Association between gut microbiota characteristics and therapeutic efficacy of multimodal combination therapy in hepatocellular carcinoma with portal vein tumor thrombus: a secondary analysis of a prospective phase II trial.

Hepatocellular carcinoma (HCC) with portal vein tumor thrombus (PVTT) carries a poor prognosis. Although SBRT combined with cadonilimab and lenvatinib has shown encouraging antitumor activity in a prior prospective phase II trial, marked interpatient heterogeneity remains. Gut microbiota may influence antitumor immunity and could help identify patients more or less likely to benefit from this regimen. To characterize baseline and longitudinal gut microbiota features in treatment-naïve HCC patients with PVTT receiving SBRT plus cadonilimab and lenvatinib, and to explore microbial signatures associated with therapeutic response. This is a complementary exploratory study nested within the aforementioned phase II trial. In this prospective multicenter exploratory cohort, 23 patients from the parent trial provided 46 paired fecal samples collected before treatment (Phase A) and after 3 cycles of cadonilimab (Phase B). Patients were categorized by best mRECIST response as responders (group C, CR/PR; n = 10) or non-responders (group D, SD/PD; n = 13), consistent with the parent trial's grouping strategy. 16S rRNA amplicon sequencing and downstream bioinformatic analyses were performed to assess microbial diversity, differential taxa, predicted KEGG functional profiles, and exploratory discriminatory performance. A total of 1,410,203 high-quality non-chimeric reads were retained and rarefied to 5,459 reads per sample. Baseline α-diversity indices were largely comparable between groups, although Faith's phylogenetic diversity was higher in responders (P = 0.0178). β-diversity did not show clear between-group separation at baseline or marked restructuring after treatment. Genus-level DESeq2 analysis identified 7 baseline differential genera. Extibacter was absent in responders but detectable in 6/13 non-responders and showed the strongest discriminatory signal (P = 0.0179). Predicted functional analysis suggested differences in KEGG level 2 modules, and [Ruminococcus]_gnavus_group was positively correlated with "immune diseases" and "carbohydrate metabolism" after BH correction (q = 0.042). In exploratory ROC analysis, baseline Extibacter yielded an AUC of 0.731, increasing to 0.777 when combined with AFP. In HCC patients with PVTT treated with SBRT plus cadonilimab and lenvatinib, the overall gut microbiota architecture remained relatively stable, whereas selected baseline genus-level features were associated with therapeutic response. Extibacter showed an exploratory non-response-associated signal that requires validation in larger independent studies. ClinicalTrials.gov, identifier NCT06040177.