- Email: [email protected]
Predictors of adverse events with intramuscular ketamine sedation in children
PEDIATRICS/ORIGINAL CONTRIBUTION
Predictors of Adverse Events With Intramuscular Ketamine Sedation in Children
From the Departments of Emergency Medicine, Loma Linda University Medical Center & Children’s Hospital, Loma Linda, CA*; University of California–Davis Medical Center, Sacramento, CA‡; Orlando Regional Medical Center, Orlando, FLII; Mammoth Hospital, Mammoth Lakes, CA§; and Martin Luther King/Drew Medical Center, Los Angeles, CA.¶
Steven M. Green, MD* Nathan Kuppermann, MD, MPH‡ Steven G. Rothrock, MDII Christopher B. Hummel, MD§ Matthew Ho, MD¶
Received for publication November 20, 1998. Revisions received May 24, 1999, and June 29, 1999. Accepted for publication July 26, 1999. Reprints not available from the authors. E-mail for correspondence: [email protected]. Copyright © 2000 by the American College of Emergency Physicians. 0196-0644/2000/$12.00 + 0 47/1/101746
Study objective: Ketamine is a safe and effective sedative for emergency department procedures in children. However, the use of ketamine sometimes is associated with airway complications, emesis, and recovery agitation. We wished to identify predictors of these adverse events that clinicians might use to risk-stratify children who are candidates for ketamine sedation. Methods: We analyzed data from 1,021 ED intramuscular ketamine sedations in children 15 years of age or younger at a university medical center and an affiliated county hospital over a 9-year period. Five potential predictor variables (age, gender, American Society of Anesthesiologists’ [ASA] risk classification, quantity of first ketamine dose, and number of ketamine doses administered) were compared between children with and without complications. We used multiple logistic regression analyses to determine the association of these 5 variables with emesis and recovery agitation, and validated these analyses with bootstrap resampling techniques. We compared children with and without airway complications using univariate statistics alone, as there were too few patients with airway complications to support a multivariate analysis. Results: No study variables had significant univariate associations with airway complications (all P values >.40). We found emesis to be associated with increasing age in multivariate analysis (odds ratio [OR] 1.25 per year, bias-corrected 95% confidence interval [CI] 1.17 to 1.34, P<.001). The incidence of emesis was 12.1% in children aged 5 years or older, and 3.5% in those younger than 5 years (∆8.6%, 95% CI 4.9% to 12.1%). Recovery agitation was associated with the presence of an underlying medical condition (ie, ASA class ≥2, OR 3.05, biascorrected 95% CI 1.65 to 7.30, P=.004) and inversely associated with increasing age (OR 0.79 per year, bias-corrected 95% CI 0.69 to 0.89, P<.001). The incidence of recovery agitation was 17.9% in ASA class 1 children and 33.3% in children in ASA class 2 or greater (∆–15.4%, 95% CI 0.0% to –30.7%). The incidence of recovery agitation was 12.1% in children aged 5
JANUARY 2000
35:1
ANNALS OF EMERGENCY MEDICINE
3 5
KETAMINE ADVERSE EVENTS IN CHILDREN Green et al
years or older, and 22.5% in those younger than 5 years (∆–10.4%, 95% CI –3.0% to –17.7%). Bootstrap resampling techniques validated the importance of the significant variables identified in the regression analyses. Conclusion: No study variable was predictive of ketamineassociated airway complications. Emesis that occurred after ketamine administration was modestly associated with increasing age. Recovery agitation was modestly associated with decreasing age and the presence of an underlying medical condition. The discriminatory power of these variables was low enough as to be unlikely to alter clinical decisions regarding patient selection for ketamine administration. No evidence of a significant ketamine dose relationship was noted for airway complications, emesis, or recovery agitation. [Green SM, Kuppermann N, Rothrock SG, Hummel CB, Ho M. Predictors of adverse events with intramuscular ketamine sedation in children. Ann Emerg Med. January 2000;35:35-42.] INTRODUCTION
The dissociative agent ketamine is a safe and effective sedative used to facilitate the performance of painful or emotionally disturbing emergency department procedures in children.1-10 We recently reported a prospective case series of 1,022 consecutive intramuscular ketamine administrations.1 Although there were no complications with sequelae in this sample, 3 types of adverse events were observed: transient airway complications (1.4%), emesis (6.7%), and generally mild recovery agitation (19.2%). We hypothesized that demographic or other clinical factors might be predictive of these 3 types of adverse events, and that emergency physicians could then use such factors to risk-stratify children when ketamine sedation is considered. Accordingly, we analyzed our 9-year experience with ketamine to identify predictors of airway complications, emesis, and recovery agitation. M AT E R I A L S A N D M E T H O D S
We previously published our ketamine protocol and the sedation characteristics of the study sample, which included 1,022 consecutive children 15 years or younger given intramuscular ketamine. This prospective study was performed in the EDs of a university medical center and an affiliated county hospital over a 9-year period.1 The study was approved by the institutional review board at each institution.
3 6
Ketamine was administered according to a specific protocol that delineated inclusion criteria, exclusion criteria, and monitoring.1 As part of routine care, nurses completed standardized ED sedation forms for the medical record including periodic vital signs, oximetry readings, and adverse effects. Treating physicians completed study data forms concurrent with patient care in 42% of the 1,022 sedations. We then reviewed the medical records (including both physician and nursing documentation) and study data forms, if completed, of all sedations using a standardized data abstraction form. We recorded the presence or absence of any adverse events, including airway complications (eg, partial airway obstruction, respiratory depression, apnea, laryngospasm), emesis, and recovery agitation (defined as any combination of agitation, crying, hallucinations, or nightmares). If any discrepancy was noted when comparing the medical record with the physician data form, we recorded the more serious account of the complication. We assumed complications to be absent if not recorded. We retrospectively judged the American Society of Anesthesiologists’ (ASA) risk classification11 based on the presenting condition and available medical history of the child. For the current analysis, we excluded one child who lacked chart documentation of weight, preventing calculation of the ketamine dose in milligrams per kilogram. No adverse events were noted in this child. A total of 1,021 ketamine sedations remained for analysis. The incidences of airway complications and emesis were similar between those children with and those without treating physician data forms. Therefore, as in the original report,1 these 2 groups were combined for further analysis (n=1,021). The incidence of recovery agitation, however, was dissimilar between these groups. Thus, in accordance with the methodology of the original report, further analysis of recovery agitation was restricted to the group with treating physician data forms (n=430) in which we judged that this adverse event was more reliably assessed. Details of the specific adverse effects have been previously reported.1 To summarize, there were 14 transient airway complications (1.4%): partial airway obstruction (n=7), laryngospasm (n=4), apnea (n=2), and respiratory depression (n=1). All were readily identified and none resulted in an adverse outcome. Emesis occurred in 68 (6.7%) children and there was no evidence of aspiration in any patient. Treating physicians noted recovery agitation in 83 (19.3%) of the 430 children with data forms. In 76 children, this agitation was graded as “mild,” and in 7
ANNALS OF EMERGENCY MEDICINE
35:1 JANUARY 2000
KETAMINE ADVERSE EVENTS IN CHILDREN Green et al
children it was graded “moderate to severe.” None of these 83 children received any specific therapy for agitation. We restricted the number of predictor variables to approximately 10% of the number of outcome observations for our multiple logistic regression analyses, in accordance with standard recommendations.12-15 Accordingly, for emesis and recovery agitation, 5 predictor variables were evaluated in the analysis. These variables were selected on the basis of their biologic plausibility of association and included age, gender, ASA risk classification, quantity of first ketamine dose (in milligrams per kilogram), and total number of doses administered. Other variables rejected from consideration as being less biologically plausible included sedation indication (eg, laceration, fracture reduction) and hospital (university versus county). Concurrent atropine administration could not be studied because essentially all subjects (98.6%) received this antisialogogue prophylactically. Similarly, concurrent benzodiazepine administration could not be studied because too few subjects in the data form group (1.2%) received this sedation adjunct. Finally, weight was not selected because it was highly correlated with age (Pearson r=0.840, P<.0001). There were few children with high ASA risk classification classes (892 class 1, 84 class 2, 36 class 3, 8 class 4, 1 class 5). Because dichotomization of this variable at the ASA level of 2 or less versus ASA level of 3 or more would not permit a sufficiently powerful analysis, we instead divided subjects as either ASA class 1 or ASA class 2 or more. Similarly, there were few children who received more than 2 ketamine doses (791 received 1 dose, 198 received 2 doses, 24 received 3 doses, and 8 received 4 doses). Accordingly, this variable was dichotomized as single or multiple doses and analyzed in this form. We examined the frequency distributions for the continuous variables (ie, age, ketamine dose); because their distributions were not bimodal, they were introduced into the multivariate analysis in their continuous forms. Because the small number of airway complications could only support a single predictor variable, we were unable to proceed with a multivariate analysis for this complication. Instead, we performed univariate analyses using the 5 predictor variables selected for emesis and recovery agitation. Categorical data were compared using χ2 or Fisher’s exact test. For continuous variables, we first performed Bartlett’s test for equal variances. If the variances between groups for a given variable were equal, we performed Student’s t test for equal variances. If the variances were not equal, we performed Student’s t test modified for unequal variances.
