Professional Documents
Culture Documents
ORIGINAL ARTICLE
Keywords Summary
pediatrics; postoperative nausea and
vomiting; risk factors Background: Few data are available in the literature on risk factors for post-
operative vomiting (POV) in children.
Correspondence Objective: The aim of the study was to establish independent risk factors for
Gilles A. Orliaguet, Department of POV and to construct a pediatric specific risk score to predict POV in chil-
Anesthesiology and Critical Care Medicine,
dren.
Centre Hospitalier Universitaire Necker –
Methods: Characteristics of 2392 children operated under general anesthesia
Enfants Malades, 149 rue de Sevres, Paris
75743, France were recorded. The dataset was randomly split into an evaluation set
Email: gilles.orliaguet@nck.aphp.fr (n = 1761), analyzed with a multivariate analysis including logistic regression
and backward stepwise procedure, and a validation set (n = 450), used to
Section Editor: Jerrold Lerman confirm the accuracy of prediction using the area under the receiver operating
characteristic curve (ROCAUC), to optimize sensitivity and specificity.
Accepted 9 April 2014
Results: The overall incidence of POV was 24.1%. Five independent risk fac-
tors were identified: stratified age (>3 and <6 or >13 years: adjusted OR 2.46
doi:10.1111/pan.12428
[95% CI 1.75–3.45]; ≥6 and ≤13 years: aOR 3.09 [95% CI 2.23–4.29]), dura-
tion of anesthesia (aOR 1.44 [95% IC 1.06–1.96]), surgery at risk (aOR 2.13
[95% IC 1.49–3.06]), predisposition to POV (aOR 1.81 [95% CI 1.43–2.31]),
and multiple opioids doses (aOR 2.76 [95% CI 2.06–3.70], P < 0.001). A sim-
plified score was created, ranging from 0 to 6 points. Respective incidences of
POV were 5%, 6%, 13%, 21%, 36%, 48%, and 52% when the risk score ran-
ged from 0 to 6. The model yielded a ROCAUC of 0.73 [95% CI 0.67–0.78]
when applied to the validation dataset.
Conclusions: Independent risk factors for POV were identified and used to
create a new score to predict which children are at high risk of POV.
and the risk of POV was not linear, age was stratified The primary end point of this study was the propor-
into a categorical variable with three groups (class 1: tion of POV during the first postoperative 24 h.
≤3 years; class 2: >3 years and <6 years or >13 years;
and class 3: ≥6 years and ≤13 years). Duration of anes-
Sample size
thesia was defined as <45 min or ≥45 min. Surgery at
risk included tympanoplasty, tonsillectomy, and strabis- The sample size of 1000 children was calculated to dem-
mus surgery. Similarly, to reduce the number of vari- onstrate a 10% difference in POV incidence between
ables, we decided to combine ‘personal history of POV,’ unexposed (POV incidence = 25%) and exposed children
‘motion sickness,’ and ‘familial history of POV’ into a (POV incidence = 35%) with a statistical power of 90%
composite variable called ‘predisposition to POV.’ In and alpha type I error of 5%, assuming a 20% prevalence
addition, we merged ‘intraoperative opioids re-injection’ of exposure to a specific risk factor. An independent sam-
and ‘postoperative use of opioids’ into a single variable ple was added for score validation (n = 500). The total
called ‘multiple opioids doses.’ sample size was 1500, and it was reached before the
expected date.
POV assessment
Statistical analysis
All patients were admitted in the postanesthesia care
unit (PACU) following anesthesia, except those requir- The evaluation dataset and the validation dataset com-
ing hospitalization in the pediatric intensive care unit. prised 1761 and 450 participants, respectively, with com-
POV was assessed in the PACU by specially instructed pleted data. Descriptive and univariate analysis included
nurses or anesthesiologists. Early POV were defined as the Student t-test or Wilcoxon rank test for quantitative
POV occurring during the stay in the PACU. All POV variables or v2 test for intergroup comparisons.
episodes occurring during the first 24 h after anesthesia Variables associated with POV (P value <0.10) were
were noted. For patients having surgery on an outpa- then computed in multiple logistic regression models
tient basis, parents were asked to fill out a prefilled letter using a backward stepwise procedure to evaluate the
and to send it to the Clinical Research Unit on the impact of potential risk factors on POV. For each pre-
second day after surgery. In case of nonreception of the dictor, crude and adjusted ORs with their 95% confi-
letter, the parents were interviewed by phone the follow- dence interval [95% CI] were calculated using regression
ing day. Because nausea is a subjective phenomenon coefficients. The performances of the model were
and small children are not able to describe it, only POV assessed on the evaluation set by calibration and dis-
were evaluated. crimination. We plotted observed outcome by decile of
predictions, which makes the plot a graphical illustra- ratio = 1.4). Premedication concerned about 82% of
tion of the Hosmer–Lemeshow goodness-of-fit test. The children, with half of the participants receiving midazo-
internal validation consisted in determining the discrimi- lam. The main types of surgery were orthopedics (24%),
native ability of the model by measuring areas under the urology (16%), ophthalmology (14%), and ENT sur-
receiver operating characteristics curve (ROCAUC). gery (12%). Three-quarter of participants received opi-
Internal validation was completed by applying the opti- oids during induction, with half of them receiving
mal model to validation set. We built POVOC score sufentanil. Maintenance of anesthesia was performed
applying the customization approach as described by with sevoflurane in 80% of the children. Maintenance
Engel et al. (6). A comparison of ROCAUC between with opioid occurred in a half of the subjects, with a
vomiting in the post operative period (VPOP) and cus- third of them receiving sufentanil. Among the 373 chil-
tomized POVOC models was performed on the valida- dren who received muscle relaxants, 36 were reversed
tion set using bootstrap method which included a 2000- with a combination of neostigmine and atropine. Dura-
fold sampling without replace. For clinical practice, a tion of anesthesia ranged from 9 to 520 min, with a
score was proposed, based on regression coefficients. median value of 60 min.
