Antagonists for Chemotherapy-Induced Nausea and Vomiting
Antagonists for Chemotherapy-Induced Nausea and Vomiting
To evaluate the effectiveness and safety of NK1R antagonists in preventing CINV, we searched for randomized controlled trials (RCTs) that compared the addition of NK1R antagonists to standard antiemetic regimens, including a 5-HT3 antagonist plus dexamethasone, for cancer patients receiving chemotherapy, regardless of its emetogenic potential. We performed an electronic database search of MEDLINE (last search, December 21, 2010), EMBASE (last search, September 30, 2010), the Cochrane Central Register of Controlled Trials—CENTRAL (last search, September 26, 2010), and LILACS (last search, August 1, 2010), and, electronically or manually, we searched all conference proceedings of the American Society of Clinical Oncology (ASCO), the American Society of Hematology (ASH) and the European Society for Medical Oncology (ESMO) between 1998 and 2010. We used a wide search strategy for the electronic database with the following combinations of keywords: neurokinin, aprepitant, casopitant, ezlopitant, netupitant, vestipitant (and their respective codes), chemotherapy-induced nausea and vomiting, nausea in cancer patients, vomiting in cancer patients, and randomized trials. The reference lists of all recovered trials and relevant reviews were also considered. Sources with the most recent data were used whenever possible. Language and publication date restrictions were not applied.
Two reviewers (LVS, FHS) screened titles and abstracts identified from the search strategy for eligibility. Both reviewers independently selected trials for inclusion according to prior agreement regarding the study population and intervention. The studies of interest were RCTs that addressed the addition of an NK1R antagonist to standard antiemetic therapy (dexamethasone plus 5-HT3 antagonist) for the prevention of CINV. Studies were eligible for inclusion in our analysis if they provided an adequate description of outcomes that could be pooled in the meta-analysis and used adequate antiemetic therapy in the control arm (dual therapy). If one of the reviewers determined that an abstract was eligible, the full text version was retrieved and selected upon concordance between LVS and FHS. Full text versions of all eligible studies were obtained for quality assessment and data extraction.
Two reviewers (LVS, FHS) independently assessed the quality of each study using the full text article, and two reviewers (JPL, LVS) independently performed the data extraction. The data extraction form (available upon request) included the following items: general information (authors, title, journal, date of publication, protocol name, and duplication of publication), methodological characteristics of the RCTs (method of randomization, allocation concealment, blinding, drop-out description, calculation of the sample size, intention-to-treat principle, funding source), study population (number of patients, types of cancer, age), chemotherapy emetogenic potential and outcomes. Any disagreement was discussed until all three reviewers reached consensus. The characteristics directly related to the risk of bias were analyzed in subgroups to test their impact on the estimation of the effect size.
The primary outcome that we tabulated was the proportion of patients who achieved a complete response (CR) during the overall period of assessment (ie, during the acute and delayed phases after chemotherapy; 0–24 hours and 24–120 hours, respectively). CR was defined as the absence of vomiting or retching and the absence of the need for rescue antiemetic drugs. CR in the acute and delayed phases was a secondary outcome. Symptoms that appeared within the first 24 hours after administration of chemotherapy were classified as acute and those that appeared from 24 to 120 hours after the administration of chemotherapy were considered as delayed. Other secondary outcomes were nausea and vomiting during the acute, delayed, and overall periods. Nausea was classified according to a 100-mm visual analogical scale (VAS), where 0mm correlates to absence of nausea and 100mm correlates to the worst possible nausea experienced by the patient. In this classification, patients with a visual analogical scale less than 5mm were classified as having "no nausea," and patients with a visual analogical scale less than 25mm were classified as having "no (clinically) significant nausea." Reported adverse events were also a secondary outcome.
RevMan 5.0 software was used to perform the meta-analysis. The Mantel–Haenszel random-effects method was used to calculate odds ratios (ORs) and the corresponding 95% confidence intervals (CIs). We measured the occurrence of events such that an OR less than 1 favored the NK1R antagonist group in the primary and secondary endpoints. The number needed to treat—or the number needed to harm in case of toxicity—to benefit one single patient was calculated as 1/[experimental event rate − control event rate].
