. Two‐sided P values were used throughout.

Results

Trial flow

A total of 1946 records were identified from initial literature searches. After the elimination of 1917 irrelevant records, 29 citations remained and were retrieved for a more detailed evaluation. Of these, two non‐randomized studies 33, 34, one non‐comparative study 35 and nine meta‐analyses 15, 16, 17, 18, 19, 20, 21, 22, 23 were excluded. Three studies with no SSI data were also excluded 36, 37, 38. One trial by Anthony et al. 39 was not included because its study group did not involve only high‐concentration oxygen intervention. Another trial by Whitney et al. 40 was excluded because of its comparison being between low‐concentration oxygen and room air. Therefore, 12 studies 6, 7, 8, 9, 10, 11, 12, 13, 14, 24, 25, 26 matched the criteria for inclusion in the meta‐analysis (Figure 1).

Figure 1.

Study flow diagram. RCT, randomized controlled trial.

Study characteristics

The major characteristics of the included studies, along with the quality assessment scores, are summarised in Table 1. All of the trials were published in English‐language journals between 2000 and 2013. Of the 12 eligible studies, 5 trials 7, 11, 24, 25, 26 were performed in America, 4 6, 8, 12, 14 in Europe, 2 9, 13 in Asia and 1 10 in 19 participating sites involving Oceania, Asia and Europe. There were a total of 6750 cases in our analysis with mean age that ranged from 27·5 years to 69 years across the studies. Perioperative prophylactic antibiotics were applied in all the trials except Williams' study 24, which did not describe this item. With regard to surgical type, 4 of the 12 trials 6, 8, 9, 13 involved patients undergoing intestinal tract surgery. Another four trials 11, 24, 25, 26 focused on caesarean delivery. The remaining four trials 7, 10, 12, 14 investigated outcomes on multiple surgical procedures as shown in Table 1. The reported length of follow‐up in all studies varied from 14 to 42 postoperative days. The trial 6 that reported SSIs within 15 days after surgery was included in subgroup analysis for 14 days of follow‐up.

Table 1.

Study Characteristics

Author	Year	Country	Study interval	Female sex		Age (years)		Cases		Comparison	Type of surgery	Prophylactic antibiotics	Follow‐up (days)	Jadad score
				High	Low	High	Low	High	Low
Greif et al. 6	2000	Austria and Germany	1996–1998	107	113	57	57	250	250	FiO2 80% versus FiO2 30%	Elective colorectal surgery	Yes	15	5
Pryor et al. 7	2004	USA	2001–2003	46	46	54	57	80	80	FiO2 80% versus FiO2 35%	Major abdominal surgery	Yes	14	5
Mayzler et al. 9	2005	Israel	2001–2002	9	7	67	69	19	19	FiO2 80% versus FiO2 30%	Elective colorectal surgery	Yes	30	2
Belda et al. 8	2005	Spain	2003–2004	77	52	64·2	62·3	148	143	FiO2 80% versus FiO2 30%	Elective colorectal surgery	Yes	14	5
Myles et al. 10	2007	Australia, China, New Zealand, Singapore, Saudi Arabia and UK	2003–2004	464	495	55·8	54·6	997	1015	FiO2 80% versus FiO2 30%	Elective or emergency non‐cardiothoracic surgery	Yes	30	5
Gardella et al. 11	2008	USA	2001–2007	69	74	31	28	69	74	FiO2≥80% versus FiO2 30%	Caesarean delivery	Yes	14	5
Meyhoff et al. 12	2009	Denmark	2006–2008	402	408	64	64	690	705	FiO2 80% versus FiO2 30%	Acute or elective laparotomy	Yes	14	5
Williams et al. 24	2009	USA	NA	77	83	NA	NA	77	83	FiO2 80% versus FiO2 30%	Caesarean delivery	NA	NA	5
Bickel et al. 13	2011	Israel	2006–2009	27	30	28·5	27·6	107	103	FiO2 80% versus FiO2 30%	Open appendectomy	Yes	14	4
Scifres et al. 25	2011	USA	2008–2010	288	297	27·5	27·8	288	297	FiO2 80% versus FiO2 25‐30%	Cesarean delivery	Yes	30	3
Thibon et al. 14	2012	France	2003–2007	208	184	52·1	51·8	226	208	FiO2 80% versus FiO2 30%	Elective abdominal, gynecological, and breast surgery	Yes	30	4
Duggal et al. 26	2013	USA	2006–2010	416	415	29·5	29·2	416	415	FiO2 80% versus FiO2 30%	Caesarean delivery	Yes	42	5

High, high‐concentration oxygen group; Low, low‐concentration oxygen group; FiO₂, fraction of inspired oxygen; NA, not available.

