emesis, granisetron, kytril, meta-analysis, ponv, postoperative nausea and vomiting
P Janicki. A Meta-analysis of the Efficacy of Granisetron 0.1 mg for Postoperative Nausea and Vomiting (PONV). The Internet Journal of Anesthesiology. 2006 Volume 12 Number 1.
This meta-analysis of published randomized clinical trials (RCTs) compared the efficacy of granisetron 0.1 mg versus ≥ 1.0 mg for prevention of postoperative nausea and vomiting (PONV). Efficacy data (proportion of patients without PONV or requiring rescue medication) were extracted and stratified by the postoperative period (i.e., early, late, and overall). Meta-analytic techniques were used to integrate findings and estimate pooled efficacy. Ten RCTs met inclusion criteria, but only 2 directly compared granisetron 0.1 mg versus 1.0 mg, which was insufficient for meta-analysis, necessitating indirect comparisons between pairwise sets of granisetron 0.1 mg and granisetron 1.0 mg with common comparators: placebo and granisetron dose levels > 1.0 mg. Granisetron 0.1 mg appears to be at least as effective as granisetron 1.0 mg during the early and overall postoperative periods. There were insufficient data for meta-analysis of any outcomes during the late postoperative period.
Conflict of Interest Disclosure
The author received in the past financial support (Independent Investigator Grant) from Roche Pharmaceuticals (maker of Kytril-granisetron).
Postoperative nausea and vomiting (PONV), a common adverse effect of general anesthesia following surgery, may result in increased medical resource utilization (e.g., increased time in the postoperative care unit, unanticipated hospital admission). Postoperative patients find vomiting more distressing than pain [1,2,3]. Patients with 3 or 4 risk factors, based on surgical procedure (e.g., gynecologic, breast, craniotomy, laparotomy, laparoscopy) and/or patient history (e.g., history of PONV and/or motion sickness, female, nonsmoking), have a 40% to 80% risk of developing PONV [3,4].
An important class of antiemetic agents used to prevent and/or treat PONV are serotonin 5-hydroxytryptamine subtype 3 receptor antagonists (5-HT3 RAs), which specifically and selectively bind to serotonin 5-HT3 receptors on afferent vagal pathways in the gastrointestinal tract and at the chemoreceptor trigger zone in the brain, resulting in decreased serotonin levels [3,5,6]. 5-HT3 RAs currently approved by the US Food and Drug Administration for prevention of PONV are dolasetron (Anzemet, Aventis Pharmaceuticals, Inc., Kansas City, MO), granisetron (Kytril, Roche Laboratories, Inc., Nutley, NJ), and ondansetron (Zofran, GlaxoSmithKline, Philadelphia, PA). Granisetron is the only 5-HT3 RA that is not metabolized by the cytochrome (CYP)2D6 pathway. Other 5-HT3 RAs that require CYP2D6 for metabolism may adversely affect patients with certain genotype polymorphisms, resulting in 5-HT3 RA metabolism that is too rapid or poor, leading to decreased therapeutic efficacy and possible failure to prevent PONV [7,8].
Granisetron is available for injection (IV) in 1 mg/ml and 0.1 mg/ml single-use vials and orally (PO) as a 1-mg tablet and 2 mg/10 ml solution [9,10].
Studies of granisetron 0.1 mg for PONV have yielded conflicting outcomes. Corman et al.  conducted a review of low-dose granisetron and, based on the scarcity of data, recommended against using granisetron 0.1 mg for prophylaxis in patients at risk of PONV. However, in a recently published double-blind randomized controlled trial (RCT), Gan et al.  observed no significant difference in the incidence of vomiting, nausea, and rescue medication, from 0 to 6 hours and from 0 to 24 hours postoperatively following abdominal hysterectomy, in patients who received granisetron 0.1 mg with dexamethasone 8 mg or ondansetron 4 mg with dexamethasone 8 mg. D'Angelo et al. , in another recently published double-blind RCT of postabdominal hysterectomy patients treated with granisetron at 0.1, 0.2, or 0.3 mg doses or placebo, reported a trend of improved efficacy compared with placebo during the first 6 hours postoperatively.
Our objective was to use meta-analytic techniques to determine and compare the efficacy of granisetron 0.1 mg and 1.0 mg for PONV. A meta-analysis is a systematic, quantitative review in which the results of at least 2 primary studies are pooled to determine the effect of a treatment. Meta-analyses, which are potentially very useful when clinical trials (primarily RCTs) yield conflicting results, historically have been frequently used to assess PONV studies [14,15,16,17,18,19], especially those addressing the efficacy of various 5-HT3 RAs [17,19,20,21,22,23].
