Leukocyte DNA Damage and Wound Infection after Nitrous Oxide Administration: A Randomized Controlled Trial

WOUND infection is a leading cause of perioperative morbidity and mortality.^1,2  In previous studies, up to 25% of patients undergoing major colorectal surgery had an episode of wound infection during the first week after surgery.¹  Given that all surgical wounds are contaminated to certain extent, it is important to institute early measures to prevent the establishment of infection.¹  In this regard, perioperative interventions, such as prophylactic administration of antibiotics and avoidance of hypothermia, have been shown to decrease the rate of wound infection.^1,3,4  There is emerging evidence that nitrous oxide administration may also contribute to surgical wound infection.⁵ 

Biochemically, nitrous oxide inhibits methionine synthase by oxidizing the cobalt atom in the cobalamin molecule,^6–8  resulting in acute depletion of intracellular tetrahydrofolate. Since dietary folate and cobalamin deficiencies are known to induce uracil misincorporation, base damage, and strand breakage during synthesis of DNA,^9–15  we hypothesized that nitrous oxide exposure may produce acute DNA damage. This may compromise host defense mechanisms and predispose patients to surgical wound infection.

The adminisration of nitrous oxide also limits the inspiratory oxygen concentration that can be delivered during surgery. Previous studies have suggested that 80% oxygen reduced wound infection,^16,17  but the resultant free radical production may lead to DNA damage.^18,19  Therefore, hyperoxia may either be beneficial or detrimental to wound healing.

Using single-cell gel electrophoresis,²⁰  the purpose of this study was to compare DNA damage in patients receiving nitrous oxide or nitrous oxide-free anesthesia for major colorectal surgery. As a secondary objective, we evaluated the effects of oxygen concentration (30 vs. 80%) on DNA damage. In an exploratory analysis, we also compared the rate of surgical wound infection in patients receiving nitrous oxide or not and with different oxygen concentrations.

This was a prospective, three-armed, parallel-group, double-blind, superiority randomized controlled trial. Our single-site study was conducted at the Prince of Wales Hospital, Hong Kong, a university-affiliated teaching hospital. Eligible patients had elective open colorectal surgery, were aged at least 18 yr, with American Society of Anesthesiologists physical status class I–IV. Patients were excluded if they had acute bowel obstruction. We also excluded patients with ongoing infection and those with fever in the 24h before surgery. Other exclusion criteria included patients with marked impairment of gaseous exchange, defined as those who required inspiratory oxygen concentration more than 40%, surgery for which primary wound closure was not anticipated, or in the opinion of the attending anesthesiologist that nitrous oxide administration was contraindicated. Patients were identified from routine surgical lists. The Joint Chinese University of Hong Kong–New Territories East Cluster Clinical Research Ethics Committee (Hong Kong Special Administration, China) approved the study protocol, and all patients gave written informed consent.

All patients received oral polyethylene glycol 2–4 l on the day before surgery for mechanical bowel preparation. Immediately before induction of anesthesia, intravenous cefuroxime 1.5g and metronidazole 500mg were administered as prophylactic antibiotics. These were continued every 8h after surgery for three more doses. Additional antibiotics were given for suspected infection. Intravenous vancomycin 500mg was used as an alternative in patients who were allergic to cefuroxime. Surgery was performed and wound was closed according to the routine practice. Povidone iodine solution was used for wound irrigation at the end of the procedure.

After enrolment and immediately before the induction of anesthesia, patients were randomly allocated from a computer-generated list, accessed through an intranet system, to receive one of the three gas mixtures for anesthesia in a 1:1:1 ratio. Randomization was stratified according to the risk of wound infection using the National Nosocomial Infections Surveillance System scale.²¹  The scale allocates 1 point for American Society of Anesthesiologists physical status III and IV, contaminated or dirty-infected surgery, and duration of surgery more than expected. The total score ranges from 0 (lowest risk) to 3 (highest risk). Before surgery, plasma folate concentration was measured using an immunoassay, as previously described.^22,23  Folate deficiency was defined as plasma folate concentration less than 7.0 nm.

