5HT 3 antagonists for prophylaxis of postoperative nausea and vomiting in breast surgery: A meta-analysisAK Singhal1, S Kannan2, VS Gota3
1 Department of Anesthesiology and Critical Care, Institute of Liver and Biliary Sciences (ILBS), New Delhi, India
2 Department of Epidemiology & Clinical Trials Unit, Advanced Centre for Treatment, Research and Education in Cancer (ACTREC), Tata Memorial Centre, Navi Mumbai, India
3 Department of Clinical Pharmacology, Advanced Centre for Treatment, Research and Education in Cancer (ACTREC), Tata Memorial Centre, Navi Mumbai, India
Correspondence Address: Source of Support: None, Conflict of Interest: None PMID: 22387645
Source of Support: None, Conflict of Interest: None
Background: Postoperative nausea and vomiting (PONV) are distressing adverse events following breast cancer surgery with an incidence of up to 80%. 5HT 3 antagonists are commonly employed as drugs of first choice for PONV although there is no clear evidence favoring one pharmacological approach over another. Aims: The objective of this meta-analysis is to compare the efficacy of 5HT 3 antagonists against all non-5HT 3 antagonism-based pharmacological approaches as a preemptive strategy for PONV in women undergoing breast surgery. Design: Meta-analysis of Randomized Controlled Trials. Materials and Methods: Literature search was conducted through PUBMED, reference lists, and Cochrane Central Register of Controlled Trials till June 2010 to identify eligible studies. Trials comparing 5-HT 3 antagonists with placebo or active controls for prophylaxis against PONV in women undergoing breast surgery were included. Two reviewers extracted the data independently. Methodological quality of each trial was assessed using Jadad score. Results: Nineteen trials were included. All trials were of good methodological quality (Jadad score >3). 5HT 3 antagonists were found superior to placebo [Odds ratio (OR)=0.18 (0.13-0.26)] or active controls [OR=0.65 (0.47-0.91)] in the prevention of PONV. 5HT 3 antagonists were also superior to placebo in preventing nausea alone [OR=0.51 (0.34-0.76)], vomiting [OR=0.31 (0.20-0.47)] and the use of rescue antiemetics [OR=0.18 (0.11-0.28)]. No significant difference was observed in the use of rescue antiemetics as compared to active controls [0.59 (0.19 to 1.86)]. Conclusion: 5HT 3 antagonists are superior to other pharmacological interventions for the prevention of PONV in patients undergoing breast surgery under general anesthesia.
Keywords: 5HT 3 antagonists, breast surgery, general anesthesia, meta-analysis, postoperative nausea and vomiting
Breast surgery is associated with a very high risk for postoperative nausea and vomiting (PONV), others being intra-abdominal surgery, gynecologic surgery, strabismus repair and otolaryngologic surgery. The incidence of PONV following breast cancer surgery is up to 80%. ,, The etiology of PONV following breast surgery performed under general anesthesia is complex and is dependent on a variety of factors which include age, obesity, history of motion sickness and/or previous PONV, operative procedure, anesthetic technique, postoperative pain, phase of menstrual cycle and psychological factors . Many pharmacological interventions like droperidol, metoclopramide, dexamethasone and 5HT 3 antagonists are available for prophylaxis of PONV. However, droperidol and metoclopramide may cause excessive sedation and extrapyramidal effects which has limited their use.  Also, droperidol has attracted a 'Black Box' warning from the US Food and Drug Administration for the risk of dysrhythmia such as prolonged QT syndrome.  5HT 3 inhibitors came into clinical use in the early 1990s but were sparingly used till recently because of high costs. A systematic review of randomized controlled trials to determine the relative efficacy and safety of ondansetron, droperidol, and metoclopramide for preventing PONV showed that ondansetron and droperidol were equally effective but superior to metoclopramide in adults. 5HT 3 antagonists are generally well tolerated as demonstrated by the safety profile of ondansetron in this meta-analysis.  The 2003 PONV guidelines firmly recommended 5HT 3 antagonists as the first-line option for the prevention of PONV for surgeries done under general anesthesia.  Although it did not address some of the issues related to clinical differences between 5HT 3 receptor antagonists, optimal dosage and timing of administration, and optimal combinations of 5HT 3 receptor antagonists with other antiemetics.  Owing to these reasons, anesthetists are inclined to use 5HT 3 antagonists for prophylaxis even in breast surgery. However, in the absence of large randomized trials, no single agent has emerged as the standard of care for preventing PONV in women undergoing breast surgery. This is evident from the fact that till recently many randomized clinical trials for prophylaxis of PONV were placebo-controlled , although use of placebo controls in PONV studies is considered unethical .  The objective of our meta-analysis is to compare the efficacy of 5HT 3 antagonists against all non-5HT 3 antagonism-based pharmacological approaches as a preemptive strategy for PONV in women undergoing breast surgery.
