A quantitative systematic review of ondansetron in treatment of established postoperative nausea and vomitingBMJ 1997; 314 doi: https://doi.org/10.1136/bmj.314.7087.1088 (Published 12 April 1997) Cite this as: BMJ 1997;314:1088
- Martin R Tramèr (), research fellowa,
- R Andrew Moore, consultant biochemista,
- D John M Reynolds, consultant clinical pharmacologistb,
- Henry J McQuay, clinical reader in pain reliefa
- a Pain Research Nuffield Department of Anaesthetics Churchill Oxford Radcliffe Hospital Oxford OX3 7LJ
- b Department of Clinical Pharmacology Radcliffe Infirmary Oxford OX2 6HE
- Correspondence to: Dr Tramèr
- Accepted 16 January 1997
Objectives: To test the evidence for a dose-response with ondansetron for treatment of postoperative nausea and vomiting and to establish whether differences in efficacy between doses are of clinical relevance.
Design: Quantitative systematic review of published randomised controlled trials.
Data sources: Seven trials from 1991 to January 1996 retrieved from a systematic literature search (Medline, reference lists, hand searching of anaesthetic journals, manufacturer's database); no restriction on language.
Main outcome measures: Estimation of efficacy (incidence of complete control of further nausea and vomiting) by using odds ratios and the “number needed to treat” method for early (within 6 hours of administration) and late (within 24 hours) periods.
Results: Four placebo controlled trials with 1043 patients studied intravenous ondansetron 1 mg, 4 mg, or 8 mg. All doses were more efficacious than placebo in preventing further episodes of nausea or vomiting. For combined data, the point estimates for the number needed to treat were between 3.1 (8 mg) and 3.8 (1 mg) for early efficacy and between 4.1 (8 mg) and 4.8 (1 mg) for late efficacy, without significant differences between doses. No difference was found between ondansetron and droperidol in two trials with 129 patients or between ondansetron and metoclopramide in one trial with 80 patients.
Conclusions: Further nausea and vomiting could be prevented with ondansetron compared with placebo in 25% of patients who had nausea or vomiting (number needed to treat, about 4). There was no evidence of a clinically relevant dose-response between 1 mg and 8 mg or a difference between ondansetron and either droperidol or metoclopramide in a limited dataset. A false impression of ondansetron's efficacy may arise because a quarter of all relevant published reports are duplicates, and reporting of study results is uncritical.
Little information exists on the efficacy of anti-emetic interventions in patients with established postoperative nausea and vomiting
To evaluate the effectiveness of ondansetron in this setting we conducted a quantitative systematic review of all relevant published randomised controlled trials
Four trials (1043 patients) compared intravenous ondansetron 1 mg, 4 mg, or 8 mg with placebo, two trials (129 patients) compared ondansetron with droperidol, and one trial (80 patients) compared ondansetron with metoclopramide
All three tested doses of ondansetron were more efficacious than placebo. There was no evidence of a clinically relevant dose-response between 1 mg and 8 mg (number needed to treat to prevent further nausea or vomiting was about 4), or a difference between ondansetron and either droperidol or metoclopramide.
Stopping further postoperative nausea and vomiting in 25% of the patients may be the best that can be achieved currently
Postoperative nausea and vomiting are unpleasant complications of surgery and anaesthesia. Although much attention has been paid to the prevention of these conditions during the past three decades,1 2 3 4 5 little information exists on the efficacy of anti-emetic interventions in patients with established postoperative nausea and vomiting.
The first clinical trials in 1991 showed that a single intravenous dose of ondansetron 8 mg was an efficacious anti-emetic compared with placebo in treating postoperative nausea and vomiting.6 7 Reports of multicentre trials, with data on hundreds of patients and comparing different doses of ondansetron with placebo, concluded that intravenous ondansetron 4 mg was the optimal dose for treating established postoperative nausea and vomiting.8 9
We tested the evidence for a dose-response with ondansetron for treatment of postoperative nausea and vomiting and aimed to establish whether differences in efficacy between doses are of clinical relevance.
