Urgent Adenotonsillectomy: An Analysis of Risk Factors Associated with Postoperative Respiratory Morbidity

The study design was retrospective and received institutional approval. Informed consent was not required. Subjects were identified from two databases. Although they have the same procedure code, urgent adenotonsillectomies have a different list code compared with elective adenotonsillectomies. This allowed the study group to be identified from the hospital computerized data log. The control group was identified from the sleep laboratory database. Children who had undergone urgent adenotonsillectomy comprised the study group. Children who had undergone a sleep study and elective adenotonsillectomy between January 1, 1999, and March 31, 2001, comprised the control group. The inclusion criteria were a diagnosis of OSAS and extubation in the operating room at the end of the adenotonsillectomy. The diagnostic criteria for OSAS were a clinical context consistent with obstructive sleep apnea plus:

1. an abnormal sleep study (see OSAS Severity) defined, in our institution, by an obstructive apnea and hypopnea index greater than 1 event per hour, or

2. an abnormal oximetry study defined by a trend graph showing at least three clusters of desaturations below 90% (fig. 1), or

3. a carbon dioxide tension (from a capillary sample) in excess of 50 mmHg during rest while awake, or

4. witnessed severe upper airway obstruction and nocturnal desaturations below 90%.

Pertinent history and operative and postoperative information were extracted from the medical record by a pediatric anesthesiologist (K. B. and/or C. H.). The start time of the operative procedure, the times of admission to and discharge from the postoperative unit, the time of opioid administration, and the time at which a respiratory complication occurred were all recorded. The duration of the operation was the combined anesthetic and surgical times. The postoperative period began on admission to the postoperative unit, either the pediatric intensive care unit or recovery room (postanesthesia care unit [PACU]), and ended with discharge from hospital. Postoperative day 1 began at 08:00 h on the day after surgery.

Both known (age, preoperative comorbidities, and nocturnal saturation nadir) 1,5,6,8and potential (anesthetic technique) risk factors that might have contributed to postoperative respiratory morbidity were recorded.

Although the anesthetic technique varied, the rationale influencing the choice of induction technique, the choice of opioid, and the use of atropine at induction, muscle relaxants and reversal of nondepolarizing muscle relaxants could not be ascertained. The decision to administer intraoperative dexamethasone was often made by the operating surgeon on the basis of a difficult tonsillar dissection and uvular edema at the end of the adenotonsillectomy.

It is our routine to extubate children awake in the operating room. Children are transported to the postoperative unit in the lateral decubitus position. It is our routine to administer oxygen by facemask, attached to a Jackson Rees circuit, placed in close proximity to the child's face for the initial postoperative period. Although the child recovers in an oxygen-enriched environment, it is not possible to know the inspired concentration of oxygen during recovery. It is our routine to administer morphine, in repeated doses, as the opioid of choice in the postoperative unit until the child is able to tolerate oral codeine. In addition, we monitor all children admitted with a diagnosis of OSAS with an oximeter on the first postoperative night.

Postoperative events of interest included both respiratory complications and medical interventions. The respiratory complications were classified as desaturation , defined as a recorded oxygen saturation less than 95%, and airway obstruction , identified in the chart record by such words as “stopped breathing” or “apnea.” It was sometimes difficult to distinguish respiratory depression from airway obstruction, but if the medical notes documented that the child was repositioned or the airway was supported, an obstructed breathing pattern was inferred.

The children were divided into two groups based on the need for medical intervention: those who required postoperative medical interventions for respiratory complications (INT group) and those who did not (non-INT group). Complications occurring intraoperatively, i.e. , before the child left the operating room, did not influence the grouping criteria.

