Gut microbiome biomarkers for colorectal cancer detection: a systematic review highlighting age as a key confounder

Citation (APA 7)

Chejara, P., & Eriksson, A. (2026). Gut microbiome biomarkers for colorectal cancer detection: a systematic review highlighting age as a key confounder. Frontiers in Oncology, 16, 1810802.

Abstract

The gut microbiome is a promising source of non-invasive biomarkers for colorectal cancer (CRC), yet the diagnostic performance remains heterogeneous, and the identified key markers often have inconsistency across studies, potentially influenced by host-specific factors. To ascertain the performance obtained by gut-based biomarkers in the field and investigate the impact of host characteristics on it, we conducted a systematic review and meta-analysis. After a comprehensive literature search on PubMed and Science Direct in January 2025, 27 studies were included in the review. On average, the studies reported Area Under the Receiver-Operator Curve (AUROC) of .89 and .80 for differentiating CRC and adenoma cases, respectively. We found a negative relationship between sample size and reported AUROC, though not statistically significant but still suggesting potential overfitting effect due to small dataset size, possibly inflating detection performance. Meta-regression analyses identified age difference between cases and controls as a significant moderator of diagnostic performance (p = 0.009), wherein cohorts with a higher age difference demonstrated a higher AUROC; in contrast, the BMI (Body Mass Index) and female ratio of study populations did not exert a significant influence. Regarding specific CRC-associated biomarkers, notable consistency was observed across studies, with an enrichment of Fusobacterium nucleatum, Parvimonas micra, and Peptostreptococcus stomatis and a depletion of protective butyrate producers like Roseburia spp. and Faecalibacterium prausnitzii. These findings confirm the robust aggregate potential of gut microbiome-based biomarkers for CRC diagnosis while highlighting that host age is a critical confounder that must be integrated into future predictive models to improve generalizability and clinical translation.