Inspiratory Oxygen Fraction and Postoperative Complications in Obese Patients: A Subgroup Analysis of the PROXI Trial

• In the Danish multicenter PROXI Trial involving 1,400 patients, no significant reduction in surgical site infection was observed when 80% oxygen was given during and 2 h after abdominal surgery compared with use of 30% oxygen in all patients, although obese patients may be at high risk

• In this planned analysis of the obese subgroup (body mass index ≥ 30 kg/m², n = 231) of the PROXI Trial, there was no significant difference in the frequency of surgical site infection or postoperative pulmonary complications

OBESITY is associated with a high risk of postoperative complications, including atelectasis,1pneumonia,2thromboembolic events, cardiac complications, wound hernia, anastomotic leak, and surgical site infection (SSI).3The risk of complications is increased further in obese patients with modified metabolic syndrome.4 

Arterial oxygenation and subcutaneous perfusion have an important role in the host defense against infection.5The bactericidal activity of neutrophils is mediated by oxidative killing,6which is dependent on the partial pressure of oxygen in the tissue. Subcutaneous oxygen tension (Psqo²) is often low during surgery, and this is significantly related to the development of SSI.7Psqo²can be increased with a high inspiratory oxygen fraction (Fio²), and two trials have found significant reduction in SSI when patients are given 80% compared with 30% oxygen during the perioperative period.8,9However, the PROXI Trial, involving 1,400 patients, recently showed no significant reduction in the frequency of SSI.10 

The arterial oxygen tension is reduced more in obese than in lean patients during general anesthesia as a consequence of atelectasis and increased shunt fraction.1,11Moreover, obesity increases the size of individual fat cells without increasing blood flow,12resulting in subnormal total blood flow in relation to body weight13and relatively hypoperfused fat tissue,11but cardiac output, circulating blood volume, and resting oxygen consumption are increased.14This contributes to a more pronounced reduction of Psqo²in obese patients,11increased risk of SSI,3and thus a potential for enhanced effect of a high Fio².

The aim of this planned subgroup analysis was to evaluate the effect of a high perioperative Fio²(80%) compared with 30% in obese patients undergoing laparotomy. We hypothesized that 80% oxygen would reduce the frequency of SSI. We also assessed the association between perioperative Fio²and the frequency of pulmonary complications and other complications.

The study was a planned subgroup analysis of the Danish multicenter PROXI Trial15and was approved by The Danish Medicines Agency and the regional ethics committee (NCT00364741, De Videnskabsetiske Komiteer for Region Hovedstaden, Hillerød, Denmark).16Written informed consent was obtained from all patients, and they were included between October 8, 2006, and October 6, 2008.

Eligible patients were 18 yr or older, had a preoperative body mass index (BMI) ≥ 30 kg/m², and were scheduled for acute or elective laparotomy.15Exclusion criteria were inability to give informed consent, chemotherapy for malignancy within 3 months, surgery performed under general anesthesia within 30 days, and preoperative arterial oxygen saturation less than 90% without supplemental oxygen as assessed by pulse oximetry.

Patients were randomized by a central interactive voice-response system at the Copenhagen Trial Unit to an Fio²of either 0.80 (the 80% oxygen group) or 0.30 (the 30% oxygen group) using the following stratification variables: study center, diabetes mellitus, and acute or elective surgery.

The trial protocol15emphasized an optimal perioperative care, including epidural analgesia, adequate temperature and glucose control, appropriate and timely prophylactic antibiotics, absence of preoperative oral bowel preparation, and standardized anesthesia without nitrous oxide. Perioperative fluids were given only to replace measured or calculated deficits aiming at body weight increase of less than 1 kg. Blood loss was replaced 1:1 with colloids, and blood transfusion was initiated if blood loss exceeded 20 ml/kg.15 

Patients were preoxygenated with an Fio²of 1.0 until tracheal intubation, after which patients were given the allocated Fio²until the end of surgery, when an Fio²of 1.0 was given immediately before extubation. The patients were ventilated to assure normocapnia.15In both groups, Fio²was increased to ensure arterial oxygen saturation above 94% and arterial oxygen tension above 9 kPa (68 mmHg). Positive end expiratory pressure was used at a level chosen by the attending anesthetist. Alveolar recruitment maneuvers were not routinely used but were allowed if the attending anesthetist thought they were clinically indicated.