Medical interventions were classified by complexity into minor and major interventions. Minor interventions (INT^minor), such as might be provided by a nurse, included oxygen therapy beyond usual period (identified from the chart record by a notation that oxygen was required beyond the initial check for vital signs) and/or repositioning of the child. Major interventions (INT^major), which required an assessment by a physician, included administration of racemic epinephrine, Ventolin, or Lasix; airway instrumentation with an oropharyngeal or nasopharyngeal airway or an endotracheal tube; and ventilation. Given the nature of a retrospective review, it was not possible to validate the appropriateness and efficacy of the medical interventions.

The time at which the medical intervention occurred relative to the time of admission to the postoperative unit was classified as (1) within the first postoperative hour, (2) between 1 and 8 postoperative hours, or (3) beyond 8 h. In addition, children experiencing multiple episodes of desaturation were identified.

Four tests were used to establish a diagnosis of OSAS: (1) polysomnography performed in the sleep laboratory, (2) cardiorespiratory sleep studies performed in the home, (3) overnight oximetry performed in the home or hospital, and/or (4) an awake capillary carbon dioxide tension.

Details of our sleep study recording system have been published elsewhere. 1,9–12Although the cardiorespiratory signals were analyzed for several variables, only the obstructive apnea and hypopnea index and the saturation nadir (Sao²nadir) are reported. The Sao²nadir was defined as the minimum validated hemoglobin oxygen saturation regardless of duration. In the cardiorespiratory study and polysomnography studies, the validity of the Sao²nadir was verified by visual inspection of a computerized data record. 10,13 

The oximetry studies present two types of graphic printouts: a trend graph and an event graph. 10The trend graph displays a 6- to 12-h summary of the oxygen saturation and pulse rate (fig. 1). Apparent but artifactual desaturations that can be produced by low pulse amplitude or movement can be assessed with the event graphs, allowing a validation of the Sao²nadir. 13 

Of necessity, we stratified the OSAS disease severity by the Sao²nadir criteria alone because most children in the study group did not have values for the obstructive apnea and hypopnea index. We defined severe OSAS by an Sao²nadir less than 80% because this threshold has been shown to be associated with an increased risk of postadenotonsillectomy respiratory complications in a pediatric sleep laboratory referral population. 1,6 

The variables were analyzed using SAS for Windows (version 8.02; SAS Institute, Cary, NC). Categorical data were summarized using frequencies, and continuous variables were reported as mean ± SD or ranges. For the purpose of the statistical analysis, adenotonsillectomy, tonsillectomy, and adenoidectomy were considered a single surgical entity.

The main outcome was the requirement for postoperative medical intervention. Potential risk factors were first evaluated with univariate analysis followed by multivariate logistic regression. Associations were tested by chi-square statistics and the corresponding Mantel-Haenszel odds ratio and 95% confidence intervals. 14The threshold for statistical significance was P < 0.05. The relation between Sao²nadir and outcome was assessed descriptively with the Lowess smoothing curve, a robust locally weighted regression-smoothing curve, to verify our definition of severe OSAS 15(fig. 2).

Differences between the control and study group were assessed with chi-square statistics for categorical variables and with a t  test for continuous variables. The threshold for statistical significance was P < 0.05.

The annual adenotonsillectomy caseloads for the years 1999 and 2000 were 843 and 761, respectively.

Of 64 children who underwent urgent adenotonsillectomy, 54 met the inclusion criteria. Nine children were excluded because the indication for adenotonsillectomy was not OSAS (peritonsillar abscess = 4 diagnostic biopsy = 1, chronic sinusitis = 1, unstated = 3). One child remained intubated after surgery and therefore was excluded from further analysis.

There were 43 boys and 11 girls. The age and weight (mean ± SD) were 4.0 ± 2.4 yr and 22.4 ± 18.7 kg, respectively. There were 50 inpatients and 4 outpatients. The median time interval between the sleep study and surgery was 2 days (range, 1–61 days).

Twenty-five children had associated medical conditions; 84% of these children required medical interventions (table 1).

