Introduction

Nausea and vomiting are distressing and debilitating side-effects of chemotherapy and radiotherapy, which if left untreated may result in physiological changes including dehydration, electrolyte imbalance and malnutrition [36]. These adverse events may impact on patients’ quality of life, hinder compliance with cancer treatment and potentially compromise treatment outcome [26]. Chemotherapy-induced nausea and vomiting (CINV) commonly occur within a few hours of treatment and may continue for several days. Symptoms are therefore classified as acute (<24 h post-chemotherapy) or delayed (>24 h post-chemotherapy). The onset and duration of nausea and vomiting do, however, vary depending on which chemotherapy agents are used.

A number of antiemetic agents are available for the management of nausea and vomiting, including 5-HT3-receptor antagonists, corticosteroids, NK1 receptor antagonists, dopamine-receptor antagonists, neuroleptics and benzodiazepines. 5-HT3-receptor antagonists are currently recommended for prophylaxis of nausea and vomiting associated with highly and moderately emetogenic chemotherapy [47]. Studies directly comparing the efficacy of different 5-HT3-receptor antagonists have failed to demonstrate clear superiority of any one agent; however, differences in their receptor specificities and pharmacological profiles may have an impact on their clinical efficacy over the full 24-h acute period [31].

To date, published studies have shown conflicting results. Large studies have shown heterogeneous results, while results of small studies need to be interpreted with caution due to lack of power or differences in trial designs and patient populations. A comprehensive meta-analysis of trials is therefore required to clarify overall outcomes. Few systematic reviews that compare the efficacy of the 5-HT3-receptor antagonists in the management of acute CINV have been completed, and none has clearly demonstrated superiority of one agent over another. These studies include a comparison of granisetron (Kytril®, Hoffmann-La Roche, Nutley, NJ, USA), ondansetron (Zofran®, GlaxoSmithKline, Research Triangle Park, NC, USA), tropisetron (Navoban®, Novartis Pharmaceuticals, North Ryde NSW, Australia) and dolasetron (Anzemet®, Aventis Pharmaceuticals, Kansas City, MO, USA), using all studies as references rather than head-to-head trials [16], and meta-analyses of randomized trials comparing granisetron with ondansetron [13], granisetron or tropisetron with ondansetron [5] and granisetron with ondansetron in high-dose cisplatin-based chemotherapy regimens [45]. One of the reviews is available as an abstract only [5], while one is in Spanish [45], and another is in German [16]. The prospective studies included in the latter publication were deemed too small to conclude whether one agent was more effective than another, and the retrospective analysis was designed to look for the optimal antiemetic treatment regimen for acute and delayed emesis, rather than comparing the 5-HT3-receptor antagonists. Importantly, none of the above reviews or meta-analyses included a comparison of all the commonly available 5-HT3-receptor antagonists.

This paper presents a meta-analysis of randomized trials of four currently available 5-HT3-receptor antagonists—dolasetron, granisetron, ondansetron and tropisetron—to attempt to clarify differences in their efficacy for prophylaxis of acute CINV. Palonosetron, a new 5-HT3-receptor antagonist with a prolonged half-life was only touched on, as the current paucity of comparative trials does not warrant a formal meta-analysis. More than half of the studies included in this meta-analysis compared granisetron with ondansetron and tested various dosing schedules in conjunction with different chemotherapy regimens. This prompted further sub-studies and dosing analyses for these two antiemetics in the acute phase.

Patients and methods

The meta-analysis was performed on the results of 44 randomized studies, including 12,343 patients, comprising direct comparisons between different 5-HT3-receptor antagonists in the treatment of acute CINV [2, 4, 6, 7, 10, 11, 18–20, 22–25, 27–29, 32, 33, 37–44, 46, 48–50, 52, 53, 55–61, 64, 65, 67, 69, 71]. The studies were identified by literature research on English language and non-English language publications referenced in MEDLINE, CANCERLIT or EMBASE and were categorized as follows: granisetron vs ondansetron, granisetron vs tropisetron, ondansetron vs tropisetron and ondansetron vs dolasetron. The trials and publications included had to fulfil the following selection criteria: treatment allocation randomization; parallel-group or crossover design (if response to the first cycle was reported or could be derived); no high-dose chemotherapy requiring blood cell transplantation; insufficient data for evaluation of acute antiemetic efficacy (e.g. only overall data for day 1 to 5 emesis provided); English, French, or German language (at least abstract). If the same trial was published repeatedly, the most recent or most detailed report was selected. Specific dose ranges were prospectively defined for inclusion in the dose-related analyses.