The first 2 h after surgery, patients randomized to the 80% oxygen group breathed a Fio²of 0.80 via  a nonrebreathing facemask with a reservoir (High Concentration Oxygen Mask, Intersurgical Ltd., Wokingham, United Kingdom) and an oxygen flow of 14 l/min and air flow of 2 l/min. The patients randomized to the 30% oxygen group received a mixture of oxygen (2 l/min) and air (14 l/min) through an identical nonrebreathing facemask.15Two hours after surgery, supplemental oxygen was administrated according to clinical practice.

The primary outcome was SSI within 14 days of surgery, defined according to the criteria by Centers for Disease Control and Prevention as superficial or deep wound infection or intraabdominal organ/space infection.17The secondary outcomes were atelectasis within 14 days, pneumonia within 14 days (according to the criteria by Centers for Disease Control and Prevention),§and respiratory failure within 14 days (defined as the need for controlled ventilation or arterial oxygen saturation less than 90% despite supplemental oxygen). Tertiary outcomes were localization of SSI, admission to the intensive care unit within 14 days (if not part of routine postoperative care), another abdominal operation for any reason within 14 days, duration of postoperative hospitalization, wound-related adverse events, any adverse event, any serious adverse event, and mortality within 30 days.

The surgical investigator assessed all outcomes daily in the postoperative period, and a follow-up visit was conducted between postoperative days 13 and 30. The attending physician examined patients with symptoms of pulmonary complications according to clinical practice, including chest radiographs or computed tomography, when relevant. All chests radiographs and computed tomography were evaluated for infiltrate and atelectasis by the attending radiologist, who was blinded to allocation. The group allocation was also blinded to patients, outcome assessors, statisticians, the surgical team, and staff on the wards.15 

Patients with the following major protocol deviations were not included in the per-protocol analysis16: did not meet the inclusion criteria, fulfilled an exclusion criterion, Fio²above 0.60 for more than 1 h in the 30% group, Fio²less than 0.60 for more than 1 h in the 80% oxygen group, oxygen mask used less than 1 h, no in-hospital evaluation of the outcomes for 4 consecutive days or more, no follow-up visit between postoperative days 13 and 30, and unblinding of outcome assessors.

The complete statistical analysis plan is described in the trial protocol.15A univariate analysis was carried out for the primary and secondary outcomes and in a multivariate analysis, and the intervention effect was assessed after adjustment for the stratification variables as well as the design variables: chronic obstructive pulmonary disease, daily smoking, and surgical incision extending above the umbilical transversal. The tertiary outcomes were reported without statistical analyses. All intervention effect estimates were reported with 95% confidence limits and a two-sided P  value of <0.05 was considered to indicate statistical significance. Analyses were performed using R version 2.8.0.∥

We estimated that the frequency of SSI would be 30% among obese patients,3and the frequency of obesity would be 20% among all patients included in the PROXI Trial, for which a total sample size of 1,400 patients was required.15We expected a high Fio²to be associated with a relative risk reduction of 50% in SSI in obese patients and calculated that we would have 80% power to detect this in our subgroup analysis with a 5% risk of type 1 error and 10% dropout.

A total of 213 patients had a BMI ≥ 30 kg/m²(fig. 1). Demographic and perioperative characteristics were similar in the two groups (table 1).18,19The median (5–95% range) BMI was 34 kg/m²(30–44) and 33 kg/m²(30–41) in patients allocated to the 80% and 30% oxygen groups, respectively.

Surgical site infection occurred in 32 of 102 (31%) versus  29 of 111 (26%) patients in the 80% and the 30% oxygen groups, respectively (P = 0.40, table 2). The difference between the 80% and 30% oxygen groups in SSI was 5% (95% CI, 7–17%); in atelectasis, the difference was 3% (95% CI, −5 to 10%); in pneumonia, 1% (95% CI, −5 to 7%); and in respiratory failure, 3% (95% CI, −3 to 10%).

