Abstract
Abstract
Background:
Whether the fraction of inspired oxygen (FIO2) influences the risk of surgical site infection (SSI) is controversial. The World Health Organization and the World Federation of Societies of Anesthesiologists offer conflicting recommendations. In this study, we evaluate simultaneously three different definitions of FIO2 exposure and the risk of SSI in a large surgical population.
Patients and Methods:
Patients with clean (type 1) surgical incisions who developed superficial and deep organ/space SSI within 30 days after surgery from January 2003 through December 2012 in five surgical specialties were matched to specialty-specific controls. Fraction of inspired oxygen exposure was defined as (1) nadir FIO2, (2) percentage of operative time with FIO2 greater than 50%, and (3) cumulative hyperoxia exposure, calculated as the area under the curve (AUC) of FIO2 by time for the duration in which FIO2 greater than 50%. Stratified univariable and multivariable logistic regression models tested associations between FIO2 and SSI.
Results:
One thousand two hundred fifty cases of SSI were matched to 3,248 controls. Increased oxygen exposure, by any of the three measures, was not associated with the outcome of any SSI in a multivariable logistic regression model. Elevated body mass index (BMI; 35+ vs. <25, odds ratio [OR] 1.78, 95% confidence interval [CI] 1.43–2.24), surgical duration (250+ min vs. <100 min, OR 1.93, 95% CI 1.48–2.52), diabetes mellitus (OR 1.37, 95% CI 1.13–1.65), peripheral vascular disease (OR 1.52, 95% CI 1.10–2.10), and liver cirrhosis (OR 2.48, 95% CI 1.53–4.02) were statistically significantly associated with greater odds of any SSI. Surgical sub-group analyses found higher intra-operative oxygen exposure was associated with higher odds of SSI in the neurosurgical and spine populations.
Conclusion:
Increased intra-operative inspired fraction of oxygen was not associated with a reduction in SSI. These findings do not support the practice of increasing FIO2 for the purpose of SSI reduction in patients with clean surgical incisions.
T
Furthermore, a recent article by Kurz et al. [4] and a Cochrane Database review [12] found insufficient evidence to recommend increased peri-operative fraction of oxygen as an SSI risk mitigating strategy [4,12]. Peri-operative hyperoxia has been associated with increased long-term risk of adverse cardiac outcomes and reduced oncologic survival after upper abdominal surgery [5,13–15]. As a result, the World Federation of Societies of Anesthesiologists called on WHO to reconsider the guideline recommendation to reflect the preponderance of evidence against the use of high-concentration peri-operative fraction of inspired oxygen (FIO2) in the prevention of SSIs [16].
A major limitation of the literature evaluating the impact of hyperoxia on SSI rates is heterogeneity of definitions and the potential for false-positive findings when definitions of hyperoxia are not predetermined a priori. To understand the impact of hyperoxia on SSI risk better, we used a case control study design to evaluate intra-operative oxygen exposure using three pre-specified definitions.
Patients and Methods
Human subjects protection
This study was approved by the Mayo Clinic Institutional Review Board. The requirement for written informed consent was waived.
Study design/patient selection
The Mayo Clinic Rochester, Infection Prevention and Control (IPAC) Healthcare Associated Infections database was used to ascertain cases of SSI involving type 1 surgical incisions from five surgical specialties (general, orthopedic, vascular, neurologic, and spine) from January 2003 through December 2012. Neurologic surgery and orthopedic spine surgery are separated because of specialty-specific differences at our institution. A type 1 (class I/clean) surgical incision is defined as an uninfected operative site closed primarily and drained with closed drainage if necessary [17]. Surgical site infections were included in the analysis if they occurred within 30 to 90 days following National Healthcare Safety Network/CDC definition. An electronic search algorithm was used to classify cases further in the IPAC by admission diagnosis, surgical diagnosis, and laboratory data. Cases identified electronically were verified subsequently by manual chart review by two independent healthcare providers. Surgical site infections were categorized into three subgroups (superficial, deep, or organ/space) and analyzed both as an overall group and as deep organ/space separately. Cases were matched to controls without SSIs by age (±5 y), gender, American Society of Anesthesiology (ASA) physical status classification (1–2 vs. 3–4), date of surgery (±2 y), and procedure code. Based on the number of suitable controls within the database, cases were matched with between one to three controls following an m:n matching algorithm to maximize the number of controls per case [18,19]. Patients under the age of 18 were excluded from analysis.