Studies were also subcategorized by the type of chemotherapy received by the patient, i.e. cisplatin-based or non-cisplatin-based. In the cases of a ‘mixed’ chemotherapy-regimen population, without results presented separately according to chemotherapy, the study was categorized according to the predominant cytostatic treatment administered by numbers of patients treated.

Antiemetic efficacy was defined as ‘complete acute response’, i.e. no nausea and vomiting, or only mild nausea within the first 24 h after chemotherapy (as defined by Smith [62] and Soukop [63]) or simply a complete absence of vomiting in the first 24 h after administration of chemotherapy (in cases of insufficient data to apply the nausea criterion).

Response rates were based on individual patients, i.e. the unit of observation taken from the publications for this meta-analysis was the ‘randomized patient’, taking into account data from the first chemotherapy cycle only, whenever possible. Pooled odds ratios were generated separately for cisplatin-based and non-cisplatin-based studies. Combined odds ratios were generated using Mantel-Haenszel pooling of contingency tables (fixed effect model as the standard approach) [68, 70]. In addition, a test for heterogeneity was performed for each combined analysis; a low p value indicates considerable variability between the results of the individual trials, more than would be expected due to chance. In these cases, the pooled therapy group comparison should be considered with care, as the treatment effect might differ with respect to patient selection, drug administration schedules and other factors varying from trial to trial. Thus, random-effects-model results were additionally presented in some of the analyses, as this approach yields more conservative statistical parameters (i.e. confidence intervals and p values) [68]. The figures provide odds ratios with 95% confidence intervals.

The following software was applied for the analyses: TESTIMATE (idv, Gauting, Germany); Comprehensive Meta-Analysis (Biostat, Englewood, NJ, USA).

Results

Granisetron vs ondansetron

Comparisons were made for: granisetron vs ondansetron (all doses), granisetron vs low-dose ondansetron, granisetron vs high-dose ondansetron, and granisetron vs ondansetron excluding dexamethasone. A total of 27 studies comparing all doses of granisetron with all doses of ondansetron were included in the meta-analysis (Table 1) [4, 7, 10, 11, 19, 20, 22, 23, 25, 27, 29, 38–40, 43, 44, 48, 49, 52, 53, 55, 57, 60, 64, 65, 69]. They comprised 16 cisplatin-based studies and 11 non-cisplatin-based studies. Overall pooled odds ratios (including all dose schedules) revealed an overall equivalence of both drugs (Fig. 1).

Table 1 Granisetron vs ondansetron: details of cisplatin-based and non-cisplatin-based studies used in this meta-analysis
Full size table
Fig. 1
figure 1

Comparison of granisetron vs ondansetron with cisplatin- and non-cisplatin-based chemotherapy studies. An odds ratio (OR) above 1 indicates superiority of ondansetron. *p < 0.05; **p < 0.01

Full size image

Subanalyses

Low-dose ondansetron, 8 mg i.v., vs granisetron, 3 mg

A subanalysis of six studies comparing low-dose ondansetron, 8 mg i.v., with granisetron, 3 mg i.v., showed a possible advantage for granisetron in non-cisplatin-based studies (p = 0.015, fixed effect model; Fig. 3b) [27, 29, 43, 53, 60, 69]. However, the latter finding is based on three trials with a total of 183 patients only, showing results with considerable variation (test of heterogeneity: p = 0.041; test of difference, random effects model: p = 0.27).