The incidence of SSI was 27% in obese patients (BMI = 30.0–34.9 kg/m²) and 32% in morbidly obese patients (BMI ≥ 35.0 kg/m², fig. 2). Atelectasis occurred in 9% versus  5%, pneumonia in 6% versus  3%, and respiratory failure in 6% versus  5% in the obese and morbidly obese patients, respectively.

Overall, 52% of the patients experienced adverse events in the follow-up period (table 3). Approximately 21% of patients had serious complications, with sepsis occurring in 3%; 30-day mortality was 2%. The median duration of postoperative hospitalization was 6 versus  5 days in the 80% and 30% oxygen groups, respectively (table 3). Anastomotic leak occurred in 2 of 31 (6%) versus  2 of 41 (5%) patients, and rupture of the abdominal fascia occurred in 9 of 102 (9%) versus  6 of 111 (5%) patients in the 80% and the 30% oxygen groups, respectively. Forty-three (20%) patients underwent abdominal reoperation. Twenty-four (11%) patients underwent second operation because of SSI, including five (2%) debridement procedures. Fifteen (7%) patients had rupture of the abdominal fascia, compared with 9 of 658 (1%) of patients of normal weight in the PROXI Trial.

The per-protocol analysis (n = 167, fig. 1) of the primary and secondary outcomes showed a result similar to that of the intention-to-treat analysis, with SSI occurring in 31 of 91 (34%) versus  24 of 76 (32%) patients in the 80% and the 30% oxygen groups, respectively (P = 0.73).

Contrary to our hypothesis, we did not find a reduction in the frequency of SSI in obese patients undergoing abdominal surgery when a high perioperative Fio²of 80% was given compared with when Fio²of 30% was given. The high perioperative Fio²was not associated with a significant increase in the frequency of pulmonary complication or other adverse events. In contrast, the primary and secondary outcomes all tended to be more common in patients allocated to 80% oxygen, but the power for the secondary outcomes was relatively low, as reflected in the wide confidence intervals. Therefore, we cannot exclude that a clinically important difference exists, but the detection of a difference in the incidence of respiratory failure between 4.5% and 8% would require a sample of nearly 2,000 patients with 80% power.

We included a total of 231 patients, of whom 73 were morbidly obese. We also analyzed the incidence of postoperative complications for the different BMI groups because it was thought there may be important differences in intervention effect, but this was not found. However, because only 34% of the patients were morbidly obese, the effect of high oxygen on this specific group needs to be investigated further.

The overall frequency of SSI was 29%, which is somewhat higher than had been reported previously in obese patients.20The frequency of SSI after laparoscopic and endoscopic bariatric surgery recently was found by Birkmeyer et al.  to be 3.2%,21whereas Merkow et al.  22found an incidence of approximately 15% after colectomy for cancer. The higher frequency in our study probably is caused by a high number of acute procedures, increased comorbidity, and intraoperative contamination. SSI was detected with thorough follow-up according to the sensitive criteria by the Centers for Disease Control and Prevention, which recently has been shown to be a suitable standard definition for identifying SSI.23Our observed frequency of SSI was comparable with a study by Cantürk et al.,  who reported SSI in 29% of 61 obese and extremely obese patients undergoing general elective surgery.3 

The frequency of SSI among the patients in the PROXI Trial (median BMI = 25 kg/m², interquartile range = 22–28 kg/m²) was 20%,10compared with the 29% found in this subgroup analysis of the obese patients (BMI ≥ 30 kg/m²). The distribution of SSI in the different weight groups among the patients in the PROXI Trial is shown in figure 2, which illustrates that the frequency of SSI appears to increase with increasing body weight. In addition to a higher frequency of SSI, the obese patients had a much higher frequency of rupture of the abdominal fascia, which underlines the compromised healing process in these patients. A recent retrospective study of 1,024 trauma patients showed a higher rate of nosocomial infections in obese patients and a significantly higher frequency of pneumonia and wound infections.24In that study, obese patients had a 4.7-fold higher risk of infection, and morbidly obese patients had an almost 6-fold higher risk of infection compared with patients with a BMI less than 30 kg/m². Obesity was still a risk factor for infection when controlling for age and comorbidity.

