e14620 Background: Immunotherapy combinations are used to improve outcomes in metastatic cancer, but knowledge of appropriate starting doses for novel combinations is lacking. Methods: Phase I-III adult combination clinical trials (≥1 drug was immunotherapy; anti-PD1, PDL1, or CTLA4) were reviewed (PubMed Jan 1, 2010 to Sep 1, 2016; ASCO 2014-2016, ASH/ESMO 2014-2015 abstracts). The safe dose used in each combination was divided by the single agent recommended dose to calculate dose percentage. Additive dose percentage was the sum of each dose percentage. Results: 84 studies (N=3,526 patients, 60 combinations) were analyzed (see Table). Eight studies had unacceptable safety: ipilimumab with cytoxan, vemurafenib, dacarbazine, dabrafenib/trametinib, or carboplatin/etoposide; nivolumab or pembrolizumab with pazopanib; and tremelimumab with sunitinib. Lowest additive dose percentages evaluated in these studies were: 160, 175, 175, 217, 300, 167, 175, and 167%; it is unclear if lower doses would have been tolerable. Combinations with ipilimumab had lower additive dose percentages. Response rates (median [interquartile range]) were higher for 3 than 2 drug combinations (53% [33%-63%] (N = 23 studies) vs. 23% [14%-39%]) (N = 60 studies) with similar rates seen for targeted, cytotoxic, or dual immunotherapy combinations. Conclusions: In 63% of studies not including ipilimumab, full doses of combinations could be given. Triplet combinations had median response rates of 53 vs. 23% for doublets (p<0.001). These findings can help inform safe starting doses for novel immunotherapy combinations in clinical trials and practice. [Table: see text]