We searched Medline (date of search 22 January 1996) back to 1991 for randomised controlled trials that evaluated the effect of ondansetron compared with a control (placebo, no treatment, or another anti-emetic) on established postoperative nausea and vomiting and reported the outcome in dichotomous form. The search was not restricted to the English language and used the combination (ondansetron and human and (emesis or nausea or vomiting)) not (chemotherapy or cancer). We identified additional reports from reference lists of retrieved reports and from review articles of postoperative nausea and vomiting and ondansetron and from hand searching locally available anaesthesia journals. We compared our database with the database of published trials provided by the manufacturer of ondansetron. We did not search for unpublished trials or consider abstracts. We did not analyse efficacy data for ondansetron as prophylaxis against postoperative nausea and vomiting.
Scoring and extraction of data
Each report was read by three of the authors independently to assess adequacy of randomisation and blinding and to assess description of withdrawals.10 These three authors met to agree consensus. Reports that were described as randomised were given one point, plus a further point if the method of randomisation was described and adequate (such as a table of random numbers). There had been an earlier agreement that trials without randomisation or with an inadequate randomisation method (without concealment of treatment allocation) would be excluded from further analysis. Reports that were described as blinded were given one point, plus a further point if the method of blinding was described and adequate (such as identical ampoules). Reports that described the number of and reasons for withdrawals were given one point. Thus the minimum score of an included randomised controlled trial was 1, the maximum score 5.
When origin of data was unclear in reviewed articles, we wrote to the principal authors for information about duplicate publication.
We took information about patients, dose and route of administration of ondansetron and control treatments, anaesthetics, surgery, incidence of postoperative nausea and vomiting in the studied population before randomisation, and study endpoints from each included report. The endpoint indicating a treatment success that was closest to complete control of postoperative nausea and vomiting (absence of further nausea or vomiting, or of both, after treatment) was extracted in dichotomous form. The incidence of this endpoint was treated as the success rate with ondansetron or control. When success rates were reported at different times after administration of ondansetron, the times nearest to the 6th and the 24th hour were used for extraction of cumulative results. Estimates of efficacy during the two time periods (0 to 6 hours and 0 to 24 hours after administration of the ondansetron) were used as indicators of early and late efficacy, respectively. Post hoc analyses, stratified data analyses (according to sex, for example), different grades of nausea, number of vomiting episodes, or number of patients needing anti-emetic rescue treatment were not considered.
The scatter of success rates with ondansetron against success rates with control11 was used as a graphical means of exploring consistency of ondansetron's efficacy and the homogeneity of the data. On such plots a scatter lying predominantly between the line of equality and the axis of the active intervention (ondansetron) would suggest consistent efficacy with the intervention, and relative homogeneity.
Significance and clinical relevance of ondansetron's efficacy compared with control were evaluated with odds ratios and number needed to treat methods12 respectively. Calculations were done by combining ondansetron arms for each dose separately, and corresponding control arms. This means that data from patients receiving placebo from studies using several different doses of ondansetron could be included in several analyses. Odds ratios were estimated with 95% confidence intervals with a fixed effect model.13 A significant improvement of ondansetron over control was assumed when the lower 95% confidence limit of the odds ratio was >1. Point estimates and 95% confidence limits of the number needed to treat were calculated.14 The number needed to treat indicated how many patients with vomiting and nausea have to be treated with ondansetron to achieve complete control of postoperative nausea and vomiting–that is, to prevent any further nausea or vomiting, or both, in one of them, who would otherwise have had further postoperative nausea and vomiting with control treatment. Absence of a significant difference between different doses of ondansetron was assumed when the 95% confidence intervals of the corresponding odds ratio or number needed to treat overlapped.
Calculations were performed with excel version 5.0 on a Power Macintosh 7100/66.