All of the included trials described the mode of randomisation in detail. Except two trials 9, 25, the description of double‐blinding was reported in all trials along with the appropriate method. The statement regarding withdrawals or dropouts was included in nine trials 6, 7, 8, 10, 11, 12, 24, 25, 26. Overall, the majority of studies included in this analysis were regarded as better quality trials according to their Jadad score.

Quantitative data synthesis

Twelve RCTs included in this analysis all reported the incidence of SSIs with a total of 3362 patients receiving high‐concentration inspired oxygen intervention and 3388 patients in the control group. Overall, an SSI incidence of 11·2% was observed in the study group and 12·7% was observed in the control group. Our pooled result showed that no significant difference was found in SSIs between the two groups, but there was high statistic heterogeneity across the studies (RR: 0·91; 95% CI: 0·72–1·14; P = 0·40; I ²= 54%) (Table 2, Figure 2). To remove statistical heterogeneity, a sensitivity analysis was carried out by excluding the trial of Pryor et al. 7 and showed the impact of high‐concentration oxygen in decreasing the incidence of SSIs as compared with the control group (RR: 0·86; 95% CI: 0·75–0·98; P = 0·02; I ²= 44%) (Table 2).

Table 2.

Meta‐analyses and subgroup analyses of surgical site infection rate comparing high‐concentration oxygen versus low‐concentration oxygen

Outcome	No. of studies	No. of participants	Statistical method	Effect estimate	P for HG	I 2 (%)	P
Total SSI	12 6, 7, 8, 9, 10, 11, 12, 13, 14, 24, 25, 26	6750	RR (M–H, Random, 95% CI)	0·91 (0·72, 1·14)	0·01	54	0·40
Sensitivity analysis	11 6, 8, 9, 10, 11, 12, 13, 14, 24, 25, 26	6590	RR (M–H, Fixed, 95% CI)	0·86 (0·75, 0·98)	0·06	44	0·02
Subgroup analysis by follow‐up
Postoperative 14 days	6 6, 7, 8, 11, 12, 13	2690	RR (M–H, Random, 95% CI)	0·88 (0·57, 1·36)	0·001	75	0·57
Postoperative 30 days	4 9, 10, 14, 25	3069	RR (M–H, Fixed, 95% CI)	0·85 (0·68, 1·07)	0·25	27	0·17
Subgroup analysis by surgical type
Intestinal tract surgery	4 6, 8, 9, 13	1039	RR (M–H, Fixed, 95% CI)	0·53 (0·37, 0·74)	0·84	0	0·0003
Caesarean delivery	4 11, 24, 25	1719	RR (M–H, Fixed, 95% CI)	1·22 (0·91, 1·65)	0·53	0	0·19
Multiple surgical procedures	4 7, 10, 12, 14	3992	RR (M–H, Random, 95% CI)	0·98 (0·71, 1·35)	0·04	63	0·90

CI, confidence interval; Fixed, fixed effects model; HG, heterogeneity; IV, inverse variance; M–H, Mantel‐Haenszel; MD, mean difference; Random, random effects model; RR, risk ratio.

Figure 2.

Meta‐analysis of total surgical site infection rate. CI, confidence interval.

Subgroup analysis

No significant difference of SSIs between groups was found in subgroup analyses by the duration of follow‐up, neither for 14 days (RR: 0·88; 95% CI: 0·57–1·36; P = 0·57; I ²= 75%) nor for 30 days (RR: 0·85; 95% CI: 0·68–1·07; P = 0·17; I ²= 27%). The overall effects of SSIs remained unchanged in another two subgroup analyses for caesarean delivery (RR: 1·22; 95% CI: 0·91–1·65; P = 0·19; I ²= 0%) and multiple surgical procedures (RR: 0·98; 95% CI: 0·71–1·35; P = 0·90; I ²= 63%). However, the reduction in SSIs was noted in the study group and with low intergroup heterogeneity in the subgroup of studies where intestinal tract surgery was performed (RR: 0·53; 95% CI: 0·37–0·74; P = 0·0003; I ²= 0%) (Table 2).

