Secondary Logo

Journal Logo

Intraoperative Hyperoxia Does Not Reduce Postoperative Pain: Subanalysis of an Alternating Cohort Trial

Cohen, Barak MD*,†; Ahuja, Sanchit MD*; Schacham, Yehoshua N. MD*,‡; Chelnick, David BS*; Mao, Guangmei PhD§; Ali-Sakr Esa, Wael MD; Maheshwari, Kamal MD, MPH*,‖; Sessler, Daniel I. MD*; Turan, Alparslan MD*,‖

doi: 10.1213/ANE.0000000000004002
Patient Safety: Original Clinical Research Report

BACKGROUND: Postoperative pain is common and promotes opioid use. Surgical wounds are hypoxic because normal perfusion is impaired. Local wound ischemia and acidosis promote incisional pain. Some evidence suggests that improving oxygen supply to surgical wounds might reduce pain. We therefore tested the hypothesis that supplemental (80% inspired) intraoperative oxygen reduces postoperative pain and opioid consumption.

METHODS: We conducted a post hoc analysis of a large, single-center alternating cohort trial allocating surgical patients having general anesthesia for colorectal surgery to either 30% or 80% intraoperative oxygen concentration in 2-week blocks for a total of 39 months. Irrespective of allocation, patients were given sufficient oxygen to maintain saturation ≥95%. Patients who had regional anesthesia or nerve blocks were excluded. The primary outcome was pain and opioid consumption during the initial 2 postoperative hours, analyzed jointly. The secondary outcome was pain and opioid consumption over the subsequent 24 postoperative hours. Subgroup analyses of the primary outcome were conducted for open versus laparoscopic procedures and for patients with versus without chronic pain.

RESULTS: A total of 4702 cases were eligible for analysis: 2415 were assigned to 80% oxygen and 2287 to 30% oxygen. The groups were well balanced on potential confounding factors. Average pain scores and opioid consumption were similar between the groups (mean difference in pain scores, −0.01 [97.5% CI, −0.16 to 0.14; P = .45], median difference in opioid consumption, 0.0 [97.5% CI, 0 to 0] mg morphine equivalents; P = .82). There were also no significant differences in the secondary outcome or subgroup analyses.

CONCLUSIONS: Supplemental intraoperative oxygen does not reduce acute postoperative pain or reduce opioid consumption.

From the *Department of Outcomes Research, Cleveland Clinic, Cleveland, Ohio

Division of Anesthesia, Critical Care and Pain Management, Tel-Aviv Medical Center, Sackler Faculty of Medicine, Tel-Aviv University, Tel-Aviv, Israel

Internal Medicine C, Sheba Medical Center, Sackler Faculty of Medicine, Tel-Aviv University, Tel-Aviv, Israel

Departments of §Quantitative Health Sciences

General Anesthesia, Cleveland Clinic, Cleveland, Ohio.

Published ahead of print 26 November 2018.

Accepted for publication November 26, 2018.

Funding: This work received internal funding. None of the authors has a personal financial interest in this analysis. B.C. is a recipient of a Fellowship Grant from the American Physicians Fellowship for Medicine in Israel.

The authors declare no conflicts of interest.

Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website.

Trial registration: number NCT01777568.

Reprints will not be available from the authors.

Address correspondence to Alparslan Turan, MD, Department of Outcomes Research, Anesthesiology Institute, Cleveland Clinic, 9500 Euclid Ave, P77, Cleveland, OH 44195. Address e-mail to

See Editorial, p

Postoperative pain prolongs hospitalization, increases health care costs, and promotes the use of both opioid and nonopioid analgesics.1–3 Pain is among surgical patients’ leading concerns, and pain increases preoperative anxiety and postoperative sympathetic response.4 Preventing and treating surgical pain is thus a clinical priority.