Search Strategy and Study Inclusion
MEDLINE and the Cochrane Library were searched for all RCTs published from 1966 through February 15, 2006, that included the terms “granisetron” or “Kytril” and excluded “chemotherapy,” “chemotherapeutic,” or “cancer” in the title. RCTs were included for analysis if they (1) pertained to prophylaxis and/or treatment of PONV, (2) used an adult population, and (3) had at least one arm in which patients underwent prophylaxis or treatment of PONV with granisetron 0.1 mg or 1.0 mg (IV or PO) or a larger dose, alone or in combination with another agent (e.g., dexamethasone).
Studies with 3 or more treatment arms were divided into pairwise comparisons (Npair) between an “intervention” group and a “control” group: patients receiving granisetron 0.1 mg or 1.0 mg (alone or in combinations with other agents) were designated as the “intervention” group, and those receiving active comparators (e.g., ondansetron, droperidol, other doses of granisetron) or placebo were designated as the “control” group. Efficacy data extracted from each treatment arm included: (1) proportion of patients without nausea, (2) proportion of patients without vomiting, and (3) proportion of patients requiring rescue medication. Efficacy data were stratified by postoperative period (defined by Habib et al. in 2004 ): early (0 to 6 hours postoperatively), late (> 6 to 24 hours), and overall (0 to 24 hours).
ORs from pairwise comparisons were pooled (ORpool), CIs calculated, and fixed and/or random effects models employed. The Mantel-Haenszel method, which assumes that effect sizes do not significantly differ among studies, was used to calculate the fixed effects model. Its validity was tested by the heterogeneity statistic Q. When the p value for Q was significant (i.e., p < 0.05), the random effects model, which accounts for both random variations within and among studies, was employed using the method of DerSimonian and Laird . Meta-analyses were performed for each efficacy outcome within each postoperative period and for safety outcomes during the overall postoperative period if and only if Npair ≥ 3 originated from ≥ 3 independent studies that reported the outcome variable. Whenever there were insufficient data to directly compare granisetron 0.1 mg and 1.0 mg, meta-analytic techniques were used to indirectly compare them by identifying similar controls (comparators), i.e., versus placebo and versus granisetron doses > 1.0 mg. The use of indirect comparisons for meta-analyses has been established . Pairwise comparisons of granisetron 0.1 mg and 1.0 mg with common comparators were pooled to determine effect sizes (ORpool). Statistical calculations were performed using MedCalc version 7.6.
Studies for Analysis
A total of 154 RCTs were identified in the initial search. Among the 89 RCTs that pertained to PONV, 79 were excluded because they used a pediatric population or did not contain a treatment arm using granisetron 0.1 or 1.0 mg IV or PO. The remaining 10 RCTs (total of 1084 patients) included in the analysis are presented in Table 1 [12,13,29,30,31,32,33,34,35,36].
Granisetron was administered IV, except for the Fujii et al. studies [31,32,34]. Granisetron 0.1 mg–based therapy was used in 4 studies [12,13,35,36], and granisetron 1.0 mg–based therapy was used in 8 studies [29,30,31,32,33,34,35,36]. The 10 RCTs had a total of 29 randomized pairwise treatment comparisons (Npair): 9 between granisetron 0.1 mg and a comparator (Npair = 3 versus placebo, Npair = 5 versus another dose of granisetron, and Npair = 1 versus ondansetron/dexamethasone) and 20 between granisetron 1.0 mg and a comparator (Npair = 7 versus placebo, Npair = 10 versus granisetron > 1.0 mg, and Npair = 3 versus other active comparators). Because there were only 2 pairwise comparisons between granisetron 0.1 mg and granisetron 1.0 mg [35,36], which are insufficient for meta-analytic direct comparison, indirect comparisons were performed.
ORs for efficacy outcomes (i.e., absence of vomiting, absence of nausea, and need for rescue medication) from pairwise comparisons within each study during the early and overall postoperative periods are presented in Table 2.
Early Postoperative Period (0 to 6 Hours)
Late Postoperative Period (6 to 24 Hours)
None of the studies with a granisetron 0.1 mg treatment arm provided data on the absence of vomiting or nausea during the late postoperative period, nor were there any direct pairwise comparisons for rescue medication use between granisetron 0.1 mg and granisetron 1.0 mg. However, 3 studies provided data on the use of rescue medication in pairwise comparisons with granisetron 0.1 mg [12,13] or granisetron 1.0 mg  during the late postoperative period. None of the effect sizes were significant: granisetron 0.1 mg versus placebo (OR = 3.13; CI = 0.74, 13.20) or granisetron 3.0 mg (OR = 1.57; CI = 0.47, 5.18); or between granisetron 0.1 mg (with dexamethasone) and ondansetron (with dexamethasone) (OR = 0.94; CI = 0.41, 2.14); or between granisetron 1.0 mg and placebo (OR = 2.11; CI = 0.18, 25.35) or ondansetron 4.0 mg (OR = 2.11; CI = 0.18, 25.35).
