Publication

• Title: Effects of a clinical metagenomics intervention on clinical outcomes, healthcare costs, and health-related quality of life in patients with sepsis or septic shock: results of the randomized-controlled DigiSep trial
• Acronym: DigiSep
• Year: 2026
• Journal published in: Intensive Care Medicine
• Citation: Brenner T, Skarabis A, Schaller SJ, von Groote T, Putensen C, Günther U, et al. Effects of a clinical metagenomics intervention on clinical outcomes, healthcare costs, and health-related quality of life in patients with sepsis or septic shock: results of the randomized-controlled DigiSep trial. Intensive Care Med. 2026.

Context & Rationale

• Background

  • Sepsis is life-threatening organ dysfunction caused by a dysregulated host response to infection, and remains a major global cause of death and disability.¹²
  • Modern sepsis guidance emphasises early antimicrobials in septic shock or highly likely sepsis, but also stresses diagnostic stewardship, avoidance of unnecessary broad-spectrum therapy, and de-escalation when microbiology allows.³
  • Conventional cultures remain central to sepsis microbiology, but they are slow, have limited sensitivity after antimicrobial exposure, and may not detect fastidious, localised, fungal, viral, or polymicrobial infection early enough to guide treatment.
  • Metagenomic next-generation sequencing (mNGS) of microbial circulating cell-free DNA offers an untargeted diagnostic approach capable of detecting bacteria, fungi, parasites, and DNA viruses from plasma or other body fluids, without requiring organism growth in culture.⁴
• Research Question/Hypothesis

  • The key gap was not diagnostic sensitivity alone, but whether adding mNGS to routine microbiology changes patient-centred outcomes, antimicrobial exposure, quality of life, and healthcare costs in ICU sepsis.
  • The preceding non-interventional Next GeneSiS trial reported higher pathogen positivity with mNGS than blood culture within 3 days of sepsis onset, and suggested that knowledge of mNGS results could have changed anti-infective therapy in approximately one-third of patients, but could not establish clinical benefit.⁵
• Why This Matters

  • DigiSep tested whether an apparently more informative diagnostic strategy could improve the trade-off between adequate early therapy and antimicrobial overuse, rather than simply increasing the number of organisms detected.

Design & Methods

• Research Question: In adult ICU patients with sepsis or septic shock diagnosed within 24 hours, does adding plasma microbial cell-free DNA mNGS to standard microbiological diagnostics improve the 28-day DOOR/RADAR score, clinical outcomes, health-related quality of life, and costs compared with standard microbiology alone?
• Study Type: Prospective, individual-patient randomised, controlled, open-label, interventional, multicentre trial in 24 German intensive care units.
• Population:
  • Adults aged ≥18 years.
  • Sepsis or septic shock according to Sepsis-3.
  • Sepsis onset within <24 hours of inclusion.
  • Written informed consent from the patient or legal representative.
  • Key exclusions: age <18 years, refusal or withdrawal of consent, anticipated ICU discharge within the first 72 hours, primary palliative treatment intent, no commitment to aggressive treatment, death deemed imminent or inevitable, or readmission during the same hospitalisation after previous enrolment.
• Intervention:
  • Standard-of-care microbiology plus mNGS-based diagnostics using the CE-IVD-marked DISQVER® test.
  • Blood for mNGS was obtained at baseline and day 3 at the same time as blood cultures; additional mNGS sampling could occur up to day 14 when further blood cultures were clinically requested.
  • Samples were transported overnight to an external reference laboratory.
  • Results were delivered through a digital portal to the commissioning physician.
  • Optional infectious diseases expert consultation was available, particularly when mNGS and routine microbiology were discordant.
  • Antimicrobial choice and treatment duration were left to local clinical teams, not mandated by the protocol.
• Comparison:
  • Standard-of-care microbiological analyses alone, according to local institutional practice.
  • Routine diagnostics included blood culture sets at baseline and day 3, with other cultures and rapid multiplex PCR permitted where part of local practice.
  • Optional infectious diseases consultation was also available in the control group.
• Blinding: Treating clinicians were not blinded. Patients were not expected to infer allocation because of their critical illness and identical study visits, but this was not a double-blind trial.
• Statistics: The protocol planned 410 randomised patients, 205 per arm, allowing 12.5% attrition, to provide 90% power to detect a common-language relative effect of p=0.6 for a lower DOOR/RADAR score in the mNGS group using a two-sided 5% significance level.⁶ The primary analysis used a Mann–Whitney U test in the intention-to-treat/full-analysis population, with multiple imputation for missing DOOR/RADAR components. DOOR/RADAR is an ordinal desirability ranking adjusted for antimicrobial exposure.⁷
• Follow-Up Period: Primary endpoint at 28 days; clinical, survival, health-related quality-of-life, and healthcare-cost follow-up to 90 and 180 days.