Nine randomised controlled trials were found in eight reports.6 7 8 9 15 16 17 18 Results from one multicentre trial with data from 500 patients treated with three different doses of ondansetron compared with placebo8 were assumed to have been published on two later occasions, in 199318 and in 1994 (first study).9 All contacted authors confirmed that one single dataset had been reported in three publications. Only data from the first publication8 was analysed for the purpose of this systematic review. Data of an abstract from a scientific meeting,19 which were identical to the second part of a full paper publication (second study),9 were not analysed. No other report was excluded from analysis. All trials except two16 17 were sponsored by the manufacturer of ondansetron.
End points and quality score
The remaining seven trials had a median score of 3 (range 2 to 4). The average incidence of postoperative nausea and vomiting before randomisation (before treatment was given) was 36% (22-46%). Four trials compared a single intravenous dose of ondansetron 1 mg, 4 mg, or 8 mg with placebo in 1043 adults (859 females) who complained of nausea or vomited after general anaesthesia.6 7 8 9 In one trial with 100 gynaecology patients intravenous ondansetron 8 mg was compared with intravenous droperidol 1.25 mg; both anti-emetics could be administered up to three times in 24 hours.15 In one trial 29 vomiting children received either ondansetron 100 μg/kg or droperidol 20 μg/kg intravenously.17 In one trial with 80 patients undergoing major abdominal surgery intravenous ondansetron 4 mg was compared with intravenous metoclopramide 10 mg.16
No recurrence of vomiting was the analysed endpoint in one trial.7 In all other trials complete control of postoperative nausea and vomiting was the analysed endpoint. Early (short term) efficacy (within 6 hours) of ondansetron was reported on five occasions.6 7 8 9 10 11 12 13 14 15 16 17 Late (long term) efficacy (within 24 hours) was reported in two placebo controlled trials8 9 and one trial with metoclopramide as the control.16 Data extracted from these reports are available from the worldwide web (http://www.jr2.ox.ac.uk/Bandolier/painres/ondR/ondR.html).
The success rate scatter, exploring the incidence of treatment success with ondansetron and placebo, suggested homogeneity of data and consistent efficacy (both early and late) of ondansetron compared with placebo, and with no obvious dose-response (fig 1).
Odds ratios showed a significant difference between each of the three doses of ondansetron and placebo for both early and late efficacy, but no difference between ondansetron and droperidol for early efficacy, and none between ondansetron and metoclopramide for both early and late efficacy (table 1). The number needed to treat point estimates for early efficacy with ondansetron compared with placebo were 3.8 for 1 mg, 3.2 for 4 mg, and 3.1 for 8 mg. Over a 24 hour observation period the number needed to treat point estimates were 4.8 for 1 mg, 3.9 for 4 mg, and 4.1 for 8 mg.
The point estimates for the number needed to treat for early efficacy with ondansetron 8 mg were 2.5 (95% confidence interval 1.7 to 4.7) and 2 (1.3 to 4.6) in two small trials with 35 and 18 treated patients respectively,6 7 compared with 3.7 (2.6 to 6.5) in a large multicentre trial with 122 treated patients.8
For all three ondansetron doses, both for early and late observation periods, the 95% confidence intervals of the estimates of efficacy (odds ratio and number needed to treat) overlapped, indicating absence of any significant difference in anti-emetic efficacy between the three doses (table 1).
Ondansetron used as treatment for established postoperative nausea and vomiting was effective compared with placebo. About a quarter of treated patients were prevented from further nausea and vomiting with a dose of 1 mg, 4 mg, or 8 mg. It is difficult to say how well ondansetron performs in this setting relative to other treatments because of the paucity of direct comparisons with other anti-emetics. Nor is it possible to confirm that response rates in men, women, and children will be the same. Most patients (82%) in these trials were women.
There was no significant difference between ondansetron and droperidol when results from the two trials using droperidol were combined. Neither was there a significant difference between ondansetron and metoclopramide in the one trial that investigated this comparison. Indirect comparison of ondansetron with other anti-emetics will be possible by comparing their relative performance against placebo. If these drugs are highly effective in treating established postoperative nausea and vomiting, this would argue against pre-emptive use of anti-emetics.
