new england

The
journal of medicine
established in 1812 june 10 , 2004 vol. 350 no. 24

A Factorial Trial of Six Interventions for the Prevention

of Postoperative Nausea and Vomiting
Christian C. Apfel, M.D., Kari Korttila, F.R.C.A., Ph.D., Mona Abdalla, Ph.D., Heinz Kerger, M.D.,
Alparslan Turan, M.D., Ina Vedder, M.D., Carmen Zernak, M.D., Klaus Danner, M.D., Ritva Jokela, M.D., Ph.D.,
Stuart J. Pocock, Ph.D., Stefan Trenkler, M.D., Markus Kredel, M.D., Andreas Biedler, M.D., Daniel I. Sessler, M.D.,
and Norbert Roewer, M.D., for the IMPACT Investigators*

abstract

background
Untreated, one third of patients who undergo surgery will have postoperative nausea From Julius-Maximilians Universitt,
and vomiting. Although many trials have been conducted, the relative benefits of pro- Wrzburg, Germany (C.C.A., M.K., N.R.);
the University of Louisville, Louisville, Ky.
phylactic antiemetic interventions given alone or in combination remain unknown. (C.C.A., D.I.S.); Helsinki University Cen-
tral Hospital, Helsinki, Finland (K.K., R.J.);
methods the London School of Hygiene and Tropi-
cal Medicine, London (M.A., S.J.P.); the
We enrolled 5199 patients at high risk for postoperative nausea and vomiting in a ran- Universittsklinik Mannheim, Mannheim,
domized, controlled trial of factorial design that was powered to evaluate interactions Germany (H.K.); Trakya University Hospital,
among as many as three antiemetic interventions. Of these patients, 4123 were ran- Edirne, Turkey (A.T.); von Bodelschwing-
sche Anstalten Bethel, Bielefeld, Germany
domly assigned to 1 of 64 possible combinations of six prophylactic interventions: (I.V.); Kreiskrankenhaus Garmisch-Parten-
4 mg of ondansetron or no ondansetron; 4 mg of dexamethasone or no dexamethasone; kirchen, Garmisch-Partenkirchen, Germany
1.25 mg of droperidol or no droperidol; propofol or a volatile anesthetic; nitrogen or (C.Z.); Westpfalz-Klinikum, Kaiserslaut-
ern, Germany (K.D.); Reiman University
nitrous oxide; and remifentanil or fentanyl. The remaining patients were randomly as- Hospital, Presov, Slovakia (S.T.); and the
signed with respect to the first four interventions. The primary outcome was nausea Universittskliniken des Saarlandes, Hom-
and vomiting within 24 hours after surgery, which was evaluated blindly. burg, Germany (A.B.). Address reprint re-
quests to Dr. Apfel at Outcomes Research
Institute, 501 East Broadway, Suite 210,
results Louisville, KY 40202, or at [email protected].
Ondansetron, dexamethasone, and droperidol each reduced the risk of postoperative
*The International Multicenter Protocol
nausea and vomiting by about 26 percent. Propofol reduced the risk by 19 percent, and to Assess the Single and Combined Ben-
nitrogen by 12 percent; the risk reduction with both of these agents (i.e., total intrave- efits of Antiemetic Interventions in a Con-
nous anesthesia) was thus similar to that observed with each of the antiemetics. All the trolled Clinical Trial of a 222222
Factorial Design (IMPACT) Investigators
interventions acted independently of one another and independently of the patients are listed in the Appendix.
baseline risk. Consequently, the relative risks associated with the combined interventions
could be estimated by multiplying the relative risks associated with each intervention. N Engl J Med 2004;350:2441-51.
Copyright 2004 Massachusetts Medical Society.
Absolute risk reduction, though, was a critical function of patients baseline risk.

conclusions
Because antiemetic interventions are similarly effective and act independently, the safest
or least expensive should be used first. Prophylaxis is rarely warranted in low-risk pa-
tients, moderate-risk patients may benefit from a single intervention, and multiple in-
terventions should be reserved for high-risk patients.

n engl j med 350;24 www.nejm.org june 10, 2004 2441

The New England Journal of Medicine

Downloaded from nejm.org on May 15, 2014. For personal use only. No other uses without permission.
Copyright 2004 Massachusetts Medical Society. All rights reserved.
 The new england journal of medicine

a nesthesia is given to more than

75 million surgical patients annually,
worldwide. Untreated, one third will have
postoperative nausea, vomiting, or both.1-3 Patients
often rate postoperative nausea and vomiting as
methods
The design of the study, the recruitment of patients
at each center, the acquisition and management of
data, the statistical analyses, the interpretation of
worse than postoperative pain.4,5 It is not surpris- the data, and the writing and editing of the manu-
ing, therefore, that prevention of postoperative nau- script were performed independently of the spon-
sea and vomiting improves satisfaction among pa- sors. The contributions of the individual authors
tients who are likely to experience them.6 Vomiting are listed in the Appendix.
increases the risk of aspiration and has been asso- After obtaining approval from the institutional
ciated with suture dehiscence, esophageal rupture, review boards of the 28 participating centers, we en-
subcutaneous emphysema, and bilateral pneumo- rolled 5199 adults who were scheduled to undergo
thoraxes.7,8 Postoperative nausea and vomiting fre- elective surgery during general anesthesia that was
quently delay discharge from postanesthesia care expected to last at least one hour. All the patients had
units, and they are the leading cause of unexpected a risk of postoperative nausea and vomiting that ex-
hospital admission after planned ambulatory sur- ceeded 40 percent, according to a simplified risk
gery.9 The annual cost of postoperative nausea and score,17 based on the presence of at least two of the
vomiting in the United States is thought to be sev- following risk factors: female sex, nonsmoker sta-
eral hundred million dollars.10,11 tus, previous history of postoperative nausea and
More than 1000 randomized, controlled trials vomiting or motion sickness, and anticipated use
have evaluated pharmacologic methods of prevent- of postoperative opioids.18,19 We excluded patients
ing and treating postoperative nausea and vomit- in whom any of the study drugs were contraindicat-
ing. Most have compared a single intervention with ed, those who had taken emetogenic or antiemetic
placebo. Serotonin (5-hydroxytryptamine type 3) drugs within the 24 hours before surgery, those who
antagonists (e.g., ondansetron), dexamethasone were expected to require postoperative mechanical
(a corticosteroid), and droperidol (a neuroleptic ventilation, and those who were pregnant or lactat-
drug) are among the best-studied antiemetic agents. ing. All the patients provided their written informed
Alternatively, the avoidance of emetogenic factors consent.
during anesthesia can reduce the baseline risk of
postoperative nausea and vomiting. This strategy protocol
includes the use of propofol instead of volatile an- The antiemetic efficacy of six individual treatments
esthetics, the substitution of nitrogen for nitrous and combinations of them was simultaneously
oxide, and the use of remifentanil, an ultrashort- evaluated according to a 26 factorial design.20 Three
acting opioid, instead of fentanyl.12,13 of the prophylactic interventions involved the use of
The limited efficacy of treatment with single an antiemetic drug: ondansetron, dexamethasone,
antiemetics14 has prompted evaluations of several or droperidol. The other three interventions con-
antiemetic strategies used in combination.15 How- sisted of the use of propofol instead of a volatile an-
ever, no previous study of postoperative nausea and esthetic, the omission of nitrous oxide, and the
vomiting has had an appropriate design or suffi- substitution of remifentanil for fentanyl. Thus, ac-
cient power to evaluate all the major pharmacolog- cording to the study design, each patient was to be
ic interventions simultaneously or to determine the randomly assigned to one of each of the following
extent to which combining multiple interventions six interventions: ondansetron (4 mg intravenously)
improves outcome. A recent consensus conference or no ondansetron; dexamethasone (4 mg intrave-
was thus unable to support a definitive statement on nously) or no dexamethasone; droperidol (1.25 mg
the benefits of combining antiemetic strategies.16 intravenously) or no droperidol; propofol or a vola-
We therefore conducted a large clinical trial of fac- tile anesthetic (i.e., isoflurane, desflurane, or sevo-
torial design with sufficient power to compare the flurane) in a 2:1 ratio; nitrogen or nitrous oxide;
efficacy of six well-established antiemetic strategies and remifentanil or fentanyl.
and to determine the extent to which efficacy could These 6 treatments lead to a possible 64 (i.e., 26)
be improved by combining two or three interven- different treatment combinations. However, propo-
tions. fol is associated with a reduced risk of postopera-