Incisional pain is multifactorial, but wound ischemia and local production and secretion of inflammatory mediators clearly contribute. Even when arterial blood is fully saturated, surgical wounds have high lactate concentrations and are therefore acidic, which, in animal models, provokes pain-related behavior.5,6 Supplemental oxygen (eg, 80% vs 30% inspired fraction) substantially increases arterial PaO2, which in turn doubles tissue oxygenation.7,8 Hyperoxia also promotes vasodilation. The combination of better oxygenation and perfusion reduces wound lactate concentration and possibly reduces surgical pain consequent to local acidosis.5

The importance of adequate wound oxygenation is perhaps most impressively demonstrated with hyperbaric oxygen, during which patients are exposed to 100% oxygen at up to 3 atmospheric pressures. The resulting markedly elevated PaO2 improves oxygenation of ischemic tissues and, in several animal models of nociception and inflammation, reduces pain.9–12 There is also some evidence suggesting that hyperbaric oxygen reduces pain in patients with diabetic ulcers, pressure ulcers, and chronic nonhealing wounds.13–16 However, whether supplemental oxygen at atmospheric pressure ameliorates pain remains unknown.

We therefore tested the hypothesis that intraoperative hyperoxia (inspired oxygen fraction [Fio2]; 80% vs 30%) reduces postoperative pain and opioid consumption during the initial 2 postoperative hours in adults recovering from noncardiac surgery. Secondarily, we tested the same hypothesis during the subsequent 24 postoperative hours. On an exploratory basis, we also considered subgroups of patients who had laparotomy versus laparoscopic or laparoscopic-assisted surgeries and patients who did or did not have preoperative chronic pain.

Back to Top | Article Outline


We conducted a post hoc analysis of a large, single-center alternating cohort trial in which adults undergoing noncardiac surgery were allocated to either 30% or 80% inspired oxygen during general anesthesia.17 The underlying alternating cohort study was approved by the Cleveland Clinic Institutional Review Board (No. 12–891) with waived consent and was registered at before patient enrollment (NCT01777568, Primary Investigator Andrea Kurz, MD, date of registration: January 29, 2013). The current subanalysis was also approved by the Cleveland Clinic Institutional Review Board (No. 17–825).

All patients having surgery in a designated operating suite at the Cleveland Clinic Main Campus during a 39-month period extending from January 2013 through March 2016 were included. Intraoperative Fio2 alternated between 30% (or the lowest Fio2 necessary to maintain ≥95% saturation) and 80% every 2 weeks. The first 2-week assignment was chosen randomly. Postoperative oxygen administration was not controlled. In addition, clinicians were instructed to provide sufficient intraoperative oxygen to maintain patients’ saturation at or above 95% and were allowed to use Fio2 as high as 100% during induction of and emergence from anesthesia. Aside from inspired oxygen, no aspect of anesthetic management was controlled, including analgesic administration. The primary outcome of the underlying trial was the incidence of surgical site infections, and the results of that analysis were published elsewhere.17

Only colorectal operations were included in the underlying study, which also a priori excluded surgeries lasting <2 hours and reoperations during the same hospitalization. For the purpose of the current subanalysis, we also excluded patients who had any kind of regional anesthesia and records with missing information on opioid administration or pain scores (eg, patients who remained intubated). All data were collected from electronic medical and anesthesia records.

Postoperative pain is routinely assessed and recorded on a scale of 0 (no pain) to 10 (worst pain imaginable) by the nursing staff every 15 minutes during the first postoperative hour, every 30 minutes thereafter while in the postanesthesia care unit (PACU), and at least every 4 hours in surgical wards. All available pain scores were collapsed to average scores over 30-minute intervals while patients were in the PACU and to average scores per 4-hour periods while patients were in surgical wards. Patients who went to intensive care units were treated in the same way, as long as pain scores were recorded. Opioid consumption was measured in intravenous morphine equivalents, using an opioid conversion table (Supplemental Digital Content 1, Table 1,

Morphometric and demographic characteristics and intraoperative management were summarized using descriptive statistics and compared between the 2 groups using absolute standardized differences (absolute difference in means or proportions divided by the pooled SD). Imbalances between the groups, defined as absolute standardized difference >0.1, were adjusted in all analyses. Chronic preoperative pain was defined for either having the diagnosis of chronic pain in the medical record problem list or daily consumption of 15 mg of oral morphine equivalents.

