Contribution of the microbiome to differential drug efficacy and clinical trial outcomes between the US and Japan.

Abstract

3086 Background: The gut microbiome is a source of patient variation that influences the efficacy of anticancer drugs (Spanogiannopoulos et al., 2022) . The microbiome also varies at a population level (Pasolli et al., 2019) .There is a limited understanding of the impact of this variation. We show for the first time that differences in drug-microbiome interactions may translate to differential clinical trial outcomes. For extensive-stage, small cell lung cancer (SCLC), irinotecan-cisplatin is a first-line treatment in Japan, based on data from Japan Clinical Oncology Group (JCOG) 9511 (Noda et al., 2002) demonstrating superior efficacy to etoposide-cisplatin (RR 87% vs 68%; PFS 12.8 vs 9.4 months). Despite employing an identical study design, the Southwest Oncology Group (SWOG) 0124 (Lara et al., 2009) failed to confirm this benefit for a US population. The reasons for this discrepancy are unclear. Methods: Microbiome analysis was performed using BioCorteX’s industry-leading knowledge graph and integrated proprietary engines v20231108_220940. 56,171 stool microbiome samples from the US adult population and 3,558 stool samples from the Japanese adult population were included. A database built on existing literature was used to identify proteins produced by organisms in these samples. These microbial-derived proteins were analysed for sequence similarity to human proteins that interact with irinotecan, etoposide and cisplatin. Interactions were ranked by the abundance of microbial genera producing a given protein. This was normalised by interaction prevalence to produce a clinical impact metric denoted the “Numbers Needed to Interact” (NNI). Results: Our results show that the NNI for irinotecan is almost 5 times lower in the US (US: 6.2; Japan: 30.9). By contrast, the NNI for both cisplatin and etoposide (US: 2.3; Japan: 1.7) is similar between countries. A pooled analysis of anticancer drugs shows that a negative Spearman’s rank correlation can be derived for NNI vs trial failure rate (ρ=-0.86, p<0.01). Analysis of the drug-microbiome interactions for irinotecan revealed 8 distinct, microbial-derived proteins with significant homology to proteins encoding human topoisomerases, including TOP1. All 8 of these molecular mimics were enriched in the US population. Only a single protein was found in Japanese stool samples. Conclusions: To our knowledge, this work represents the first demonstration that drug-microbiome interactions may contribute to the efficacy of anticancer drugs, as well as to differential trial outcomes between countries. Proteins produced by the microbiome that functionally mimic host proteins may interfere with the pharmacodynamics of cancer therapeutics, including irinotecan. This interaction is found 5 times more frequently in US stool samples than Japanese samples and may contribute to the observed differences in outcomes in JCOG9511 and SWOG0124.