Statistical analysis was performed with R program (Bos- For ambulatory patients, the final completion rate for
ton, MA, USA www.r-project.org) (7) and specific pack- parental reports (prefilled letter or phone call) was 88%.
ages (EPICALCTM, PROC, EPITOOLS, PREDICTABEL; R Core The global incidence of POV was 24.1%, with 414
Team, R Foundation for Statistical Computing, Vienna, patients (23.5%) in the evaluation set and 118 patients
Austria). (26.2%) in the validation set (P = 0.25).
References
1 Murat I, Constant I, Maud’huy H. Perioper- 8 Lerman J. Surgical and patient factors 17 Busoni P, Sarti A, Crescioli M et al. Motion
ative anaesthetic morbidity in children: a involved in postoperative nausea and vomit- sickness and postoperative vomiting in chil-
database of 24,165 anaesthetics over a 30- ing. Br J Anaesth 1992; 69: 24S–32S. dren. Paediatr Anaesth 2002; 12: 65–68.
month period. Pediatr Anesth 2004; 14: 9 Oddby E, Englund S, Lonnqvist PA. Postop- 18 Apfel CC, Philip BK, Cakmakkaya OS et al.
158–166. erative nausea and vomiting in paediatric Who is at risk for postdischarge nausea and
2 Eberhart LH, Morin AM, Guber D et al. ambulatory surgery: sevoflurane versus vomiting after ambulatory surgery? Anesthe-
Applicability of risk scores for postoperative spinal anaesthesia with propofol sedation. siology 2012; 117: 475–486.
nausea and vomiting in adults to paediat- Paediatr Anaesth 2001; 11: 337–342. 19 Elhakim M, Abd-Elfattah H, El-Din DN
ric patients. Br J Anaesth 2004; 93: 10 Cohen MM, Cameron CB, Duncan PG. et al. Midazolam as an antiemetic in patients
386–392. Pediatric anesthesia morbidity and mortality receiving epidural morphine for postopera-
3 Eberhart LH, Geldner G, Kranke P et al. in the perioperative period. Anesth Analg tive pain relief. J Opioid Manag 2009; 5:
The development and validation of a risk 1990; 70: 160–167. 189–195.
score to predict the probability of postopera- 11 Karlsson E, Larsson LE, Nilsson K. Postan- 20 Riad W, Altaf R, Abdulla A et al. Effect of
tive vomiting in pediatric patients. Anesth aesthetic nausea in children. Acta Anaesthesi- midazolam, dexamethasone and their combi-
Analg 2004; 99: 1630–1637. ol Scand 1990; 34: 515–518. nation on the prevention of nausea and vom-
4 Friedman DS, Repka MX, Katz J et al. 12 Murray DJ, Schmid CM, Forbes RB. Anes- iting following strabismus repair in children.
Prevalence of amblyopia and strabismus in thesia for magnetic resonance imaging in Eur J Anaesthesiol 2007; 24: 697–701.
white and African American children aged 6 children: a low incidence of protracted post- 21 Riad W, Marouf H. Combination therapy in
through 71 months the Baltimore Pediatric procedure vomiting. J Clin Anesth 1995; 7: the prevention of PONV after strabismus
Eye Disease Study. Ophthalmology 2009; 232–236. surgery in children: granisetron, ondanse-
116: 2128–2134 e2121–2122. 13 Apfel CC, Laara E, Koivuranta M et al. A tron, midazolam with dexamethasone. Mid-
5 Robaei D, Rose KA, Ojaimi E et al. Causes simplified risk score for predicting postopera- dle East J Anesthesiol 2009; 20: 431–436.
and associations of amblyopia in a popula- tive nausea and vomiting: conclusions from 22 Splinter WM, MacNeill HB, Menard EA
tion-based sample of 6-year-old Australian cross-validations between two centers. Anes- et al. Midazolam reduces vomiting after ton-
children. Arch Ophthalmol 2006; 124: 878– thesiology 1999; 91: 693–700. sillectomy in children. Can J Anaesth 1995;
884. 14 Gan TJ. Risk factors for postoperative nau- 42: 201–203.
6 Engel JM, Junger A, Hartmann B et al. sea and vomiting. Anesth Analg 2006; 102: 23 Apfel CC, Kranke P, Greim CA et al. What
Performance and customization of 4 1884–1898. can be expected from risk scores for predict-
prognostic models for postoperative onset of 15 Sinclair DR, Chung F, Mezei G. Can postop- ing postoperative nausea and vomiting? Br
nausea and vomiting in ear, nose, and erative nausea and vomiting be predicted? J Anaesth 2001; 86: 822–827.
throat surgery. J Clin Anesth 2006; 18: Anesthesiology 1999; 91: 109–118. 24 Apfel CC, Korttila K, Abdalla M et al. A
256–263. 16 Thomas M, Woodhead G, Masood N et al. factorial trial of six interventions for the pre-
7 Ihaka R, Gentleman R. R: a language for Motion sickness as a predictor of postopera- vention of postoperative nausea and vomit-
data analysis and graphics. J Comput Graph tive vomiting in children aged 1–16 years. ing. N Engl J Med 2004; 350: 2441–2451.
Stat 1996; 5: 299–314. Pediatr Anesth 2007; 17: 61–63.