The primary outcome of average pain score and total opioid consumption during the initial 2 hours after admission to PACU was analyzed in a “joint hypothesis testing” framework, described elsewhere.18 Using this framework, pain or opioid consumption was deemed superior only if both were noninferior and ≥1 was superior. Noninferiority was defined if the upper limit of the 95% 2-sided CI for the median difference of total opioid consumption was less than the noninferiority delta of 5 mg of morphine, and for the difference in means of pain score, less than the noninferiority delta of 1. These delta values were chosen to be clinically relevant. The methodology is graphically described in Supplemental Digital Content 2, Figure 1,

Total opioid consumption for the 2 inspired oxygen groups was assessed using Wilcoxon rank sum test because the distribution of opioid consumption is not normal, with 20% of data at zero. The comparison of pain scores was assessed using a mixed-effects model, which allowed us to simultaneously assess the effect across the time points if there is no group-time interaction or else at specific times if there is an interaction detected. As a second step, superiority testing was conducted in the same direction. Noninferiority test was 1 tailed and assessed at the .025 significance level in each direction. Superiority testing was 1 tailed in the direction of noninferiority, with a significance level of .0125 (ie, .025/2, Bonferroni).

The secondary analysis used similar methods applied to the pain scores and opioid consumption from 2 hours after PACU admission until 26 postoperative hours. The noninferiority delta for opioid consumption was increased to 10 mg of morphine for this analysis (to be more clinically relevant when assessing a 24 rather than a 2-hour period), and the delta for pain scores remained 1 point. Stratified subgroup analyses on the primary outcome were performed for the following groups: (1) open versus laparoscopic or laparoscopic-assisted surgeries; and (2) existence versus absence of preoperative chronic pain, which was defined by either diagnosis of preoperative chronic pain in the medical record problem list or consumption of ≥15 mg of oral morphine equivalents before surgery.

The underlying trial enrolled 8097 patients, with 5749 being eligible for analysis. We expected that each group would have 2500 eligible patients meeting the inclusion/exclusion criteria of this study. Power estimation was based on the analysis of superiority on opioid consumption because it required a larger sample size than noninferiority test or test on pain scores. We planned to assess the difference in opioid consumption in a linear regression model, using a logarithmic transformation of IV morphine equivalent as appropriate. Given a sample size of 2500 patients in each group and assuming coefficient of variation (SD/mean) of 2 for both groups, we would have about 90% power at the .025 significance level to detect opioid consumption decrease corresponding to a ratio of means of <0.88 comparing 80% to 30% oxygen groups (ie, decrease of ≥12% in mean opioid consumption). When we analyzed the actual opioid consumption from 0 to 2 hours, we found that the log-transformed data were not normally distributed, with about 20% patients having no opioid consumption. We therefore decided to use the nonparametric Wilcoxon rank sum test to assess the treatment effect. We conducted a post hoc power analysis assuming that 20% of patients in the control group have zero opioid consumption and that 25%, 35%, and 20% have opioid consumption in the higher 3 ordinal categories. A shift in the distribution to 30%, 22%, 30%, and 18%, respectively, in the 4 ordinal categories in the treatment group would be clinically meaningful. Then, a total of 1840 patients would give 90% power in a test of superiority.

Back to Top | Article Outline


A total of 5749 qualifying surgeries performed at Cleveland Clinic from January 28, 2013, to March 11, 2016, were retrieved from the underlying trial. After excluding ineligible cases (829 cases with regional anesthesia and 218 others with incomplete pain or opioid data), 4702 surgeries were analyzed (2415 in the 80% oxygen group and 2287 in the 30% oxygen group; Figure 1). Patient characteristics by oxygen group were summarized in Table 1.

Table 1.

Table 1.

Figure 1.

Figure 1.

The groups were comparable on all measured potential confounding factors (absolute standardized difference <0.1 for all), and no adjustments were needed. The median intraoperative time-weighted average Fio2 was 44% (interquartile range: 39%–55%) in patients assigned to 30% or the lowest tolerated Fio2 and 81% (interquartile range: 77%–82%) in those assigned to 80% inspired oxygen. Most patients received only minimal oxygen enrichment through a low-flow nasal cannula while in the PACU (median [interquartile range] time-weighted average Fio2, 27% [27%–27%]).

Average pain scores and opioid consumption during the first 2 hours and the following 24 hours are presented by study group in Table 2. Average opioid consumption is presented in Figure 2. The changes in pain scores by group over time are plotted in Figure 3.

Table 2.