The statistical heterogeneity of trial results was assessed by the χ test and expressed as I plus the corresponding P value. Heterogeneity was considered substantial if I² was 50% or greater (29,30) and, if encountered, a plausible explanation was intensively pursued. If a reasonable cause for heterogeneity between trials was found, a separate analysis was then performed to explore the impact of this factor on the estimation of the effect size. If the cause was not apparent and if heterogeneity was generated by divergent data (ie, data favoring one or other treatment), the data would then not be pooled.
Publication bias was evaluated by Egger's test using Comprehen sive Meta-Analysis software, version 2.0. The estimation for the impact of publication bias was done using the "Trim and Fill" method. This method consists in representing the "missing" (or unpublished) studies in a funnel plot graph and then calculating the impact these studies would have in the estimation of the effect size.
If a given study had more than one interventional arm, we chose to combine all intervention groups to avoid multiple counting of the same individuals in the control arm (so called unit-of-analysis error). For example, if a trial had two interventional arms (with different doses of NK1R antagonists) and one control arm (with no NK1R antagonists), the number of subjects and events in both interventional arms was added and then compared with the number of subjects and events in the control arm for each endpoint.
Predefined subgroup analyses were undertaken in clinically relevant subsets to evaluate the impact of these subgroups on the estimation of the effect size. The following comparisons were carried out: chemotherapy emetogenic potential (highly vs moderately emetogenic chemotherapy), NK1R antagonist treatment length (1 day vs more than 1 day), route of administration (oral vs intravenous vs both), the characteristics of the control group (presence vs absence of 5-HT3 antagonist instead of placebo, in the control group only, during the delayed phase), type of drug (aprepitant vs casopitant vs other), and dexamethasone dose modification because of pharmacological interaction with the NK1R antagonist. Sensitivity analyses based on methodological quality parameters were performed to test for possible variations in estimates of ORs between subgroups.
Additional exploratory analyses using Spearman's rank correlation coefficient (rs) were performed to test for potential correlations between responses in the acute and delayed phases. Statistical tests for heterogeneity were one-sided; statistical tests for effect estimates and publication bias were two-sided.
Methods
Search Strategy
To evaluate the effectiveness and safety of NK1R antagonists in preventing CINV, we searched for randomized controlled trials (RCTs) that compared the addition of NK1R antagonists to standard antiemetic regimens, including a 5-HT3 antagonist plus dexamethasone, for cancer patients receiving chemotherapy, regardless of its emetogenic potential. We performed an electronic database search of MEDLINE (last search, December 21, 2010), EMBASE (last search, September 30, 2010), the Cochrane Central Register of Controlled Trials—CENTRAL (last search, September 26, 2010), and LILACS (last search, August 1, 2010), and, electronically or manually, we searched all conference proceedings of the American Society of Clinical Oncology (ASCO), the American Society of Hematology (ASH) and the European Society for Medical Oncology (ESMO) between 1998 and 2010. We used a wide search strategy for the electronic database with the following combinations of keywords: neurokinin, aprepitant, casopitant, ezlopitant, netupitant, vestipitant (and their respective codes), chemotherapy-induced nausea and vomiting, nausea in cancer patients, vomiting in cancer patients, and randomized trials. The reference lists of all recovered trials and relevant reviews were also considered. Sources with the most recent data were used whenever possible. Language and publication date restrictions were not applied.
Selection Criteria
Two reviewers (LVS, FHS) screened titles and abstracts identified from the search strategy for eligibility. Both reviewers independently selected trials for inclusion according to prior agreement regarding the study population and intervention. The studies of interest were RCTs that addressed the addition of an NK1R antagonist to standard antiemetic therapy (dexamethasone plus 5-HT3 antagonist) for the prevention of CINV. Studies were eligible for inclusion in our analysis if they provided an adequate description of outcomes that could be pooled in the meta-analysis and used adequate antiemetic therapy in the control arm (dual therapy). If one of the reviewers determined that an abstract was eligible, the full text version was retrieved and selected upon concordance between LVS and FHS. Full text versions of all eligible studies were obtained for quality assessment and data extraction.