Fifty-one children met the laboratory diagnostic criteria for OSAS (see Methods). In four children, a diagnosis of OSAS was made by clinical criteria alone. Fourteen children underwent a sleep study, 12 of which were a home cardiorespiratory sleep study and 2 of which were laboratory polysomnography. Forty-two children had oximetry studies. Thirty-one children had severe OSAS defined by a Sao²nadir less than 80%. Five of nine children had a preoperative capillary partial pressure of oxygen (Pco²) during rest in excess of 50 mmHg (range, 39–71 mmHg).

In the study group, 53 children underwent adenotonsillectomy, and 1 underwent adenoidectomy (patient 5 in table 2had a diagnostic bronchoscopy in addition to adenotonsillectomy). The start time (24-h clock; mean ± SD) in the study group was 13:59 ± 3.0 h (group non-INT = 13:22 ± 2.4 h, group INT = 14:22 ± 3.4 h). The duration of surgery (mean ± SD) was 47.2 ± 14.9 min (group non-INT = 42.3 ± 12.3, group INT = 50.4 ± 15.7; P = 0.05).

Although children in the study group were booked for urgent adenotonsillectomy, all fasted according to our guidelines for elective surgery. No child received a rapid-sequence induction. An inhalational induction (sevoflurane = 21, halothane = 2) was used in 41% of the children. The majority (93%) of the children received propofol during induction of anesthesia. Twenty-seven children received a muscle relaxant (rocuronium = 81%, succinylcholine = 19%). Rocuronium, when used, was given in a dose of 0.5 mg/kg. The tracheae of all children were intubated. All children received maintenance anesthesia with isoflurane. Forty-one children received rectal acetaminophen (32.8 ± 6.8 mg/kg). Intraoperative antiemetics were given to three children (droperidol = 2, dimenhydrinate = 1). The majority (98%) received an intraoperative opioid. The opioid dose (mean ± SD) was 1.04 mg/kg for codeine (n = 1), 0.08 ± 0.04 mg/kg for morphine (n = 12), 1.44 ± 0.62 μg/kg for fentanyl (n = 30), and 0.12 ± 0.04 μg/kg for sufentanil (n = 10). Thirteen children received intraoperative dexamethasone in a dose of 0.26 ± 0.18 mg/kg. Residual neuromuscular blockade in 7 of the 22 children who received rocuronium was reversed with neostigmine.

Eight children were admitted to the pediatric intensive care unit, and 47 were admitted to the PACU. Forty-seven of 54 children were assessed as having pain and were given an opioid (morphine = 44, fentanyl = 1, codeine = 2). The dose (mean ± SD) of morphine was 0.08 ± 0.04 mg/kg. Twenty-six children who received postoperative opioids required medical intervention. The onset of respiratory complications, in 12 children, occurred within 2 h of morphine administration.

Of 75 children who underwent a sleep study and adenotonsillectomy, 44 had a positive sleep study and were admitted to hospital on the night after surgery (adenotonsillectomy = 41, adenoidectomy = 1, tonsillectomy = 2). There were 25 boys and 19 girls. The age and weight (mean ± SD) were 3.5 ± 1.9 yr and 18.9 ± 8.4 kg, respectively. The preoperative Sao²nadir was 86.0 ± 6.5%. Seven children had a preoperative Sao²nadir less than 80%. The obstructive apnea and hypopnea index was 14.2 ± 17.0 events/h. The median time interval between the sleep study and surgery was 60 days (range, 2–365 days).The median time interval between the sleep study and surgery, in children whose Sao²nadir was less than 80%, was 4 days.

An inhalational induction (sevoflurane = 35, halothane = 2, unknown = 12) was used in all children. An induction dose of propofol was used in 93% of the children. A minority (22%) of the children received muscle relaxants (rocuronium = 8, succinylcholine = 2). The tracheae of all children were intubated. All children were given intraoperative opioids. One child received intramuscular codeine (0.98 mg/kg). The opioid dose (mean ± SD) was 0.07 ± 0.03 mg/kg for morphine (n = 10), 1.23 ± 0.54 μg/kg for fentanyl (n = 27), and 0.14 ± 0.05 μg/kg for sufentanil (n = 6). Three children received dexamethasone.