Table 2.

Figure 2.

Figure 2.

Figure 3.

Figure 3.

When testing whether 80% oxygen provides better analgesia than 30% oxygen, we confirmed noninferiority for both opioid consumption and pain scores. However, neither was superior. The estimated median difference in opioid consumption was 0.0 (97.5% CI, 0 to 0; P = .82), and the mean difference in pain scores was −0.01 (97.5% CI, −0.18 to 0.16; P= .45), comparing 80% to 30% oxygen (Table 2; Figures2–3). Similarly, when testing the hypothesis in the opposite direction, we did not find 30% oxygen to provide better pain control than 80% oxygen (Supplemental Digital Content 3, Table 2,

For the secondary analysis (pain and opioid consumption from 2 to 26 postoperative hours), both outcomes were noninferior, with 80% vs 30% inspired oxygen, but again, neither was superior (Table 2; Figures2–3). When testing the opposite direction (30% vs 80% oxygen), the results were similar (Supplemental Digital Content 3, Table 2, Supplemental oxygen also offered no benefit in subgroups defined by type of surgery or existence of chronic preoperative pain (Table 3).

Table 3.

Table 3.

Back to Top | Article Outline


Most perioperative trials of supplemental oxygen were designed to test the hypothesis that hyperoxia reduces the risk of surgical site infection. Initial trials reported that supplemental oxygen halved the incidence of surgical site infection,19,20 while most subsequent trials have shown little or no benefit.21–23 A meta-analysis of high-quality trials also suggests that supplemental oxygen does not reduce infection risk.24 In contrast, there is little information about the potential effect of supplemental oxygen on surgical pain, although some animal data are suggestive, as are results from hyperbaric oxygen treatments.

We now report the effects of supplemental oxygen on surgical pain in >4700 patients who had colorectal surgery with general anesthesia. Intraoperative hyperoxia did not improve analgesia or reduce opioid consumption during the initial 2 postoperative hours, nor during the subsequent 24 hours. Results were similar in predefined subgroups defined by type of surgery and preexisting chronic pain.

The hyperoxic intervention was restricted to the intraoperative period, and postoperative oxygen administration was left to the discretion of the PACU staff. More than 75% of the patients only received minimal oxygen enrichment in the PACU. It is therefore possible that a minor beneficial effect of intraoperative hyperoxia was rapidly cancelled by the comparable and minor amount of oxygen that was provided for both groups postoperatively. However, even during the initial 2 postoperative hours, which were our a priori primary outcome, there was no effect on pain or opioid consumption. It remains possible that supplemental intraoperative oxygen improved wound oxygenation, ameliorated acidosis and decreased the release of pain-provoking mediators, but if so, the potential effect was too short-lived to be of any clinical consequence.

Patients suffering episodes of intense pain usually have more frequent pain score assessment. We compensated by collapsing pain scores into predefined time periods. Scores within each time block thus accurately represent pain within the relevant period, and extra measurements do not influence pain scores in other periods.

Our study is designed as a post hoc analysis of a prospective alternating cohort trial. The underlying trial tested the possible association between intraoperative Fio2 and wound-related outcomes and was published elsewhere.17 This novel design allowed rapid and relatively inexpensive enrollment of a large number of participants. We note that treatment allocation was not truly randomized on an individual basis. In this respect, the design resembles cluster randomization, with clusters distributed in time rather than in space. Nevertheless, this design effectively achieved the purpose of randomization (eliminating selection bias and confounding), and our results show remarkable balance between the groups on a long list of observed confounders. Similarly to randomized trials, we cannot fully guarantee balance on unobserved confounding factors, but considering the achieved balanced on observed confounders, it seems unlikely that such imbalance truly exists. Furthermore, although this design does not include formal blinding, we relied on data collected by nursing staff (pain scores and opioid consumption), and it seems unlikely that measurements were biased by nurses who were presumably indifferent to intraoperative Fio2.

