Rapid and accurate antimicrobial susceptibility testing (AST) is critical for guiding therapy in life-threatening infections such as sepsis, but current diagnostics rely on blood culture, delaying results by 48–72 h. This time lag forces empirical use of broad-spectrum antibiotics, increasing risks of treatment failure and antimicrobial resistance. We developed an approach combining bacterial growth-independent nanomotion detection with magnetic bead–based enrichment directly from blood. Using machine-learning classifiers trained on nanomotion spectra, we established ten antibiotic-specific models across 1,337 blood culture samples containing 390 clinical isolates of Enterobacterales (e.g., Escherichia coli, Klebsiella pneumoniae, Enterobacter cloacae) and Pseudomonas aeruginosa. These models achieved 94–98% sensitivity and specificity with fixed 2-h readouts. In parallel, SepsiSTAT enrichment enabled rapid recovery of viable bacteria and yeast at clinically relevant concentrations from 10 mL blood, with time-to-positivity averaging 5.5 h, i.e., over 7 h faster and more predictable than conventional blood culturing. The workflow was afterwards adapted to 1 mL input volumes, compatible with blood volume sampling constraints in fragile patient groups such as geriatric or neonatal patients. In spiked E. coli blood samples, ceftriaxone susceptibility was determined within 9–11 h of collection, nearly two days earlier than current standards. By integrating enrichment with nanomotion AST, we introduce a low-volume, label-free, phenotypic diagnostic platform capable of delivering actionable results within clinically meaningful timeframes. This approach holds potential to improve antibiotic stewardship, enable earlier transition from empiric to targeted therapy, and expand diagnostic access for vulnerable populations such as neonates.
Jóźwiak et al. (Fri,) studied this question.