2442 n engl j med 350;24 www.nejm.org june 10 , 2004

The New England Journal of Medicine

Downloaded from nejm.org on May 15, 2014. For personal use only. No other uses without permission.
Copyright 2004 Massachusetts Medical Society. All rights reserved.
 a factorial trial of antiemetic strategies

tive nausea and vomiting,21 so to ensure sufficient mental oxygen, and pain was ameliorated with the
power to quantify the effect of antiemetics in the use of nonsteroidal antiinflammatory medications
propofol subgroup, we assigned twice as many pa- administered intraoperatively. The patients who had
tients to propofol as to volatile anesthetics (for a 2:1 been assigned to receive intraoperative remifentanil
randomization ratio). Therefore, permuted blocks were given 50 g of morphine per kilogram or an
of 96 (23322) patients were generated. Each cen- equivalent opioid at the end of surgery. The need
ter received four blocks with a unique computerized for a postoperative opioid was left to the discretion
randomization, stored in sequentially numbered, of the anesthesiologist, and the dose was adjusted
sealed, opaque envelopes. according to clinical needs. Patients who requested
The envelopes were opened after consent was antiemetic therapy or who had an emetic episode
obtained, just before the induction of general anes- were given 4 mg of ondansetron; if symptoms per-
thesia. The anesthesiologists responsible for intra- sisted, 4 mg of dexamethasone and 1.25 mg of dro-
operative management were not blinded to the treat- peridol were added.
ment, but they were not involved in the postoperative
assessment. Supplemental oxygen may22,23 or may measurements
not24,25 have an antiemetic effect. Consequently, at Our primary outcome measure was the incidence of
three centers patients were randomly assigned to any nausea, emetic episodes (retching or vomiting),
30 percent oxygen in nitrous oxide, 30 percent oxy- or both (i.e., postoperative nausea and vomiting)
gen in nitrogen, or 80 percent oxygen in nitrogen, during the first 24 postoperative hours. After the
in a randomization ratio of 1:1:1. As a result, a min- 2nd and 24th postoperative hours, trained investi-
imum of 144 (348) patients were required per gators who were fully blinded to the intraoperative
block. To provide sufficient power, each center management and random treatment assignments
agreed to study 288 patients, twice as many as the recorded the number of emetic episodes and the
minimum. time each one occurred. At both these time points,
The patients were given premedication with a patients orally rated their worst nausea episode dur-
benzodiazepine. Three minutes before the induc- ing the preceding interval on an 11-point scale,
tion of anesthesia, they received either a bolus of where 0 represented no nausea and 10 the most
fentanyl (100 to 200 g) or an infusion of remifen- severe nausea possible.
tanil (0.25 g per kilogram of body weight per min-
ute), according to the treatment to which they had statistical analysis
been assigned. Anesthesia was induced with intra- Different sample-size estimations were performed
venous propofol (Disoprivan or Diprivan, Astra- and indicated that about 5000 patients would be
Zeneca) at a dose of 2 to 3 mg per kilogram, and tra- needed for the analysis of interactions involving as
cheal intubation was facilitated with rocuronium. many as three factors, whereas the number of pa-
Normocapnic mechanical ventilation was insti- tients required for the analysis of two-factor inter-
tuted with the assigned gas combination. Anesthe- actions or of single factors was considerably small-
sia was maintained with either propofol (starting at er.20 An interaction was defined as present if the
about 80 g per kilogram per hour) or a standard- effect of two factors in combination was signifi-
ized concentration of a volatile anesthetic. If the cantly different from the separate effects of each
heart rate or blood pressure deviated by more than factor multiplied together on an odds-ratio scale.
20 percent from the preoperative value, an intrave- For each of the six randomized treatments, the
nous bolus of fentanyl (50 to100 g) was given or the numbers of patients who had postoperative nausea
rate of remifentanil infusion was increased slightly. and vomiting were compared with the use of chi-
In addition, the concentration of volatile anesthetics square tests for each main effect, and reductions
or the propofol infusion rate could be adjusted as in the relative risk of nausea and vomiting were
clinically appropriate. In the designated patients, estimated. Logistic-regression analyses were used
4 mg of dexamethasone (if assigned) and 1.25 mg to quantify the relative effects of the six interven-
of droperidol (if assigned) were given intravenously tions as odds ratios and to identify potential two-
within 20 minutes after the start of anesthesia,10,26 or three-factor interactions by a stepwise forward-
and 4 mg of ondansetron (if assigned) was given in- inclusion algorithm. This analysis was repeated
travenously during the last 20 minutes of surgery.27 to compensate for the specified covariates (female
Postoperatively, the patients received supple- sex, nonsmoking status, age, a history of postoper-