We chose to test the effect of the intervention on postoperative analgesia by using the “joint hypothesis testing” method. This was done because both opioid consumption and reported pain are clinically important when measuring postoperative analgesic effect. Specifically, it is reasonable to assume that nursing practice “pushes” the majority of patients toward an “acceptable” pain score by administering opioids as much as necessary. Therefore, testing only pain scores as the primary outcome, while ignoring a difference in opioid administration, might not represent an important difference in actual postoperative pain. The joint hypothesis test of superiority requires that the intervention (Fio2 of 80%) demonstrates noninferiority (namely, being “as good as” the control group) in both pain scores and total opioid consumption, followed by superiority (being “better than” the control group) on ≥1 of the 2 modalities. While we were able to show noninferiority of 80% oxygen on both modalities, superiority was not demonstrated. Therefore, according to our predefined statistical plan, we concluded that the null hypothesis claiming that intraoperative Fio2 of 80% has a similar analgesic effect to that of 30% oxygen should not be rejected.

The median Fio2 administered to patients in the 30% group was 44%. Although this might seem as a deviation from the study protocol, it represents a common and expected clinical practice that was actually described in the original study design. Namely, care givers were instructed to use sufficient oxygen to keep patients’ saturation at 95% or higher. Although this decreased the magnitude of difference in Fio2 between the study groups, we believe our results are relevant as they represent the real-life clinical practice of anesthesia providers attempting to use the lowest required Fio2.

In summary, wound hypoxia causes local acidosis and release of mediators that promote pain. Supplemental oxygen appears to be analgesic in animals and in humans given hyperbaric oxygen. We therefore compared 30% or 80% inspired oxygen on pain and opioid consumption in 4702 patients. Intraoperative hyperoxia did not reduce pain or opioid consumption in adults recovering from noncardiac surgery, either during the initial 2 hours or the subsequent 24 hours. Oxygen administration should not be considered as an analgesic adjuvant in surgical patients.

Back to Top | Article Outline


Name: Barak Cohen, MD.

Contribution: This author helped design the study, contribute the initial draft, and review and approve the final manuscript.

Name: Sanchit Ahuja, MD.

Contribution: This author helped design the study, review and make substantive contributions to the manuscript draft, and approve the final manuscript.

Name: Yehoshua N. Schacham, MD.

Contribution: This author helped design the study, review and make substantive contributions to the manuscript draft, and approve the final manuscript.

Name: David Chelnick, BS.

Contribution: This author helped review and make substantive contributions to the manuscript draft and approve the final manuscript.

Name: Guangmei Mao, PhD.

Contribution: This author helped contribute the data extraction and analysis and review and approve the final manuscript.

Name: Wael Ali-Sakr Esa, MD.

Contribution: This author helped review and make substantive contributions to the manuscript draft and approve the final manuscript.

Name: Kamal Maheshwari, MD, MPH.

Contribution: This author helped review and make substantive contributions to the manuscript draft and approve the final manuscript.

Name: Daniel I. Sessler, MD.

Contribution: This author helped design the study, contribute the initial draft, and review and approve the final manuscript.

Name: Alparslan Turan, MD.

Contribution: This author helped design the study, contribute the initial draft, and review and approve the final manuscript.

This manuscript was handled by: Richard C. Prielipp, MD, MBA.

