Background Early concurrent antibiotic usage in patients receiving immune checkpoint inhibitors (ICIs) is linked to decreased progression-free survival (PFS) and overall survival (OS), likely mediated by lower gut microbial diversity and skewed taxonomic abundance. The relationship of late antibiotic exposure to ICI efficacy has not been explored.

Methods Data from a single-arm, Phase 1/2 study (Study 1108) were retrospectively analyzed in 945 patients with metastatic disease in 18 tumor types (largest subsets: non-small-cell lung cancer [NSCLC], 31.5% and urothelial cancer [UC], 20.8%) receiving durvalumab 10 mg/kg Q2W between 8/2012–6/2017. Early antibiotic exposure was defined as 30 days prior to until 30 days after durvalumab initiation. Late exposure was any time >30 days after durvalumab initiation. Demographics, infection type, and antibiotic data (route, duration, class) were collected. Median PFS and OS were compared between no antibiotics (n=525), early antibiotics (n=239), and late antibiotics (n=181) by Kaplan-Meier methods and log-rank tests. The Cox proportional hazards model was used for multivariable adjustments. In a translational in vivo analysis, Balb/c mice implanted with CT26 tumors were prospectively treated with oral levofloxacin 1 mg/mL for 7 days either 1 week prior to, concurrently with, or 1 week after initiating anti-PD-L1 or control IgG therapy (n=40 per group). The primary endpoint was tumor growth kinetics.

Results ß-lactams (51.3%) and fluoroquinolones (39.3%) were most commonly prescribed overall. Early antibiotic exposure was associated with reduced mOS compared to no exposure (7.2 vs 9.8 months, p=0.049). Unexpectedly, patients with late antibiotic exposure had markedly improved mOS versus those with no exposure (19.8 vs 9.8 months, p<0.001), which persisted after adjusting for baseline tumor volume, demographics, treatment-induced immune-related adverse events, and neutrophil to lymphocyte ratio (table 1). In NSCLC and UC cohorts, these results were preserved. To account for time-on-treatment bias, an independent model using antibiotic start time as a time-dependent covariate also showed improved benefit with late antibiotics (HR 0.91, 95% CI 0.87–0.95, p<0.001). Compared to control IgG, mice receiving prior or concurrent levofloxacin with anti-PD-L1 had similar survival, while mice receiving late levofloxacin with anti-PD-L1 had improved survival (p=0.004).

Conclusions This large retrospective analysis is consistent with previous studies showing associations of early antibiotics with shorter survival during ICI therapy. For the first time, we report that late antibiotics do not negatively impact survival and may actually benefit survival. Future studies will evaluate immune phenotyping and microbiome characterization to clarify mechanistic underpinnings. These findings will require confirmation in prospective clinical studies.

Trial Registration clinicaltrials.gov NCT01693562

Ethics Approval This study was conducted according to the Declaration of Helsinki and approved by the independent ethics committee/institutional review board at each participating center.

Consent Informed consent was obtained from all patients.