n engl j med 350;24 www.nejm.org june 10, 2004 2443

The New England Journal of Medicine

Downloaded from nejm.org on May 15, 2014. For personal use only. No other uses without permission.
Copyright 2004 Massachusetts Medical Society. All rights reserved.
 The new england journal of medicine

ative nausea and vomiting or motion sickness, use tions of interventions, which ranged from 59 per-
of postoperative opioids, type of surgery, and study cent among patients who were given volatile anes-
center). A two-sided P value of less than 0.05 was thesia, nitrous oxide, fentanyl, and no antiemetics
considered to indicate statistical significance. (26 of 44 of these patients had nausea and vomit-
ing) to 17 percent among patients who received pro-
results pofol, nitrogen, remifentanil, ondansetron, dexa-
methasone, and droperidol (17 of 102 of these
Patients were recruited from February 2, 2000, un- patients had nausea and vomiting). Nausea occurred
til July 30, 2002, at 28 centers; 5199 patients under- in 1617 patients (31 percent) and vomiting in 734
went factorial randomization to ondansetron or (14 percent). Among the patients who had symp-
no ondansetron, dexamethasone or no dexameth- toms, the median and mean ratings for the maximal
asone, droperidol or no droperidol, and propofol or nausea level were 5 and 5.7, respectively, and the
a volatile anesthetic. Outcome data were incomplete median and mean numbers of emetic episodes were
for 38 patients, leaving 5161 patients (99 percent) 1 and 1.5, respectively. According to bivariate analy-
for whom complete outcome data were available. ses, each antiemetic reduced the incidence of post-
One center each did not randomize with respect to operative nausea and vomiting by about 26 percent,
carrier gas (424 patients), use of remifentanil or fen- propofol reduced it by about 19 percent, and nitro-
tanyl (191 patients), or both of these factors (181 gen reduced it by about 12 percent (Table 1). The
patients). Three centers randomly assigned a total rates of hypotension, use of intraoperative vasocon-
of 280 patients to 80 percent oxygen in nitrogen (as strictors, and shivering were similar with each anti-
a third alternative to 30 percent oxygen in nitrogen emetic. Propofol was associated with less frequent
or in nitrous oxide). A total of 4123 patients were use of intraoperative vasoconstrictors (15 percent)
thus randomly assigned with respect to all six pri- than were volatile anesthetics (20 percent, P=0.001).
mary factors, and outcome data were incomplete The use of remifentanil rather than fentanyl did not
in 37 of them (1 patient with incomplete data was significantly reduce the incidence of postoperative
among those not randomly assigned with respect nausea and vomiting, but it was associated with in-
to carrier gas), leaving 4086 patients (99 percent) for creased use of intraoperative vasoconstrictors (21
whom complete outcome data could be analyzed percent, vs. 13 percent with fentanyl; P<0.001) and
(Fig. 1). an increased incidence of shivering (6.7 percent,
Of the 5161 patients, 81.5 percent were women, vs. 3.3 percent with fentanyl; P<0.001).
81.2 percent were nonsmokers, 54.5 percent had a Increasing the number of antiemetics adminis-
history of postoperative nausea and vomiting or mo- tered reduced the incidence of postoperative nausea
tion sickness, and 78.1 percent received postopera- and vomiting from 52 percent when no antiemetics
tive opioids. Hernia repair was performed in 2.8 per- were used to 37 percent, 28 percent, and 22 percent
cent of the patients, cholecystectomy in 7.7 percent, when one, two, and three antiemetics, respectively,
hysterectomy in 16.9 percent, thyroid surgery in 5.9 were administered (Fig. 2). This corresponds to a
percent, breast surgery in 2.8 percent, hip replace- 26 percent reduction in the relative risk of nausea
ment in 3.5 percent, knee arthroscopy in 2.2 per- and vomiting for each additional antiemetic used
cent, arm or hand surgery in 2.5 percent, head and (95 percent confidence interval, 23 percent to 30
neck surgery (including ophthalmic surgery) in 9.0 percent). Furthermore, there were no significant
percent, gynecologic surgery other than hysterecto- differences among the antiemetics (chi-square=
my in 28.2 percent, other bone surgery in 6.6 per- 0.01, 2 df; P=1.00) or among any pair of antiemet-
cent, and other types of general surgery in 11.7 per- ics (chi-square=0.42, 2 df; P=0.81).
cent. The baseline characteristics were similar The effects of the anesthetic interventions and
among the patients randomly assigned to each their combinations were explored in the 4086 pa-
intervention; more detailed information can be tients who were randomly assigned with respect to
found in Table S1 of the Supplementary Appendix all six interventions. The average incidence of post-
(available with the full text of this article at www. operative nausea and vomiting was 41 percent
nejm.org). among those given a volatile anesthetic and nitrous
Overall, 1731 of 5161 patients (34 percent) had oxide, 34 percent among those given a volatile an-
postoperative nausea and vomiting. This reflects the esthetic and nitrogen, 32 percent among those given
average incidence among all 64 possible combina- propofol and nitrous oxide, and 29 percent among

2444 n engl j med 350;24 www.nejm.org june 10 , 2004

The New England Journal of Medicine

Downloaded from nejm.org on May 15, 2014. For personal use only. No other uses without permission.
Copyright 2004 Massachusetts Medical Society. All rights reserved.
 Random Outcome Data Outcome Data Analysis of Analysis of
Assignment Incomplete Analyzable Four Factors Six Factors

Yes Yes
N=2591 N=2576
Ondansetron
N=15 N=5161
N=5199
No No
N=2608 N=2585
N=23

Yes Yes
N=2617 N=2596
Dexamethasone
N=21 N=5161
N=5199
No No
N=2582 N=2565
N=17 N=5161
Yes Yes
N=2586 N=2573
Droperidol
N=13 N=5161
N=5199
No No

n engl j med 350;24

N=2613 N=2588
N=25

Propofol Propofol
Randomly N=3455 N=3427
Consented Maintenance
assigned N=28 N=5161 N=4086
N=5262 N=5199 Volatile Volatile
N=5199
anesthetic anesthetic

www.nejm.org
N=1744 N=1734
N=10
Withdrew
N=424
N=63 Nitrogen Nitrogen
N=2164 N=2146
N=18 N=4277

The New England Journal of Medicine

Carrier gas Nitrous oxide Nitrous oxide

june 10, 2004

N=4594 N=2150 N=2131
a factorial trial of antiemetic strategies

N=19
Oxygen Oxygen
N=280 N=280
N=181
N=0

Copyright 2004 Massachusetts Medical Society. All rights reserved.

