IntroductionColorectal cancer (CRC) has been linked with gut microbiota dysbiosis, thereby fostering the discovery of disease-associated microbial signatures. However, the use of covariate–adjusted methods accounting for confounders effects on microbiome shifts is far from being a gold standard approach, leading to spurious microbial-disease associations. This study aimed to characterize stool microbiota alterations along the adenoma-carcinoma sequence in a Portuguese cohort, and assess the robustness of candidate microbial features.MethodsStool samples from healthy individuals (HC), adenoma patients (AP), and CRC patients at early (SI/II) and advanced (SIII/IV) stages, were analyzed using 16S rRNA gene amplicon sequencing. Microbial features’ profiles were compared with publicly available datasets from France, Ireland, Italy, and USA/Canada. Differential abundance (DA) analysis was conducted through multiple DA models (ANCOM-BC2, fastANCOM, MaAsLin2, limma voom, and LEfSe) and statistical approaches, including covariate (age, sex, and body mass index (BMI)) adjustment, prevalence filtering, and analysis of matched subsets.Results and discussionAmong the altered taxa along de adenoma-carcinoma sequence, Sutterella and Desulfovibrio were generally enriched across different DA tools and matched subset analyses, while Parasutterella and Adlercreutzia were depleted. Sutterella, Parasutterella, and Adlercreutzia, were already altered in adenoma and early-stage CRC samples, maintaining or increasing their effect sizes towards advanced CRC stages. This pattern highlights the potential of these taxa as early microbial signatures of colorectal cancer progression. In contrast, Fusobacterium, which is a widely accepted CRC biomarker, was not detected after covariate adjustment. Cross-cohort comparison revealed a limited reproducibility of microbial features (i.e., Dialister was enriched in the Portuguese and French cohorts; Parasutterella and Terrisporobacter were depleted in the Portuguese and USA/Canada datasets), being most of them cohort-specific.ConclusionOverall, our results strengthen the importance of covariate control towards the identification of potentially robust and reproducible microbial markers of CRC onset and progression.