Overall Postoperative Period (0 to 24 Hours)
Because there were insufficient studies for meta-analysis of direct pairwise comparisons of granisetron 0.1 mg with granisetron 1.0 mg for any outcome measure, indirect comparisons were conducted between the sets of granisetron 0.1 mg and granisetron 1.0 mg with comparators common to each set: placebo and granisetron dose levels > 1.0 mg (i.e., 2.0, 3.0, and 4.0 mg).
By indirect comparison, the effect of treatment with granisetron 0.1 mg versus placebo for prevention of early postoperative vomiting had a CI that did not overlap and appeared to be superior to that of granisetron 1.0 mg versus placebo. The combined set of granisetron 0.1 mg and granisetron 1.0 mg (0.1/1.0 mg) versus placebo was heterogeneous (p < 0.0001) but yielded a significant effect size favoring granisetron 0.1/1.0 mg with or without trial data from Fujii et al.
The effect sizes (ORpool) for prevention of nausea during the early postoperative period for granisetron 0.1 mg and 1.0 mg, respectively, versus placebo were both significant and favored granisetron therapy. By indirect comparison, the effect sizes for granisetron 0.1 mg versus placebo and for granisetron 1.0 mg versus placebo had overlapping CIs (Figure 2).
The combination set of granisetron 0.1/1.0 mg versus placebo had a significant effect size favoring granisetron 0.1/1.0 mg that was similar to both component sets but was heterogeneous with (p = 0.028) or without Fujii et al. study data (p = 0.015). There were insufficient data to meta-analyze pairwise comparisons of granisetron 0.1 mg and granisetron 1.0 mg versus placebo for the use of rescue medication during the early postoperative period.
There were insufficient data to assess the effect sizes (ORpool) of granisetron 0.1 mg versus granisetron doses > 1.0 mg for prevention of nausea during the early postoperative period. The effect size for granisetron 1.0 mg versus granisetron > 1.0 mg was not significant. As with absence of vomiting, indirect comparison between these sets for prevention of nausea was not possible. The effect size for granisetron 0.1/1.0 mg was homogeneous (p = 0.06) but nonsignificant. There were insufficient data to determine the effect sizes of granisetron 0.1 mg and/or granisetron 1.0 mg versus granisetron >1.0 mg for rescue medication during the early postoperative period.
Similar to absence of vomiting, both sets of granisetron 0.1 mg versus placebo and granisetron 1.0 mg versus placebo for absence of nausea yielded positive, medium-sized effect sizes, regardless of inclusion of Fujii et al. studies. Granisetron 0.1/1.0 mg versus placebo also yielded a positive effect size favoring granisetron, regardless of whether Fujii et al. studies were included. By indirect comparison, granisetron 0.1 mg versus placebo and granisetron 1.0 mg versus placebo had overlapping CIs, regardless of whether Fujii et al. studies were included (Figure 2).
There were insufficient data to calculate ORpool for pairwise comparisons of granisetron 0.1 mg versus placebo for the use of rescue medication during the overall postoperative period. Granisetron 1.0 mg versus placebo favored granisetron therapy, regardless of whether Fujii et al. studies were included, as did the combined set of granisetron 0.1/1.0 mg versus placebo.
When granisetron 0.1/1.0 mg versus granisetron doses > 1.0 mg were compared for prevention of vomiting, treatment favored granisetron > 1.0 mg when Fujii et al. studies were included but was not significant when Fujii et al. studies were excluded. Granisetron 0.1/1.0 mg versus granisetron > 1.0 mg for rescue medication favored granisetron > 1.0 mg, but the set appeared to be heterogeneous (p < 0.0001).
Our intent was to perform a meta-analysis directly comparing granisetron 0.1 mg with granisetron 1.0 mg for prevention and/or treatment of PONV during the early, late, and overall postoperative periods and for safety. We identified 10 RCTs in which granisetron 0.1 mg or granisetron 1.0 mg was used for PONV. Our criterion for meta-analysis was that efficacy or safety data was reported in at least 3 independent studies. However, in only 2 studies [35,36] were data available for direct comparison between RCTs comparing granisetron 0.1 mg– and granisetron 1.0 mg–based treatment for PONV.