Yes Remifentanil
N=2403 N=2386
Remifentanil
N=17 N=4789
N=4827
No (fentanyl) Fentanyl
N=2424 N=2403
N=21

Downloaded from nejm.org on May 15, 2014. For personal use only. No other uses without permission.
N=191

Figure 1. Study Design.

According to the factorial design of the study, patients were simultaneously randomly assigned to several interventions. A total of 5199 patients were randomly assigned to four interven-
tions, and data could be analyzed for 5161 patients. One center each was unable to randomize for nitrogen (424 patients), remifentanil or fentanyl (191 patients), or both of these factors
(181 patients). Thus, a total of 4123 patients were randomly assigned according to the factorial design with respect to all six interventions, and data could be analyzed for 4086 patients.

2445
 The new england journal of medicine

Table 1. Risk of Postoperative Nausea and Vomiting According to Patients Randomly Assigned Interventions.

Percent Relative P
Intervention Received Intervention Risk (95% CI)* Value
Yes No
no. with PONV/total no. (%)
Ondansetron (vs. no ondansetron) 735/2576 (28.5) 996/2585 (38.5) 26.0 (31.5 to 19.9) <0.001
Dexamethasone (vs. no dexamethasone) 739/2596 (28.5) 992/2565 (38.7) 26.4 (31.9 to 20.4) <0.001
Droperidol (vs. no droperidol) 742/2573 (28.8) 989/2588 (38.2) 24.5 (30.2 to 18.4) <0.001
Propofol (vs. inhalational anesthetic) 1066/3427 (31.1) 665/1734 (38.4) 18.9 (25.0 to 12.3) <0.001
Nitrogen as carrier gas (vs. nitrous oxide) 668/2146 (31.1) 755/2131 (35.4) 12.1 (19.3 to 4.3) 0.003
Remifentanil (vs. fentanyl) 827/2386 (34.7) 792/2403 (33.0) 5.2 (2.9 to 13.8) 0.21

* CI denotes confidence interval.

P values were calculated by the chi-square test.
The numbers shown are the numbers of patients who had postoperative nausea, vomiting, or both (PONV) divided by
the total numbers of patients randomly assigned to the specified intervention for whom complete outcome data could
be analyzed. The data are based on all 5161 randomly assigned patients who completed the study, with the exceptions of
the data for carrier gas (4277 patients) and for remifentanil versus fentanyl (4789 patients).

those given propofol and nitrogen. Figure 3 shows curred; in contrast, 79 of the 482 men who did not
these incidences broken down according to the receive droperidol had nausea or vomiting (16 per-
number of antiemetics. There was no significant cent), as compared with 80 of the 472 men who did
interaction between propofol and nitrogen (chi- receive this agent (17 percent) (odds ratio, 1.04; 95
square=0.94, 2 df, by the likelihood ratio test; percent confidence interval, 0.74 to 1.46; P=0.82).
P=0.33). Although the type of volatile anesthetic The results based on analyses of data from 4086
(isoflurane, sevoflurane, or desflurane) was not a patients remained essentially unchanged when data
randomized factor, it had no significant effect on from all 5161 patients were considered or when po-
the incidence of postoperative nausea and vomiting tential confounders were included in the statistical
in a multivariate model (P=0.30). The incidence of models (Table 2). Detailed results for the 4086 pa-
postoperative nausea and vomiting was 31 percent tients in the 64 groups are given in Table S2 of the
among the patients who received 80 percent oxygen Supplementary Appendix. Given the finding that to-
in nitrogen and 24 percent among those who re- tal intravenous anesthesia or the use of any anti-
ceived 30 percent oxygen in nitrogen (P=0.07). emetic independently reduced the risk of postoper-
Multivariate logistic analyses of data from all ative nausea and vomiting by about 26 percent, the
5161 patients and of data from the 4086 patients incidence of postoperative nausea and vomiting for
assigned with respect to all six treatments are shown five different initial risks was calculated for as many
in Table 2. This analysis found no significant inter- as four interventions (Table 3).
actions among the treatments. When the interac-
tions between treatments and potentially confound- discussion
ing factors (e.g., the type of surgery) were analyzed,
only one significant interaction was detected: an in- The large enrollment and the factorial design of our
teraction between droperidol and sex (P=0.003). trial allowed simultaneous evaluation of the anti-
Droperidol significantly reduced the risk of post- emetic efficacy of three antiemetic interventions and
operative nausea and vomiting among women, but three anesthetic interventions and of all possible
not among men: 910 of the 2106 women who did combinations of two or three interventions. All the
not receive droperidol had nausea or vomiting (43 tested antiemetics appeared to be similarly effective.
percent), as compared with 662 of the 2101 women Ondansetron and other 5-hydroxytryptamine type 3
who did receive this agent (32 percent) (odds ratio, antagonists are considered relatively safe, but they
0.61; 95 percent confidence interval, 0.53 to 0.69; are more expensive than droperidol and dexameth-
P<0.001), and the effect was independent of men- asone. However, low-dose droperidol can cause dys-
strual-cycle phase or whether menopause had oc- phoria,28,29 and the Food and Drug Administration

2446 n engl j med 350;24 www.nejm.org june 10 , 2004

The New England Journal of Medicine

Downloaded from nejm.org on May 15, 2014. For personal use only. No other uses without permission.
Copyright 2004 Massachusetts Medical Society. All rights reserved.
 a factorial trial of antiemetic strategies

recently added a black box warning to the drugs

60
labeling to indicate that it may be associated with
torsade de pointes; however, there is little evidence
that antiemetic doses trigger this condition.30 No 50

