Patients with obstructive sleep apnea (OSA) experience episodic upper airway obstruction during sleep, which causes symptoms of disruptive snoring and daytime sleepiness. In addition, these patients are at increased risk for hypertension, heart failure, myocardial infarction, and stroke when compared with the general population (1,2). Symptomatic OSA producing at least a moderate degree of sleepiness affects approximately 2% and 4% of middle-aged women and men, respectively. The overall prevalence of sleep-disordered breathing is even more frequent, estimated to be approximately 10% and 25% for middle-aged women and men, respectively. Many of these cases are unrecognized by the patients themselves or their health care providers (3,4).
The significance of an OSA diagnosis as a risk factor for perioperative complications is unclear. Most data come from studies of uvulopalatopharyngoplasty, a surgical procedure used to treat OSA. In general, OSA patients are at risk for postoperative airway obstruction, desaturation, and cardiac complications (5–11), although it is difficult to separate the effects of the surgery itself from the effects of the underlying OSA. For other surgical (nonairway) procedures, there are several case reports of postoperative respiratory complications, including airway obstruction and oxygen desaturation, in patients with OSA. However, only one prior study has actually quantified perioperative risk. Gupta et al. (12) found an increase (approximately twofold) in the frequency of adverse perioperative outcomes in OSA patients undergoing hip or knee replacement compared with a group of matched control patients. The risk posed by OSA in patients undergoing more minor procedures, such as outpatient surgery, is unknown. Indeed, it could be questioned whether patients with the diagnosis of OSA should be managed as outpatients.
The purpose of this study was to determine whether the preoperative diagnosis of OSA, confirmed by polysomnography, is a significant independent risk factor for unanticipated hospital admission or readmission in patients scheduled for outpatient surgery. The medical records of two cohorts of patients were retrospectively examined: a cohort with OSA diagnosed with polysomnography according to defined criteria, and a matched control cohort.
This study was approved by the Mayo IRB. Sleep laboratory and billing databases were used to identify all patients with a diagnosis of OSA billed by the Mayo Clinic Rochester Sleep Laboratory over the period of August 1993 to March 2000. This list was then cross-matched with a database to identify those patients who had undergone outpatient surgical procedures after the date of OSA diagnosis under general or major regional (central neuraxial) anesthesia in the period from January 1997 through December 2000. Otorhinolaryngologic procedures were excluded. For patients who had undergone more than one procedure, records from the first eligible surgical procedure after diagnosis were examined. Records were examined to confirm that the patient met standard sleep laboratory polysomnographic criteria for OSA (including a respiratory disturbance index >10) (2,3,13).
The control patients were selected from an anesthesia database, excluding patients with diagnoses consistent with sleep-disordered breathing, such as hypersomnia with sleep apnea. By using a 1:1 matched-set study design, controls were matched for type of anesthesia, age (±5 yr), sex, body mass index (BMI) (±5 kg/m2), surgical procedure (by primary procedure code), and date of surgery (±1 yr). For identified controls, the first outpatient nonotorhinolaryngologic surgical procedure within 1 yr of the case procedure was examined. If no control could be identified with the above criteria, the matching intervals for date of surgery, age, and BMI were relaxed until a match was identified.
Abstracted preoperative data included demographic data, respiratory disturbance index measured by polysomnography, use of continuous positive airway pressure (CPAP), surgical procedure, comorbid disease as previously defined (14), and ASA physical status. Intraoperative data included type and duration of anesthesia, method of airway management (including provider assessment of intubation difficulty according to a three-point scale), primary anesthetic, use of muscle relaxants, type and amount of narcotics administered, and notation of the following adverse events: bronchospasm, airway obstruction, and “other” adverse events (intended to encompass major cardiorespiratory complications). Postoperative data included duration of postanesthesia care unit (PACU) admission and hospital stay, adverse events occurring in the PACU (reintubation or airway obstruction), and the occurrence of unplanned hospital admission. Adverse events that occurred over the period from the time of anesthetic induction to the time of PACU discharge were quantified under the category “perioperative adverse event.” Any intraoperative or PACU adverse event that resulted in admission was also included as an “unanticipated hospital admission,” and the indication for admission was appropriately noted.