Because it was not possible to construct meta-analyses that directly compared outcomes reported for granisetron 0.1 mg and 1.0 mg, we used indirect comparisons for meta-analyses (as established by Song et al. ). Because of the dearth of data and because all but one RCT  employed more than 2 treatment groups, studies with 3 or more treatment groups were separated into pairwise comparisons between granisetron 0.1 mg– and granisetron 1.0 mg–based therapy and (placebo or active) comparators. Where possible, meta-analyses were performed to indirectly compare granisetron 0.1 mg and granisetron 1.0 mg treatments (“intervention”) with common comparators (“control”), namely placebo, and granisetron doses > 1.0 mg. Our criteria for meta-analysis were amended to also include data from at least 3 pairwise comparisons (taken from at least 3 independent studies). Pairwise sets for indirect comparison that appeared to be homogeneous (p ≥ 0.05) were considered more reliable than those that appeared to be heterogeneous (p < 0.05), which might include confounding factors and/or dissimilar characteristics.
During the early postoperative period, the overall effect sizes (ORpool) for granisetron 0.1 mg versus placebo appeared to be superior to granisetron 1.0 mg for prevention of vomiting, although the latter set was heterogeneous, implying that the granisetron 1.0 mg versus placebo set might include confounding factors. Compared with granisetron doses > 1.0 mg for prevention of vomiting, the combination set of granisetron 0.1/1.0 appeared superior. For prevention of nausea, the overall effect sizes (ORpool) for the granisetron 0.1 mg and granisetron 1.0 mg sets were both superior to placebo and had overlapping CIs. However, granisetron 0.1 mg was heterogeneous, implying possible confounding factors. There were insufficient data for comparisons with granisetron doses > 1.0 mg for prevention of nausea or need of rescue medication. Based on these findings, granisetron 0.1 mg appears to be at least as effective (and possibly more effective) than granisetron 1.0 mg during the early postoperative period. There were insufficient data for meta-analysis of any outcomes during the late postoperative period.
During the overall postoperative period, compared with placebo, the overall effect sizes (ORpool) for the granisetron 0.1 mg and granisetron 1.0 mg sets, which were both homogeneous, had overlapping CIs for prevention of vomiting and for prevention of nausea. The ORpool for the granisetron 0.1/1.0 mg set, which was superior to placebo for need for rescue medication, appeared to be homogeneous. Compared with granisetron doses > 1.0 mg, the ORpool for granisetron 0.1/1.0 mg was inferior to granisetron > 1.0 mg for prevention of vomiting and need for rescue medication but was superior to granisetron > 1.0 mg for prevention of nausea. However, the set for rescue medication appeared to be heterogeneous, implying that granisetron 0.1 mg and granisetron 1.0 mg may be dissimilar. Based on these findings, granisetron 0.1 mg appears to be at least as effective as granisetron 1.0 mg during the overall postoperative period.
Our findings accord with a recently published retrospective cohort study of 400 patients, in which the bulk were administered granisetron 0.1 mg perioperatively for prophylaxis against PONV . Most patients in the study experienced “excellent control,” including those in the highest-risk groups.
Limitations of our meta-analysis included publication bias, deficiencies in reported efficacy data, insufficient studies to perform direct meta-analyses of granisetron 0.1 mg versus granisetron 1.0 mg, and use of study aggregate data (which has significantly less power than a meta-analysis based on individual patient data). Insufficient studies necessitated the use of pairwise comparisons (i.e., direct comparison of 2 treatment arms in studies with 3 or more treatment arms). There were insufficient data to analyze heterogeneity among studies and pairwise comparisons into subsets or to perform multiple regression (multivariate) analyses to account for potential confounding factors (e.g., time of administration). Some conclusions were drawn on data from only 3 studies. Further limitations of this meta-analysis arose from the inclusion of RCTs with slightly different timing of administration of PONV medication (i.e., granisetron administered pre-, intra-, and postoperatively in response to PONV) and those in which the intervention (or control) treatment were accompanied by the use of dexamethasone for 2-drug PONV prophylaxis. These facts should be taken into consideration during critical evaluation of the results of this meta-analysis.
Granisetron 0.1 mg appears to be at least as effective as granisetron 1.0 mg for PONV at preventing vomiting and nausea during the early and overall postoperative periods. There are insufficient data to determine whether this comparative effect is also true for use of rescue medication. The use of granisetron 0.1 mg may be associated with a decreased incidence of headaches and hepatic enzyme increase, as well as substantial cost savings. Additional studies directly comparing granisetron 0.1 mg with granisetron 1.0 mg for PONV may provide more data, enabling a higher-powered meta-analysis to address this important issue.
The author wishes to acknowledge Nelson Erlick, DPM, MS, for his statistical analyses and editorial assistance in the preparation of this manuscript.
Piotr K. Janicki, M.D., Ph.D. Department of Anesthesiology Pennsylvania State University College of Medicine Milton S. Hershey Medical Center 500 University Drive, H 187 Hershey, PA 17033-0850 Phone: 717-531-7021 Fax: 717-531-6221 E-mail: firstname.lastname@example.org