Measurement of intra-operative oxygen exposure
Fraction of inspired oxygen data were abstracted from our intra-operative electronic medical record database (Perioperative DataMart), which records inspired oxygen concentration at two-minute intervals, obtained from the anesthesia machine. Intra-operative oxygen exposure was defined a priori as follows: (1) nadir FIO2, (2) percentage of time where FIO2 was greater than 50%, and (3) cumulative hyperoxia exposure, calculated as the area under the curve (AUC) for time when FIO2 was greater than 50%. All three measures were assessed as both continuous and categorical measures, whereas nadir FIO2 was categorized by quartiles as 21%–29%, 30%–37%, 38%–47%, and 48%–100%; percentage of time when FIO2 was greater than 50% was categorized as 0%, >0% to <30%, 30% to <75%, and 75%–100%; and cumulative hyperoxia exposure was categorized as 0, >0 to <5, 5 to <21, and ≥21 FIO2%*h > 50%. Patients for whom FIO2 was not available for at least 95% of the duration from incision to closure, or whose matched case or control had unavailable FIO2 data, were excluded from analysis.
Demographic and covariate data
Baseline demographic data including Charlson comorbidity index components and risk factors for SSI were abstracted from the electronic medical record using predefined search criteria [18].
Analytical approach
Patient demographics and intra-operative oxygen exposure are presented as n (%) for categorical data and as mean (± standard deviation) or median (interquartile range) for parametric and non-parametric continuous data, respectively. Although descriptive statistics for unadjusted data are presented for cases versus controls, statistical comparisons are not shown because unadjusted comparisons do not account for the unbalanced case/control matching and thus are not meaningful. Stratified univariable and multivariable logistic regression models assessed associations between intra-operative oxygen exposure and any SSI and deep organ/space SSI, as well as the associations between patient characteristics and any SSI. Because a large number of confounding factors were anticipated, matching was performed based on the aforementioned categorical parameters and the remaining confounders were controlled for via regression modeling as previously described [20]. All multivariable logistic regression models included the same covariates, selected based on a p value ≤0.10 in univariable models of any SSI including all surgical specialties. Exceptions include sub-group analyses where numbers were too low and excluded to avoid spurious findings. P values were not corrected for the pre-planned comparisons. All statistical analyses were performed with SAS version 9.4, (SAS Institute, Cary, NC).
Results
Baseline comparison of cases and controls
One thousand two hundred fifty cases of SSI were matched to 3,248 controls. Among the cases, 603 had superficial SSIs, 249 had deep space SSIs, 397 had organ/space SSIs, and one had an SSI of unknown type. Patient characteristics were similar and are summarized in Table 1. Area under the curve for FIO2 hours greater than 50%, nadir FIO2, and percentage of time with FIO2 >50% between the two groups are shown in Table 2.
Body mass index was unavailable for 226 controls and 73 cases.
SD = standard deviation; IQR = interquartile range; BMI = body mass index (kg/m2); COPD = chronic obstructive pulmonary disease; RBC = red blood cells.
FIO2 = fraction of inspired oxygen; IQR = interquartile range; AUC = area under the curve.
Matched univariable logistic regression analysis of oxygen exposure for any SSI
Among the five surgical specialties and all incision types, increased cumulative hyperoxia exposure (per increase of 10 FIO2%-hrs) was associated with increased odds of SSI (OR 1.06, 95% CI 1.03–1.08, p < 0.001). This association was not demonstrated among the other two oxygen exposure measures: nadir FIO2 (48%–100% vs. 21%–29%; OR 1.12, 95% CI 0.91–1.38, p = 0.27) or percentage of time with FIO2 >50% (FIO2 >50% for 75%–100% vs. 0% of operative time; OR 1.21, 95% CI 0.99–1.48, p = 0.06).