The study was performed in accordance with the Preferred Reporting in Systematic review and Meta-Analysis (PRISMA) statement for conducting systematic reviews.  An initial list of published studies till June 2010 was obtained by searching Pubmed and the Cochrane Central Register of Controlled Trials database, using the terms (MeSH as well as text search) "prevention," "PONV," "antiemetics" separately for "breast surgery," and "mastectomy." The list was expanded by a manual search of reference lists from all articles, review articles, correspondence, and abstracts related to PONV in breast surgery. The search was limited to 'randomized controlled trials' and articles published in the English language. The search was done by AKS and VG independently.
Inclusion and exclusion criteria
Articles that met the following criteria were included in the meta-analysis: The study was a randomized controlled trial, patients underwent breast surgery under general anesthesia, the trial should have compared at least one 5HT 3 antagonist against either placebo or an active control (any pharmacological intervention for PONV apart from 5HT 3 antagonists), antiemetic therapy was administered prophylactically and not just in the treatment of PONV. Trials that were evaluating different anesthesia techniques or non-pharmacological interventions for prevention of PONV were excluded from the meta-analysis. The literature search identified 19 articles that met the selection criteria [Figure 1]. Trials with three or more treatment arms were considered separately for the respective comparisons.
The primary outcome for the meta-analysis was incidence of PONV, which was reported in all studies. Secondary outcomes included incidence of nausea alone, vomiting and the use of rescue medications for severe PONV. Subgroup analysis was carried out for age (≤50 versus >50 years) and duration of surgery (<2, 2-3 and >3 h) and the timing of antiemetic prophylaxis (at induction vs. end of anesthesia). If the authors have categorized retching as a separate outcome, we have included retching under nausea. , In most studies, these outcomes were determined for 24 h after surgery. In three trials PONV data was considered for 2, 3 and 48 h respectively due to unavailability of 24-h data. ,, In the study by Lee et al., nausea and vomiting outcomes were considered for 12 h and PONV data for 24 h as per the data available. 
Data extraction form was designed a priori which included information regarding publication details of the studies (author's name, journal, year, etc.,), interventions compared with respect to dose, timing, etc., demographic data like age and weight, and outcome measures including incidence of nausea, vomiting and PONV. Data was independently extracted by two review authors (VG and AKS). Studies with different drug doses within the therapeutic range, different routes of administration of 5HT 3 antagonists, variable time of antiemetic prophylaxis (preoperative, after induction of general anesthesia, or immediately before emergence), were included in the meta-analysis. Three studies compared more than one regimen of a 5HT 3 antagonist against placebo. ,, In such cases, the 'optimal' regimen was included in the meta-analysis. For example, Chan et al., compared intravenous (i.v.) tropisetron 5 mg and 2 mg against placebo.  The antiemetic efficacy of tropisetron 5 mg was superior to 2 mg. Fujii et al., observed that granisetron administered at 40 μg/kg was as effective as a higher dose of 80 μg/kg, but 20 μg/kg was no better than placebo.  A similar observation was made in another study comparing granisetron 1, 2 and 4 mg, where the antiemetic efficacy of granisetron 2 mg was similar to that of 4 mg and 1 mg was clearly inferior.  Therefore the tropisetron 5 mg, granisetron 40 μg/kg, and granisetron 2 mg arms respectively were included for this analysis.
Methods of analysis and inclusion criteria were specified in advance and documented in a protocol. The design and conduct of the selected trials were evaluated by SK and VG using a quality score adopted from Jadad et al.  Out of the 23 studies retrieved for detailed evaluation, four studies were excluded [Figure 1]. Data was analyzed using Rev Man software (Ver. 5.0). Outcomes of interest were compared between 5HT 3 antagonists and placebo/active control using odds ratios (ORs) with 95% confidence interval (CI). Statistical heterogeneity was assessed by I 2 , a widely used measure by Cochrane Collaboration. Random effects model was used for all the estimates. Publication bias was assessed using funnel plot. L'Abbe plot and regression analysis was used to assess if the baseline risk was a significant source of heterogeneity. Influence of individual trials on the overall outcome was assessed using sensitivity analysis.
