Immune signature-based uncoupling of checkpoint inhibitor efficacy and toxicity.

Personalized escalation and de-escalation of immune checkpoint inhibitor (ICI) regimens may help to overcome upfront resistance and mitigate the risk for immune-related adverse events (irAEs). Here, we examined the association between pathological response and irAEs per ICI regimen. Meta-analysis of neoadjuvant ICI trials in melanoma illustrated a pattern of increased toxicity and efficacy with the addition and/or higher dosing of anti-CTLA-4 to anti-PD-1. We subgrouped anti-PD-1, low-dose anti-CTLA-4 + anti-PD-1, high-dose anti-CTLA-4 + anti-PD-1, and anti-PD-1 + anti-LAG-3 cohorts according to the baseline interferon-gamma (IFN-γ) signature and analyzed these for response and toxicity rates. Whereas in IFN-γ high subgroups the addition of (high-dose) anti-CTLA-4 increased toxicity but not efficacy, the addition of high-dose anti-CTLA-4 to anti-PD-1 increased efficacy in the IFN-γ low subgroup, while toxicity remained low. Our findings suggest that baseline immune signatures may be used to separate risk for toxicity from risk for non-response, with implications for patient stratification and treatment regimens.