Obstructive sleep apnea (OSA), the most prevalent breathing disturbance in sleep,1 affects 2–26% of the general population, depending on sex, age, and the criteria of diagnosis.2 A growing body of literature suggests that OSA has been implicated in the development of cardiorespiratory diseases.310 OSA also has a significant impact on both acute and chronic cardiorespiratory function.11

Although no large epidemiologic studies have been undertaken to determine the prevalence of OSA in the general surgical population, reports of several studies suggest that the prevalence of OSA in the surgical population might be higher than in the general population, with variability among different surgical populations. In particular, as many as 29/41 (71%) of patients undergoing bariatric surgery were found to have OSA.12 In the general surgical population, 24% of surgical patients were identified by the Berlin questionnaire as being at a high risk of having OSA.13

Physiological studies show that the upper airway muscles are more sensitive to neuromuscular blocking drugs than either the diaphragm or the peripheral muscles.14 Patients with OSA are highly vulnerable to medications that suppress pharyngeal muscle activity.1517 In addition to compromising upper airway muscle function, general anesthesia and narcotics can depress the patients’ ventilatory responses to obstruction and inhibit their normal arousal and awakening responses to hypoxia and hypercapnia.18 A high incidence of early and late postoperative nocturnal hypoxemic episodes in healthy patients undergoing surgery has been reported in a number of studies.1921

However, the currently available literature on the occurrence of perioperative respiratory complications among surgical patients with OSA is too limited to make any major correlations. The objective of this study was to test the hypothesis that OSA is a risk factor for the development of postoperative complications.

Methods

We undertook a retrospective matched cohort study of a large cross section of surgical patients at the University Health Network, Toronto. Following research ethics board approval, the study subjects were selected from the hospital administrative database. The institutional research ethics board waived the requirement for individual informed consent.

Surgical patients in the OSA group were selected based on the following criteria: (1) age > 18 yr; (2) diagnosis with OSA when discharged from hospital (the discharge diagnosis of OSA was coded according to the international classification of disease (ICD-9) codes; (3) undergoing elective surgery under general, regional, or local anesthesia with monitored anesthesia care at the University Health Network during the period January 1, 1990 to December 31, 2005. Exclusion criteria included patients who were undergoing surgical procedures involving the upper airway, including tonsillectomy, septoplasty, uvuloplasty, uvulopalatoplasty, uvulopharyngoplasty, or uvulopalatopharyngoplasty, as these procedures were most likely to have been indicated to cure the primary disease process (OSA) and, thus, would be impossible to match. The non-OSA patients (non-OSA group) were based on a one-to-one match with the OSA patients and were selected from a cohort of surgical patients without an OSA diagnosis at the time of hospital discharge. The match criteria included gender, age difference < 5 yr, same type of surgery, and <5 yr between the two surgery dates.

The primary outcome variable was the incidence of the total postoperative complications. Data collection was implemented through chart review. The charts of the selected patients were reviewed by a research anesthesiologist (S.V.). The following data were collected: demographic data (including gender and age at the time of surgery), ASA physical status, pre-existing medical conditions, concurrent medications, type of surgery and anesthesia, postoperative complications and therapeutic interventions. The definitions of postoperative complications are shown in the Appendix.

Statistical considerations

In a previously published paper of a similar design,22 the incidence of postoperative complications was 39% among OSA patients and 18% among non-OSA patients. Assuming a similar incidence of postoperative complications in our study, with α = 0.05 and β = 0.9, we required 132 pairs of matching OSA and non-OSA patients.

The data collected were entered into a specifically designed Microsoft Access™ database (Microsoft Corporation, Redmond, WA, USA). SAS® 9.1.3 for Windows® (SAS Institute Inc., Cary, NC, USA) was used for data analysis. The demographic data and prevalence of the pre-existing co-morbidities were summarized. The incidence of postoperative complications and related treatments were compared between the OSA patients and the matched non-OSA patients. Testing of differences between the two groups was undertaken using the McNemar test for categorical data and Student’s paired t tests for numerical data.