Anesthesia was induced with propofol 1–2.5mg/kg. All patients received sevoflurane targeted to achieve a bispectral index (Covidien, Mansfield, MA) value between 40 and 60. Intraoperative analgesia was provided by remifentanil infusion 0.1–0.5 μg⋅kg⁻¹⋅min⁻¹ and intravenous morphine 0.1–0.15mg/kg, 30min before completion. Muscle relaxation was facilitated by rocuronium. The lungs were ventilated through a tracheal tube using the assigned gas mixture to achieve normocarbia, defined as end-tidal carbon dioxide tension between 35 and 45 mmHg. In the operating room, anesthetic flow meters and monitoring devices were shielded, so that attending surgeons were not aware of the assigned gas mixture.

When surgery was completed, 100% oxygen was provided until tracheal extubation. In the subsequent 24h, patients breathed through facemasks at the assigned inspired oxygen concentration (30 or 80%). This was then taken away and all patients breathed room air afterward. Supplemental oxygen was provided during and after surgery, if necessary, to maintain oxygen saturation more than 92%. Intraoperative normothermia (>35.5°C) was maintained using forced air warming devices and fluid warmers. The administration of intravenous fluid or blood component therapy was determined by the attending anesthesiologists to maintain the hemoglobin concentration of 9g/dl or more. The anesthetic record was then sealed in an opaque “confidential” envelope after the patient left the operating room until 30 days after surgery. Postoperative pain relief was provided by a patient-controlled analgesia device delivering intravenous bolus doses of morphine 20–40 µg/kg at 10-min lockout intervals. This was changed to oral paracetamol as appropriate.

While in hospital, patients were monitored for outcome daily, specifically surgical wounds were examined by ward medical staff who were unaware of the allocated group identity. Wound healing characteristics were rated using the ASEPSIS (Additional treatment, Serous discharge, Erythema, Purulent exudate, Separation of deep tissues, Isolation of bacteria, and duration of inpatient Stay) scale.²⁴  A score of more than 20 was suggestive of wound infection. In addition, wound infection was classified according to the Centers for Disease Control and Prevention criteria.²⁵  Other infective complications, including urinary tract infection and pneumonia, were specifically sought. Definitions of outcomes are listed in table 1. Discharged patients were contacted at 30 days after surgery. If patients or their relatives indicated the occurrence of an outcome, the medical records were retrieved for documentation.

Immediately before induction of anesthesia and 24h after the end of surgery, 10-ml venous blood samples were collected for the measurement of DNA damage in circulating leukocytes using single-cell gel electrophoresis as previously described.²⁰  Briefly, DNA supercoils of cells that contain strand breaks and base damage are relaxed and unwound under low-current electrophoresis. The broken ends, that are negatively charged, are pulled toward the anode. This results in a comet-shaped tail behind the nuclear head that can be observed and quantified under microscopy with fluorescence staining (fig. 1).

We used the CometAssay (Trevigen Inc., Gaithersburg, MD) for our study. Whole blood of 0.1ml was diluted with 0.3ml ice-cold 0.1% phosphate buffer saline. We mixed 6 μl phosphate buffer saline-diluted whole blood with 60 μl of 0.5% low-melting point agarose (Sigma-Aldrich, Munich, Germany). The mixture was added to the sample area of the CometSlide (Trevigen Inc.). Cells were lysed with a prechilled solution containing 10% dimethyl sulfoxide and were incubated at 4°C for 2h. Postlysis slides were immersed in freshly prepared sodium hydroxide (NaOH) 300 mm for 30min. Electrophoresis was carried out at 4°C for 30min at 300 mA and 25V in alkaline electrophoresis solution (NaOH, 200 mm, pH >13). The slides were then neutralized with Tris buffer (pH 7.5) and were fixed in 70% ethanol. DNA was stained by SYBR Green I (Trevigen Inc.). All assays were performed in duplicates. The extent of DNA damage was assessed visually by digital fluorescence microscope at 10-fold magnification. Images of 100 randomly selected nuclei (50 nuclei from each of two replicated slides) were analyzed using the CometAssay IV software (Perceptive Instruments, Haverhill, Suffolk, United Kingdom). Two observers, who were blinded to the treatment allocation, reviewed one of the two replicated slides in a random manner. The final result for each sample was the combined score of the two observers. DNA damage was expressed as the percentage of DNA intensity in the comet tail. The assay was calibrated using the CometAssay control cells (Trevigen Inc.).