The increased risk of SSI among obese patients may also be caused by impaired immune system,25larger wound area, and longer operating time.2Another explanation is that the perioperative Psqo²is significantly reduced in obese patients during major abdominal surgery.11Even during supplemental oxygen administration, Psqo²remains at a lower than normal level of partial pressure, which is associated with a substantial risk of infection.11This increased risk is in accordance with the results of our trial and suggests that factors other than perioperative Fio²influence Psqo², such as hypercapnia,26body temperature,27supplemental fluid,28and the use of vasopressors and epidural analgesia.29However, these factors were all similar in the two groups. It is possible that long-term supplemental postoperative oxygen can reduce the incidence of SSI because it has been shown to significantly increase Psqo²when administered for an average period of 13 h after surgery.30However, the first hours after bacterial contamination traditionally have been recognized as critical for establishing the infection.31 

Atelectasis is an important perioperative pulmonary complication. Two mechanisms contribute to atelectasis formation: compression and absorption.32Atelectasis leads to a ventilation-perfusion mismatch33and may predispose the patient to other pulmonary complications. Within minutes, ventilation with 100% oxygen results in significantly larger areas of atelectasis than does ventilation with 80% oxygen.34However, one trial by Akça et al.  of 30 nonobese patients showed only a small, nonsignificant difference in the degree of atelectasis when an inspiratory perioperative Fio²of 80% was given compared with 30% oxygen.35The frequency of atelectasis among the patients in the PROXI Trial was 8% versus  7% in the 80% and 30% oxygen groups, respectively, and 6% in both the 80% and the 30% oxygen groups experienced pneumonia.10Another subgroup study of the PROXI Trial involving 35 patients showed no significant difference in change in oxygenation index or functional residual capacity when 80% oxygen was administered than when 30% oxygen was used.36 

Less evidence of the effect of ventilation with high oxygen is available in obese patients, but morbidly obese patients are more prone to perioperative atelectasis formation, and the atelectasis remains unresolved for a longer period after surgery than occurs in nonobese patients.1One recent study of 142 moderately obese patients found a minimal reduction in postoperative lung function when patients were given 80%, compared with 40%, oxygen during minor peripheral surgery.37The study showed a tendency toward better spirometry values in the low oxygen group and better arterial saturation during the first 2 h. We were not able to find any significant differences in the incidence of atelectasis, pneumonia, or other respiratory complications when given 80% oxygen compared with 30% oxygen. Moreover, there were no significant differences between obese and morbidly obese patients with regard to the secondary outcomes.

We examined only patients with pulmonary symptoms, and a chest radiograph or computed tomography was taken when relevant. Therefore, a difference in the frequency of subclinical atelectases could have been overlooked in this study, but if present, these subclinical atelectases did not result in significant differences in pneumonia or respiratory failure. One recent study found that a recruitment maneuver followed by positive end-expiratory pressure reduced atelectasis and improved oxygenation in morbidly obese patients.38However, a positive end-expiratory pressure of 10 cm H²O alone did not reduce atelectasis. We did not measure the use of recruitment maneuvers followed by positive end expiratory pressure; however, we believe the use of this combination was limited and thus not a bias to the results.

We used the preoperative calculated BMI as inclusion criteria. This formula could have been a limitation, particularly if applied to individuals with a great muscle mass, in whom body fat percent may be a more accurate measurement in regard to SSI.39For pulmonary complications, an evaluation of body fat distribution as upper body fat distribution or central obesity may have been a more accurate measurement.40,41It is also possible that some patients have had a significant amount of ascites or tumor mass removed during surgery, resulting in a falsely high BMI. However, BMI is easy to measure and allows for comparison with other trials. The change in body weight at the first postoperative day was similar in the two groups (table 1).

We included a large number of obese patients undergoing open abdominal surgery, including 18% acute procedures. We had a thorough follow-up with assessment of adverse events in all patients. Thus, we believe the results of this study are generalizable to a general obese surgical population undergoing laparotomy, including acute laparotomy and gynecologic cancer surgery.

In summary, our trial, which included more than 200 obese patients undergoing acute or elective abdominal surgery, did not find a reduction in the frequency of SSI when 80% oxygen was given. Moreover, we did not find a significant association between perioperative Fio²and the incidence of pulmonary complications or other adverse events.