Publication bias

The funnel plot of the studies based on the incidence of SSIs is shown in Figure 3. The distribution of studies in the funnel plot was relatively symmetrical, suggesting that publication bias was not present.

Figure 3.

Funnel plot for publication bias. RR, risk ratio.

Discussion

SSIs represent a pervasive problem for surgical patients and surgeons all over the world. The majority of SSIs are superficial, involving skin and subcutaneous tissue, whereas deep, organ and space infections are frequently related with severer morbidity, higher health care cost and greater risk of mortality 41. To reduce the occurrence of SSIs, many preventive measures have been investigated by medical workers. One of these is hyperoxygenation, which has been recognised as a potential preventive approach for many years 6.

The primary defense against wound pathogens after contamination is the oxidative killing of neutrophils, which depends on the production of bactericidal superoxide radicals from oxygen 42. Because surgical wounds inevitably disrupt the local vascular supply, the tissue oxygen tension is often low, and this may weaken the effect of oxidative killing.

Theoretically, a higher inspiratory oxygen concentration may increase perioperative arterial and wound oxygen tension 43. In 2000, Greif et al. 6 published a study and reported a significant reduction (5·2% versus. 11·2%) in the incidence of SSIs when 80% rather than 30% oxygen was administered in patients undergoing colorectal resection during the perioperative period. This finding was supported by Belda and his research team 8. Moreover, Myles et al. 10 demonstrated the superior capability of supplemental inspired oxygen in decreasing the frequency of SSIs (7·7% versus. 10%) in patients experiencing non‐cardiothoracic surgery. Nevertheless, this positive evidence was gradually reported by other medical researchers. Several trials have found that there was no relationship between the incidence of SSIs and supplemental inspired oxygen when patients underwent intestinal tract surgery, caesarean delivery or multiple surgical procedures 9, 11, 12, 13, 14, 24, 25, 26. Conversely, a trial by Pryor et al. 7 observed that the routine use of high‐concentration oxygen in a general surgical population resulted in a higher rate of SSIs (25·0% versus. 11·3%) as compared with low‐concentration oxygen.

In this study, our pooled results failed to demonstrate the advantage of supplemental inspired oxygen in reducing SSIs when all types of surgery were considered together. This finding was similar to the results of three published meta‐analyses 18, 20, 21. However, it should be noted that a significant heterogeneity was presented among all included RCTs. Then, the sensitivity analysis detected that a trial from Pryor et al. 7 has no standardised perioperative care, and this may be recognised as a possible explanation for the heterogeneity observed across studies. Our sensitivity analysis, in accordance with the other systematic reviews 17, 22, provided support that perioperative supplemental oxygen reduces SSIs in all patient populations. The subgroup analysis in this study further clarified that the patients receiving intestinal tract surgery would benefit most from such perioperative intervention, as previous meta‐analyses concluded 15, 16, 19.

The major limitation of our study is the intergroup heterogeneity, which deserves discussion. Clinical heterogeneity among studies principally includes the definition of SSIs, the duration of follow‐up, the type of surgical procedures and the patient populations. According to the definitions of the Centers for Disease Control and Prevention 44, all SSIs occurring within 30 days of surgery were classified as superficial incisional, deep incisional and organ/space infection with corresponding criteria. The lack of a uniform definition of SSIs across all eligible studies may have contributed more or less to the heterogeneity. Besides, half of the studies in our analysis had less than 30 days of follow‐up, and this may influence the overall effect of supplemental oxygen on reducing SSIs. Given the presence of heterogeneity in our study, which is common to all meta‐analytical studies, we carried out the sensitivity analysis and subgroup analysis to maximise the ability to infer causation between intervention and outcome. Future workers should pay more attention to implementing standardised procedures to evaluate the clinical effect of supplemental oxygen on the occurrence of SSIs.

In conclusion, our meta‐analysis provided no direct support for high‐concentration inspired oxygen to reduce the incidence of SSIs in patients undergoing all types of surgery. Interestingly, our sensitivity analysis did find a significant reduction of SSIs in supplemental oxygen groups, and one subgroup analysis demonstrated that the patients receiving intestinal tract surgery would benefit most from such intervention. Nowadays, the clinical guidelines, which advocate that patients having surgery are kept optimally perfused and oxygenated during surgery, are considerably important for medical workers. We believe that our meta‐analysis would provide scientific evidence for the writers of these guidelines when they decide what concentration of inspired oxygen is optimal for patients having surgery.