The primary endpoint of the study was the perioperative change in DNA damage. Secondary endpoints included rates of surgical wound infection, other major postoperative complications, and recovery times. This was defined as the time from the end of surgery until hospital discharge, return of bowel function, and resumption of enteral feeding.

Changes in the percentage of DNA in tail were compared among groups using generalized linear models with repeated measures, and were correlated with surgical wound infection with multiple logistic regression models. The correlations between DNA damage and duration of anesthesia were compared among groups after Fisher Z transformation.

Complication rates among groups were compared among groups using the likelihood-ratio χ² statistics. Continuous data were tested with one-way ANOVA or Kruskal–Wallis test, as appropriate. Recovery times were calculated by Kaplan–Meier analysis and were compared among groups using log-rank test. A P value less than 0.05 was considered statistically significant. All P values reported were two-tailed. We used IBM SPSS Statistics version 20 (IBM Corporation, Somers, NY) for all statistical analyses except for the comparison of correlation coefficients among groups which was performed using SAS procedure COMPCORR (SAS Institute, Cary, NC).

We estimated the sample size based on a difference among groups in the extent of DNA damage by 20% with a variation of 30–40%. On the basis of these assumptions, we calculated that 30 patients per group were required to achieve 90% power at 5% α error.

The study recruited patients between November 2, 2009 and July 15, 2011. A total of 93 patients were enrolled into the study. Two patients were excluded after randomization because their surgeries were cancelled (fig. 2). The remaining 91 patients completed the 30-day follow-up. Patient characteristics are summarized in table 2. Approximately 60% of patients were men and above 60% were older than 65 yr. The median (interquartile range) duration of anesthesia was 2.8 (2.1–3.7) h and was not different among groups. Sixteen patients (5 in the 70% nitrous oxide group, 6 in the 30% oxygen group, and 5 in the 80% oxygen group) had colonoscopies within 2 weeks of surgery. These procedures were performed during sedation with pethidine 1.5mg/kg and diazepam 0.1mg/kg. None of the patients had exposure to nitrous oxide during this period.

Administration of nitrous oxide reduced the doses of sevoflurane required to maintain a bispectral index value between 40 and 60 (table 3). Anesthetic delivery was otherwise comparable among groups. Recovery times in patients receiving nitrous oxide and 30 or 80% oxygen are summarized in table 4. The median (95% CI) time to hospital discharge after nitrous oxide administration of 9 (4.46–10.54) days was similar to those of receiving 30% oxygen, 7 (4.95–10.05) days, or 80% oxygen, 7 (4.09–12.01); log-rank test; P = 0.27. The times to first passage of flatus and to tolerate solid food were similar among groups.

The percentages of DNA intensity in tail in the preoperative samples were similar among groups (P = 0.25). After surgery, in patients receiving nitrous oxide, a two-fold increase in the percentage of DNA in tail was observed (P = 0.0003), but this did not occur in the 30 (P = 0.181) or 80% oxygen (P = 0.419) groups (fig. 3). The changes in DNA damage was exposure dependent (fig. 4), such that the longer duration of nitrous oxide exposure, the larger the extent of DNA damage, r = 0.33, P = 0.029. There was, however, no correlation between changes in the percentage of DNA in tail and duration of anesthesia in patients receiving 30% oxygen, r = 0.02, or 80% oxygen, r = −0.04. The correlation coefficient in the nitrous oxide group (r = 0.33) was significantly higher than that in the 30% oxygen (r = 0.02) or 80% oxygen groups (r = −0.04); P = 0.042.