Fig. 2
figure 2

Comparison of a granisetron, 3 mg (or 40 μg/kg) vs granisetron, 1 mg (or 10 μg/kg) (odds ratios) with cisplatin based chemoteraphy studies and b granisetron, 2/3 mg, vs ondansetron, 24/32 mg, with cisplatin- and non-cisplatin-based chemotherapy studies. An odds ratio (OR) above 1 indicates superiority of ondansetron. *p < 0.05; **p < 0.01

Full size image

High-dose ondansetron, 24 or 32 mg, vs granisetron, 2 or 3 mg

Thirteen studies comparing higher-dose ondansetron (24 or 32 mg) with granisetron (2 or 3 mg), administered orally or intravenously, revealed an overall equivalence of these doses for both cisplatin-based and non-cisplatin-based studies (Fig. 2b) [4, 7, 10, 11, 22, 38–40, 52, 55, 57, 60, 65]. The results of this subgroup were predominantly driven by two large trials [22, 57], both of which used oral granisetron. Excluding studies using oral antiemetic regimens, seven cisplatin-based studies were compared (n = 1,117); six of these showed a non-significant trend in favour of granisetron (common odds ratio, 0.83; 95% confidence interval, 0.63–1.09), and a high degree of homogeneity between studies was apparent (test for heterogeneity p = 1.0) [4, 39, 40, 52, 55, 60].

Ondansetron was not compared with granisetron, 1 mg, due to a lack of data with this dose.

Granisetron vs ondansetron with the exclusion of dexamethasone

Trials comparing ondansetron with granisetron that did not include administration of dexamethasone were also compared [7, 20, 25, 29, 38–40, 43, 44, 48, 49, 52, 53, 55, 60, 64, 65, 69]. Results showed an overall equivalence of both drugs, similar to the analysis of all the ondansetron vs granisetron studies, some of which included dexamethasone.

Dosing study analyses

Ondansetron, 24 or 32 mg, vs ondansetron, 8 mg

Six cisplatin-based trials were included in this analysis, with ten of the studies identified initially excluded because of lack of randomization or lack of separate reporting of the first cycle in a crossover trial (Table 2) [6, 33, 56, 60, 61, 67]. The fixed-effect model meta-analysis indicates a significant result (p = 0.012) in favour of higher-dose ondansetron regimens (24 or 32 mg; Fig. 3a). However, a highly significant heterogeneity test (p = 0.0023) indicates that the pooled results should be interpreted with caution due to differences between the individual trials. Consequently, a random-effects model was applied and resulted in a higher p value of 0.19.

Table 2 Ondansetron, 8 mg vs 24 or 32 mg; and granisetron 1 mg (or 10 μg/kg) vs granisetron 3 mg (or 40 μg/kg): details of cisplatin-based studies used in this meta-analysis
Full size table
Fig. 3
figure 3

Subanalysis: comparison of a ondansetron, 24/32 mg, vs ondansetron, 8 mg b comparison of high-dose granisetron, 3 mg, vs low-dose ondansetron (8 mg). An odds ratio (OR) above 1 indicates superiority of the low dose. *p < 0.05; **p < 0.01; ***p < 0.001

Full size image

Granisetron, 1 mg (or 10 μg/kg), vs granisetron, 3 mg (or 40 μg/kg)

Six cisplatin-based trials were included in the comparison of low (1 mg) and high (3 mg) granisetron doses (Table 2), six studies having been excluded due to study design or inadequately reported content [32, 41, 49, 50, 58, 59]. No significant differences in efficacy were seen between the doses (Fig. 2a), and a high degree of homogeneity (test for heterogeneity p = 0.69) was observed between trials.

Granisetron vs tropisetron

The antiemetic efficacy of granisetron was compared with tropisetron in 12 studies (Table 3) [2, 4, 11, 19, 28, 39, 43, 44, 46, 53, 69, 71]. Granisetron was superior to tropisetron (odds ratio > 1.0) in six of the seven cisplatin-based trials, and in four of five non-cisplatin-based studies, although only one of these studies [2] showed a statistically significant benefit of granisetron (Fig. 4a). Pooled analysis of all cisplatin and non-cisplatin-based studies showed a significant advantage in favour of granisetron (p = 0.018), with no major heterogeneity detected (test for heterogeneity p = 0.88).