Differences in age, weight, and sex were not significant. A greater proportion of children in the study group had an associated medical condition compared with the control group (45.5 vs.  13.6%; P < 0.01). Asthma was the most frequent associated medical condition in both groups. Although all of the children in the control group met our diagnostic criteria for OSAS, the disease severity was greater in the study group as evidenced by a lower preoperative saturation nadir (76.4 ± 11.1 vs.  86.0 ± 6.5%; P < 0.05) compared with the control group. More children in the study group had a preoperative Sao²less than 80% than in the control group (57.4 vs.  15.9%; P < 0.001).

The average start times in the study and control group were 13:57 ± 3.0 and 11:03 ± 2.3 h, respectively. The surgical procedure was longer in the study group compared with the control group (46.9 ± 14.9 vs.  40.3 ± 7.4 min, respectively; P < 0.05).

Certain aspects of the anesthetic technique differed between the two groups. Fewer children received reversal of rocuronium (P < 0.05) compared with the control group (table 3). The postoperative morphine doses in the study and control groups were 0.08 ± 0.04 and 0.09 ± 0.04 mg/kg, respectively.

Sixty percent (n = 33) of the children in the study group required medical intervention in the postoperative period (table 1). (Two children experienced postextubation laryngospasm in the operating room, one had no other complications in the postoperative period, and the other required a minor medical intervention.) Forty-nine percent (n = 16) of the medical interventions were required within the first postoperative hour. A third (n = 19) of the children experienced multiple episodes of desaturation in the postoperative period.

Five children were diagnosed with pneumonia in the postoperative period. Evidence for a diagnosis of pneumonia was fever, sputum, an abnormal chest radiograph, and initiation of antibiotic therapy. Three children had a preoperative Sao²nadir of 80%, and two had an elevated preoperative capillary Pco²(71 and 51 mmHg). Radiographic evidence of bilateral infiltrates in one child was consistent with pulmonary aspiration, although no documentation of such was noted.

Twenty-two children required a minor medical intervention. All patients in group INT^minorexperienced postoperative oxygen desaturation within 5.8 ± 4.8 h of admission to the postoperative unit. One child desaturated within 1 h of admission to the PACU and returned to the operating room for control of postadenotonsillectomy bleeding. She had no further desaturations.

Eleven children required a major medical intervention (tables 1, 3, and 4). Seven children received racemic epinephrine. Six required tracheal reintubation. All children with preoperative hypercarbia required a major medical intervention. No child in the study group was given naloxone. Details of the postoperative course for each child are given in tables 3 and 4.

Three children required medical intervention (naloxone administration = 2, oropharyngeal airway = 1) during emergence from anesthesia. Based on their postoperative course, two children were assigned to group non-INT, and one was assigned to group INT^major. Sixteen children (36.4%) in the control group required medical intervention in the postoperative period (INT^minor= 13, INT^major= 3). Six percent of the children experienced a major medical intervention (oropharyngeal airway = 1, Ventolin = 1, medical doctor assessment = 1). Twelve children experienced multiple desaturations during sleep on the first postoperative night and required oxygen. One child (2%) was diagnosed with pneumonia in the postoperative period. The postoperative Sao²nadir in the PACU was 82.0 ± 16.4%.

The data were analyzed with univariate analysis for the known (age, associated medical conditions, OSAS severity) 1,5,6and potential (administration of atropine at induction, dexamethasone, reversal of muscle relaxants) risk factors (table 5). Multivariate logistic regression identified an associated medical condition, an Sao²nadir less than 80%, and intraoperative dexamethasone administration as risk factors for postadenotonsillectomy respiratory morbidity. Atropine administration at induction reduced the risk of postoperative respiratory morbidity (table 6). A secondary analysis of data excluding dexamethasone administration, which may be a rogue risk factor (see Discussion), yielded similar results.