Key Results

This trial was not stopped early. There was no prespecified interim analysis; 409 patients were recruited, 20 were excluded after randomisation/enrolment, and 389 were included in the final analysis: 200 in the mNGS plus standard-care group and 189 in the standard-care group.

• The primary DOOR/RADAR endpoint was not statistically significantly improved: 3.21 ± 1.54 with mNGS plus standard care versus 3.49 ± 1.51 with standard care, mean difference −0.28; 95% CI −0.58 to 0.03; P=0.149.
• The clearest clinical signals were shorter mechanical ventilation and faster shock resolution, but these were secondary, unadjusted, exploratory findings.
• Important subgroup signals were exploratory and multiplicity-unadjusted: among patients receiving source control, DOOR/RADAR was 2.93 ± 1.49 with mNGS versus 3.43 ± 1.53 with standard care, mean difference −0.50; 95% CI −0.94 to −0.06; P=0.030; in the small catheter-removal subgroup, 2.73 ± 1.62 versus 4.19 ± 1.10, mean difference −1.46; 95% CI −2.57 to −0.34; P=0.028.

Internal Validity

• Randomisation and Allocation:
  • Randomisation was individual, 1:1, centre-stratified, and used an internet-based tool with permuted block lengths.
  • Personnel assigning patients had no access to the allocation sequence, although prior allocations could be viewed during the trial.
  • The final analysed groups were reasonably balanced despite modest numerical imbalance: 200 mNGS patients versus 189 controls.
• Dropout or Exclusions:
  • 409 patients were enrolled/randomised, but 20 were excluded after allocation: 11 control and 9 intervention.
  • Reasons included not fulfilling Sepsis-3 criteria (1 in each group), private health insurance (1 control, 4 intervention), lack of informed consent (4 control, 5 intervention), withdrawal of consent (3 control, 2 intervention), and expected ICU transfer within <72 hours (1 control, 0 intervention).
  • This makes the final analysis closer to a modified intention-to-treat/full-analysis-set approach than a strict all-randomised ITT analysis, although the number and balance of exclusions make major selection distortion unlikely.
• Performance and Detection Bias:
  • The trial was open-label to treating clinicians.
  • The primary endpoint included objective elements such as mortality and kidney replacement therapy, but also clinician- and system-dependent elements such as ICU length of stay and antimicrobial duration.
  • Key secondary endpoints, especially ventilation duration and catecholamine discontinuation, are vulnerable to unblinded bedside practice variation.
• Protocol Adherence:
  • Diagnostic separation was achieved: mNGS results were available to clinicians in the intervention group and not available for usual-care decision-making in the control group.
  • Therapeutic separation was weak: documented mNGS-driven therapy change occurred in only 14.3% of evaluable intervention-group patients (20/140), despite retrospective expert recommendation suggesting that treatment modification would have been appropriate in 32.3% of control patients.
  • Optional infectious diseases consultation was rarely used: 2 intervention-group patients and 0 control-group patients.
• Baseline Characteristics:
  • Groups were clinically similar: age 64.1 ± 14.3 versus 64.3 ± 15.6 years; male sex 60.0% versus 62.4%; septic shock 75.5% versus 78.8%; SOFA score 8.5 ± 2.7 versus 8.8 ± 2.6.
  • Infection source was also similar: outpatient-acquired infection 59.5% versus 58.2%, pulmonary/thoracic focus 39.0% versus 41.8%, and intra-abdominal/biliary/gastrointestinal focus 27.5% versus 24.3%.
  • Patients were sufficiently sick for a plausible intervention effect, with approximately three-quarters in septic shock.
• Heterogeneity:
  • The trial included 24 ICUs, multiple infection sources, community and nosocomial sepsis, and local variation in microbiology, rapid PCR availability, and antimicrobial practice.
  • Centre-stratified randomisation mitigated but did not remove dilution from heterogeneous local practice and heterogeneous mechanisms of sepsis.
• Timing:
  • The intervention targeted early sepsis, with enrolment within 24 hours and scheduled mNGS sampling at baseline and day 3.
  • The send-in laboratory model delayed actionable information: transport logistics took 1.59 ± 0.97 days and overall mNGS time-to-result was 2.61 ± 0.97 days.
  • Once samples arrived at the laboratory, 98.2% were reported within 24 hours, making transport and workflow rather than sequencing itself the dominant delay.
• Dose:
  • The “dose” of diagnostics was substantial: baseline plus day-3 mNGS sampling, with further sampling up to day 14 when additional blood cultures were requested.
  • The “dose” of implementation support was insufficient: expert consultation was optional and rarely used; therapy was not standardised after mNGS reporting.
• Separation of the Variable of Interest:
  • Analytical separation was large: at day 3, mNGS positivity was 45.9% versus 4.7% for blood culture.
  • Clinical treatment separation was limited: anti-infective treatment duration was 14.8 ± 8.1 days versus 15.5 ± 8.2 days, mean difference −0.7 days; 95% CI −2.60 to 1.20; P=0.467.
  • Clinical outcome separation was most apparent in ventilation duration, 6.6 ± 9.4 versus 9.3 ± 10.6 days, and shock resolution, 6.9 ± 7.4 versus 8.8 ± 8.5 days.
• Key Delivery Aspects:
  • The trial tested a pragmatic diagnostic-service model rather than an integrated mandatory antimicrobial-stewardship pathway.
  • This distinction is central: a diagnostic test cannot improve outcomes unless the result is timely, trusted, interpreted correctly, and acted upon.
• Outcome Assessment:
  • DOOR/RADAR is conceptually attractive because it jointly ranks survival, organ support, length of stay, and antimicrobial exposure.
  • However, its composite nature may obscure opposing effects: the trial showed large diagnostic-yield differences and some secondary clinical signals without a statistically significant primary ranking effect.
  • Health-related quality-of-life data were incomplete at follow-up, especially for VAS components, and cost data were available for only 33.2% of patients.
• Statistical Rigor:
  • The statistical analysis plan was finalised before database closure.
  • The primary endpoint analysis matched the trial design and used multiple imputation for missing DOOR/RADAR components.
  • Secondary endpoints were not adjusted for multiplicity; therefore, findings with P values close to 0.05, including ventilation duration, shock resolution, and 90-day EQ-5D-5L, should be interpreted as hypothesis-generating.

