Baseline gut microbiota and short-term response to acupressure combined with thermal stimulation in maintenance hemodialysis patients with functional constipation: an exploratory responder analysis

Background Constipation is common in patients undergoing maintenance hemodialysis (MHD), and individual responses to non-pharmacological interventions may vary substantially. This exploratory study investigated whether baseline gut microbiota characteristics were associated with short-term response to acupressure combined with thermal stimulation in MHD patients with functional constipation. Methods This responder analysis was based on a randomized two-sequence crossover study. Clinical data and fecal samples obtained immediately before each participant entered the 14-day intervention period were analyzed. Participants with an increase of ≥1 bowel movement (BM) per week were classified as responders (R), and those with <1 BM/week as non-responders (NR). Baseline gut microbiota were characterized using 16S rRNA gene sequencing. Microbial diversity, taxonomic composition, and genus-level differential abundance were evaluated. ANCOM-BC2 was performed with adjustment for baseline BM/week, sex, and diabetes mellitus, followed by Benjamini–Hochberg false-discovery-rate correction and pseudocount sensitivity analysis. Results Twenty-six participants were included, of whom 9 (34.6%) met the predefined response criterion. Across all participants, weekly BM frequency increased from 1.25 (1.00, 1.50) to 2.00 (1.00, 3.00) (P = 0.0003), while PAC-SYM total and stool-symptom subscale scores decreased (P = 0.0011 and P = 0.0013, respectively). Alpha-diversity did not differ significantly between R and NR groups, and beta-diversity showed no overall community separation (ANOSIM R=−0.044, P = 0.6751). Lactonifactor was nominally higher in responders in the unadjusted analysis (P = 0.015) but was not significant after multiple-testing correction (FDR = 0.86). In adjusted ANCOM-BC2 analysis, Veillonella (β=5.319, FDR = 0.0347) and Lactobacillus (β=3.621, FDR = 0.0459) were associated with responder status; however, neither association was robust to pseudocount sensitivity analysis. PICRUSt2 identified no significant between-group differences in predicted functional pathways. Conclusion Baseline microbial diversity and community structure were similar between responders and non-responders. Veillonella and Lactobacillus showed adjusted associations with response, but neither was robust in sensitivity analysis. These findings should therefore be regarded as exploratory rather than predictive. Larger longitudinal studies are needed to validate these findings and determine their clinical relevance.