JANUARY 2000
35:1
ANNALS OF EMERGENCY MEDICINE
We performed separate multiple logistic regression analyses using the selected variables for emesis and recovery agitation. Those variables retaining an association with the adverse event of interest with P less than or equal to .05 were considered to have an independent association with that adverse event, and those with .05 less than P less than or equal to .10 to trend toward an association. We calculated the likelihood ratios and area under the receiver operating characteristic curve for each model, and performed goodness of fit analyses using the HosmerLemeshow test.16 To validate each model, we performed 2 separate bootstrap resampling procedures for each model with 200 iterations each. Bootstrapping refers to a process in which random samples of a database are drawn with replacement.17,18 This method can be used to obtain conservative estimates of confidence intervals (CIs) and standard errors, as well as to assess the stability of a model. In the first bootstrap procedure, we obtained 95% bias-corrected CIs of the predictor variables in each regression analysis. In the second bootstrap procedure, we reported the multivariate analysis on 200 random bootstrap samples of our data and identified how many times each predictor variable was identified as independently associated with the given outcome (P<.05). We performed regression diagnostics to identify goodness of fit for each model and overly influential covariate patterns.16 We inspected graphs of the change in deviance residual resulting from the deletion of subjects with a particular covariate pattern versus the predicted probabilities. We also inspected plots of the decrease in Hosmer-Lemeshow goodness-of-fit statistic caused by deleting patients with a given covariate pattern versus predicted probabilities, weighted by a measure of the coefficient vector change caused by deleting patients with a particular covariate pattern.19 We performed all analyses using Stata 5.0 statistical software (Stata Corporation, College Station, TX). R E S U LT S
No predictor variable had significant univariate associations with airway complications (Table 1). Age was a significant independent predictor of emesis (Table 2). A histogram of the incidence of emesis by age is shown in Figure 1. The incidence of emesis was 12.1% in children 5 years and older and 3.5% in those younger than 5 years (∆8.6%, 95% CI 4.9% to 12.1%). There was no trend or significant association between quantity of first ketamine dose and emesis, although the
3 7
KETAMINE ADVERSE EVENTS IN CHILDREN Green et al
bias-corrected 95% CI of the odds ratio (OR) did not include unity. Bootstrapping identified this variable as significantly associated with emesis less than one third of the time. The mean first ketamine dose was 4.03±0.37 mg/kg in children with emesis and 3.95±0.70 mg/kg in those without emesis (∆0.08, 95% CI –0.02 to 0.17). Administration of multiple doses trended toward predicting emesis, although the bias-corrected 95% CI for the OR did include unity. Bootstrapping identified this variable as significantly associated with emesis less than half of the time. The incidence of emesis was 6.2% in children receiving single ketamine doses and 8.3% in those receiving multiple doses (∆–2.1%, 95% CI –6.0% to 1.9%). Significant independent predictors of recovery agitation in the multivariate analysis were age and ASA class of 2 or more (Table 3). A histogram of the incidence of recovery agitation by age is shown in Figure 2. The incidence of recovery agitation was 12.1% in children aged 5 years and older and 22.5% in those younger than 5 years (∆–10.4%, 95% CI –3.0% to –17.7%). Seven of the 430 children were described as having “moderate to
severe” recovery agitation, 1 and their median age was 2.6 years (range 1.8 to 6.5 years). The incidence of recovery agitation was 17.9% in ASA class 1 children and 33.3% in children in ASA class 2 or more (∆–15.4%, 95% CI 0.0% to –30.7%). For the multivariate models, visual inspection of plots of the change in deviance residuals and Hosmer-Lemeshow goodness-of-fit statistic caused by deletion of different covariate patterns did not identify poor model fit. Likewise, inspection of plots of the Hosmer-Lemeshow goodnessof-fit statistic weighted by the change in coefficient vector did not identify overly influential covariate patterns. DISCUSSION
In this study, we report the largest analysis of predictors of ketamine adverse events ever reported in any setting. Importantly, we found that no variable could predict the adverse event of greatest concern—airway complications. We found certain parameters to be associated with emesis and recovery agitation, although the discriminatory power of these variables was relatively modest.
Table 1.
Univariate analysis of airway complications (n=1,021).
Variable Mean age (y±SD) Gender (male, %) ASA risk classification ≥2 (%) Mean quantity of first dose (mg/kg±SD) Multiple doses administered (%)
Airway Complication (n=14)
No Airway Complication (n=1,007)
Difference in Mean or Percentage (95% CI)
4.56±2.96 50.0 21.4 3.87±0.85 21.4
4.54±2.80 59.7 12.5 3.96±0.68 22.5
0.02 (–1.46 to 1.50) –9.7 (–36.0 to 16.7) 8.9 (–12.7 to 30.1) –0.09 (–0.45 to 0.27) –1.1 (–22.8 to 20.5)
.982 .464 .405 .629 1.000
P Value
Frequency Chosen as Independent Predictor Using Bootstrapping (%)
P Value
Table 2.
Multiple logistic regression model of emesis (n=1,021).*
Variable Age (y) Gender (male) ASA risk classification ≥2 Quantity of first dose (mg/kg) Multiple doses administered
Odds Ratio
Bias-Corrected 95% CI
1.25 1.19 0.82 1.30 1.66
1.17–1.34 0.72–2.19 0.23–1.62 1.02–1.67 0.91–2.83
<.001 .528 .610 .121 .084
100 8.5 8.5 28.5 42
*
The area under the model receiver operator curve is 0.715. The model demonstrated satisfactory goodness-of-fit, with the Hosmer-Lemeshow P=.905. The likelihood ratio test for this model versus the constant-only model was χ2 33.28, P<.001.