Conditional logistic regression was used to adjust for potential confounding variables.23,24 The selection of potential risk factors was based mainly on the clinical relevance to the postoperative outcome. The correlation among the potential risk factors was also checked. When there was a correlation between two or more risk factors with a correlation coefficient >0.5, only one risk factor was retained. The chosen potential risk factors were used as independent variables, and the frequency of patients with one or more postoperative adverse events was used as the dependent variable for conditional logistic regression analysis. A backward automatic selection with P < 0.2 was used in conditional logistic regression. The appropriateness of the model was assessed based on fit statistics and testing for a global null hypothesis with β = 0.

Results

Two hundred ninety-four patients with an OSA diagnosis underwent different types of surgery; however, 19 of these were excluded from our study due to having upper airway surgical procedures. To establish matched pairs, 275 OSA patient records were compared with the records of patients without diagnosed OSA. Two hundred forty patients with OSA were successfully matched with 240 patients without diagnosed OSA. The analysis is based on the data from the 240 matched pairs.

The OSA group and the matched non-OSA group were similar with respect to gender distribution and type of surgery (Tables 1, 2), and both groups were similar in age (Table 1). However, compared with the matched non-OSA patients, the OSA patients had a higher prevalence of pre-existing co-morbidities, including obesity, hypertension, gastroesophageal reflux disease, diabetes, hypothyroidism, asthma, and chronic obstructive pulmonary disease. A larger proportion of the OSA patients were ASA physical status III and IV, and they had higher New York Heart Association (NYHA) classification scores. The OSA patients also had a higher mean body mass index (BMI), and 150 (63%) of them were on home continuous positive airway pressure (CPAP).

Table 1 Demographic data
Table 2 Type of surgery and anesthesia

The types of anesthesia used were similar in the two groups (Table 2). However, a higher percentage of the OSA patients were ranked as class III and IV on laryngoscopic viewing (Table 3). More patients with OSA had difficult tracheal intubations (20% OSA group vs 10% non-OSA group; P = 0.003). There was no significant difference in the percentage of patients who were receiving opioid analgesics postoperatively, and mean doses of postoperative opioid were similar in the two groups (data not shown).

Table 3 Airway assessment and tracheal intubation

Postoperative complications

Table 4 summarizes the occurrence of postoperative complications in the OSA and non-OSA groups. There was a significantly greater overall incidence of postoperative complications in the OSA group compared with the non-OSA group (44% vs 28%, respectively; P < 0.01). The major contributor to the higher occurrence of postoperative complications in the OSA group was the increased incidence of respiratory complications (33% OSA group vs 22% non-OSA group). Desaturation with SaO2 < 90% was the most common complication (17% OSA group vs 8% non-OSA group). The majority of the postoperative complications occurred after patients were transferred to the ward (25% OSA group vs 16% non-OSA group). Although the groups had similar numbers of minor cardiovascular and neurological complications, two patients in the OSA group and one patient in the non-OSA group suffered from cardiac arrest. The details of these patients are summarized in Table 5.

Table 4 Postoperative complications
Table 5 Detailed information of patients suffering cardiac arrest

The OSA group experienced an increased number of treatments (Table 6). More OSA patients required prolonged oxygen therapy (23% OSA group vs 15% non-OSA group; P < 0.05). In the OSA group, 13% of patients required additional monitoring vs 6% of patients in the non-OSA group (P < 0.01). There were more intensive care unit (ICU) admissions among the OSA patients (40% OSA group vs 28% non-OSA group; P < 0.01), whereas the majority of ICU admissions were planned. Forty-nine percent of OSA patients received CPAP postoperatively.

Table 6 Postoperative treatment and follow-up

Independent risk factors of postoperative complications

The following medical conditions were included in the conditional logistic regression analysis according to their clinical relevance as potential risk factors for the occurrence of postoperative complications: the diagnosis of OSA and co-morbidities, including hypothyroidism, stroke, coronary artery disease, diabetes, heart failure, asthma, chronic obstructive pulmonary disease (COPD), ASA physical status III–IV, and BMI > 35 kg · m−2. In the final model obtained through a backward automatic selection with P < 0.2, the following medical conditions were retained in the model as risk factors for the occurrence of postoperative complications: the diagnosis of OSA, a high score on ASA physical status, pre-existing stroke, and asthma. Only the diagnosis of OSA and stroke were significant (P < 0.05). The P value for the hypothesis of β = 0 was 0.001 (Wald test) and <0.001 (Likelihood ratio test and Score test). The hazard ratios and P values of the risk factors are summarized in Table 7.

