A systemic search of the relevant literature was performed without language limitation but restricted to RCTs available in English. We mainly explored MEDLINE using PUBMED and SCOPUS, the Cochrane database library using references found in reviews, or initially identified articles which were also used for further articles related to the topic. Published articles contained key words such as "ramosetron," "PONV," "nausea," "vomiting," "postoperative," "postoperative nausea and vomiting," and "antiemetics" in their titles or abstracts, and electronic searches were conducted until November 2010. The search strategy consisted of a combination of free text words as follows: "postoperative OR postanaesthetic OR postanesthetic OR surgical" AND "nausea OR emesis OR vomiting OR retching" AND "ramosetron OR nausea". A copy of each paper was retrieved for final assessment of eligibility.

Inclusion criteria were as follows: (1) the study was published as a full report of randomized controlled trials that tested the efficacy and safety of ramosetron compared with placebo or any other drugs as control in the prophylaxis of PONV following general anesthesia; (2) the placebo was identical in appearance and had no active drugs; and (3) the subjects were human (adults and children), and all participants irrespective of sex, age, surgery, or age were considered. We did not include data from abstracts, posters, case reports, comments or letters to the editor, reviews, and animal studies. No attempts were made to obtain unpublished studies, nor did we request unpublished data from the company developing the drug.

Nausea was defined as subjectively having unpleasant feeling associated with being sick or awareness of the urge to vomit. Vomiting was defined as a forceful expulsion of gastric contents or retching that was defined as labored spasmodic, rhythmic contraction of the respiratory muscles without expulsion of gastric contents. PONV included both nausea and vomiting [15].

Many studies revealed evidence for different underlying PONV risk factors in children and the need for higher doses of prophylactic antiemetics in children, necessitating different preoperative risk scores than those used for adults [16]. Therefore, we separated the relevant outcome data for children and adults and analyzed accordingly.

We retrieved patient information, including the type of surgery, drug dose, and route of administration, comparators, study end-points, adverse effects, and rescue medication. The outcome data of PONV at three time points were extracted: early period (0-6 h), late period (≥ 6 h), and the overall period (0-24/48 h). Antiemetic effects of early and late periods were only analyzed for 0.3 mg ramosetron compared data because of inconsistencies or absences in reporting. When PONV events were presented at different times, we selected cumulative values near postoperative 6 hours and 24/48 hours. PONV events, including retching, were the primary outcomes, and incidence of adverse effects and the use of rescue medication were secondary outcomes. The primary and secondary outcomes were separately compiled. A single investigator examined and scored the retrieved reports using the five-point Oxford scale for validity assessment (randomization 0-2, blinding 0-2, description of withdrawals and dropouts 0-1) [17].

Collected data were recorded on formalized sheets, and values originally provided as percentages were converted into actual corresponding numbers (of patients and incidence) for analysis where needed. We computed the relative risk (RR) with corresponding 95% confidence intervals (CIs) for dichotomous outcome data using a fixed-model, if no heterogeneity was present. Otherwise, a random-effects model was applied. Heterogeneity was judged with the I²-test, assuming heterogeneity if an I² value of more than 25% was observed. When the 95% CI did not include the value 1.0, we assumed a statistically significant difference between an intervention and control. Subgroup analysis of the following items was performed to investigate relevant clinical heterogeneity: different doses, time of application, route of administration, and active-controlled versus placebo-controlled trials. We calculated the number needed to treat or to harm (NNT) as a useful estimate measure of the clinically significant effect. NNT is the number of patients that must be treated with an experimental intervention to achieve a particular result (beneficial or harmful) in one of them which would not have been the case had they all received the control intervention (placebo). A potential publication bias was explored with asymmetry of overall period PONV data using a funnel plot. Statistical analysis was performed using MIX 2.0 for Windows that was developed and validated by Leon Bax of the Department of Medical Informatics of Kitasato University in Japan [18,19]. Incidences of PONV with various doses of ramosetron were compared using Fisher's exact test using SPSS 13.0 for Windows, and P < 0.05 was considered significant.

The incidence of overall PONV events (0-24/48 hours after anesthesia) was 56 ± 15 (%, mean ± SD) with placebo, 49 ± 16 with 0.15 mg ramosetron, 16 ± 7 with 0.3 mg ramosetron, and 12 ± 5 with 0.6 mg ramosetron. There was no significant difference between placebo and 0.15 mg ramosetron or between 0.3 mg and 0.6 mg ramosetron. The most frequently used regimen of ramosetron was 0.3 mg IV as a minimum effective dose; increasing the dose to 0.6 mg provided no further benefit. In children, 6 µg/kg IV was the most effective dose (Table 1) [5,20-26].

All trials (studies 4, 5, 16) showed a benefit of 0.3 mg ramosetron in the prevention of early PON (RR: 0.4; 95% CI, 0.3-0.7), early POV (RR: 0.3; 95% CI 0.2-0.6), and late POV (RR: 0.3; 95% CI, 0.1-0.6), but not late PON (RR: 0.6; 95% CI, 0.3-1.2). Ramosetron at this dose was compared with placebo in 9 trials of overall PONV, 5 trials of rescue-need, 5 trials of adverse effects in adults, and 2 trials of overall PONV and rescue-need in children (studies 20, 21, Table 1). All results in overall PONV and rescue-need showed a significant effect of 0.3 mg ramosetron in adults and 6 µg/kg in children because the 95% CI did not include the value of 1.0 (Fig. 2). Comparing 0.3 mg ramosetron and placebo, RR in adults was 0.5 (95% CI, 0.3-0.6), and NNT was 3.3 in overall PONV. In rescue-needed trials, RR was 0.5 (95% CI, 0.3-0.8), and NNT was 5. In trials with adverse effects, RR was 1.2 (95% CI, 0.8-1.6), and NNT was 33 without significant difference (Fig. 3). As no heterogeneity was found except in overall PONV when comparing 0.3 mg ramosetron and placebo, a subgroup analysis of different time points of ramosetron administration was performed. However, different time points did not affect heterogeneity.

0.3 mg ramosetron was compared with 3 mg granisetron in 4 trials of overall PONV and 3 trials of adverse effects in adults (Table 1). RR was 0.6 (95% CI, 0.3-1.2) in overall PONV and 1 (95% CI, 0.5-1.8) in adverse effects. Therefore, no significant difference was found between 0.3 mg ramosetron and 3 mg granisetron.

As several trials compared ramosetron with other active drugs, combined data analysis was performed with respect to the route of administration (e.g. per oral and PCA), different time points of application (e.g. before and after induction), and different doses of active drugs (e.g. ondansetron and dexamethasone). Due to insufficient or missing appropriate data to combine or compare, no further analysis was conducted (Table 1).

Several trials reported adverse effects, including headache, dizziness, constipation/diarrhea, muscle pain, and drowsiness/sedation. No weighting was used for different grades of adverse effects. Headache was the most often observed adverse effect, but all adverse events were transient and clinically insignificant. There were no severe adverse events, and no trials presented any statistically significant difference between ramosetron and placebo or granisetron.