Quality Assessment and Data Extraction
Two reviewers (LVS, FHS) independently assessed the quality of each study using the full text article, and two reviewers (JPL, LVS) independently performed the data extraction. The data extraction form (available upon request) included the following items: general information (authors, title, journal, date of publication, protocol name, and duplication of publication), methodological characteristics of the RCTs (method of randomization, allocation concealment, blinding, drop-out description, calculation of the sample size, intention-to-treat principle, funding source), study population (number of patients, types of cancer, age), chemotherapy emetogenic potential and outcomes. Any disagreement was discussed until all three reviewers reached consensus. The characteristics directly related to the risk of bias were analyzed in subgroups to test their impact on the estimation of the effect size.
Outcomes of Interest
The primary outcome that we tabulated was the proportion of patients who achieved a complete response (CR) during the overall period of assessment (ie, during the acute and delayed phases after chemotherapy; 0–24 hours and 24–120 hours, respectively). CR was defined as the absence of vomiting or retching and the absence of the need for rescue antiemetic drugs. CR in the acute and delayed phases was a secondary outcome. Symptoms that appeared within the first 24 hours after administration of chemotherapy were classified as acute and those that appeared from 24 to 120 hours after the administration of chemotherapy were considered as delayed. Other secondary outcomes were nausea and vomiting during the acute, delayed, and overall periods. Nausea was classified according to a 100-mm visual analogical scale (VAS), where 0mm correlates to absence of nausea and 100mm correlates to the worst possible nausea experienced by the patient. In this classification, patients with a visual analogical scale less than 5mm were classified as having "no nausea," and patients with a visual analogical scale less than 25mm were classified as having "no (clinically) significant nausea." Reported adverse events were also a secondary outcome.
Statistical Analysis and Synthesis
RevMan 5.0 software was used to perform the meta-analysis. The Mantel–Haenszel random-effects method was used to calculate odds ratios (ORs) and the corresponding 95% confidence intervals (CIs). We measured the occurrence of events such that an OR less than 1 favored the NK1R antagonist group in the primary and secondary endpoints. The number needed to treat—or the number needed to harm in case of toxicity—to benefit one single patient was calculated as 1/[experimental event rate − control event rate].
The statistical heterogeneity of trial results was assessed by the χ test and expressed as I plus the corresponding P value. Heterogeneity was considered substantial if I² was 50% or greater (29,30) and, if encountered, a plausible explanation was intensively pursued. If a reasonable cause for heterogeneity between trials was found, a separate analysis was then performed to explore the impact of this factor on the estimation of the effect size. If the cause was not apparent and if heterogeneity was generated by divergent data (ie, data favoring one or other treatment), the data would then not be pooled.
Publication bias was evaluated by Egger's test using Comprehen sive Meta-Analysis software, version 2.0. The estimation for the impact of publication bias was done using the "Trim and Fill" method. This method consists in representing the "missing" (or unpublished) studies in a funnel plot graph and then calculating the impact these studies would have in the estimation of the effect size.
If a given study had more than one interventional arm, we chose to combine all intervention groups to avoid multiple counting of the same individuals in the control arm (so called unit-of-analysis error). For example, if a trial had two interventional arms (with different doses of NK1R antagonists) and one control arm (with no NK1R antagonists), the number of subjects and events in both interventional arms was added and then compared with the number of subjects and events in the control arm for each endpoint.
Predefined subgroup analyses were undertaken in clinically relevant subsets to evaluate the impact of these subgroups on the estimation of the effect size. The following comparisons were carried out: chemotherapy emetogenic potential (highly vs moderately emetogenic chemotherapy), NK1R antagonist treatment length (1 day vs more than 1 day), route of administration (oral vs intravenous vs both), the characteristics of the control group (presence vs absence of 5-HT3 antagonist instead of placebo, in the control group only, during the delayed phase), type of drug (aprepitant vs casopitant vs other), and dexamethasone dose modification because of pharmacological interaction with the NK1R antagonist. Sensitivity analyses based on methodological quality parameters were performed to test for possible variations in estimates of ORs between subgroups.
Additional exploratory analyses using Spearman's rank correlation coefficient (rs) were performed to test for potential correlations between responses in the acute and delayed phases. Statistical tests for heterogeneity were one-sided; statistical tests for effect estimates and publication bias were two-sided.
Source...