Inferring the role of the microbiome on survival in patients treated with immune checkpoint inhibitors: causal modeling, timing, and classes of concomitant medications

Abstract

ABSTRACTThe microbiome has been shown to affect the response to Immune Checkpoint Inhibitors (ICIs) in a small number of cancers. Here, we sought to more broadly survey cancers to identify those in which the microbiome will play a role using retrospective analyses. We created a causal model for the relationship between medications, the microbiome and ICI response and used it to guide the abstraction of electronic health records of 690 patients who received ICI therapy for advanced cancer. Medications associated with changes to the microbiome including antibiotics, corticosteroids, proton pump inhibitors, histamine receptor blockers, non-steroid anti-inflammatories and statins were abstracted. We tested the effect of medication timing on overall survival (OS) and evaluated the robustness of medication effects in each cancer. Finally, we compared the size of the effect observed for antibiotics classes to taxa correlated with ICI response and a literature review of culture-based antibiotic susceptibilities. Of the medications assessed, only antibiotics and corticosteroids significantly associated with lower OS. The hazard ratios (HRs) for antibiotics and corticosteroids were highest near the start of ICI treatment but remained significant when given prior to ICI. Antibiotics and corticosteroids remained significantly associated with OS even when controlling for multiple factors such as Eastern Cooperative Oncology Group performance status, Charlson Comorbidity Index score, and stage. When grouping antibiotics by class, β-lactams showed the strongest association with OS across all tested cancers. The timing and strength of these effects after controlling for confounding factors are consistent with role for the microbiome in response to ICIs.