Incidence of Postoperative
Nausea and Vomiting (%)
studies have identified complications associated
with the antiemetic dose of dexamethasone, al- 40

though even meta-analyses may have insufficient

power to detect rare complications.31 The combi- 30 Ond Dex Dro
nation of low cost and apparent safety makes dexa-
methasone at a dose of 4 mg an attractive first-line 20 Ond, Ond, Dex,
Dex Dro Dro
agent for prophylaxis against postoperative nausea
and vomiting. 10
Bivariate analysis indicated that substituting
propofol for a volatile anesthetic reduced the risk 0
0 1 2 3
of postoperative nausea and vomiting by about 19
No. of Antiemetics
percent, whereas substituting nitrogen for nitrous
oxide reduced the risk by about 12 percent. Com-
Figure 2. Incidence of Postoperative Nausea and Vomiting Associated
bining these two anesthetic management strategies with the Various Combinations of Antiemetic Drugs.
(i.e., total intravenous anesthesia) thus reduced the The data shown represent outcomes in 5161 patients. Solid circles represent
risk by about as much as any single antiemetic. In the average value for each number of prophylactic antiemetics, and open
contrast, the use of remifentanil instead of fentanyl symbols the incidence for each antiemetic or combination of antiemetics.
Ond denotes ondansetron, Dex dexamethasone, and Dro droperidol. I bars
did not significantly reduce the risk of nausea and represent 95 percent confidence intervals.
vomiting.
The relative risk reduction associated with each
intervention was apparently independent for a wide
range of absolute risks. Thus, interventions that re-
duce the relative risk to a similar extent will provide Volatile Volatile
Anesthetic Anesthetic Propofol Propofol
the greatest absolute risk reduction in patients most and N2O and Air and N2O and Air
likely to have postoperative nausea and vomiting. 70

For example, a single intervention in a patient with

60
an 80 percent risk of postoperative nausea and
Incidence of Postoperative

vomiting will reduce the risk to 59 percent; the abso-

Nausea and Vomiting (%)

50
lute risk reduction is 21 percent, which translates
into a number needed to treat of about five to pre- 40
vent nausea and vomiting in one patient. Conversely,
the absolute risk reduction in a patient with a base- 30

line risk of 10 percent is only about 3 percent; this

20
corresponds to a number needed to treat of about
40, which would probably not justify the expense 10
and risk of prophylactic treatment. The efficacy of
an intervention thus depends critically on patients 0
baseline risk. 0 1 2 3 0 1 2 3 0 1 2 3 0 1 2 3

Interestingly, there were no significant interac- No. of Antiemetics

tions among the antiemetic interventions, among
Figure 3. Incidence of Postoperative Nausea and Vomiting According to
the anesthetic interventions, or among the anti- the Combinations of Anesthetic Strategies and the Number of Antiemetic
emetics and the anesthetics. The resulting relative Treatments Given.
risk of nausea and vomiting associated with a com- The data shown represent outcomes in 4086 patients. Because assignment to
bination of interventions can thus be directly calcu- remifentanil did not contribute significantly to the risk of postoperative nau-
sea and vomiting, data pertaining to remifentanil are not shown. Although the
lated as the product of the individual relative risks. graph suggests that there is an interaction between nitrogen (air) and propo-
As a consequence, the absolute risk reduction pro- fol in patients who received no antiemetic agents, statistical analyses did not
vided by a second or third intervention is less than confirm this impression. I bars represent 95 percent confidence intervals.
that provided by the initial intervention (irrespective

n engl j med 350;24 www.nejm.org june 10, 2004 2447

The New England Journal of Medicine

Downloaded from nejm.org on May 15, 2014. For personal use only. No other uses without permission.
Copyright 2004 Massachusetts Medical Society. All rights reserved.
 The new england journal of medicine

Table 2. Results of Multiple Logistic-Regression Analysis and Odds Ratios for Postoperative Nausea and Vomiting.*

Patients Assigned with Respect Patients Assigned with Respect

to Four Interventions to Six Interventions
Variable (N=5161) (N=4086)

Odds Ratio (95% CI) P Value Odds Ratio (95% CI) P Value
Intervention
Ondansetron 0.56 (0.500.64) <0.001 0.56 (0.480.65) <0.001
Dexamethasone 0.57 (0.500.65) <0.001 0.57 (0.490.66) <0.001
Droperidol 0.58 (0.510.67) <0.001 0.56 (0.480.66) <0.001
Propofol 0.69 (0.600.79) <0.001 0.71 (0.610.83) <0.001
Carrier gas
Nitrogen (30% oxygen) 0.81 (0.700.93) 0.003 0.83 (0.720.97) 0.02
Nitrogen (80% oxygen) 0.99 (0.701.40) 0.96
Remifentanil 0.96 (0.841.10) 0.56 0.94 (0.811.09) 0.39
Covariate
Female sex 3.13 (2.334.20) <0.001 2.87 (2.083.95) <0.001
Interaction of droperidol and male sex 1.85 (1.262.72) 0.002 1.97 (1.293.00) 0.002
Nonsmoking 1.57 (1.321.87) <0.001 1.57 (1.291.91) <0.001
History of postoperative nausea and 1.70 (1.491.95) <0.001 1.80 (1.542.09) <0.001
vomiting or motion sickness
Operation
Hernia repair 1.04 (0.651.67) 0.87 1.08 (0.661.78) 0.75
Cholecystectomy 1.49 (1.082.06) 0.015 1.44 (1.012.05) 0.04
Hysterectomy 1.78 (1.352.35) <0.001 1.94 (1.432.63) <0.001
Thyroid surgery 1.22 (0.861.72) 0.27 1.23 (0.851.78) 0.27
Breast surgery 0.74 (0.481.17) 0.20 0.76 (0.461.24) 0.27
Orthopedic surgery 0.91 (0.671.24) 0.55 0.91 (0.651.29) 0.60
Head and neck surgery 1.08 (0.711.65) 0.72 0.88 (0.551.42) 0.61
Other gynecologic surgery 0.91 (0.691.19) 0.48 0.98 (0.731.32) 0.91
Duration of anesthesia (per hr) 1.20 (1.121.28) <0.001 1.15 (1.061.24) 0.001
Use of postoperative opioids 2.14 (1.752.61) <0.001 2.01 (1.592.53) <0.001

* The data are presented as odds ratios, which describe the effects of the interventions or covariates as compared with the
effects when the intervention or covariate is absent. CI denotes confidence interval.
Analyses were adjusted for the study center as a potentially confounding factor.