Matched univariable logistic regression analysis of comorbid conditions and surgical specialty for any SSI
In univariable analysis, increased odds of any SSI among all surgical specialties was associated with BMI (OR 1.04, 95% CI 1.03–1.05, p < 0.001), Charlson comorbidity index score (OR 1.07, 95% CI 1.04–1.10, p < 0.001), surgical duration (per increase of 10 min) (OR 1.02, 95% CI 1.02–1.03, p < 0.001), diabetes mellitus (OR 1.65, 95% CI 1.39–1.95, p < 0.001), congestive heart failure (CHF; OR 1.48 95% CI 1.09–2.00, p = 0.01), peripheral vascular disease (PVD; OR 1.82, 95% CI 1.34–2.45, p < 0.001), chronic obstructive pulmonary disease (COPD; OR 1.36, 95% CI 1.05–1.76, p = 0.02), cerebrovascular disease (OR 1.41, 95% CI 1.08–1.83, p = 0.01), cirrhosis (OR 2.23, 95% CI 1.41–3.53, p < 0.001), and chronic renal disease (OR 1.50, 95% 1.19–1.90, p < 0.001). Blood product administration was also associated with increased odds of any SSI (red blood cells, OR 1.41, 95% CI 1.15–1.72, p < 0.001; plasma, OR 1.73, 95% CI 1.12–2.67, p = 0.01; platelets, OR 1.87, 95% CI 1.19–2.97, p = 0.007). Last, compared with general surgery, vascular surgical procedures (OR 4.55, 95% CI 2.10–9.85, p < 0.001), orthopedic surgical procedures (OR 3.76, 95% CI 2.13–6.62, p < 0.001), and spine surgical procedures (OR 2.83, 95% CI 1.42–5.66, p = 0.003) were also associated with increased odds of SSI.
Matched multivariable logistic regression for any SSI and deep organ/space SSI
In a multivariable logistic regression model, increased oxygen exposure, by any of the three measures, was not associated with the outcome of any SSI. Elevated BMI (35+ vs. <25, OR 1.78, 95% CI 1.43–2.24), surgical duration (250+ vs. <100 min, OR 1.93, 95% CI 1.48–2.52), DM (OR 1.37, 95% CI 1.13–1.65), PVD (OR 1.52, 95% CI 1.10–2.10), and liver cirrhosis (OR 2.48, 95% CI 1.53–4.02) remained statistically significantly associated with greater odds of any SSI (Fig. 1). However, a higher nadir FIO2 was associated with increased deep organ/space SSI (nadir FIO2 48%–100% vs. 21%–29%, OR = 1.37, 95% CI 1.02–1.84, p = 0.04). Surgical sub-group analyses found higher intra-operative oxygen exposure to be associated with higher odds of SSI in the neurosurgical and spine populations. Odds of any SSI were greater in the neurosurgical population with higher nadir FIO2 (nadir FIO2 48%–100% vs. 21%–29%, OR 2.35, 95% CI 1.15–4.79, p = 0.02) and percent of operative time with an FIO2 >50% (FIO2 >50% for 75%–100% vs. 0% of operative time, OR 1.94, 95% CI 1.01–3.73, p = 0.048). Odds of deep organ/space SSI were greater in the neurosurgical population with higher nadir FIO2 (48%–100% vs. 21%–29%, OR 2.82, 95% CI 1.14–6.99, p = 0.03) and in the spine surgery population with higher cumulative oxygen exposure (FIO2%*h > 50% ≥ 21 vs. 0, OR 2.61, 95% CI 1.16–5.90, p = 0.02) and higher nadir FIO2 (48%–100% vs. 21%–29%, OR 2.40, 95% CI 1.15–5.03, p = 0.02).
Multivariable logistic regression accounting for matching for any surgical site infection (SSI) among all surgical specialities. (
Discussion
We tested whether intra-operative oxygen exposure was associated with SSI by applying three pre-specified definitions of intra-operative oxygen exposure and found that greater intra-operative oxygen exposure was not associated with lower odds of SSI. The lack of protective effect of FIO2 on SSI risk persisted across measures of intra-operative oxygen exposure. In all, our findings suggest that increasing FIO2for the sole purpose of SSI reduction is not an effective SSI mitigation strategy.
Increasing FIO2 to improve blood oxygen content and subsequent oxygen delivery to tissues is a plausible mechanism to mitigate SSI and is part of the basis of hyperbaric oxygen therapy use for patients with chronic wounds [21]. Resistance to infection may depend on the partial pressure of oxygen in the wound tissue, and the ability to resist infection may be improved by increasing arterial oxygen tension beyond that required to saturate hemoglobin [22]. Previous data indicate that both traumatic and surgical wounds are hypoxic relative to normal tissue. In fact, oxygen tension in wounds is often less than 30 mm Hg because of disruption of local vascular supply by vessel injury and/or thrombosis [23].
