### Methods

#### Search Strategy and Selection Criteria

We systematically searched the PubMed, Embase, and Cochrane Central Register of Controlled Trials databases using a combination of the terms "neurokinin-1 receptor antagonist," "aprepitant," "casopitant," "fosaprepitant," "netupitant," and "rolapitant" to find relevant articles published up to February 2016. A manual search through reference lists of relevant reviews was additionally performed. Three authors carried out the literature retrieval independently **(**YZ, YY, ZZ). Eligible studies met the following criteria: 1) they were randomized control trials (RCTs) or prospective studies that evaluated NK-1RA-based triple antiemetic regimens in the prophylaxis of CINV; 2) efficacy and/or toxicity measures were available; 3) NK-1RA was used at the standard dose. Studies failing to meet these inclusion criteria were excluded.

#### Outcomes Measures, Data Extraction, and Quality Assessment

The outcomes of antiemetic efficacy were the proportions of patients with complete responses (CRs) and no clinically significant nausea in the acute (0–24 hours after chemotherapy), delayed (>24–120 hours after chemotherapy), and overall (0–120 hours) phases. The toxic outcome was defined as treatment-related adverse events (TRAEs). The assessment of efficacy and toxicity occurred during the first cycle of chemotherapy.

The data on trial name, therapeutic and antiemetic regimens, and clinical outcomes were extracted by two investigators independently (YZ and ZZ). HEC, such as anthracycline plus cyclophosphamide (AC) or cisplatin, and MEC, such as carboplatin or oxaliplatin, were defined according to the National Comprehensive Cancer Network Antiemesis Guideline Version 2, 2016.^{[11]} Cochrane risk of bias was used to assess the quality of all included studies by another two reviewers (YL and YY).^{[12]} Discrepancies were discussed by all investigators to reach a consensus. More details on this can be found in Supplementary Figure 1 https://academic.oup.com/jnci/article/109/2/djw217/2572051/Neurokinin-1-Receptor-Antagonist-Based-Triple#supplementary-data (available online).

#### Statistical Analyses

First, we conducted pair-wise meta-analyses using a random-effects model to synthesize studies comparing the same pair of antiemetic treatments. The results were reported as pooled odds ratios (ORs) with corresponding 95% confidence intervals (CIs). Statistical heterogeneity across studies was assessed with a forest plot and the inconsistency statistic (I^{2}). Statistical significance was set at a *P* value of .05. All calculations were performed using REVIEW MANAGER (version 5.2 for Windows; the Cochrane Collaboration, Oxford, UK). The statistical methods to detect funnel plot asymmetry were the rank correlation test of Begg and Mazumdar and the regression asymmetry test of Egger.^{[13,14]}

Secondly, a random-effects network within a Bayesian framework using the Markov chain Monte Carlo methods was built using ADDIS 1.15.^{[15,16]} We networked binary clinical outcomes within studies and specified the relations among the ORs across studies to make comparisons of different antiemetic treatments in terms of efficacy and/or toxicity. *P* values of less than .05 and 95% confidence intervals were used to assess statistical significance. The inconsistency within this multiple treatment comparison was evaluated by a variance calculation as previously described.^{[16]} All statistical tests were two-sided.

J Natl Cancer Inst. 2016;109(2) © 2016 Oxford University Press