Table 1 presents definitions of hospital admission indications. More than one indication for admission was recorded if appropriate. The category “other problem” was reserved for two indications: urinary retention and cases performed late in the day, thus requiring that the patient stay overnight because the outpatient surgery unit was closed.
The primary end-point was frequency of unplanned hospital admission (including readmission within 24 h). This study used a 1:1 matched-pair design whereby each OSA patient was matched to a non-OSA control. All analyses performed took into account the 1:1 matched-pair design. Conditional logistic regression analyses were performed to determine whether patient and procedural characteristics differed significantly between groups. Regression was also used to assess whether OSA is an independent risk factor for unplanned hospital admission. Because exact matching for age, BMI, and date of surgical procedure was not achieved, we included these variables as covariates in the logistic regression analysis. Two-sided tests were performed, with P values ≤0.05 considered statistically significant.
The admission rate for patients scheduled as outpatients at our institution is approximately 9%. Because this includes patients receiving monitored anesthesia care, the rate for patients receiving only general or major regional anesthesia would presumably be more frequent. In general, if the rate of admission for the control group is 10% or more, a sample size of 234 in each group provides statistical power of >85% to detect a doubling in the rate of admission for OSA patients (e.g., 10% versus 20%).
A total of 234 patients were included in the OSA group. An equal number of control patients were matched exactly for sex, type of anesthesia, and surgical procedure. For the other matching variables, surgery date was matched within ±1 yr for 92% of the pairs, with a median difference (OSA patients minus controls) of 28.5 days (range, −2.4 to 2.8 yr). BMI was matched within ±5 kg/m2 for 90% of the pairs, with a median difference of 2.3 kg/m2 (range, −5.9 to 11.9 kg/m2). Age was matched within ±5 yr for 94% of the pairs, with median difference of 0 yr (range, −30 to 24 yr).
Overall BMI was slightly higher in the OSA group (Table 2). Patients with OSA were more likely to have insulin-dependent diabetes, hypertension, and a higher ASA physical status. Baseline characteristics of the two groups were otherwise similar (Table 2).
The method of anesthesia was not different between the two groups, with the exception that OSA patients were more likely to be endotracheally intubated and less likely to have a laryngeal mask airway placed for airway management (Table 3). There were no significant differences between the two groups in the type of anesthetics, muscle relaxants, or narcotics used. Three patients in the OSA group used CPAP in the postoperative period while in hospital.
Fifty-six (23.9%) of the OSA patients and 44 (18.8%) of the non-OSA patients experienced an unplanned hospital admission, a difference that was not significant (odds ratio, 1.4; 95% confidence interval, 0.8–2.5;P = 0.316) (Table 4). The most common indication for unplanned hospital admission was related to surgical factors (Table 5). Most of these events represented a change in the surgical procedure to a more invasive one than originally planned. The other common reasons for admission included pain and “other problem,” with most of the latter admissions due to urinary retention. Four patients with OSA and two without OSA were admitted for “other respiratory indication.” Among the OSA patients in this category, one was admitted for pain control and physical therapy but was later noted to have bronchospasm that required therapy. The other three patients developed episodic arterial oxygen desaturation after PACU discharge sufficient to require supplemental oxygen overnight, in the opinion of the attending physician. In the control group, two patients experienced episodic arterial oxygen desaturation sufficient to require supplemental oxygen.
There was no difference between groups in the frequency of perioperative adverse events (occurring during surgery or in the PACU) noted in the anesthesia records (Table 4). Five OSA patients experienced such an event: one patient had airway obstruction during surgery because of difficult mask ventilation (and was subsequently successfully intubated), three patients had temporary postoperative upper airway obstruction after extubation, and one patient had intraoperative bronchospasm and postoperative upper airway obstruction and was admitted for observation. No patient required reintubation. Three control patients had a perioperative adverse event: one patient had paroxysmal atrial fibrillation with ST segment depression during surgery, one patient had an anaphylactic reaction with hypotension and flushing, and one patient had laryngospasm with pulmonary edema and subsequent admission for ventilatory support.
We found that patients with the diagnosis of OSA scheduled for outpatient surgery were not admitted more frequently than their age-, sex-, and BMI-matched counterparts without this diagnosis. When admission did occur, the reason for admission was usually unrelated to respiratory or cardiovascular events. In addition, patients with OSA in this study were no more likely than control patients to have perioperative morbidity, such as oxygen desaturation or airway obstruction, noted in their perioperative records.