3 8
ANNALS OF EMERGENCY MEDICINE
35:1 JANUARY 2000
KETAMINE ADVERSE EVENTS IN CHILDREN Green et al
We found no variable to be predictive of airway complications, and observed no trends to suggest any effect that might be detected with a larger sample size. The rarity of airway complications in the sample (1.4%) limited the power of these comparisons and precluded a multivariate analysis. There are no other studies evaluating predictors of ketamine-associated airway complications with which to contrast our data. The existing ketamine literature consists primarily of case series in various clinical settings wherein occasional airway complications are individually described. Respiratory depression and apnea have been reported when ketamine is administered to infants younger Figure 1.
Histogram of the incidence of emesis by age.
than 3 months (especially when ill), or when central nervous system injuries, masses, or abnormalities are present.4,20-22 Neonates and small infants have greater difficulty maintaining a patent airway with ketamine just as they do with other sedatives. Ketamine elevates intracranial pressure and accordingly can exacerbate concurrent central nervous system conditions.4,20-22 These situations are thus well accepted as contraindications to ketamine, and there would appear to be few if any indications for emergency physicians to administer ketamine in such children. Our protocol excluded children with these conditions from this study. Procedures involving the posterior pharynx are generally considered to represent a relative contraindication to the use of ketamine. The dissociative state preserves and slightly exaggerates protective airway reflexes such as coughing and swallowing, and laryngospasm has been reported in association with posterior pharyngeal stimulation.4,20-22 As previously reported, 1 of our 4 cases of laryngospasm began during oropharyngeal suctioning.1 Despite this being a relative contraindication in our protocol, treating physicians elected to administer ketamine to 3 children in our data set for procedures involving the posterior pharynx (removal of posterior pharyngeal foreign bodies in all 3 cases), and these procedures were accomplished without adverse events.1 Although likely a risk factor, available evidence suggests that pharyngeal procedures should be a relative exclusion criterion rather than an absolute one. Respiratory depression and apnea are well recognized when ketamine is administered by rapid intravenous infusion, and it is generally accepted that intravenous doses should be given over at least 1 minute to avoid such responses.4,20-22
Table 3.
Multiple logistic regression model of recovery agitation (n=430).*
Variable Age (y) Gender (male) ASA risk classification ≥2 Quantity of first dose (mg/kg) Multiple doses administered
Odds Ratio
Bias-Corrected 95% CI
P Value
0.79 0.87 3.05 0.86 1.34
0.69–0.89 0.47–1.46 1.65–7.30 0.49–1.73 0.78–2.41
<.001 .575 .004 .596 .336
Frequency Chosen as Independent Predictor Using Bootstrapping (%) 99.5 8 79.5 14 18
*
The area under the model receiver operator curve is 0.673. The model demonstrated satisfactory goodness-of-fit, with the Hosmer-Lemeshow P=.363. The likelihood ratio test for this model versus the constant-only model was χ2=26.31, P<.001.
JANUARY 2000
35:1
ANNALS OF EMERGENCY MEDICINE
3 9
KETAMINE ADVERSE EVENTS IN CHILDREN Green et al
Beyond the existing exclusion criteria already accounted for in patient selection, airway complications with ketamine are unusual and appear unpredictable. Accordingly, clinicians must be prepared to deal with them in any given sedation situation. We believe that the routine use of continuous pulse oximetry, the presence of a nurse dedicated to patient monitoring, and the immediate availability of a physician with airway expertise permit rapid detection and treatment of such rare events. It is unlikely that prospective randomized studies comparing the incidence of airway complications between ketamine sedation regimens (eg, dose, route, patient groups) will ever be performed. We calculate that 7,216 total subjects would be required for such a study to detect a 50% relative difference in airway complications from our 1.4% baseline incidence (α=.05, β=.2). We found that increasing age was predictive of emesis. There are no other studies evaluating predictors of ketamine-associated emesis with which to contrast our data. Although age was the strongest predictor of emesis identified in our analysis, even in the highest-risk age range (10 to 15 years) the incidence was only 17%. As previously reported, 1 there were no adverse outcomes attributable to ketamine-associated emesis in this sample, although vomiting was believed to have provoked one episode of transient laryngospasm. In 30 years of extensive use worldwide, there are no documented
Figure 2.
Histogram of the incidence of recovery agitation by age.
4 0
cases in the medical literature of clinically significant ketamine-associated aspiration syndrome in a patient lacking established contraindications to this drug. 4 We believe it unlikely that clinicians will alter their decisions regarding ketamine use based on the modestly higher rates of emesis noted in older children. Prophylactic antiemetic therapy could be considered in older children, although the efficacy of such an intervention is not known. Although the quantity of the first ketamine dose and administration of multiple doses were not significantly associated with emesis, we found other weak evidence suggesting a relationship in both. For first dose, the OR bias-corrected 95% CI by a small margin did not overlap unity, and for the number of doses there was a statistical trend toward an association (P=.084). Despite this, the observed difference in first dose between subjects with and without emesis was only 0.08 mg/kg, and the observed difference in rates of emesis between these groups was only 2.1%. We believe that both of these differences are clinically unimportant. Because bootstrapping selected these 2 variables as independent predictors of emesis less than half of the time, we suspect that these findings are either the result of chance alone or represent associations below the threshold of clinical significance. A limitation of this analysis is that we did not collect data on time of last meal, and it is possible that shorter fasting periods may be predictive of emesis. Prospective studies are needed to evaluate this potential relationship. Of note, a pooled comparison of data from multiple ketamine studies did not demonstrate an association between fasting state and emesis. 4 We found that younger age and the presence of an underlying medical condition (ie, ASA class ≥2) were predictors of recovery agitation. Several prior studies, primarily in adult populations, have examined risk factors for recovery agitation or hallucinations. These studies have reported that female gender, age older than 10 years, rapid intravenous administration, excessive noise or stimulation during recovery, prior personality disorders, and patients who normally dream frequently represent situations of greater risk. 4,20-22 Our data dispute a female predisposition to recovery agitation in the pediatric population and provide strong evidence against an increased risk of recovery reactions in the 10- to 15-year-old age group. We did not systematically evaluate the other risk factors previously described in the literature, and thus our data do not provide additional information in those areas.