Table 7 Risk factors for postoperative complicationsa

Postoperative complications and need for CPAP in OSA patients

To explore the relationship between CPAP utilization and the occurrence of postoperative complications in OSA patients, we examined the frequency of postoperative complications in OSA patients using different combinations of CPAP use at home and after surgery (Table 8). Of the 150 OSA patients who were on home CPAP, only 63% of patients received postoperative CPAP. Of the 90 patients who were not on home CPAP, 27% (24/90) received postoperative CPAP. This group of patients had the highest incidence of postoperative complications, mainly desaturation with SaO2 < 90%. The overall incidence of postoperative complications was 44% in OSA patients and 28% in non-OSA patients.

Table 8 Use of CPAP and postoperative complications in OSA patients

Discussion

This retrospective cohort study demonstrates that patients with OSA have a higher incidence of postoperative complications compared with matched non-OSA surgical patients. Oxygen desaturation with SpO2 < 90% was the most common complication. The OSA patients who were not receiving home CPAP therapy before surgery and required CPAP after surgery had the highest incidence of postoperative complications. There was a need for a greater number of treatments and interventions in the OSA group, including prolonged oxygen therapy and additional monitoring. Multivariable logistic regression analysis demonstrated that the diagnosis of OSA and pre-existing stroke are risk factors for the occurrence of postoperative complications.

We recognize that this study has a number of important limitations. First, this was a retrospective analysis with all outcomes collected post hoc. The results must be considered as merely associations and causality cannot be implied. Second, patients with OSA were identified by using ICD-10 codes that are entered by non-physician coders after the patient has been discharged from hospital. Since the completion of this study, we have undertaken to quantitate the sensitivity and positive predictive value of ICD-10 codes to identify patients with OSA. During the year April 2008 to April 2009, there were 5542 elective non-cardiac surgical patients assessed prior to surgery. Of the 379 patients diagnosed with OSA (6.8%), 36 had an ICD-10 code which specified a preoperative diagnosis of OSA. Thus, the sensitivity of ICD-10 codes is 0.09 with a positive predictive value of 0.86. We are uncertain why only 10% of OSA patients were identified; perhaps our analysis preferentially identified a group of patients with severe OSA. Thus, our results may be applicable only to patients with moderate and severe OSA. Third, and related to the above point, there were no polysomnographic data available for the patients with OSA; thus, we could not quantitate the severity of OSA. Finally, our matching process was constructed so that age, gender, type of surgery, and surgery date were matched. However, this process did not match for individual patient characteristics, and our results show that there were more associated co-morbidities in the OSA patients, including higher ASA status, NYHA status, hypertension, diabetes, asthma, and gastroesophageal reflux disease (GERD). Thus, we cannot exclude the possibility that the associated co-morbidities may account for the higher complication rates rather than the diagnosis of OSA itself. Despite the above limitations, this study indicates that patients diagnosed with OSA had an increased incidence of postoperative adverse events, mainly oxygen desaturation with SpO2 < 90%. After adjusting for other confounding factors using conditional logistic regression, OSA remained a significant risk factor with an odds ratio of 2.00 (1.25–3.19).

