Saccharomyces cerevisiae and Bacillus subtilis show modest, node-dependent effects on insulin sensitivity in periparturient cows.

Periparturient immunometabolic adaptations yield small, node-dependent shifts in endotoxin proxies and insulin-sensitivity indices. Direct-fed microbials (DFM) are mechanistically plausible but often produce modest metabolic readouts. Test whether Saccharomyces cerevisiae (SC), Bacillus subtilis (BS), or SC + BS alter endotoxin (blood, milk) and metabolic families versus Control (Ctrl) across prespecified nodes, and evaluate prespecified equivalence where applicable. A 4-arm trial enrolled 20 Holsteins from -60 to + 21 d around calving (Ctrl, SC, BS, SC+BS). Blood and milk were sampled at multiple timepoints. Mixed models estimated marginal means and contrasts; BH-FDR controlled multiplicity. Endotoxin was log-scale modeled (back-transformed); insulin used identity scale. TOST/ROPE applied where predefined. Serum AUC was computed for pre- and post-calving windows with robust checks. After BH-FDR, no significant differences were found between treatment arms for endotoxin or metabolic families. Endotoxin geometric means overlapped at -10, + 10, and + 20 d. QUICKI met prespecified equivalence at multiple nodes; endotoxin and RQUICKI did not. AUC results were consistent across estimators. SC, BS, and SC + BS did not alter endotoxin/metabolic families. QUICKI met equivalence at several nodes. Findings indicate small, localized effects in early lactation, underscoring need for clear reporting, multiplicity control, and prespecified equivalence in future studies.