We report a 20.3% incidence of major respiratory complications after urgent adenotonsillectomy (study group) compared with an incidence of 6.1% in the control group. Although significant differences in the severity of the OSAS preclude extensive comparison between the study and control groups, the low incidence of postoperative complications in the control group and in a previous publication 1confirms our clinical impression that the complication rate in the study group was excessive. In the study group, six children (11.1%) required postoperative reintubation, and five children (9.3%) developed postoperative pneumonia. Both early and late complications occurred. The onset of respiratory complications was often delayed and, in some children, began more than 8 h after surgery (table 1). Two clinical scenarios characterized the postoperative period. The clinical course in group INT^minorwas one of episodic and repeated desaturation. The clinical course of children in group INT^majorsuggested worsening airway obstruction between admission to the postoperative unit and the need for a major medical intervention (tables 2 and 4).

Several factors may have contributed to the outcome, including patient factors (the presence of associated medical conditions and the severity of the OSAS) and other factors.

The odds ratio for a medical intervention if an associated medical condition was present was 8.15 (95% confidence interval 1.81–36.7). Although this risk factor has been reported elsewhere, 1,5,6,16the majority of children experiencing major respiratory complications reported by McColley et al. , 5Rosen et al. , 6and McGowan et al.  16had cardiorespiratory, neuromuscular, or craniofacial comorbidities. Of note, the majority of children experiencing major postoperative complications in our study had either no comorbidity (n = 4) or asthma (n = 2).

Disease severity in OSAS was an independent risk factor as evidenced by the fact that in the study group, an Sao²nadir less than 80% had an odds ratio of 5.54 (95% confidence interval 1.15–26.72). Therefore, the high incidence of respiratory morbidity in the study group is explained in part by the fact that a higher proportion of the study population (57.4%) had severe OSAS, defined by a preoperative Sao²nadir less than 80%, compared with the control group (16%) and our previous study population (26%). 1Despite that the saturation nadir represents a single data point in the overnight oximetry study, the saturation nadir is a predictor of postoperative morbidity. Furthermore, although we treated the Sao²nadir as a categorical variable with a cutoff threshold of 80%, inspection of figure 2suggests that the Sao²nadir may in fact represent a continuum of risk for postoperative respiratory complications, and this finding merits prospective study.

However, the complexity of the respiratory complications cannot be explained by the severity of the OSAS or coexisting medical conditions because the children in study group INT^majorwere older, with fewer comorbidities and a higher preoperative Sao²nadir than children in study group INT^minor(table 1). Therefore, other factors, including preoperative preparation, intraoperative management, and postoperative care, may have influenced the outcome.

Support for the practice of urgent adenotonsillectomy based on abnormal oximetry was derived from experimental evidence suggesting that intermittent hypoxia adversely affects the developing brain and our disquiet in waiting for surgical intervention. 17However, a pitfall in the use of oximetry alone in the diagnosis of OSAS is that lower respiratory tract infections, 18asthma, 19and pulmonary hypertension, 20which coexist with OSAS, may also promote desaturation during sleep. It was not possible to comment on the adequacy of the preoperative preparation in the study group, but a 2-day time interval between the sleep test and surgery may not have been sufficiently long to allow for optimal preoperative preparation. The impact of aggressive preoperative preparation, including administration of antibiotics, bronchodilators, and oral and/or nasal steroids, 21,22as a risk-reduction strategy to decrease respiratory morbidity after urgent adenotonsillectomy merits prospective study. For example, one child who was diagnosed with preoperative pneumonia and hypercapnia was managed with preoperative antibiotics, dexamethasone, oxygen, and a nasopharyngeal airway for several days before urgent adenotonsillectomy and experienced no postoperative respiratory complications.