Observations such as these may not translate to acute, clean type 1 surgical incisions because of important pathophysiologic differences in healing processes [24]. Furthermore, the partial pressure of oxygen at the tissue level is difficult to measure with current technology, making the adequacy of oxygen delivery difficult to gauge. It is possible that cases with increased intra-operative FIO2 may have had supra-therapeutic partial pressure of oxygen leading to cellular damage, rather than improved immune function. Unfavorable consequences of hyperoxia such as free radical induced cellular damage, cellular apoptosis, reduced antibacterial function of macrophages, and microbial adaptation and resistance due to mutation and capsular expression may have mitigated any benefit of increased tissue oxygenation [25–32].
The present study demonstrates positive associations between oxygen exposure and deep tissue/organ space SSI, which were strongest in the neurosurgical and spine surgery cohorts. These associations were unexpected and were also present with increased nadir and duration and dose hyperoxia as measured by the AUC definition of oxygen exposure. These results are similar to those reported by Pryor et al. [8] who reported an increase in SSI in patients with a sustained intra-operative FIO2 of 80% undergoing intra-abdominal procedures with class II surgical wounds [8]. Maintenance of tissue oxygenation and end-organ perfusion is of utmost importance and efforts to maintain adequate oxygen delivery must remain a primary peri-operative goal. As such, we would not recommend efforts to restrict FIO2 to prevent SSI at the expense of end-organ oxygenation based upon the results of the present work.
A notable strength of the present investigation is the more comprehensive assessment of intra-operative oxygen exposure. Previous studies that have attempted to determine an association between intra-operative FIO2 and SSI have commonly used two discrete levels of FIO2 as the parameter of interest: percentage of surgical time FIO2 was greater than 50% and intra-operative nadir FiO2 [1–3,5–9]. In this study, in addition to these commonly used measures, duration and dose as reported with the AUC of oxygen exposure was also evaluated. This methodology has recently been used to study the association between hypothermia and SSI and transfusion requirements, and intraoperative hypotension and perioperative ischemic outcomes [33], but to our knowledge, using this more comprehensive definition of intra-operative oxygen exposure has not been used previously to study SSI [18,33,34]. The study further benefited from a large, multi-specialty sample and the utilization of innovative and validated electronic search strategies (Perioperative DataMart, IPAC) to ensure the accurate identification of the study population and detailed characterization of their baseline demographics and clinical conditions as well as surgical exposures and associated physiologic responses.
Conclusions from the present study are limited by its single-center, retrospective design. As with all observational investigations, confounding and bias are important considerations, which we attempted to address as effectively as possible with a matched study design, restriction to type 1 incisions, and rigorous statistical adjustments. We did not match the SSI population to controls using all of the accepted risk factors identified by the American College of Surgeons National Surgical Quality Improvement Program (NSQIP). However, in addition to our matching criteria (ASA physical status, year of surgery, gender, procedural code, and wound class), we adjusted for imbalances in most of the relevant measured parameters (e.g., DM, BMI, COPD, PVD, surgical duration, blood product administration, etc.). We were unable to adjust for the Centers for Medicare and Medicaid Services and the Joint Commission's Surgical Care Improvement Project (SCIP) initiatives as theses were implemented in stages during the study period and all available measures were note reported in our database. However, previous work has failed to show a reduced incidence of SSI with improved SCIP compliance [18]. Similarly, we were unable to adjust for blood glucose control because of limitations with our dataset. Fraction of inspired oxygen data were not corrected for blood oxygen saturation, which limits the associations found. It is difficult to know if an increased FIO2 was associated with the outcome of SSI or whether the FIO2 was high because of a low arterial oxygen saturation. Last, SSIs diagnosed at an outside center were not included in our analysis, because these would not have been captured in our IPAC database.
Conclusion
Increased intra-operative FIO2 was not associated with a reduction in SSI. Increasing FIO2 for the sole purpose of SSI reduction does not appear to be an effective strategy in patients with type 1 surgical incisions.
Footnotes
Acknowledgments
Funding support was provided by the Mayo Clinic Small Grants Program and The Robert D. and Patricia E. Kern Center for the Science of Health Care Delivery and were used for design and conduct of the study; collection, management, analysis and interpretation of the data; and preparation, review, or approval of the manuscript.
Authors Disclosure Statement
All authors have no competing financial interests.