ANNALS OF EMERGENCY MEDICINE
35:1 JANUARY 2000
KETAMINE ADVERSE EVENTS IN CHILDREN Green et al
Our inverse association of recovery agitation to age is surprising. Existing literature describes recovery reactions as rare in children and relatively common (up to 30%) in adults,4,20-22 so an increasing relationship with age would have been expected. It is possible that the preprocedure level of agitation may explain this discrepancy. Infants and toddlers are typically anxious before ED sedation and it would not be surprising if some degree of agitation might persist through recovery. Older children, however, are more likely to be calm before sedation after reassurance from their parents and ED staff. Our study is limited in that we did not collect data on the degree of agitation before sedation. This should be the subject of future study. The basis for our observed association of recovery agitation with an underlying medical condition is unclear. Possibly, children with many previous physician visits associate medical contact with physical discomfort (eg, phlebotomy, injections) and thus may exhibit more preprocedure and postprocedure fear and agitation than corresponding healthy children. A limitation of this data set is that only 5 of the 430 children in this portion of the analysis received concurrent benzodiazepines, and thus these agents were not sufficiently represented to permit us to assess their relationship to recovery agitation. Concurrent benzodiazepine use is widely believed to diminish ketamineassociated recovery reactions, 20-22 although the value of these agents has never been established in children. A prospective study specifically focused on this question is needed. This report is also limited in that we did not collect data on agitation or behavior changes after ED discharge. A recent randomized study comparing a sedation regimen of ketamine and midazolam with fentanyl and midazolam found similar incidences of adverse effects between regimens after ED discharge, except that vomiting was more common in the ketamine group. 10 How clinicians should translate these findings into clinical practice is not apparent. Although agitation is not infrequent after ketamine administration (19.2% overall), when it occurs it is rated by treating physicians as “mild” 92% of the time and rarely leads to specific treatment. 1 Dose-related adverse effect relationships are well documented with benzodiazepines and narcotics. 5 The assumption of a similar ketamine dose response has led some physicians to advocate relatively low doses of ketamine in the hope of minimizing adverse events. 23,24 Our data do not demonstrate evidence of a significant
JANUARY 2000
35:1
ANNALS OF EMERGENCY MEDICINE
ketamine dose relationship with the 3 adverse effects studied. Doses high enough to consistently generate the dissociative state (4 mg/kg IM) appear to be no less safe than lower doses. Dose-related adverse effects with ketamine are not described in the literature. No prior studies have compared airway complications or emesis by dose. Two retrospective studies failed to note significant differences in adult recovery reactions by dose. 24,25 A comparison of data pooled from multiple studies did not demonstrate an association between dose and adverse effects.4 In fact, other authors have anecdotally reported their perception that recovery agitation occurs in an inverse relationship with dose. 26-28 This study is subject to the usual limitations of studies with chart review components, including dependence on the quality of medical record documentation. We believe this factor had minimal effect on the current study, as the information we abstracted from records was largely objective and not likely prone to misinterpretation or bias. We based the single subjective item (recovery agitation) solely on the concurrently recorded impression of treating physicians. The incidence of recovery agitation was dissimilar between the groups with and without treating physician data forms. This disparity suggests that the majority of such reactions are not judged sufficiently substantial to warrant comment in the medical record. Airway complications and emesis, however, were recorded with similar frequency in sedations with and without data forms, and we believe that our data accurately depict such events. It is possible that treating physicians and nurses may be prone to underreport serious adverse events. To avoid identifying any such events in this study would have required their omission from the study data form as well as from all physician and nursing notes in the medical record. Although we believe it unlikely, we cannot exclude the possibility that some adverse events may have been missed. We believe that the most important implications from our data are that children at risk for rare airway complications cannot be predicted beyond existing exclusion criteria, and that there is no evidence of a ketamine dose relation to such events. Although the modest association of age with emesis, and age and underlying medical condition with recovery agitation are of interest, these factors exhibit low enough discriminatory value that they are unlikely to alter clinical decisions regarding patient selection for ketamine administration.
4 1
KETAMINE ADVERSE EVENTS IN CHILDREN Green et al
We thank Sean P Bush, MD, for his review of the manuscript and many helpful suggestions.
REFERENCES
27. Phillips LA, Seruvatu SG, Rika PN. Anaesthesia for the surgeon-anaesthetist in difficult situations. Anaesthesia. 1970;25:36-45. 28. Bennett JA, Bullimore JA. The use of ketamine hydrochloride anaesthesia for radiotherapy in young children. Br J Anaesth. 1973;45:197-201.
1. Green SM, Rothrock SG, Lynch EL, et al. Intramuscular ketamine for pediatric sedation in the emergency department: safety profile with 1,022 cases. Ann Emerg Med. 1998;31:688-697. 2. Green SM, Rothrock SG, Harris T, et al. Intravenous ketamine for pediatric sedation in the emergency department: safety profile with 156 cases. Acad Emerg Med. 1998;5:971-976. 3. Green SM, Nakamura R, Johnson NE. Ketamine sedation for pediatric procedures: part 1, a prospective series. Ann Emerg Med. 1990;19:1024-1032. 4. Green SM, Johnson NE. Ketamine sedation for pediatric procedures: part 2, review and implications. Ann Emerg Med. 1990;19:1033-1046. 5. Sacchetti A, Schafermeyer R, Gerardi M, et al. Pediatric analgesia and sedation. Ann Emerg Med. 1994;23:237-250. 6. Petrack EM, Marx CM, Wright MS. Intramuscular ketamine is superior to meperidine, promethazine, and chlorpromazine for pediatric emergency department sedation. Arch Pediatr Adolesc Med. 1996;150:676-681. 7. Pruitt JW, Goldwasser MS, Sabol SR, et al. Intramuscular ketamine, midazolam, and glycopyrrolate for pediatric sedation in the emergency department. J Oral Maxillofac Surg. 1995;53:13-17. 8. Dachs RJ, Innes GM. Intravenous ketamine sedation of pediatric patients in the emergency department. Ann Emerg Med. 1997;29:146-150. 9. Qureshi F, Mellis PT, McFadden MA. Efficacy of oral ketamine for providing sedation and analgesia to children requiring laceration repair. Pediatr Emerg Care. 1995;11:93-97. 10. Kennedy RM, Porter FL, Miller JP, et al. Comparison of fentanyl/midazolam with ketamine/midazolam for pediatric orthopedic emergencies. Pediatrics. 1998;102:956-963. 11. Dripps RD, Lamong A, Eckenhoff JE. The role of anesthesia in surgical mortality. JAMA.1961;178:261-266. 12. Wears RL, Lewis RJ. Statistical models and Occam’s razor [editorial]. Acad Emerg Med. 1999;6:93-94. 13. Harrell FE, Lee KL, Matchar DB, et al. Regression models for prognostic prediction: advantages, problems, and suggested solutions. Cancer Treat Rep. 1985;69:1071-1077. 14. Harrell FE, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996;15:361-387. 15. Concato J, Feinstein AR, Holford TR. The risk of determining risk with multivariable models. Ann Intern Med. 1993;118:201-210. 16. Hosmer DW, Lemeshow S. Applied Logistic Regression. New York, NY: John Wiley & Sons; 1989:135-145. 17. Efron B, Gong G. A leisurely look at the bootstrap, the jackknife and cross-validation. Am Stat. 1983;37:36-48. 18. Efron B, Tibshirani R. Statistical data analysis in the computer age. Science. 1991;253:390395. 19. Pregibon D. Logistic regression diagnostics. Ann Statistics. 1981;9:705-724. 20. White PF, Way WL, Trevor AJ. Ketamine—its pharmacology and therapeutic uses. Anesthesiology. 1982;56:119-136. 21. Corssen G, Reves JG, Stanley TH. Intravenous Anesthesia and Analgesia. Philadelphia, PA: Lea & Febiger; 1988:99-174. 22. Reich DL, Silvay G.Ketamine: an update on the first twenty-five years of clinical experience. Can J Anaesth. 1989;36:186-197. 23. Epstein FB. Ketamine dissociative sedation in pediatric emergency medical practice. Am J Emerg Med. 1993;11:180-182. 24. Liang HS, Liang HG. Minimizing emergence phenomena: subdissociative dosage of ketamine in balanced surgical anesthesia. Anesth Analg. 1975;54:312-316. 25. Knox JW, Bovill JG, Clarke RS, et al. Clinical studies of induction agents—ketamine. Br J Anaesth. 1970;42:875-885. 26. Faithfull NS, Haider R. Ketamine for cardiac catheterisation. Anaesthesia. 1971;26:318-323.