The trial flow is shown in [Figure 1]. We selected 23 trials for analysis but subsequently excluded four trials. ,, The details of the 19 trials involving a total of 2053 patients included in the meta-analysis are summarized in [Table 1]. There were eight two-arm trials, nine three-arm trials and two four-arm trials. In all there were 15 studies leading to 19 comparisons (n=1393) of 5HT 3 antagonists with placebo and seven studies leading to nine comparisons (n=661) of 5HT 3 antagonists with active controls. Analysis was done separately for placebo and active controls.
Study characteristics and methodological quality
Eleven out of 16 trials administered antiemetic prophylaxis at or around induction and five at the end of surgery while three trials did not provide this information. Eleven trials had adequate methods of randomization. ,,,,,,,,,, All studies included blinded assessment with a majority of them (17/19) being double-blinded, and 16 of these trials described the method of double-blinding adequately. Time points of measurements for outcomes such as PONV and use of rescue antiemetic were considerably uniform (24 h) except for three studies as described earlier. The inhalation agents used were isoflurane (15 trials), ,,,,, [17,],,,,,,,,, sevoflurane (3 trials) ,, and halothane (1 trial).  Opioids were used during intubation and maintenance in 15 trials. ,,,,,,,,,,,,,,,
A. Comparison of 5HT 3 antagonists versus placebo
A total of 15 studies comparing 5HT 3 antagonists versus placebo (six for ondansetron, eight for granisetron and four for all other 5HT 3 antagonists) were available. Ondansetron, granisetron and other 5HT 3 antagonists were compared separately with placebo in addition to a pooled analysis involving all 5HT 3 antagonists for PONV, nausea alone, vomiting and use of rescue medication.
1. Incidence of PONV
Out of the six studies comparing ondansetron with placebo, greater efficacy of ondansetron in the prevention of PONV was observed in four studies [Figure 2]. ,,, The 95% CI of the remaining two studies spanned the "no difference" line (an OR of 1.0). , The overall result indicated a statistically significant reduction in the incidence of PONV with ondansetron (0.20 (0.09 to 0.46), n=411, 6 trials). Similar results were obtained for studies comparing granisetron with placebo (0.21 (0.13 to 0.35), n=470, 8 trials). Other 5HT 3 antagonists also showed superior efficacy for the prevention of PONV (0.13 (0.09 to 0.19), n=512, 4 trials). The overall odds of PONV were reduced significantly by 82% with 5HT 3 antagonists (0.18 (0.13 to 0.26), n=1393, 15 trials) [Figure 2]. 5HT 3 antagonists offered similar but significant protection against PONV irrespective of the age of the patients, duration of surgery and timing of prophylaxis [Table 2]. There was significant heterogeneity among the trials (I 2 =44%, P=0.02). Two studies comparing ondansetron and placebo significantly contributed to the heterogeneity. ,
The L'Abbe plot was used to examine the correlation of baseline risk with treatment effect. The regression analysis indicated that the baseline risk of PONV influenced the efficacy of 5HT 3 antagonists (slope= −2.79 (−5.38 to −0.19)). The high negative value of the slope also indicated that even a small increase in baseline risk can lead to a significant benefit from prophylaxis [Figure 3].
2. Incidence of nausea alone, vomiting, use of rescue antiemetic and adverse events
The odds of nausea alone were reduced significantly by 49% with 5HT 3 antagonists (0.51 (0.34 to 0.76), n=791, 10 trials). However, only granisetron showed a statistically significant effect (0.44 (0.22, 0.86), n=320, 6 trials) . The 5HT 3 antagonists were effective in preventing vomiting (0.31 (0.20 to 0.47), n=1294, 14 trials) and reducing the requirement of rescue antiemetics (0.18 (0.11 to 0.28), n=1332, 15 trials) . Granisetron and ondansetron alone had significant effect on the prevention of vomiting, whereas the other 5HT 3 antagonists were not as effective. 5HT 3 antagonists did not cause a significantly higher incidence of adverse events as compared to placebo [Table 3].