To date, no differences in outcome have been attributed to anesthetic management. 5,6,16,23We report that half of the complications occurred in the initial postoperative hours, a time period potentially influenced by anesthetic technique. Two aspects of anesthetic management may have affected outcome: the administration of atropine at induction and the decision not to reverse the intermediate nondepolarizing relaxant, rocuronium. Although administration of intraoperative dexamethasone was identified as a risk factor by statistical criteria (table 5), we suspect a selection bias because the decision to administer dexamethasone was based on clinical criteria (see Methods).

Atropine administration may be beneficial as a risk-reduction strategy in children with severe OSAS undergoing adenotonsillectomy.

The higher usage of neuromuscular blockade and the lower usage of neostigmine to reverse residual neuromuscular blockade in the study group compared with the control group are notable (table 3). This may be of clinical significance because not reversing nondepolarizing muscle relaxants has been reported to increase the risk of perioperative respiratory complications. 24The child with OSAS may be vulnerable to residual neuromuscular blockade because OSAS is associated with an impairment in neuromuscular control of the upper airway, persisting despite adenotonsillectomy. 25,26 

Respiratory complications began within 2 h of postoperative opioid administration in 12 children (table 1). Although no study has shown that opioid administration affects outcome after adenotonsillectomy, all studies to date report a mixed population of OSAS, ranging from mild to severe disease. 1,5,6,23In a developmental experimental model, a history of intermittent recurrent hypoxia has been shown to increase μ-opioid binding in the brainstem, suggesting an up-regulation of these receptors. 27Such up-regulation may produce increased sensitivity to opioid drugs. If a similar phenomenon occurred in central opioid systems of children with severe OSAS, then this might be a mechanism whereby the margin of safety for opioid administration in these children would be reduced.

Other factors may also have affected outcome in the study group. Newland et al.  28reported that the time of day is a risk factor for anesthesia related cardiac arrest. Bell et al.  29speculated that uneven staffing patterns may contribute to the increased hospital mortality on weekends. In this regard, 5 of the 11 children who required a major medical intervention were admitted to the PACU after 13:00 h, coincident with a decrease in the medical and nursing staffing patterns. In addition, the nurse and physician–to–patient ratio is highest in a pediatric intensive care unit setting, and this may have influenced the complexity of the medical interventions because airway obstruction in children admitted to the pediatric intensive care unit was treated with racemic epinephrine, not reintubation (tables 2 and 4).

We report a high incidence of postoperative respiratory morbidity associated with urgent adenotonsillectomy, including an 11.1% incidence of reintubation and a 9.3% incidence of postoperative pneumonia. The rationale for urgent adenotonsillectomy based on documentation of a very low preoperative saturation nadir requires critical review. The temptation to react to the profound nocturnal desaturation with urgent adenotonsillectomy intervention should be balanced against the fact that these children are at increased risk for postoperative complications. To date, the most important predictors of postadenotonsillectomy respiratory morbidity are the presence of an associated medical condition and OSAS disease severity. Therefore, the preoperative evaluation of the child with OSAS should include an assessment of these risk factors. Overnight oximetry, a readily available test, may enable evidence-based decisions in the treatment of children with severe OSAS who require adenotonsillectomy.

The authors thank Melvin D. Schloss, M.D., F.R.C.S.C. (Director, Division of Otolaryngology, Montreal Children's Hospital, Montreal, Quebec, Canada), for his critical review of the manuscript; Xun Zhang, Ph.D. (Consultant, Biostatistics Laboratory, Montreal Children's Hospital Research Institute, Montreal, Quebec, Canada), for assistance with the statistical analysis; and Roula Cacolyris (Administrative Coordinator, Division of Pediatric Anesthesia, McGill University Health Centre/Montreal Children's Hospital, Montreal, Quebec, Canada) for her help in the preparation of the manuscript.