Table 3 Granisetron vs tropisetron, tropisetron vs ondansetron and dolasetron vs ondansetron: details of cisplatin-based and non-cisplatin-based studies used in these meta-analyses
Full size table
Fig. 4
figure 4

Comparison of a granisetron vs tropisetron, b tropisetron vs ondansetron and c dolasetron vs ondansetron (odds ratios) with cisplatin- and non-cisplatin-based chemotherapy studies. An odds ratio (OR) above 1 indicates superiority of a granisetron or (b and c) ondansetron. *p < 0.05

Full size image

Ondansetron vs tropisetron

The efficacy of ondansetron and tropisetron was compared in 11 studies (Table 3) [4, 11, 19, 28, 29, 39, 42–44, 53, 69]. Comparisons of these agents in six cisplatin-based studies revealed a non-significant advantage for patients receiving ondansetron compared with tropisetron. However, in non-cisplatin-based studies, a weak trend in the opposite direction was observed. Overall, the differences were neither significant nor of major clinical relevance (Fig. 4b).

Ondansetron vs dolasetron

Only three studies, one cisplatin-based and two non-cisplatin-based, were available for comparisons between ondansetron and dolasetron (Table 3) [18, 24, 37]. No difference in efficacy was seen in the cisplatin-based study [24], although a significant advantage for ondansetron was observed in one of the two non-cisplatin-based studies (p = 0.01; Fig. 4c). The meta-analytic technique cannot add much to this limited and heterogeneous body of evidence.

Granisetron vs dolasetron

Only three published studies were identified for such a comparison [3, 9, 66] and only one of these met the study inclusion criteria [3]. Therefore, a meta-analysis of these two agents was not possible.

Ondansetron and dolasetron vs palonosetron

Currently, there have only been two studies published that compare ondansetron with palonosetron, one in the moderately emetogenic and one in the highly emetogenic setting, and one study for the comparison of dolasetron with palonosetron (Table 4) [1, 17, 21]. However, there is inadequate data available for a meta-analysis.

Table 4 Palonosetron vs ondansetron and dolasetron: details of cisplatin-based and non-cisplatin-based studies and efficacy results
Full size table

Discussion

Current antiemetic guidelines recommend the use of 5-HT3-receptor antagonists as part of combination therapy for management of moderately to highly emetogenic chemotherapy regimens, but do not differentiate between agents, suggesting that they have similar efficacies at equivalent doses [47]. With a high degree of variability seen between small antiemetic studies and conflicting results observed with larger studies (e.g. Gralla et al. [22] and IGARR [27]), a comprehensive combined analysis is warranted. This meta-analysis was therefore designed to evaluate the comparative efficacy of four commonly used 5-HT3-receptor antagonists in the treatment of acute CINV.

Overall analysis of all studies (cisplatin-based and non-cisplatin-based) demonstrated that for acute CINV:

  1. 1.

    granisetron and ondansetron have comparable efficacies at equally potent doses in cisplatin- and non-cisplatin-based studies

  2. 2.

    3 mg granisetron may have superior antiemetic efficacy compared with ondansetron dosed at 8 mg (in moderately emetogenic regimens) and is comparable to ondansetron, 24 or 32 mg

  3. 3.

    no difference in efficacy was found between different doses of granisetron

  4. 4.

    high-dose ondansetron may be more effective than low-dose ondansetron for cisplatin-based chemotherapy, although the results of this analysis are affected by the high degree of variation between the individual trials

  5. 5.

    granisetron is superior to tropisetron in both cisplatin and non-cisplatin-based studies but reached statistical significance only when the results of all studies were combined

  6. 6.

    ondansetron may have similar efficacy to tropisetron (as found in the pooled analysis) but may be superior to dolasetron, although this is based on analysis of just three studies.

Our findings are supported by those of previous analyses, which have suggested comparable efficacy between ondansetron and granisetron [5, 13, 16, 45] and between tropisetron and ondansetron [5].