of which combination is chosen). A 70 percent re- solute risk, but to a lesser extent than will the initial
duction in the relative risk of postoperative nausea intervention. Combining prophylactic interventions
and vomiting is thus the best that can be expected, therefore markedly increases costs and the likeli-
even when total intravenous anesthesia is used in hood of adverse effects while providing progressive-
combination with three antiemetics. ly less additional absolute benefit. Multiple interven-
Because each tested antiemetic agent and the tions should thus generally be reserved for patients
use of total intravenous anesthesia reduced the rel- at high risk for postoperative nausea and vomiting
ative risk of nausea and vomiting to a similar extent, or those in whom nausea and vomiting would be es-
the logical sequence is to use the least expensive or pecially dangerous.
safest intervention first. Additional interventions In analyses based on the entire study population,
that cost more or that are associated with a greater droperidol decreased the risk of postoperative nau-
chance of adverse effects will further reduce the ab- sea and vomiting as much as did the other anti-

2448 n engl j med 350;24 www.nejm.org june 10 , 2004

The New England Journal of Medicine

Downloaded from nejm.org on May 15, 2014. For personal use only. No other uses without permission.
Copyright 2004 Massachusetts Medical Society. All rights reserved.
 a factorial trial of antiemetic strategies

emetics, but when sex was considered, no signifi-

Table 3. Estimated Incidence of Postoperative Nausea and Vomiting
cant benefit was found in men. Such a finding has as a Function of Baseline Risk, on the Basis of the Assumption That Each
not been described in previous studies, presumably Intervention Reduces the Relative Risk by 26 Percent.
because many studies have been restricted to wom-
Baseline Risk
en and because studies that included both sexes (No Intervention)* Estimated Incidence of Postoperative Nausea and Vomiting
were too small to detect the interaction. Estrogen
or other hormonal factors seem unlikely to be the One Two Three Four
Intervention Interventions Interventions Interventions
cause, since the effectiveness of droperidol was in-
percent
dependent from menstrual-cycle phase and meno-
10% 7 5 4 3
pause (data not shown). It is possible that dopamine
is a more important trigger in women than in men. 20% 15 11 8 6
It is also possible that the lack of efficacy of droperi- 40% 29 22 16 12
dol in men is simply a spurious finding resulting 60% 44 33 24 18
from multiple testing. 80% 59 44 32 24
It is well known that the incidence of postoper-
ative nausea and vomiting varies considerably ac- * The baseline risk levels of 10 percent, 20 percent, 40 percent, 60 percent, and
80 percent reflect the presence of 0, 1, 2, 3, and 4 risk factors, respectively, ac-
cording to the type of surgery conducted. However, cording to a simplified risk score.17
with the exception of hysterectomy and possibly
cholecystectomy, the relative risk was similar for all
types of surgery when corrected for major risk fac- operative nausea and vomiting and to reserve sero-
tors including sex, nonsmoking status, a history of tonin antagonists as a rescue treatment.
postoperative nausea and vomiting, and the use of Supported by a grant (1518 TG 72) from the Klinik fr Anaesthe-
postoperative opioids. As a consequence, risk mod- siologie, Julius-Maximilians Universitt, Wrzburg, Germany; a
Helsinki University Central Hospital State Allocation grant (TYH
els that include the type of surgery1,32 do not pro- 0324) from Helsinki University Central Hospital, University of
vide greater predictive power than a simplified mod- Helsinki, Helsinki, Finland; AstraZeneca, Wedel, Germany;
el.18,19 Since no interactions were detected between GlaxoSmithKline, Hamburg, Germany; the Gheens Foundation,
Louisville, Ky.; the Joseph Drown Foundation, Los Angeles; the
the interventions and the type of surgery, it is not Commonwealth of Kentucky Research Challenge Trust Fund, Lou-
necessary to repeat studies of postoperative nausea isville, Ky.; and a Health Grant (GM 061655) from the National In-
stitutes of Health, Bethesda, Md.
and vomiting for various types of surgery.14,33 Presented in part at meetings of the European Society of Anaes-
Management techniques such as total intrave- thesiologists, Glasgow, Scotland, June 2003; the German Society of
nous anesthesia cannot be used once postoperative Anesthesiologists, Munich, Germany, April 2003; and the American
nausea and vomiting have begun. Dexamethasone, Society of Anesthesiologists, San Francisco, October 2003.
Dr. Apfel reports having served as a paid speaker for and having
similarly, prevents postoperative nausea and vomit- received grant support from AstraZeneca (Wedel, Germany) and
ing only when given near the beginning of surgery, GlaxoSmithKline (Munich, Germany), having served as a consult-
probably by reducing surgery-induced inflamma- ant for GlaxoSmithKline, and having served as a consultant for Merck
(Whitehouse, N.Y.) and Helsinn Healthcare (Lugano, Switzerland)
tion.34 Moreover, rescue treatments are ineffec- with respect to antiemetics unrelated to the products discussed in
tive when the same drug has already been used pro- this article. Dr. Korttila reports having served as a consultant for
phylactically.35 Postoperative treatment options are Pharmacia/Pfizer (New London, Conn.) and Helsinn Healthcare
and as a paid speaker for Pharmacia/Pfizer. Dr. Apfel and Dr. Sessler
thus limited when compared with the broader range report having received a grant from Pharmacia/Pfizer for a grant-
of prophylactic options, suggesting that prophylax- writing workshop. Dr. Sessler reports having received grants from
is may be preferable to the treatment of established Progenics (Tarrytown, N.Y.), Cardinal Health (McGraw Park, Ill.),
Pfizer (New London, Conn.), Ocean Optics (Dunedin, Fla.), and
postoperative nausea and vomiting. A reasonable Scott Laboratories (Lubbock, Tex.) for work unrelated to the field of
treatment strategy would be to use dexamethasone antiemetic interventions.
and total intravenous anesthesia as first-line and forWe are indebted to Nancy L. Alsip, Ph.D., University of Louisville,
editorial assistance.
second-line methods of prophylaxis against post-

ap p e n d i x
The IMPACT investigators are as follows: Steering Committee C.C. Apfel (Outcomes Research Institute and Department of Anesthesiology,
University of Louisville, Louisville, Ky., and the Klinik und Poliklinik fr Anaesthesiologie, Universitt Wrzburg, Wrzburg, Germany); K.
Korttila (Department of Anesthesiology and Intensive Care, University of Helsinki, Helsinki, Finland); and A. Biedler (Klinik fr Ansthesi-
ologie und Intensivmedizin, Universittskliniken des Saarlandes, Homburg, Germany). Data Management and Monitoring C.C. Apfel, E.
Kaufmann, M. Kredel, A. Schmelzer, and J. Wermelt (Klinik und Poliklinik fr Anaesthesiologie, Universitt Wrzburg, Wrzburg, Germa-
ny); and G. Link (Database Engineering, Rimpar, Germany). Manuscript Preparation and Data Analyses C.C. Apfel (Outcomes Research In-
stitute and Department of Anesthesiology, University of Louisville, Louisville, Ky., and the Klinik und Poliklinik fr Anaesthesiologie, Uni-

