**Background:** Motility, stool characteristics, and microbiota composition are expected to modulate probiotics' passage through the gut, but their effects on persistence after intake cessation remain uncharacterized. This pilot, open-label study aimed to characterize probiotic fecal detection parameters (onset, persistence, and duration) and their relationship with whole gut transit time (WGTT), and to explore correlations with fecal microbiota composition.
**Methods:** Thirty healthy adults (30.4 ± 13.3 years, 12 male/18 female) were enrolled in a 10-week study conducted in Florida between December 2019 and February 2020. Participants received a multi-strain probiotic (30 × 10^9 CFU/capsule/day for 14 days) containing L. helveticus R0052 (45%), L. paracasei HA-108 (17%), B. breve HA-129 (16%), B. longum R0175 (6%), and S. thermophilus HA-110 (16%). Probiotic intake was flanked by 4-week washout periods, with 18 stool collections throughout the study. Strain-specific detection was performed by qPCR using strain-specific primers. WGTT was measured using 80% recovery of radio-opaque markers (Sitzmarks, 1 capsule/day from day 0 to day 2). Baseline fecal microbiota was profiled via 16S rRNA V3-V4 region amplicon sequencing. Machine learning (ExtraTrees algorithm) was used to classify WGTT subgroups and identify discriminant taxa.
**Key Results:** The average WGTT was 54.5 hours (95% CI, 47.7-61.3). Three distinct WGTT subgroups were identified: Fast (mean 28.2 hr, n=5), Intermediate (mean 47.3 hr, n=14), and Slow (mean 75.6 hr, n=11). All strains were detected in feces approximately 1-2 days after first intake, with no significant difference in onset between strains (P=0.533). For R0052, HA-108, and HA-129, persistence after intake cessation was not significantly different (~3-6 days). However, B. longum R0175 persisted significantly longer (mean 8.5 ± 1.7 days, P<0.01 vs R0052), driven by 6 of 13 Intermediate subgroup participants in whom R0175 persisted ≥15 days. Machine learning classified the three WGTT subgroups with 89.9% overall accuracy (100% for Intermediate vs Slow). Taxa including Roseburia, Eggerthella, Acidaminococcus, Prevotella, Sutterella, and Bifidobacterium were less abundant in the Slow subgroup, while Oscillospira, Lachnospira, Alistipes, Rikenellaceae, Akkermansia, and Turicibacter were more abundant. Within the Intermediate subgroup, machine learning distinguished participants with longer (≥15 days) vs shorter (<5 days) R0175 persistence with 100% accuracy, identifying 31 discriminating features. Longer R0175 persistence was associated with lower baseline abundance of Bifidobacterium, Coprococcus, Erysipelotrichaceae, and Ruminococcaceae, and higher abundance of Blautia, Roseburia, Parabacteroides, and Eggerthella.
**Clinical Implications:** These results support the notion that host-specific parameters such as WGTT and microbiota composition should be considered when designing studies involving probiotics, especially for optimizing washout duration in crossover studies and defining enrollment criteria or supplementation regimens. The finding that B. longum R0175 persistence is influenced by both WGTT and specific microbiota composition suggests that some individuals may have a more 'permissive' microbiome for certain probiotic strains. The study is limited by its small cohort size (n=30), lack of dietary control, and assessment of WGTT only during the first 3 days of probiotic intake. The use of qPCR rather than culture-based methods may overestimate persistence duration by detecting non-viable bacteria.