Postoperative complications are summarized in table 5. By using ASEPSIS criteria (score >20), the rate of surgical wound infection was higher in patients receiving nitrous oxide compared with those in the 30 or 80% oxygen groups, odds ratios (95% CIs): 4.29 (1.38–13.28); P = 0.036. The risk was largely unchanged after adjustment for age, gender, smoking status, and National Nosocomial Infections Surveillance System score, adjusted odds ratio: 4.51 (1.42–14.60); P = 0.041. When diagnosis was based on Centers for Disease Control and Prevention criteria, there was little effect of nitrous oxide on wound infection, odds ratio (95% CI): 3.36 (0.87–12.96); P = 0.205. Patients with wound infection stayed in the hospital longer than those who did not, 20.8 vs. 8.3 days, mean difference (95% CI): 12.5 (1.6–23.5) days, P = 0.028.

An increase in DNA damage after surgery was associated with higher risk of surgical wound infection (Centers for Disease Control and Prevention criteria), odds ratio (95% CI): 1.07 (1.01–1.13), P = 0.03. This risk estimate was unchanged after adjustment for age, gender, National Nosocomial Infections Surveillance System score, and lowest intraoperative temperature, adjusted odds ratio (95% CI): 1.07 (1.01–1.14); P = 0.04. By using APSESIS criteria, the risk of wound infection remained substantially higher in patients with DNA damage, adjusted odds ratio (95% CI): 1.19 (1.07–1.34); P = 0.003. The rates of other complications, including urinary tract infection, pneumonia, myocardial infarction, and death were similar among groups.

In this randomized controlled trial of adult patients undergoing major elective open colorectal surgery, we found that nitrous oxide administration was associated with DNA damage in peripheral leukocytes after surgery. The DNA damage was exposure dependent and was more apparent when nitrous oxide exposure lasted longer than 2h. We also identified DNA damage as a potential mechanism for postoperative wound infection after nitrous oxide anesthesia. When the percentage of DNA intensity in tail was doubled after nitrous oxide exposure, the risk of wound infection was increased by more than two-fold. In contrast, changes in inspiratory concentration of oxygen (30 or 80%) have no measurable effect on DNA damage.

A number of studies have reported genotoxic effects of nitrous oxide. In cellular experiments, cell proliferation was inhibited by nitrous oxide in mitogen-treated mononuclear cells in the peripheral blood.²⁶  In another experiment, chemotaxis was reduced in harvested leukocytes that were exposed to nitrous oxide in vitro.²⁷  Interestingly, in a cell-culture study, nitrous oxide also accelerated the growth of microorganisms.²⁸  This study, together with previous laboratory findings,^26,27  highlights the worst case scenario where host defense is impaired in the presence of bacterial overgrowth. In humans, several studies have reported an increase in DNA damage among operating room personnel who are regularly exposed to nitrous oxide compared with those healthcare workers working in other areas of the hospital.^29–31  However, none of these reports established the link between experimental findings and patient outcomes. Therefore, the clinical relevance of these changes after nitrous oxide exposure is unclear. In the current study, we have identified an association between DNA damage after nitrous oxide administration and postoperative wound infection within 30 days after surgery. It should be understood that this finding does not imply a causal relationship. Although DNA damage is one of the hallmarks in apoptosis,³²  its immunosuppressive effect remains unclear. Future study should explore the mechanisms of DNA damage to produce wound infection.