4 2
ANNALS OF EMERGENCY MEDICINE
35:1 JANUARY 2000
Predictors of Adverse Events With Intramuscular Ketamine Sedation in Children
From the Departments of Emergency Medicine, Loma Linda University Medical Center & Children’s Hospital, Loma Linda, CA*; University of California–Davis Medical Center, Sacramento, CA‡; Orlando Regional Medical Center, Orlando, FLII; Mammoth Hospital, Mammoth Lakes, CA§; and Martin Luther King/Drew Medical Center, Los Angeles, CA.¶
Steven M. Green, MD* Nathan Kuppermann, MD, MPH‡ Steven G. Rothrock, MDII Christopher B. Hummel, MD§ Matthew Ho, MD¶
Received for publication November 20, 1998. Revisions received May 24, 1999, and June 29, 1999. Accepted for publication July 26, 1999. Reprints not available from the authors. E-mail for correspondence: [email protected]. Copyright © 2000 by the American College of Emergency Physicians. 0196-0644/2000/$12.00 + 0 47/1/101746
Study objective: Ketamine is a safe and effective sedative for emergency department procedures in children. However, the use of ketamine sometimes is associated with airway complications, emesis, and recovery agitation. We wished to identify predictors of these adverse events that clinicians might use to risk-stratify children who are candidates for ketamine sedation. Methods: We analyzed data from 1,021 ED intramuscular ketamine sedations in children 15 years of age or younger at a university medical center and an affiliated county hospital over a 9-year period. Five potential predictor variables (age, gender, American Society of Anesthesiologists’ [ASA] risk classification, quantity of first ketamine dose, and number of ketamine doses administered) were compared between children with and without complications. We used multiple logistic regression analyses to determine the association of these 5 variables with emesis and recovery agitation, and validated these analyses with bootstrap resampling techniques. We compared children with and without airway complications using univariate statistics alone, as there were too few patients with airway complications to support a multivariate analysis. Results: No study variables had significant univariate associations with airway complications (all P values >.40). We found emesis to be associated with increasing age in multivariate analysis (odds ratio [OR] 1.25 per year, bias-corrected 95% confidence interval [CI] 1.17 to 1.34, P<.001). The incidence of emesis was 12.1% in children aged 5 years or older, and 3.5% in those younger than 5 years (∆8.6%, 95% CI 4.9% to 12.1%). Recovery agitation was associated with the presence of an underlying medical condition (ie, ASA class ≥2, OR 3.05, biascorrected 95% CI 1.65 to 7.30, P=.004) and inversely associated with increasing age (OR 0.79 per year, bias-corrected 95% CI 0.69 to 0.89, P<.001). The incidence of recovery agitation was 17.9% in ASA class 1 children and 33.3% in children in ASA class 2 or greater (∆–15.4%, 95% CI 0.0% to –30.7%). The incidence of recovery agitation was 12.1% in children aged 5
JANUARY 2000
35:1
ANNALS OF EMERGENCY MEDICINE
3 5
KETAMINE ADVERSE EVENTS IN CHILDREN Green et al
years or older, and 22.5% in those younger than 5 years (∆–10.4%, 95% CI –3.0% to –17.7%). Bootstrap resampling techniques validated the importance of the significant variables identified in the regression analyses. Conclusion: No study variable was predictive of ketamineassociated airway complications. Emesis that occurred after ketamine administration was modestly associated with increasing age. Recovery agitation was modestly associated with decreasing age and the presence of an underlying medical condition. The discriminatory power of these variables was low enough as to be unlikely to alter clinical decisions regarding patient selection for ketamine administration. No evidence of a significant ketamine dose relationship was noted for airway complications, emesis, or recovery agitation. [Green SM, Kuppermann N, Rothrock SG, Hummel CB, Ho M. Predictors of adverse events with intramuscular ketamine sedation in children. Ann Emerg Med. January 2000;35:35-42.] INTRODUCTION
The dissociative agent ketamine is a safe and effective sedative used to facilitate the performance of painful or emotionally disturbing emergency department procedures in children.1-10 We recently reported a prospective case series of 1,022 consecutive intramuscular ketamine administrations.1 Although there were no complications with sequelae in this sample, 3 types of adverse events were observed: transient airway complications (1.4%), emesis (6.7%), and generally mild recovery agitation (19.2%). We hypothesized that demographic or other clinical factors might be predictive of these 3 types of adverse events, and that emergency physicians could then use such factors to risk-stratify children when ketamine sedation is considered. Accordingly, we analyzed our 9-year experience with ketamine to identify predictors of airway complications, emesis, and recovery agitation. M AT E R I A L S A N D M E T H O D S
We previously published our ketamine protocol and the sedation characteristics of the study sample, which included 1,022 consecutive children 15 years or younger given intramuscular ketamine. This prospective study was performed in the EDs of a university medical center and an affiliated county hospital over a 9-year period.1 The study was approved by the institutional review board at each institution.
3 6
Ketamine was administered according to a specific protocol that delineated inclusion criteria, exclusion criteria, and monitoring.1 As part of routine care, nurses completed standardized ED sedation forms for the medical record including periodic vital signs, oximetry readings, and adverse effects. Treating physicians completed study data forms concurrent with patient care in 42% of the 1,022 sedations. We then reviewed the medical records (including both physician and nursing documentation) and study data forms, if completed, of all sedations using a standardized data abstraction form. We recorded the presence or absence of any adverse events, including airway complications (eg, partial airway obstruction, respiratory depression, apnea, laryngospasm), emesis, and recovery agitation (defined as any combination of agitation, crying, hallucinations, or nightmares). If any discrepancy was noted when comparing the medical record with the physician data form, we recorded the more serious account of the complication. We assumed complications to be absent if not recorded. We retrospectively judged the American Society of Anesthesiologists’ (ASA) risk classification11 based on the presenting condition and available medical history of the child. For the current analysis, we excluded one child who lacked chart documentation of weight, preventing calculation of the ketamine dose in milligrams per kilogram. No adverse events were noted in this child. A total of 1,021 ketamine sedations remained for analysis. The incidences of airway complications and emesis were similar between those children with and those without treating physician data forms. Therefore, as in the original report,1 these 2 groups were combined for further analysis (n=1,021). The incidence of recovery agitation, however, was dissimilar between these groups. Thus, in accordance with the methodology of the original report, further analysis of recovery agitation was restricted to the group with treating physician data forms (n=430) in which we judged that this adverse event was more reliably assessed. Details of the specific adverse effects have been previously reported.1 To summarize, there were 14 transient airway complications (1.4%): partial airway obstruction (n=7), laryngospasm (n=4), apnea (n=2), and respiratory depression (n=1). All were readily identified and none resulted in an adverse outcome. Emesis occurred in 68 (6.7%) children and there was no evidence of aspiration in any patient. Treating physicians noted recovery agitation in 83 (19.3%) of the 430 children with data forms. In 76 children, this agitation was graded as “mild,” and in 7
ANNALS OF EMERGENCY MEDICINE
35:1 JANUARY 2000
KETAMINE ADVERSE EVENTS IN CHILDREN Green et al
children it was graded “moderate to severe.” None of these 83 children received any specific therapy for agitation. We restricted the number of predictor variables to approximately 10% of the number of outcome observations for our multiple logistic regression analyses, in accordance with standard recommendations.12-15 Accordingly, for emesis and recovery agitation, 5 predictor variables were evaluated in the analysis. These variables were selected on the basis of their biologic plausibility of association and included age, gender, ASA risk classification, quantity of first ketamine dose (in milligrams per kilogram), and total number of doses administered. Other variables rejected from consideration as being less biologically plausible included sedation indication (eg, laceration, fracture reduction) and hospital (university versus county). Concurrent atropine administration could not be studied because essentially all subjects (98.6%) received this antisialogogue prophylactically. Similarly, concurrent benzodiazepine administration could not be studied because too few subjects in the data form group (1.2%) received this sedation adjunct. Finally, weight was not selected because it was highly correlated with age (Pearson r=0.840, P<.0001). There were few children with high ASA risk classification classes (892 class 1, 84 class 2, 36 class 3, 8 class 4, 1 class 5). Because dichotomization of this variable at the ASA level of 2 or less versus ASA level of 3 or more would not permit a sufficiently powerful analysis, we instead divided subjects as either ASA class 1 or ASA class 2 or more. Similarly, there were few children who received more than 2 ketamine doses (791 received 1 dose, 198 received 2 doses, 24 received 3 doses, and 8 received 4 doses). Accordingly, this variable was dichotomized as single or multiple doses and analyzed in this form. We examined the frequency distributions for the continuous variables (ie, age, ketamine dose); because their distributions were not bimodal, they were introduced into the multivariate analysis in their continuous forms. Because the small number of airway complications could only support a single predictor variable, we were unable to proceed with a multivariate analysis for this complication. Instead, we performed univariate analyses using the 5 predictor variables selected for emesis and recovery agitation. Categorical data were compared using χ2 or Fisher’s exact test. For continuous variables, we first performed Bartlett’s test for equal variances. If the variances between groups for a given variable were equal, we performed Student’s t test for equal variances. If the variances were not equal, we performed Student’s t test modified for unequal variances.