B. Comparison of 5HT 3 antagonists versus active controls
A total of 661 patients from seven studies were included in this meta-analysis comparing 5HT 3 antagonist with active controls. ,,,,,, All 5HT 3 antagonists except i.v. dolasetron  showed greater efficacy of 5HT 3 antagonists than active controls in the prevention of PONV (0.65 (0.47 to 0.91), n=661, 7 trials). There was no significant heterogeneity between the trials [Figure 4]. Five studies with a total of 387 patients were available for the analysis of rescue antiemetic requirement. 5HT 3 antagonists did not differ significantly from active controls in reducing the need for rescue antiemetics (0.59, (0.19 to 1.86), n=387, 5 trials). The results could not be analyzed separately for ondansetron, granisetron and other 5HT 3 antagonists because of the small number of trials available.
C. Comparison of granisetron alone versus combination
Three studies comparing granisetron alone with a combination of granisetron and dexamethasone , or granisetron and droperidol  were available for analysis. Each study individually showed greater efficacy of the combination as compared to granisetron alone. The pooled OR was found to be (0.25 (0.12 to 0.53), n=266, 3 trials) [Figure 5]. The analysis was restricted to PONV only.
Funnel plots were used to assess the publication bias for comparisons of 5HT 3 antagonists with placebo and active controls. No asymmetry was detected in either comparison [Figure 6]. Three studies were lying outside the 95% CI limits of funnel plot [Figure 6]a. ,, When the analysis was repeated after excluding these trials, the calculated OR for PONV incidence changed to (0.16 (0.12 to 0.21), n=909, 14 trials) with a non-significant heterogeneity (I 2 =0%, P=0.82).
In order to assess the influence of individual trials on the overall outcome, we conducted the sensitivity analysis by omitting one trial at a time. One of the trials by Fujii et al., appeared to influence the outcome of comparison between 5HT 3 antagonists and active controls.  The CI was found to straddle unity (0.72 (0.50 to 1.03)) when this study was omitted from the analysis [Figure 7].
This systematic review is to our knowledge, the first to quantify the effects of commonly used 5HT 3 antagonists for the prevention of PONV in breast surgery. The first-line use of 5HT 3 receptor antagonists for PONV prophylaxis in patients categorized as moderate to high-risk was promoted in the 2003 PONV guidelines.  In our analysis, we compared ondansetron, granisetron and other 5HT 3 antagonists separately against placebo for all outcomes. The results demonstrate strong evidence for the role of ondansetron and granisetron in the prevention of PONV, and preliminary evidence for other 5HT 3 antagonists like ramosetron, tropisetron and dolasetron. There is also some evidence for combining 5HT 3 antagonists with dexamethasone or droperidol. O ther 5HT 3 antagonists like ramosetron, tropisetron and dolasetron offered better protection against PONV (87%) as compared to ondansetron or granisetron (79% each) [Figure 2].
A combination of granisetron with dexamethasone or droperidol was more effective than granisetron alone [Figure 5]. Involvement of multiple mechanisms in the pathogenesis of PONV suggests that combination therapy with different classes of antiemetics may be more effective than 5HT 3 antagonists administered alone. The 2007 guidelines for the management of PONV from the Society for Ambulatory Anesthesia recommend drugs with different mechanisms of action to be used in adults at moderate risk for PONV.  That droperidol blocks a different receptor type (dopamine) as compared to granisetron, may explain the synergism. Besides, the greater anti-nausea efficacy of droperidol makes it an ideal candidate for combination with certain 5HT 3 antagonists which have better anti-vomiting than anti-nausea efficacy (e.g. ondansetron).  The precise mechanism of 5HT 3 antagonists plus dexamethasone in increasing the response rate is not known. Recently, a multimodal approach employing dexamethasone and intraoperative ondansetron in 200 consecutive patients was found to be highly effective with only 3.7% patients suffering severe PONV.  Future trials should explore combination treatments for better control of PONV in women undergoing breast surgery.
The conclusions must be considered in the light of several potential limitations. First, although we attempted to obtain all available data from trials examining the pharmacological interventions for the prevention of PONV in breast surgery, there is still likely to be some reporting bias. Analysis of publication bias using the funnel plot did not reveal any bias, although it may not be reliable for small-scale trials.  Second, we have included only pharmacological interventions in this meta-analysis as it is routinely practiced in almost all centers across the world. Other interventions like acupressure, acupuncture, transcutaneous electrical nerve stimulation, and electro-acupoint stimulation have shown benefits in isolated studies. ,, It is possible that these interventions are comparable to 5HT 3 antagonists.  Third, we have compared 5HT 3 antagonists against an 'average' antiemetic agent. The small number of trials with active controls (n=9) precluded separate analysis with each comparator individually.
