Objective:

To compare the pathogen-detection performance and clinical relevance of paired plasma and whole-blood droplet digital PCR (ddPCR) in critically ill adults with clinically suspected sepsis.

Approach:

• Study Design: This prospective multicenter cohort study was conducted in the intensive care units of 3 tertiary hospitals from December 8, 2022, through March 30, 2025.
• Patient Enrollment: The study enrolled 223 adults aged 18 years or older with suspected infection, new-onset organ dysfunction, and at least 1 specified clinical or laboratory feature suggestive of infection.
• Sample Collection: Conventional blood cultures and 2 additional peripheral blood samples were collected simultaneously. One sample underwent whole-blood processing, and the other was centrifuged for plasma ddPCR.
• ddPCR Methodology: Plasma and whole-blood assays used a predefined multiplex panel covering 17 bacterial targets and the Candida genus. Whole-blood processing included erythrocyte lysis, cellular disruption, and chemical and thermal lysis to improve recovery of cell-associated microbial DNA.
• Clinical Interpretation: Blood culture–negative, ddPCR-positive results were classified using structured adjudication based on microbiological, clinical, laboratory, radiologic, and treatment-response evidence.

Key Findings:

• Blood cultures were positive in 62 of 223 patients (27.8%); 4 cultured pathogens were not included in the ddPCR panel.
• Among 58 culture-positive cases involving panel pathogens, plasma and whole-blood ddPCR detected the corresponding pathogen in 54 (93.1%) and 55 (94.8%) cases, respectively.
• Among 161 blood culture–negative patients, plasma ddPCR supported 38 adjudicated definite sepsis cases, whereas whole-blood ddPCR supported 46, corresponding to additional diagnostic yields of 23.6% and 28.6%.
• Overall pathogen-level detection was numerically higher with whole blood than plasma (91.6% vs 81.1%), but the difference was not statistically significant.
• Whole-blood ddPCR detected the Candida genus significantly more frequently than plasma ddPCR (87.5% vs 31.3%; P = .03), although only 16 Candida detection events were evaluated.
• Plasma and whole-blood ddPCR had sensitivities of 87.1% and 90.3%, specificities of 80.7% and 71.4%, and negative predictive values of 94.2% and 95.0%, respectively; performance did not differ significantly between matrices.
• Concordant positivity in both matrices was associated with higher procalcitonin than single or double negativity and with higher lactate, SOFA, and APACHE II values than double negativity; however, 28-day survival did not differ significantly across ddPCR detection patterns.
• The laboratory ddPCR workflow usually required approximately 2–4 hours.

Interpretation:

Plasma and whole-blood ddPCR produced broadly comparable results and provided microbiological information beyond blood culture, particularly among culture-negative patients. Whole blood did not demonstrate overall diagnostic superiority but showed a potential advantage for Candida detection, possibly because it retains cell-associated microbial DNA removed during plasma preparation. Matrix selection should therefore reflect the suspected pathogen and clinical context, and ddPCR findings should be interpreted alongside cultures and other clinical evidence rather than used alone to establish active infection.

Limitations:

• The moderate sample size may limit the robustness and generalizability of subgroup findings.
• The adjudication framework incorporated ddPCR findings when classifying some ddPCR-positive, blood culture–negative cases, creating a risk of incorporation bias and overestimated diagnostic performance.
• All participants had received empirical antimicrobial treatment before enrollment, and its matrix-specific effects could not be formally assessed.
• The study was not designed to determine whether ddPCR-guided decisions improved antimicrobial use or patient outcomes.
• The predefined panel contained only 18 pathogen targets and missed organisms outside its coverage.
• Only 16 Candida detection events were identified, so the apparent whole-blood advantage requires confirmation in larger fungal-sepsis cohorts.
• Associations between concordant ddPCR positivity and disease severity were observational and do not establish causation.

Conclusion:

Plasma and whole-blood ddPCR both complemented conventional blood culture in critically ill patients with suspected sepsis, with only modest overall differences between matrices. Whole-blood ddPCR detected Candida more frequently in this cohort, but this finding is hypothesis-generating. Larger multicenter studies are needed to validate pathogen-specific matrix effects, refine specimen selection, and determine how ddPCR should be integrated into diagnostic and antimicrobial-stewardship workflows.