Sleep-disordered breathing in general, and OSA in particular, is relatively common (15). Several factors should make patients with OSA particularly susceptible to airway obstruction in the perioperative period. Anesthetics and analgesics themselves interfere with the function of the upper airway muscles, so that even healthy patients can develop postoperative obstruction; patients with OSA may be particularly susceptible to this effect. Also, sleep architecture is abnormal in the postoperative period, disturbed by drug effects and pain, which can also interfere with the regulation of upper airway muscles. Finally, trauma from airway manipulation can result in edema that worsens airway obstruction.
Although it is plausible that patients with OSA are at increased risk for postoperative airway obstruction and related respiratory complications after nonairway surgery, few studies evaluate this assumption. Case reports describe patients with significant postoperative airway obstruction who were subsequently diagnosed with OSA (16–18). Postoperative arterial desaturation as measured by oximetry occurs with increased frequency in patients considered to be at risk for OSA (19,20). For example, Isono et al. (20) found that preoperative apnea and hypoxia were correlated with postoperative hypoxemia. However, Rosenberg et al. (21) could find no correlation between preoperative snoring habits and postoperative hypoxemia. Interpretation of all these studies is limited by a general lack of diagnostic criteria for OSA. The best evidence for the role of OSA as a risk factor for postoperative complications comes from Gupta et al. (12). They used a methodology similar to ours, examining the records of patients with OSA documented by polysomnography (either before or after surgery) who underwent hip and knee replacement, and a group of patients matched for age, sex, surgeon, type of operation, and type of anesthesia. Patients were not matched for BMI, which tended to be higher in OSA patients. When all complications (including cardiac, neurologic, and other) were analyzed as a group, the authors found a significantly more frequent rate in OSA patients. However, the only individual complication that was significantly different between groups was unplanned intensive care unit admission; in particular, the frequency of acute hypercapnia, episodic hypoxemia, or reintubation was not different between groups.
To better evaluate OSA specifically as a risk factor for perioperative morbidity, we attempted to match as many factors as possible between groups. This was successful for most pairs, and those variables that could not be matched within parameters were included as covariates in analysis. Consistent with the typical characteristics of patients with OSA, most were obese, with BMI slightly higher in the OSA group. Other features of anesthetic management, including the use of opioids and choice of anesthetic technique, were similar between groups, with the exception of small but statistically significant difference in airway management. It appears that providers were more reluctant to use the laryngeal mask airway in patients with OSA. Although this was not a primary goal of our study, it was interesting to note that the frequency of difficult intubation did not differ between groups. Thus, it is possible the suspected relationship between difficult intubation and OSA noted in prior studies (22,23) is related more to obesity than specifically to OSA; our study design is not sufficient to answer this question.
We chose to examine unanticipated hospital admission as our primary outcome, reasoning that this would be a sensitive indicator of serious morbidity and would have a direct bearing on the question of whether patients with OSA could be managed as outpatients. However, this variable was not specific to anesthesia-related indications, because most patients were admitted for surgical indications. Nonetheless, the number of admissions for reasons possibly related to anesthesia, such as nausea, prolonged regional block, or respiratory problems, was identical (n = 12) in both groups. Patients with OSA tended to require admission for pain control more frequently, for reasons that are unclear. One could speculate that providers were more reluctant to use opioids in OSA patients, but the rate of long-acting opioid use in both groups was relatively infrequent and not different. The overall rate of unanticipated admission was relatively more frequent in both groups—more than the overall rate at our institution (9%). This may reflect the fact that we did not study patients who received monitored anesthesia care (who are included in the overall institutional rate) or that characteristics common to both groups (such as obesity or other comorbidity) may increase the admission rate. Notably, however, the cause of admission was usually unrelated to cardiopulmonary or upper airway factors in either the control or the OSA group.
In addition to the factors responsible for admission, the frequency of adverse events occurring during surgery was small and did not differ between groups. Furthermore, no adverse event led to mortality or permanent morbidity in either group. With 0 observed events in 234 patients, the upper 95% confidence interval for this finding is 1.6%. Although this is reassuring, it does not exclude the possibility of significant morbidity in this population.