We did not consider the variation in the doses used since these were almost similar in the different studies [Table 1].Ondansetron 4 mg was used in all but one study.  A meta-analysis of 48 trials of ondansetron vs. placebo involving more than 12,000 patients demonstrated that 4 mg was as effective as 8 mg.  Similarly, the dose of granisetron ranged from 1-3 mg or 40 μg·kg -1 i.v. A study by Wilson and colleagues demonstrated no difference in efficacy between 1 mg and 3 mg single i.v. dose of granisetron.  Dolasetron was used in 12.5 mg  and 50 mg (27) dose i.v. The similar efficacy of the two regimens has been reported previously,  and 12.5 mg is generally recommended for PONV prophylaxis.  Ramosetron was used at two different doses, 0.1 mg  and 0.3 mg.  Clear recommendation regarding the dose of ramosetron is not available. Therefore both doses were included in the analysis.
The subgroup analysis included categories based on age, duration of surgery and timing of antiemetic prophylaxis with respect to general anesthesia. The anticipated benefits of PONV prophylaxis increases with increase in the baseline risk for PONV. Younger age and longer duration of surgery are associated with a higher risk of PONV. , Each 30-min increase in duration of surgery increases the risk of PONV by 60%. There is some controversy regarding the timing of antiemetic prophylaxis. The PONV guidelines (2003) and the American Society of Health-system Pharmacists Guidelines are in disagreement regarding the time of administration of 5HT 3 antagonists.  Administration of ondansetron at the end of surgery has been found to reduce the need for rescue antiemetics in the post-anesthesia care unit.  Another study showed no difference in the efficacy of dolasetron with respect to time of administration.  Our analysis showed that age of the patients, duration of surgery and the timing of antiemetic prophylaxis did not have significant impact on the antiemetic efficacy of 5HT 3 antagonists. Since all 5HT 3 antagonists were grouped together in this analysis, the efficacy of individual drugs as regards to timing could not be evaluated.
The findings of our meta-analysis are contrary to the conclusions of a Cochrane meta-analysis of drugs for preventing PONV involving 737 studies and 103,237 patients.  The primary outcome measure of this meta-analysis was incidence of PONV at 24 h, which is not unlike the outcome of primary interest in our analysis. The authors concluded that "compared to placebo, eight drugs prevented postoperative nausea and vomiting: Droperidol, metoclopramide, ondansetron, tropisetron, dolasetron, dexamethasone, cyclizine and granisetron. Publication bias makes evidence for differences among these drugs unreliable". However, the results were clearly in favor of 5HT 3 antagonists wherever a head-to-head comparison was available with other interventions. For example, granisetron was superior to metoclopramide and droperidol for all outcomes of PONV, ondansetron was superior to metoclopramide in preventing nausea and vomiting and superior to droperidol in preventing vomiting, whereas dolasetron was superior to metoclopramide in preventing vomiting and the need for rescue antiemetics. However, equal efficacy was observed with respect to dexamethasone and cyclizine. The observed differences between the two reviews could be attributed to the following factors: 1) Subgroup analysis was not conducted for different types of surgeries according to risk of PONV in the Cochrane meta-analysis. This might have clearly brought out the superiority of 5HT 3 antagonists in high-risk surgeries since we could conclusively demonstrate that the benefit of 5HT 3 antagonism correlates positively with the baseline risk for PONV. 2) We did not observe any publication bias in our analysis as indicated in the funnel plot [Figure 6].
Our search was very comprehensive since we used both web-based as well as cross-reference approaches for identifying the studies, although selecting trials published in English language only may have introduced a selection bias. The data was reviewed and extracted independently by two reviewers to minimize bias. All trials included in the analysis were methodologically sound. Thus, the findings of this meta-analysis are generalizable to all women undergoing breast surgery. One of the trials by Fujii et al., however, influenced the overall outcome against active comparators,  as evident from our sensitivity analysis. It could be because of a small number of trials available for this comparison (n=9).
To conclude, this meta-analysis provides sufficient evidence to adopt 5HT 3 antagonists as first-line drugs for the prophylaxis of PONV in patients undergoing breast surgery under general anesthesia. 5HT 3 antagonists are effective irrespective of age of the patients, the duration of surgery or the timing of administration. In the face of existing evidence, placebo-controlled trials are not ethically justifiable in this patient population.
The authors are grateful to Dr. Tejpal Gupta for his opinion and suggestions during the design of the study and preparation of the manuscript.
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6], [Figure 7]
[Table 1], [Table 2], [Table 3]