JANUARY 2000
35:1
ANNALS OF EMERGENCY MEDICINE
We performed separate multiple logistic regression analyses using the selected variables for emesis and recovery agitation. Those variables retaining an association with the adverse event of interest with P less than or equal to .05 were considered to have an independent association with that adverse event, and those with .05 less than P less than or equal to .10 to trend toward an association. We calculated the likelihood ratios and area under the receiver operating characteristic curve for each model, and performed goodness of fit analyses using the HosmerLemeshow test.16 To validate each model, we performed 2 separate bootstrap resampling procedures for each model with 200 iterations each. Bootstrapping refers to a process in which random samples of a database are drawn with replacement.17,18 This method can be used to obtain conservative estimates of confidence intervals (CIs) and standard errors, as well as to assess the stability of a model. In the first bootstrap procedure, we obtained 95% bias-corrected CIs of the predictor variables in each regression analysis. In the second bootstrap procedure, we reported the multivariate analysis on 200 random bootstrap samples of our data and identified how many times each predictor variable was identified as independently associated with the given outcome (P<.05). We performed regression diagnostics to identify goodness of fit for each model and overly influential covariate patterns.16 We inspected graphs of the change in deviance residual resulting from the deletion of subjects with a particular covariate pattern versus the predicted probabilities. We also inspected plots of the decrease in Hosmer-Lemeshow goodness-of-fit statistic caused by deleting patients with a given covariate pattern versus predicted probabilities, weighted by a measure of the coefficient vector change caused by deleting patients with a particular covariate pattern.19 We performed all analyses using Stata 5.0 statistical software (Stata Corporation, College Station, TX). R E S U LT S
No predictor variable had significant univariate associations with airway complications (Table 1). Age was a significant independent predictor of emesis (Table 2). A histogram of the incidence of emesis by age is shown in Figure 1. The incidence of emesis was 12.1% in children 5 years and older and 3.5% in those younger than 5 years (∆8.6%, 95% CI 4.9% to 12.1%). There was no trend or significant association between quantity of first ketamine dose and emesis, although the
3 7
KETAMINE ADVERSE EVENTS IN CHILDREN Green et al
bias-corrected 95% CI of the odds ratio (OR) did not include unity. Bootstrapping identified this variable as significantly associated with emesis less than one third of the time. The mean first ketamine dose was 4.03±0.37 mg/kg in children with emesis and 3.95±0.70 mg/kg in those without emesis (∆0.08, 95% CI –0.02 to 0.17). Administration of multiple doses trended toward predicting emesis, although the bias-corrected 95% CI for the OR did include unity. Bootstrapping identified this variable as significantly associated with emesis less than half of the time. The incidence of emesis was 6.2% in children receiving single ketamine doses and 8.3% in those receiving multiple doses (∆–2.1%, 95% CI –6.0% to 1.9%). Significant independent predictors of recovery agitation in the multivariate analysis were age and ASA class of 2 or more (Table 3). A histogram of the incidence of recovery agitation by age is shown in Figure 2. The incidence of recovery agitation was 12.1% in children aged 5 years and older and 22.5% in those younger than 5 years (∆–10.4%, 95% CI –3.0% to –17.7%). Seven of the 430 children were described as having “moderate to
severe” recovery agitation, 1 and their median age was 2.6 years (range 1.8 to 6.5 years). The incidence of recovery agitation was 17.9% in ASA class 1 children and 33.3% in children in ASA class 2 or more (∆–15.4%, 95% CI 0.0% to –30.7%). For the multivariate models, visual inspection of plots of the change in deviance residuals and Hosmer-Lemeshow goodness-of-fit statistic caused by deletion of different covariate patterns did not identify poor model fit. Likewise, inspection of plots of the Hosmer-Lemeshow goodnessof-fit statistic weighted by the change in coefficient vector did not identify overly influential covariate patterns. DISCUSSION
In this study, we report the largest analysis of predictors of ketamine adverse events ever reported in any setting. Importantly, we found that no variable could predict the adverse event of greatest concern—airway complications. We found certain parameters to be associated with emesis and recovery agitation, although the discriminatory power of these variables was relatively modest.
Table 1.
Univariate analysis of airway complications (n=1,021).
Variable Mean age (y±SD) Gender (male, %) ASA risk classification ≥2 (%) Mean quantity of first dose (mg/kg±SD) Multiple doses administered (%)
Airway Complication (n=14)
No Airway Complication (n=1,007)
Difference in Mean or Percentage (95% CI)
4.56±2.96 50.0 21.4 3.87±0.85 21.4
4.54±2.80 59.7 12.5 3.96±0.68 22.5
0.02 (–1.46 to 1.50) –9.7 (–36.0 to 16.7) 8.9 (–12.7 to 30.1) –0.09 (–0.45 to 0.27) –1.1 (–22.8 to 20.5)
.982 .464 .405 .629 1.000
P Value
Frequency Chosen as Independent Predictor Using Bootstrapping (%)
P Value
Table 2.
Multiple logistic regression model of emesis (n=1,021).*
Variable Age (y) Gender (male) ASA risk classification ≥2 Quantity of first dose (mg/kg) Multiple doses administered
Odds Ratio
Bias-Corrected 95% CI
1.25 1.19 0.82 1.30 1.66
1.17–1.34 0.72–2.19 0.23–1.62 1.02–1.67 0.91–2.83
<.001 .528 .610 .121 .084
100 8.5 8.5 28.5 42
*
The area under the model receiver operator curve is 0.715. The model demonstrated satisfactory goodness-of-fit, with the Hosmer-Lemeshow P=.905. The likelihood ratio test for this model versus the constant-only model was χ2 33.28, P<.001.