n engl j med 350;24 www.nejm.org june 10, 2004 2449

The New England Journal of Medicine

Downloaded from nejm.org on May 15, 2014. For personal use only. No other uses without permission.
Copyright 2004 Massachusetts Medical Society. All rights reserved.
 The new england journal of medicine

versitt Wrzburg, Wrzburg, Germany); D.I. Sessler (Outcomes Research Institute and Departments of Anesthesiology and
Pharmacology, University of Louisville, Louisville, Ky.); S.J. Pocock and M. Abdalla (Statistics Unit, London School of Hygiene and Tropical
Medicine, London). Site Investigators A. Turan (Department of Anesthesiology and Reanimation, Trakya University, Edirne, Turkey); E.
Kaufmann, P. Kranke, M. Kredel, N. Roewer, A. Schmelzer, and J. Wermelt (Klinik und Poliklinik fr Ansthesiologie, Julius-Maximilians
Universitt, Wrzburg, Germany); R.M. Jokela, A. Soikkeli, and K. Korttila (Department of Anesthesiology and Intensive Care, Helsinki
University Central Hospital, Helsinki, Finland); C. Isselhorst, B. Fritz, and H. Kerger (Klinik fr Ansthesiologie und Intensivmedizin, Uni-
versittsklinik Mannheim, Mannheim, Germany); O. Danzeisen, C. Heringhaus, I. Schramm, and S. Spieth (Department of Anesthesiolo-
gy, Universitt Freiburg, Freiburg, Germany); L. Eberhart and K. Werthwein (Department of Anesthesiology, Universitt Ulm, Ulm, Germa-
ny); W. Leidinger, J.N. Meierhofer, U. Ruppert, and C. Zernak (Abteilung fr Ansthesiologie, Operative Intensivmedizin, und
Blutprodukte, Kreiskrankenhaus Garmisch-Partenkirchen, Garmisch-Partenkirchen, Germany); A. Bacher (Klinik fr Ansthesie und All-
gemeine Intensivmedizin, Universitt Wien, Vienna); H. Bartsch and H. Forst (Department of Anesthesiology and Surgical Intensive Care,
Zentralklinikum Augsburg, Augsburg, Germany); B. Book, W. Hoeltermann, and C. Prause (Department of Anesthesiology, Klinikum Lin-
gen, Lingen, Germany); E. Palencikova and S. Trenkler (Department of Anesthesiology, Reiman University Hospital, Presov, Slovakia); H.
Bause, H. Bordon, and K. Stoecklein (Department of Anesthesiology, Allgemeines Krankenhaus Altona, Hamburg-Altona, Germany); F.
Bach, D. Buschmann, F. Mertzlufft, and I. Vedder (Klinik fr Ansthesiologie und Operative Intensivmedizin, von Bodelschwingsche An-
stalten Bethel, Bielefeld, Germany); C. Frenkel and A. Paura (Department of Anesthesiology, Klinikum Lneburg, Lneburg, Germany); K.
Danner, C. Madler, and B. Steinbrecher (Institut fr Ansthesiologie und Notfallmedizin, Westpfalz-Klinikum, Kaiserslautern, Germany);
A. Kimmich, E. Schneider, and M. Trick (Department of Anesthesiology, Universittsklinik Tbingen, Tbingen, Germany); A. Biedler, D.
Detzel, and W. Wilhelm (Klinik fr Ansthesiologie und Intensivmedizin, Universittskliniken des Saarlandes, Homburg, Germany); M.
Koivuranta (Department of Anesthesiology, Central Lapland Hospital, Rovaniemi, Finland); M. Hinojosa, M. Lucas, S. Muoz, R. Rincon,
and P. Vila (Department of Anesthesiology, Hospital Universitario Germans Trias i Pujol, Badalona, Barcelona, Spain); M. Hergert and F.
Liebenow (Department of Anesthesiology and Intensive Care, Klinikum Schwerin, Schwerin, Germany); H.-B. Hopf and S. Pohl (Depart-
ment of Anesthesiology, Kreisklinik Langen, Langen, Germany); G. Frings (Anesthesiology Unit, Wedau-Kliniken, Duisburg); G. Fritz and
C. Hoehne (Department of Anesthesiology, Charit, Campus Virchow-Klinikum, Berlin); H. Feierfeil and J. Motsch (Department of Anes-
thesiology, Universitt Heidelberg, Heidelberg, Germany); A. Goebel (Department of Anesthesiology, Eichhof Krankenhaus, Lauterbach,
Germany); S. Alahuhta, T. Kangas-Saarela, and P. Karjaleinen (Department of Anesthesiology, Oulu University Hospital, Oulu, Finland); R.
Sneyd (Department of Anaesthetics, Derriford Hospital, Plymouth, United Kingdom); and U. Koschel and M. Lange (Department of Anes-
thesiology, Waldkrankenhaus Rudolf Elle, Eisenberg, Germany).