Beyond the usual limitations of any retrospective analysis, several other limitations of this study must be recognized. The most obvious is that the control patients did not have polysomnography to exclude the diagnosis of OSA. Population-based studies have estimated the prevalence of undiagnosed OSA to be between 2% and 4%(4,24). Data from Duran et al. (24) suggest that the prevalence in men and women with a BMI >30 kg/m2 may be two to three times more frequent. Therefore, it is possible that the rate of undiagnosed OSA in our control group could approximate 10%. Inclusion of patients with undiagnosed OSA in the control group could potentially obscure an increased rate of adverse events in the OSA group. Thus, it is important to interpret this study as evaluating the perioperative risk in patients presenting for surgery with the diagnosis of OSA confirmed by polysomnography. Related to this issue, it must be understood that we studied the subset of patients with OSA who were diagnosed by polysomnography and were probably appropriately treated (as indicated by the increased frequency of preoperative CPAP use). Thus, the results may not be typical of all patients with OSA, who may not be appropriately treated. The results may also not apply to patients undergoing airway surgery.
The fact that these patients were diagnosed with OSA also may have affected their treatment and subsequent outcomes in the perioperative period. Providers may have altered their practices given the diagnosis. We found little evidence of systematic differences in anesthetic management, other than the difference in airway management technique noted above. Many of the patients were treated before (but not after) surgery with CPAP. This likely optimized their preoperative physical status. The response of patients with undiagnosed OSA, or OSA that has not been adequately treated, may be quite different.
Another possible study limitation involves selection bias in the scheduling of outpatient surgery. Patients with OSA may be more likely to be scheduled as inpatients rather than outpatients because of concerns regarding perioperative risk. Thus, we may have selected a relatively healthy population of patients diagnosed with OSA. To estimate whether the diagnosis of OSA affects surgical scheduling (inpatient versus outpatient), we examined the scheduling of the five most common surgical procedures in our groups over the period of study, using the criteria of sleep laboratory billing and recorded OSA diagnosis to define OSA patients. For each procedure, we compared the type of surgical scheduling in OSA and non-OSA groups (Table 6). OSA patients were less likely to be scheduled as outpatients for only one of these procedures, with other procedures showing an opposite tendency. Thus, we find little evidence of a bias to schedule patients with the diagnosis of OSA as inpatients in our practice.
The overall rate of unanticipated admission at our institution (9%) is more frequent than the national average and may reflect the nature of the patient population at our tertiary referral center. It may also reflect a lower threshold for admission given the relative convenience of inpatient resources in our setting compared with a freestanding outpatient facility. Thus, these results may not apply to other types of outpatient facilities, although we note that because of the matched-pairs study design, such systematic differences in admission practices are taken into account.
Finally, we note that because of the nature of the surgical procedures, requiring relatively small use of long-acting opioids, we did not examine in this study one important concern in current practice. Case reports suggest that these patients may be particularly sensitive to opioid-induced ventilatory depression (25), even when opioids are administered via the epidural space (17). This issue requires further study, but our results should not be interpreted as justifying any diminution of vigilance in this setting.
In conclusion, we find that in patients scheduled for outpatient surgery in a large academic practice, the diagnosis of OSA confirmed by polysomnography is not an independent risk factor for unanticipated hospital admission or for other perioperative adverse events. These results suggest that patients with recognized and treated OSA can be safely and appropriately managed as outpatients for less-invasive surgical procedures.
The authors thank Janet Beckman for secretarial support, Paul Decker for statistical support, Cameron Harris for assistance with the sleep laboratory function database, and Marie Nepper for assistance with recovery of patient histories.
1. Partinen M, Jamieson A, Guilleminault C. Long-term outcome for obstructive sleep apnea syndrome patients: mortality. Chest 1988; 94: 1200–4.
2. Shepard JW Jr. Hypertension, cardiac arrhythmias, myocardial infarction, and stroke in relation to obstructive sleep apnea. Clin Chest Med 1992; 13: 437–58.
3. Strollo PJ Jr, Rogers RM. Obstructive sleep apnea. N Engl J Med 1996; 334: 99–104.
4. Young T, Palta M, Dempsey J, et al. The occurrence of sleep-disordered breathing among middle-aged adults. N Engl J Med 1993; 328: 1230–5.