3 8
ANNALS OF EMERGENCY MEDICINE
35:1 JANUARY 2000
KETAMINE ADVERSE EVENTS IN CHILDREN Green et al
We found no variable to be predictive of airway complications, and observed no trends to suggest any effect that might be detected with a larger sample size. The rarity of airway complications in the sample (1.4%) limited the power of these comparisons and precluded a multivariate analysis. There are no other studies evaluating predictors of ketamine-associated airway complications with which to contrast our data. The existing ketamine literature consists primarily of case series in various clinical settings wherein occasional airway complications are individually described. Respiratory depression and apnea have been reported when ketamine is administered to infants younger Figure 1.
Histogram of the incidence of emesis by age.
than 3 months (especially when ill), or when central nervous system injuries, masses, or abnormalities are present.4,20-22 Neonates and small infants have greater difficulty maintaining a patent airway with ketamine just as they do with other sedatives. Ketamine elevates intracranial pressure and accordingly can exacerbate concurrent central nervous system conditions.4,20-22 These situations are thus well accepted as contraindications to ketamine, and there would appear to be few if any indications for emergency physicians to administer ketamine in such children. Our protocol excluded children with these conditions from this study. Procedures involving the posterior pharynx are generally considered to represent a relative contraindication to the use of ketamine. The dissociative state preserves and slightly exaggerates protective airway reflexes such as coughing and swallowing, and laryngospasm has been reported in association with posterior pharyngeal stimulation.4,20-22 As previously reported, 1 of our 4 cases of laryngospasm began during oropharyngeal suctioning.1 Despite this being a relative contraindication in our protocol, treating physicians elected to administer ketamine to 3 children in our data set for procedures involving the posterior pharynx (removal of posterior pharyngeal foreign bodies in all 3 cases), and these procedures were accomplished without adverse events.1 Although likely a risk factor, available evidence suggests that pharyngeal procedures should be a relative exclusion criterion rather than an absolute one. Respiratory depression and apnea are well recognized when ketamine is administered by rapid intravenous infusion, and it is generally accepted that intravenous doses should be given over at least 1 minute to avoid such responses.4,20-22
Table 3.
Multiple logistic regression model of recovery agitation (n=430).*
Variable Age (y) Gender (male) ASA risk classification ≥2 Quantity of first dose (mg/kg) Multiple doses administered
Odds Ratio
Bias-Corrected 95% CI
P Value
0.79 0.87 3.05 0.86 1.34
0.69–0.89 0.47–1.46 1.65–7.30 0.49–1.73 0.78–2.41
<.001 .575 .004 .596 .336
Frequency Chosen as Independent Predictor Using Bootstrapping (%) 99.5 8 79.5 14 18
*
The area under the model receiver operator curve is 0.673. The model demonstrated satisfactory goodness-of-fit, with the Hosmer-Lemeshow P=.363. The likelihood ratio test for this model versus the constant-only model was χ2=26.31, P<.001.
JANUARY 2000
35:1
ANNALS OF EMERGENCY MEDICINE
3 9
KETAMINE ADVERSE EVENTS IN CHILDREN Green et al
Beyond the existing exclusion criteria already accounted for in patient selection, airway complications with ketamine are unusual and appear unpredictable. Accordingly, clinicians must be prepared to deal with them in any given sedation situation. We believe that the routine use of continuous pulse oximetry, the presence of a nurse dedicated to patient monitoring, and the immediate availability of a physician with airway expertise permit rapid detection and treatment of such rare events. It is unlikely that prospective randomized studies comparing the incidence of airway complications between ketamine sedation regimens (eg, dose, route, patient groups) will ever be performed. We calculate that 7,216 total subjects would be required for such a study to detect a 50% relative difference in airway complications from our 1.4% baseline incidence (α=.05, β=.2). We found that increasing age was predictive of emesis. There are no other studies evaluating predictors of ketamine-associated emesis with which to contrast our data. Although age was the strongest predictor of emesis identified in our analysis, even in the highest-risk age range (10 to 15 years) the incidence was only 17%. As previously reported, 1 there were no adverse outcomes attributable to ketamine-associated emesis in this sample, although vomiting was believed to have provoked one episode of transient laryngospasm. In 30 years of extensive use worldwide, there are no documented
Figure 2.
Histogram of the incidence of recovery agitation by age.
4 0
cases in the medical literature of clinically significant ketamine-associated aspiration syndrome in a patient lacking established contraindications to this drug. 4 We believe it unlikely that clinicians will alter their decisions regarding ketamine use based on the modestly higher rates of emesis noted in older children. Prophylactic antiemetic therapy could be considered in older children, although the efficacy of such an intervention is not known. Although the quantity of the first ketamine dose and administration of multiple doses were not significantly associated with emesis, we found other weak evidence suggesting a relationship in both. For first dose, the OR bias-corrected 95% CI by a small margin did not overlap unity, and for the number of doses there was a statistical trend toward an association (P=.084). Despite this, the observed difference in first dose between subjects with and without emesis was only 0.08 mg/kg, and the observed difference in rates of emesis between these groups was only 2.1%. We believe that both of these differences are clinically unimportant. Because bootstrapping selected these 2 variables as independent predictors of emesis less than half of the time, we suspect that these findings are either the result of chance alone or represent associations below the threshold of clinical significance. A limitation of this analysis is that we did not collect data on time of last meal, and it is possible that shorter fasting periods may be predictive of emesis. Prospective studies are needed to evaluate this potential relationship. Of note, a pooled comparison of data from multiple ketamine studies did not demonstrate an association between fasting state and emesis. 4 We found that younger age and the presence of an underlying medical condition (ie, ASA class ≥2) were predictors of recovery agitation. Several prior studies, primarily in adult populations, have examined risk factors for recovery agitation or hallucinations. These studies have reported that female gender, age older than 10 years, rapid intravenous administration, excessive noise or stimulation during recovery, prior personality disorders, and patients who normally dream frequently represent situations of greater risk. 4,20-22 Our data dispute a female predisposition to recovery agitation in the pediatric population and provide strong evidence against an increased risk of recovery reactions in the 10- to 15-year-old age group. We did not systematically evaluate the other risk factors previously described in the literature, and thus our data do not provide additional information in those areas.