refer enc es
1. Gan TJ. Postoperative nausea and vom- ment of postoperative nausea and vomiting. A, Roewer N. Comparison of predictive
iting can it be eliminated? JAMA 2002; Anesthesiology 2000;92:931-3. models for postoperative nausea and vomit-
287:1233-6. 12. Apfel CC, Kranke P, Katz MH, et al. Vol- ing. Br J Anaesth 2002;88:234-40.
2. Kovac AL. Prevention and treatment of atile anaesthetics may be the main cause for 20. Apfel CC, Korttila K, Abdalla M, et al.
postoperative nausea and vomiting. Drugs early but not delayed postoperative nausea An international multicenter protocol to
2000;59:213-43. and vomiting: a randomized controlled trial assess the single and combined benefits of
3. Watcha MF. Postoperative nausea and of factorial design. Br J Anaesth 2002;88: antiemetic interventions in a controlled clini-
emesis. Anesthesiol Clin North Am 2002; 659-68. cal trial of a 2x2x2x2x2x2 factorial design
20:709-22. 13. Divatia JV, Vaidya JS, Badwe RA, Hawal- (IMPACT). Control Clin Trials 2003;24:736-
4. Koivuranta M, Laara E, Snare L, Alahuh- dar RW. Omission of nitrous oxide during 51.
ta S. A survey of postoperative nausea and anesthesia reduces the incidence of postop- 21. Sneyd JR, Carr A, Byrom WD, Bilski AJ.
vomiting. Anaesthesia 1997;52:443-9. erative nausea and vomiting: a meta-analy- A meta-analysis of nausea and vomiting fol-
5. Macario A, Weinger M, Carney S, Kim A. sis. Anesthesiology 1996;85:1055-62. lowing maintenance of anaesthesia with
Which clinical anesthesia outcomes are im- 14. Tramer MR. A rational approach to the propofol or inhalational agents. Eur J Anaes-
portant to avoid? The perspective of pa- control of postoperative nausea and vomit- thesiol 1998;15:433-45.
tients. Anesth Analg 1999;89:652-8. ing: evidence from systematic reviews. I. Ef- 22. Greif R, Laciny S, Rapf B, Hickle RS,
6. Darkow T, Gora-Harper ML, Goulson ficacy and harm of antiemetic interventions, Sessler DI. Supplemental oxygen reduces the
DT, Record KE. Impact of antiemetic selec- and methodological issues. Acta Anaesthe- incidence of postoperative nausea and vom-
tion on postoperative nausea and vomiting siol Scand 2001;45:4-13. iting. Anesthesiology 1999;91:1246-52.
and patient satisfaction. Pharmacotherapy 15. Scuderi PE, James RL, Harris L, Mims 23. Goll V, Akca O, Greif R, et al. On-
2001;21:540-8. GR III. Multimodal antiemetic management dansetron is no more effective than supple-
7. Schumann R, Polaner DM. Massive sub- prevents early postoperative vomiting after mental intraoperative oxygen for prevention
cutaneous emphysema and sudden airway outpatient laparoscopy. Anesth Analg 2000; of postoperative nausea and vomiting.
compromise after postoperative vomiting. 91:1408-14. Anesth Analg 2001;92:112-7.
Anesth Analg 1999;89:796-7. 16. Gan TJ, Meyer T, Apfel CC, et al. Con- 24. Purhonen S, Turunen M, Ruohoaho U-M,
8. Bremner WG, Kumar CM. Delayed sur- sensus guidelines for management of post- Niskanen M, Hynynen M. Supplemental oxy-
gical emphysema, pneumomediastinum and operative nausea and vomiting. Anesth An- gen does not reduce the incidence of post-
bilateral pneumothoraces after postopera- alg 2003;97:62-71. operative nausea and vomiting after ambula-
tive vomiting. Br J Anaesth 1993;71:296-7. 17. Apfel CC, Laara E, Koivuranta M, Greim tory gynecologic laparoscopy. Anesth Analg
9. Gold BS, Kitz DS, Lecky JH, Neuhaus CA, Roewer N. A simplified risk score for 2003;96:91-6.
JM. Unanticipated admission to the hospital predicting postoperative nausea and vomit- 25. Joris JL, Poth NJ, Djamadar AM, et al.
following ambulatory surgery. JAMA 1989; ing: conclusions from cross-validations be- Supplemental oxygen does not reduce post-
262:3008-10. tween two centers. Anesthesiology 1999;91: operative nausea and vomiting after thy-
10. Hill RP, Lubarsky DA, Phillips-Bute B, et 693-700. roidectomy. Br J Anaesth 2003;91:857-61.
al. Cost-effectiveness of prophylactic anti- 18. Pierre S, Benais H, Pouymayou J. Apfels 26. Liu K, Hsu CC, Chia YY. The effect of
emetic therapy with ondansetron, droperi- simplified score may favourably predict the dose of dexamethasone for antiemesis after
dol, or placebo. Anesthesiology 2000;92: risk of postoperative nausea and vomiting. major gynecological surgery. Anesth Analg
958-67. Can J Anaesth 2002;49:237-42. 1999;89:1316-8.
11. Watcha MF. The cost-effective manage- 19. Apfel CC, Kranke P, Eberhart LHJ, Roos 27. Tang J, Wang B, White PF, Watcha MF,

2450 n engl j med 350;24 www.nejm.org june 10 , 2004

The New England Journal of Medicine

Downloaded from nejm.org on May 15, 2014. For personal use only. No other uses without permission.
Copyright 2004 Massachusetts Medical Society. All rights reserved.
 a factorial trial of antiemetic strategies

Qi J, Wender RH. The effect of timing of on- 30. Gan TJ, White PF, Scuderi PE, Watcha study postoperative nausea and vomiting.
dansetron administration on its efficacy, MF, Kovac A. FDA black box warning re- Acta Anaesthesiol Scand 2002;46:921-8.
cost-effectiveness, and cost-benefit as a pro- garding use of droperidol for postoperative 34. Wang JJ, Ho ST, Tzeng JI, Tang CS. The
phylactic antiemetic in the ambulatory set- nausea and vomiting: is it justified? Anes- effect of timing of dexamethasone adminis-
ting. Anesth Analg 1998;86:274-82. thesiology 2002;97:287. tration on its efficacy as a prophylactic anti-
28. Melnick B, Sawyer R, Karambelkar D, 31. Henzi I, Walder B, Tramer MR. Dexa- emetic for postoperative nausea and vomit-
Phitayakorn P, Uy NT, Patel R. Delayed side methasone for the prevention of postopera- ing. Anesth Analg 2000;91:136-9.
effects of droperidol after ambulatory gen- tive nausea and vomiting: a quantitative 35. Kovac AL, OConnor TA, Pearman MH,
eral anesthesia. Anesth Analg 1989;69:748- systematic review. Anesth Analg 2000;90: et al. Efficacy of repeat intravenous dosing
51. 186-94. of ondansetron in controlling postoperative
29. Lim BS, Pavy TJ, Lumsden G. The anti- 32. Sinclair DR, Chung F, Mezei G. Can nausea and vomiting: a randomized, dou-
emetic and dysphoric effects of droperidol postoperative nausea and vomiting be pre- ble-blind, placebo-controlled multicenter
in the day surgery patient. Anaesth Intensive dicted? Anesthesiology 1999;91:109-18. trial. J Clin Anesth 1999;11:453-9.
Care 1999;27:371-4. 33. Apfel CC, Roewer N, Korttila K. How to Copyright 2004 Massachusetts Medical Society.

n engl j med 350;24 www.nejm.org june 10, 2004 2451

The New England Journal of Medicine

Downloaded from nejm.org on May 15, 2014. For personal use only. No other uses without permission.
Copyright 2004 Massachusetts Medical Society. All rights reserved.