Other anesthetic agents may also induce DNA damage. Sevoflurane was reported to increase oxidative stress and DNA strand breakage in peripheral leukocytes.^33,34  However, other studies have shown no change in the amount of DNA damage after sevoflurane exposure.^35,36  This discrepancy may be attributed to the differences in patient populations and techniques in quantifying DNA damage. In our study, we standardized the anesthetic regimens, so that a similar depth of anesthesia was provided to all patients. Given the anesthetic sparing effect, the dose of sevoflurane administered was reduced by 48% in the nitrous oxide group. However, despite a decrease in sevoflurane delivery, we observed an increase in DNA damage with nitrous oxide. This finding, together with the lack of change in DNA damage in the 30 and 80% oxygen groups, suggested that sevoflurane is unlikely to produce important DNA damage.

We observed an increased rate of surgical wound infection with nitrous oxide administration. This finding is consistent with a previous study.⁵  The Evaluation of Nitrous oxide In the Gas Mixture for Anaesthesia Trial recruited relatively unselected adult surgical patients. Nitrous oxide increased the risk of wound infection compared with patients receiving 80% oxygen, relative risk (95% CI): 1.17 (1.02–1.33); P = 0.04. In contrast, a randomized controlled trial in colorectal surgery showed no detrimental effect of nitrous oxide on wound infection.³⁷  However, the later study was stopped early, and it is plausible that the sample size is insufficient to detect a 33% increase in the rate of wound infection (from 15 to 20%). It should be emphasized that our current study was not designed to detect the difference in rate of wound infection. A much larger study, such as the Evaluation of Nitrous oxide In the Gas Mixture for Anaesthesia-II Trial (ClinicalTrial.gov identifier NCT00430989), evaluating 7,000 patients is required to answer this question definitively.

Supplemental oxygen has been shown to reduce wound infection in some studies,^16,17  but not others,³⁸  including a large definitive trial.³⁹  Although oxygen reactive species are required during bactericidal activity,^40–42  excessive oxygen free radicals during hyperoxia may induce DNA and protein damage.^18,19  In the current study, we sought to clarify the effect of hyperoxia on DNA damage by including a third group of patients receiving 80% oxygen. In contrast to nitrous oxide administration, our data showed that supplemental oxygen did not produce detrimental effect on genomic stability.

We used the CometAssay to measure in vivo DNA damage. This avoided introducing unknown effect to the cells during ex vivo treatment (e.g., mitogen stimulation) as in previous studies.²⁷  We attempted to minimize variation in measurements by standardizing the procedure, and performed duplicate measurements for each sample. We also blinded the laboratory personnel from the study group allocation. We expressed DNA damage using the percentage of DNA in tail as the surrogate marker because it is commonly reported.^{9,20,29,34,43}  The amount of DNA in tail indicates single-strand break and correlates with the extent of DNA damage. There are other measures of DNA damage such as the tail length and tail moment. However, the clinical relevance of these measures remains unclear.

There are certain limitations to our study. We measured DNA damage in circulating leukocytes but did not study cell types in other tissues. Although the CometAssay is a sensitive method to detect DNA damage at a single-cell level, it may have missed minor damage less than 50 strand breaks per cell. We found a difference in the rates of wound infection among groups, but the number of events was few, and the CIs were wide. Therefore, this secondary finding could be due to random error and should be interpreted with caution. Furthermore, we studied patients undergoing colorectal surgery at high risk of postoperative surgical wound infection. It is, however, unclear whether the same observations could apply to patients at lower baseline risk of wound infection undergoing minor surgery. Similarly, it is inappropriate to extrapolate our findings to other patient population such as pregnant patients undergoing cesarean section or children undergoing less extensive surgery.

In summary, this randomized controlled trial demonstrated that nitrous oxide administration for over 2h is associated with DNA damage. The extent of DNA damage correlates with postoperative wound infection. Whether there are other adverse consequences of this effect remains unclear, but this certainly deserves further study.