Most previously published studies examining postoperative complications in OSA patients focused on patients who underwent upper airway surgeries.2531 Only a few studies have evaluated postoperative complications in patients who underwent other types of surgery.22,3236 Our study complements this growing body of literature by demonstrating that OSA patients undergoing different types of surgery have an increased rate of postoperative complications, mainly oxygen desaturation. In a retrospective study of 101 patients with OSA who were undergoing hip replacement or knee replacement, Gupta et al. 22 demonstrated that patients with OSA have an increased incidence of total postoperative complications and major complications. The patients with OSA in that study also had a higher percentage of total and unplanned intensive care unit (ICU) transfers and a longer hospital stay.22 In another retrospective study of 37 patients with OSA undergoing cardiac surgery, Kaw et al. 32 found that there was an increased incidence of postoperative encephalopathy and infection rates (mostly mediastinitis). Hwang et al. 35 showed that surgical patients with the clinical features of OSA and oxygen desaturation index ≥5 on home nocturnal oximetry before surgery had a significantly higher rate of postoperative complications. In another study conducted by Chung et al.,34 the patients who had apnea-hypopnea index >5 on preoperative polysomnography had a higher incidence of postoperative complications. These studies all support that OSA in surgical patients is associated with an increased incidence of postoperative complications.

In this study, hypoxemia (oxygen desaturation) was the most common postoperative complication in patients with OSA. Three other studies have also demonstrated a similar result.22,34,35 However, a recent study of 31 OSA and 9 non-OSA morbidly obese patients did not find a difference in the number of hypoxemic episodes during the first 24 hr after surgery.36 This similarity of observed events may have been due to the high percentage of patients in both groups receiving oxygen therapy during the first 24 hr after surgery.36 Recently, we found that the apnea-hypopnea index (AHI) and the oxygen-desaturation index are greater among OSA patients on the third postoperative night compared with either the first postoperative night or preoperatively.37

In Gupta’s retrospective study on 101 matched pairs of patients undergoing hip or knee replacement, major complications (cardiac events and complications needing ICU transfer or urgent respiratory support) were significantly higher in patients with OSA.22 Our study did not show an increase in serious cardiopulmonary complications in OSA patients vs matched non-OSA patients. The cardiac arrests in two OSA patients, one during tracheal intubation and the other after extubation, were related to problems with airway management. This further supports the notion that airway management in OSA patients can be difficult. The cardiac arrest that occurred in one non-OSA patient was related to myocardial infarction.

Our results suggest that patients with OSA require more perioperative care than their non-OSA counterparts. Patients with OSA required prolonged oxygen therapy and additional monitoring and were admitted to ICU more frequently. Gupta’s study also demonstrated that more OSA patients required ICU care, both planned and unplanned.22 The unplanned ICU transfers occurred mainly in patients whose OSA was undiagnosed at the time of surgery. In our study, it is likely that patients were diagnosed before surgery, although we cannot exclude the possibility that OSA was diagnosed as a result of a perioperative event. The departmental policy that existed during the study period did not require OSA patients to be monitored more intensively or to be admitted to ICU. However, the anesthesiologist may have been aware of OSA when considering the need for ICU. This might account for the higher number of patients requiring ICU transfer in the OSA group and no increase in the unplanned ICU admission rate.

The recent ASA practice guideline for the perioperative management of patients with OSA recommends that the CPAP or nasal intermittent positive pressure ventilation (NIPPV) should be administered postoperatively to the OSA patients on home CPAP.38 As a matter of routine in most hospitals, a patient with OSA on home CPAP would be ordered to receive CPAP after surgery. However, our data suggests this guideline is not routinely observed, as only 63% of OSA patients on home CPAP received CPAP postoperatively. This observation suggests the need for education programs to ensure greater consistency of care for patients with OSA.

The patients with OSA who were not receiving home CPAP included two different patient populations, i.e., patients who did not require CPAP because of mild OSA and patients who were non-compliant with their CPAP devices. In our study, 27% of OSA patients not using home CPAP required CPAP postoperatively. These patients had the highest incidence of postoperative complications (Table 8). This finding is likely due to CPAP having been initiated in response to an adverse event, which may also have led to the chart abstracters coding for the disease. However, we must again reiterate that this is a retrospective study, and it is difficult to identify the exact reasons for the high incidence of postoperative complications in this group of patients.

In conclusion, these results suggest that one aspect of the preoperative assessment should focus on the evaluation and diagnosis of OSA. Patients who are non-compliant with their treatment of CPAP are at an increased risk of postoperative oxygen desaturation. These results await further prospective assessment.