Postoperative nausea and vomiting seem to have many causes,5 and it is perhaps naive to think that an anti-emetic, working at one specific receptor, should be universally effective. Given this multiple causation, from patient related factors through to the effects of anaesthesia, surgery, and opioids, preventing further postoperative nausea and vomiting in a quarter of the patients may be the best that can be achieved currently.
This quantitative analysis did, however, fail to show a significant dose-response for intravenous ondansetron between the 1 mg, 4 mg, and 8 mg tested. Although higher doses had lower point estimates for the number needed to treat, particularly for early efficacy, the differences between doses were not significant, as indicated by an overlap of the 95% confidence intervals of both odds ratio and the number needed to treat. This cannot be dismissed on the grounds of a clinically relevant difference minimised by lack of statistical power, because differences between numbers needed to treat were minor.
This inability to show a dose-response is hard to explain. The bulk (>900/1043 patients) of the data came from two large multicentre trials, and figure 1) shows little graphic evidence of heterogeneity. The two smaller trials6 7 reported higher early efficacy with ondansetron 8 mg (number needed to treat 2 to 2.5) than the large multicentre trial (3.7).8 One explanation for the failure to show a dose-response is that the minimum effective dose for treatment of established postoperative nausea and vomiting is less than the lowest dose (1 mg) studied, so that lower doses could be tested.
Two clinical messages emerge from this analysis. The first is that the number needed to treat for intravenous ondansetron compared with placebo to treat established postoperative nausea and vomiting is about 4. This means that 1 in 4 patients with nausea or vomiting treated with ondansetron will be prevented from further nausea and vomiting, who would otherwise have continued to have nausea or to vomit with placebo. The trials comparing ondansetron with droperidol or metoclopramide showed no difference in efficacy. We do not know if this is the best anti-emetic control that can be achieved.
The second message relates to anti-emetics as prophylaxis rather than as treatment for established postoperative nausea and vomiting. The justification of prophylactic postoperative anti-emetics was queried 35 years ago by Adriani and colleagues.20 They noted that no more than a quarter of patients in the recovery room vomited in the immediate postanaesthetic period and that most of this vomiting was short lived and subsided spontaneously without anti-emetics. Similar average incidence of postoperative nausea and vomiting has been reported repeatedly, both in large randomised controlled trials21 and in case series,22 23 24 although it may be higher in specific clinical settings, such as paediatric strabismus surgery.25 In the ondansetron trials analysed here the average incidence of postoperative nausea and vomiting was 36% before starting treatment, suggesting that these trials accurately reflect common clinical practice.
If the incidence is only about 30% and treatment is effective then arguably prophylaxis is unnecessary on grounds of adverse effects and cost. The humanitarian argument is that it is unacceptable to wait and see if a patient is going to vomit or develop nausea before starting a treatment. It is also widely believed that it may be more difficult to treat established postoperative nausea and vomiting than to prevent it,26 although there is no substantial evidence to support this view. The pivotal answers to resolve the debate will be the relative effectiveness of treatment and prophylaxis of postoperative nausea and vomiting.
We are concerned that data from a large, sponsored, multicentre trial were published three times.8 9 18 Inclusion of the two duplicates in the analysis would have increased the number of analysed reports by a quarter and doubled the number of analysed patients. Systematic reviewers are at risk of failing to recognise duplicates of an original report.27 The danger is that unrecognised duplicates will bias the estimates of an intervention's efficacy. Two duplicates were published in journal supplements,8 9 and the quality of supplement reports may be lower than reports in the parent journals.28 Both supplement articles declared that intravenous ondansetron 4 mg was the optimal dose to treat postoperative nausea and vomiting, although there was no good evidence to support this.8 9 Subsequent uncritical repetitions26 29 underline the potential influence of such unchallenged assertions.
MRT holds an Overseas Research Student Award.
Funding: The review was funded by Pain Research Funds.
Conflict of interest: HJMcQ has been a consultant to Glaxo, SmithKline Beecham, and other pharmaceutical companies. DJMR is a consultant to Janssen-Cilag.