5. Burgess LP, Derderian SS, Morin GV, et al. Postoperative risk following uvulopalatopharyngoplasty for obstructive sleep apnea. Otolaryngol Head Neck Surg 1992; 106: 81–6.
6. Lee WC, Skinner DW, Prichard AJ. Complications of palatoplasty for snoring or sleep apnoea. J Laryngol Otol 1997; 111: 1151–4.
7. Riley RW, Powell NB, Guilleminault C, et al. Obstructive sleep apnea surgery: risk management and complications. Otolaryngol Head Neck Surg 1997; 117: 648–52.
8. Mickelson SA, Hakim I. Is postoperative intensive care monitoring necessary after uvulopalatopharyngoplasty? Otolaryngol Head Neck Surg 1998; 119: 352–6.
9. Terris DJ, Fincher EF, Hanasono MM, et al. Conservation of resources: indications for intensive care monitoring after upper airway surgery on patients with obstructive sleep apnea. Laryngoscope 1998; 108: 784–8.
10. Mooe T, Gullsby S, Rabben T, Eriksson P. Sleep-disordered breathing: a novel predictor of atrial fibrillation after coronary artery bypass surgery. Coron Artery Dis 1996; 7: 475–8.
11. Dominguez Ortega L, Carnevali-Ruiz D, Diaz Gallego E. Sleep apnea and the risk for perioperative myocardial infarction. Ann Intern Med 1993; 119: 953.
12. Gupta RM, Parvizi J, Hanssen AD, Gay PC. Postoperative complications in patients with obstructive sleep apnea syndrome undergoing hip or knee replacement: a case-control study. Mayo Clin Proc 2001; 76: 897–905.
13. Hoffstein V, Szalai JP. Predictive value of clinical features in diagnosing obstructive sleep apnea. Sleep 1993; 16: 118–22.
14. Warner DO, Warner MA, Offord KP, et al. Airway obstruction and perioperative complications in smokers undergoing abdominal surgery. Anesthesiology 1999; 90: 372–9.
15. Benumof JL. Obstructive sleep apnea in the adult obese patient: implications for airway management. J Clin Anesth 2001; 13: 144–56.
16. Vidhani K, Langham BT. Obstructive sleep apnoea syndrome: is this an overlooked cause of desaturation in the immediate postoperative period? Br J Anaesth 1997; 78: 442–3.
17. Ostermeier AM, Roizen MF, Hautkappe M, et al. Three sudden postoperative respiratory arrests associated with epidural opioids in patients with sleep apnea. Anesth Analg 1997; 85: 452–60.
18. Lamarche Y, Martin R, Reiher J, Blaise G. The sleep apnoea syndrome and epidural morphine. Can Anaesth Soc J 1986; 33: 231–3.
19. Gentil B, Lienhart A, Fleury B. Enhancement of postoperative desaturation in heavy snorers. Anesth Analg 1995; 81: 389–92.
20. Isono S, Sha M, Suzukawa M, et al. Preoperative nocturnal desaturations as a risk factor for late postoperative nocturnal desaturations. Br J Anaesth 1998; 80: 602–5.
21. Rosenberg J, Rasmussen GI, Wojdemann KR, et al. Ventilatory pattern and associated episodic hypoxaemia in the late postoperative period in the general surgical ward. Anaesthesia 1999; 54: 323–8.
22. Hiremath AS, Hillman DR, James AL, et al. Relationship between difficult tracheal intubation and obstructive sleep apnoea. Br J Anaesth 1998; 80: 606–11.
23. Esclamado RM, Glenn MG, McCulloch TM, Cummings CW. Perioperative complications and risk factors in the surgical treatment of obstructive sleep apnea syndrome. Laryngoscope 1989; 99: 1125–9.
24. Duran J, Esnaola S, Rubio R, Iztueta A. Obstructive sleep apnea-hypopnea and related clinical features in a population-based sample of subjects aged 30 to 70 yr. Am J Respir Crit Care Med 2001; 163: 685–9.
25. Samuels SI, Rabinov W. Difficulty reversing drug-induced coma in a patient with sleep apnea. Anesth Analg 1986; 65: 1222–4.