It has been suggested that ondansetron, at a dose of 8 mg, is suboptimal in controlling acute emesis after non-cisplatin-based chemotherapy (e.g. cyclophosphamide) [12, 14, 16]. The results from the meta-analysis support this, demonstrating that in non-cisplatin-based studies, ondansetron, 8 mg i.v., appears to be inferior to granisetron, 3 mg i.v., although this observation comes from an analysis of small studies that were underpowered. No comparisons were made with the commonly used dose of 1 mg i.v. granisetron due to a lack of studies comparing this dose with ondansetron. The difference observed between high-dose granisetron and low-dose ondansetron may be due to the delayed onset of acute emesis induced by moderately emetogenic chemotherapy regimens combined with the differences in the pharmacological profiles of the two drugs. Granisetron’s longer duration of action within the first 24 h is likely due to its higher affinity to the 5-HT3 receptor and its longer half-life [34, 35, 51]. Strong receptor binding and a longer half-life might be of clinical importance as suggested by studies with palonosetron. Multiple daily doses or a single high dose of ondansetron may therefore be necessary to maintain antiemetic efficacy for 24 h for some patients [8, 47, 72, 73]. In addition, dosing-study analyses demonstrated high-dose ondansetron (24 or 32 mg) are significantly more effective than the low dose (8 mg) for acute antiemetic control. However, the reliability of this finding in the meta-analysis should be treated with caution, as it is affected by significant heterogeneity in the results of the individual trials. Indeed, due to the issue of heterogeneity in some parts of the meta-analysis, it may have been useful to assess individual patient charts, and/or perform further subgroup analyses. However, this was not technically feasible. A high degree of heterogeneity, particularly in the moderately emetogenic setting, reflects increased variation between studies, especially in terms of chemotherapy regimens and individual patient risk factors. For example, factors such as low alcohol intake, previous history of motion sickness, being female [15] and impairment of social activities [30, 54] may lead to a patient being at high risk of developing emesis despite receiving moderately emetogenic therapy. This individual variation is less apparent with cisplatin, as the emetogenicity of this agent tends to mask other variables. Differences between trials also include the trial methodology (e.g. open or blinded), how the endpoints were defined and whether cycles or patients were the unit of observation. It has been suggested that industry sponsorship may have an impact on trial outcome. Whereas a meta-analysis may be a useful tool to assess such influences or bias, this particular analysis is inappropriate due to the weak-to-moderate effects detected.

Dosing schedules are an important variable in terms of cost-effectiveness and it has been suggested that an efficient antiemetic therapy requiring a simple dosing regimen would result in reduced burden on healthcare resources, including time and materials, and may result in an overall reduction in cost. The results of this analysis reinforce the importance of choosing the most appropriate dose of setron, particularly bearing in mind the chemotherapy regimen, to maximize efficacy.

The combined analysis of granisetron vs tropisetron showed a significant advantage for granisetron (p = 0.018) in controlling acute CINV. Interestingly, pharmacogenetic differences have been observed that conceivably affect the metabolism of these agents. Granisetron is the only setron that is not metabolized by the cytochrome P450 (CYP) enzyme, 2D6. This enzyme is subject to genetic polymorphism, resulting in different rates of drug elimination. This was demonstrated in a study by Kaiser et al. [31] who showed that in patients treated with ondansetron or tropisetron, ultra-rapid metabolizers had a significantly higher frequency of vomiting (p < 0.001 within 4 h; p < 0.03 between 5 and 24 h) in the acute phase than all other patients. The difference was more notable in tropisetron-treated patients.

The findings from this meta-analysis do not entirely reflect antiemetic guidelines. In general, the analysis reveals comparable acute antiemetic efficacy between the four 5-HT3-receptor antagonists, with a few exceptions. A significant advantage for granisetron over tropisetron was revealed in a pooled analysis of all chemotherapy regimens studied. The commonly used low-dose ondansetron (8 mg) appears to be not as effective as high-dose ondansetron (24/32 mg) in cisplatin-based studies, or granisetron (3 mg) in non-cisplatin-based studies. However, no studies have been conducted so far comparing a granisetron dose of 1 mg i.v. (or 2 mg orally), which is most commonly used in clinical practice, with low-dose ondansetron. From these results, there seems to be a dose-response relationship for ondansetron, but not for granisetron. Our findings have obvious implications for clinical practice. Whereas the 5-HT3-receptor antagonists included in our analysis appear to have the same antiemetic potential (with the exception of tropisetron), equivalent efficacy is realized only when using equally effective doses. This is of particular importance for any pharmacoeconomic evaluation of antiemetics. To conclude, the emetogenicity of the chemotherapy, the antiemetic drug choice and the dosing regimen of the antiemetic used are all important considerations for administration of the most appropriate and cost-effective 5-HT3-receptor antagonist for the management of acute CINV.