Back to Top | Article Outline


1. Schug SA, Large RG. Economic considerations in pain management. Pharmacoeconomics. 1993;3:260–267.
2. Practice guidelines for acute pain management in the perioperative setting. A report by the American Society of Anesthesiologists Task Force on Pain Management, Acute Pain Section. Anesthesiology. 2012;116:248–273.
3. Breivik H. Postoperative pain management: why is it difficult to show that it improves outcome? Eur J Anaesthesiol. 1998;15:748–751.
4. Apfelbaum JL, Chen C, Mehta SS, Gan TJ. Postoperative pain experience: results from a national survey suggest postoperative pain continues to be undermanaged. Anesth Analg. 2003;97:534–540.
5. Kim TJ, Freml L, Park SS, Brennan TJ. Lactate concentrations in incisions indicate ischemic-like conditions may contribute to postoperative pain. J Pain. 2007;8:59–66.
6. Woo YC, Park SS, Subieta AR, Brennan TJ. Changes in tissue pH and temperature after incision indicate acidosis may contribute to postoperative pain. Anesthesiology. 2004;101:468–475.
7. Hopf HW, Hunt TK, West JM, et al. Wound tissue oxygen tension predicts the risk of wound infection in surgical patients. Arch Surg. 1997;132:997–1004.
8. Akça O, Melischek M, Scheck T, et al. Postoperative pain and subcutaneous oxygen tension. Lancet. 1999;354:41–42.
9. Thompson CD, Uhelski ML, Wilson JR, Fuchs PN. Hyperbaric oxygen treatment decreases pain in two nerve injury models. Neurosci Res. 2010;66:279–283.
10. Sümen G, Cimşit M, Eroglu L. Hyperbaric oxygen treatment reduces carrageenan-induced acute inflammation in rats. Eur J Pharmacol. 2001;431:265–268.
11. Chung E, Zelinski LM, Ohgami Y, Shirachi DY, Quock RM. Hyperbaric oxygen treatment induces a 2-phase antinociceptive response of unusually long duration in mice perspective: hyperbaric oxygen treatment of mice can induce a 2-phase antinociceptive response. J Pain. 2010;11:847–853.
12. Wilson HD, Wilson JR, Fuchs PN. Hyperbaric oxygen treatment decreases inflammation and mechanical hypersensitivity in an animal model of inflammatory pain. Brain Res. 2006;1098:126–128.
13. Kiralp MZ, Yildiz S, Vural D, Keskin I, Ay H, Dursun H. Effectiveness of hyperbaric oxygen therapy in the treatment of complex regional pain syndrome. J Int Med Res. 2004;32:258–262.
14. Bennett MH, French C, Schnabel A, Wasiak J, Kranke P, Weibel S. Bennett MH. Normobaric and hyperbaric oxygen therapy for the treatment and prevention of migraine and cluster headache. In: Cochrane Database of Systematic Reviews. 2015:Chichester, UK: John Wiley & Sons, Ltd; CD005219.
15. Kranke P, Bennett MH, Martyn-St James M, Schnabel A, Debus SE. Kranke P. Hyperbaric oxygen therapy for chronic wounds. In: Cochrane Database of Systematic Reviews. 2012.Chichester, UK: John Wiley & Sons, Ltd.
16. Löndahl M, Katzman P, Nilsson A, Hammarlund C. Hyperbaric oxygen therapy facilitates healing of chronic foot ulcers in patients with diabetes. Diabetes Care. 2010;33:998–1003.
17. Kurz A, Kopyeva T, Suliman I, et al. Supplemental oxygen and surgical-site infections: an alternating intervention controlled trial. Br J Anaesth. 2018;120:117–126.
18. Mascha EJ, Turan A. Joint hypothesis testing and gatekeeping procedures for studies with multiple endpoints. Anesth Analg. 2012;114:1304–1317.
19. Greif R, Akça O, Horn EP, Kurz A, Sessler DI; Outcomes Research Group. Supplemental perioperative oxygen to reduce the incidence of surgical-wound infection. N Engl J Med. 2000;342:161–167.
20. Belda FJ, Aguilera L, García de la Asunción J, et al.; Spanish Reduccion de la Tasa de Infeccion Quirurgica Group. Supplemental perioperative oxygen and the risk of surgical wound infection: a randomized controlled trial. JAMA. 2005;294:2035–2042.
21. Myles PS, Leslie K, Chan MT, et al.; ENIGMA Trial Group. Avoidance of nitrous oxide for patients undergoing major surgery: a randomized controlled trial. Anesthesiology. 2007;107:221–231.
22. Meyhoff CS, Wetterslev J, Jorgensen LN, et al.; PROXI Trial Group. Effect of high perioperative oxygen fraction on surgical site infection and pulmonary complications after abdominal surgery: the PROXI randomized clinical trial. JAMA. 2009;302:1543–1550.
23. Duggal N, Poddatoori V, Poddatorri V, Noroozkhani S, Siddik-Ahmad RI, Caughey AB. Perioperative oxygen supplementation and surgical site infection after cesarean delivery: a randomized trial. Obstet Gynecol. 2013;122:79–84.
24. Cohen B, Schacham YN, Ruetzler K, et al. Effect of intraoperative hyperoxia on the incidence of surgical site infections: a meta-analysis. Br J Anaesth. 2018;120:1176–1186.

Supplemental Digital Content

Back to Top | Article Outline
Copyright © 2019 International Anesthesia Research Society