ANNALS OF EMERGENCY MEDICINE
35:1 JANUARY 2000
KETAMINE ADVERSE EVENTS IN CHILDREN Green et al
Our inverse association of recovery agitation to age is surprising. Existing literature describes recovery reactions as rare in children and relatively common (up to 30%) in adults,4,20-22 so an increasing relationship with age would have been expected. It is possible that the preprocedure level of agitation may explain this discrepancy. Infants and toddlers are typically anxious before ED sedation and it would not be surprising if some degree of agitation might persist through recovery. Older children, however, are more likely to be calm before sedation after reassurance from their parents and ED staff. Our study is limited in that we did not collect data on the degree of agitation before sedation. This should be the subject of future study. The basis for our observed association of recovery agitation with an underlying medical condition is unclear. Possibly, children with many previous physician visits associate medical contact with physical discomfort (eg, phlebotomy, injections) and thus may exhibit more preprocedure and postprocedure fear and agitation than corresponding healthy children. A limitation of this data set is that only 5 of the 430 children in this portion of the analysis received concurrent benzodiazepines, and thus these agents were not sufficiently represented to permit us to assess their relationship to recovery agitation. Concurrent benzodiazepine use is widely believed to diminish ketamineassociated recovery reactions, 20-22 although the value of these agents has never been established in children. A prospective study specifically focused on this question is needed. This report is also limited in that we did not collect data on agitation or behavior changes after ED discharge. A recent randomized study comparing a sedation regimen of ketamine and midazolam with fentanyl and midazolam found similar incidences of adverse effects between regimens after ED discharge, except that vomiting was more common in the ketamine group. 10 How clinicians should translate these findings into clinical practice is not apparent. Although agitation is not infrequent after ketamine administration (19.2% overall), when it occurs it is rated by treating physicians as “mild” 92% of the time and rarely leads to specific treatment. 1 Dose-related adverse effect relationships are well documented with benzodiazepines and narcotics. 5 The assumption of a similar ketamine dose response has led some physicians to advocate relatively low doses of ketamine in the hope of minimizing adverse events. 23,24 Our data do not demonstrate evidence of a significant
JANUARY 2000
35:1
ANNALS OF EMERGENCY MEDICINE
ketamine dose relationship with the 3 adverse effects studied. Doses high enough to consistently generate the dissociative state (4 mg/kg IM) appear to be no less safe than lower doses. Dose-related adverse effects with ketamine are not described in the literature. No prior studies have compared airway complications or emesis by dose. Two retrospective studies failed to note significant differences in adult recovery reactions by dose. 24,25 A comparison of data pooled from multiple studies did not demonstrate an association between dose and adverse effects.4 In fact, other authors have anecdotally reported their perception that recovery agitation occurs in an inverse relationship with dose. 26-28 This study is subject to the usual limitations of studies with chart review components, including dependence on the quality of medical record documentation. We believe this factor had minimal effect on the current study, as the information we abstracted from records was largely objective and not likely prone to misinterpretation or bias. We based the single subjective item (recovery agitation) solely on the concurrently recorded impression of treating physicians. The incidence of recovery agitation was dissimilar between the groups with and without treating physician data forms. This disparity suggests that the majority of such reactions are not judged sufficiently substantial to warrant comment in the medical record. Airway complications and emesis, however, were recorded with similar frequency in sedations with and without data forms, and we believe that our data accurately depict such events. It is possible that treating physicians and nurses may be prone to underreport serious adverse events. To avoid identifying any such events in this study would have required their omission from the study data form as well as from all physician and nursing notes in the medical record. Although we believe it unlikely, we cannot exclude the possibility that some adverse events may have been missed. We believe that the most important implications from our data are that children at risk for rare airway complications cannot be predicted beyond existing exclusion criteria, and that there is no evidence of a ketamine dose relation to such events. Although the modest association of age with emesis, and age and underlying medical condition with recovery agitation are of interest, these factors exhibit low enough discriminatory value that they are unlikely to alter clinical decisions regarding patient selection for ketamine administration.
4 1
KETAMINE ADVERSE EVENTS IN CHILDREN Green et al
We thank Sean P Bush, MD, for his review of the manuscript and many helpful suggestions.
REFERENCES
27. Phillips LA, Seruvatu SG, Rika PN. Anaesthesia for the surgeon-anaesthetist in difficult situations. Anaesthesia. 1970;25:36-45. 28. Bennett JA, Bullimore JA. The use of ketamine hydrochloride anaesthesia for radiotherapy in young children. Br J Anaesth. 1973;45:197-201.
1. Green SM, Rothrock SG, Lynch EL, et al. Intramuscular ketamine for pediatric sedation in the emergency department: safety profile with 1,022 cases. Ann Emerg Med. 1998;31:688-697. 2. Green SM, Rothrock SG, Harris T, et al. Intravenous ketamine for pediatric sedation in the emergency department: safety profile with 156 cases. Acad Emerg Med. 1998;5:971-976. 3. Green SM, Nakamura R, Johnson NE. Ketamine sedation for pediatric procedures: part 1, a prospective series. Ann Emerg Med. 1990;19:1024-1032. 4. Green SM, Johnson NE. Ketamine sedation for pediatric procedures: part 2, review and implications. Ann Emerg Med. 1990;19:1033-1046. 5. Sacchetti A, Schafermeyer R, Gerardi M, et al. Pediatric analgesia and sedation. Ann Emerg Med. 1994;23:237-250. 6. Petrack EM, Marx CM, Wright MS. Intramuscular ketamine is superior to meperidine, promethazine, and chlorpromazine for pediatric emergency department sedation. Arch Pediatr Adolesc Med. 1996;150:676-681. 7. Pruitt JW, Goldwasser MS, Sabol SR, et al. Intramuscular ketamine, midazolam, and glycopyrrolate for pediatric sedation in the emergency department. J Oral Maxillofac Surg. 1995;53:13-17. 8. Dachs RJ, Innes GM. Intravenous ketamine sedation of pediatric patients in the emergency department. Ann Emerg Med. 1997;29:146-150. 9. Qureshi F, Mellis PT, McFadden MA. Efficacy of oral ketamine for providing sedation and analgesia to children requiring laceration repair. Pediatr Emerg Care. 1995;11:93-97. 10. Kennedy RM, Porter FL, Miller JP, et al. Comparison of fentanyl/midazolam with ketamine/midazolam for pediatric orthopedic emergencies. Pediatrics. 1998;102:956-963. 11. Dripps RD, Lamong A, Eckenhoff JE. The role of anesthesia in surgical mortality. JAMA.1961;178:261-266. 12. Wears RL, Lewis RJ. Statistical models and Occam’s razor [editorial]. Acad Emerg Med. 1999;6:93-94. 13. Harrell FE, Lee KL, Matchar DB, et al. Regression models for prognostic prediction: advantages, problems, and suggested solutions. Cancer Treat Rep. 1985;69:1071-1077. 14. Harrell FE, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996;15:361-387. 15. Concato J, Feinstein AR, Holford TR. The risk of determining risk with multivariable models. Ann Intern Med. 1993;118:201-210. 16. Hosmer DW, Lemeshow S. Applied Logistic Regression. New York, NY: John Wiley & Sons; 1989:135-145. 17. Efron B, Gong G. A leisurely look at the bootstrap, the jackknife and cross-validation. Am Stat. 1983;37:36-48. 18. Efron B, Tibshirani R. Statistical data analysis in the computer age. Science. 1991;253:390395. 19. Pregibon D. Logistic regression diagnostics. Ann Statistics. 1981;9:705-724. 20. White PF, Way WL, Trevor AJ. Ketamine—its pharmacology and therapeutic uses. Anesthesiology. 1982;56:119-136. 21. Corssen G, Reves JG, Stanley TH. Intravenous Anesthesia and Analgesia. Philadelphia, PA: Lea & Febiger; 1988:99-174. 22. Reich DL, Silvay G.Ketamine: an update on the first twenty-five years of clinical experience. Can J Anaesth. 1989;36:186-197. 23. Epstein FB. Ketamine dissociative sedation in pediatric emergency medical practice. Am J Emerg Med. 1993;11:180-182. 24. Liang HS, Liang HG. Minimizing emergence phenomena: subdissociative dosage of ketamine in balanced surgical anesthesia. Anesth Analg. 1975;54:312-316. 25. Knox JW, Bovill JG, Clarke RS, et al. Clinical studies of induction agents—ketamine. Br J Anaesth. 1970;42:875-885. 26. Faithfull NS, Haider R. Ketamine for cardiac catheterisation. Anaesthesia. 1971;26:318-323.
4 2
ANNALS OF EMERGENCY MEDICINE
35:1 JANUARY 2000