Conclusion on Internal Validity: Internal validity is moderate for the clinical-effect question and strong for the diagnostic-yield question. Randomisation, multicentre conduct, prespecified analyses, and balanced baseline characteristics are strengths, but open-label care, post-randomisation exclusions, delayed result availability, and weak mNGS-guided treatment adoption substantially limit causal inference about patient outcome benefit.

External Validity

• Population Representativeness:
  • The trial population is highly relevant to critical care: adults with early ICU sepsis or septic shock, with high illness severity and common infection foci.
  • The cohort was broadly typical of high-acuity sepsis practice in high-income ICUs: mean age approximately 64 years, septic shock in approximately three-quarters, and pulmonary or intra-abdominal sources predominating.
  • Patients with imminent death, primary palliation, no aggressive-treatment intent, or expected rapid ICU discharge were excluded, limiting applicability at both extremes of illness severity.
• Applicability:
  • Generalisation is strongest to German or similar high-resource ICUs with access to blood cultures, local microbiology, courier logistics, digital reporting, and specialist antimicrobial advice.
  • Translation to resource-limited settings is limited by platform cost, laboratory infrastructure, sequencing capacity, and the need for interpretive antimicrobial stewardship.
  • The tested model was an external-reference-laboratory send-in pathway; hospitals with on-site sequencing or mandatory infectious diseases review may achieve earlier action and greater treatment separation.
  • Settings already using intensive infectious diseases consultation, rapid multiplex PCR, and mature antimicrobial stewardship may see smaller incremental benefit.

Conclusion on External Validity: DigiSep is generalisable to high-income ICU sepsis systems considering mNGS implementation, but less generalisable to centres without rapid logistics, sequencing infrastructure, or embedded stewardship. The findings apply most directly to mNGS as a pragmatic add-on diagnostic, not to mNGS embedded within a mandatory expert-guided antimicrobial pathway.

Strengths & Limitations

• Strengths:
  • First randomised trial evaluating mNGS-guided diagnostics for clinical outcomes in ICU patients with sepsis or septic shock.
  • Multicentre design across 24 ICUs, improving practice representativeness within a high-resource system.
  • Clinically meaningful primary endpoint integrating outcome desirability and antimicrobial exposure.
  • Prespecified statistical analysis plan and no interim stopping.
  • Broad outcome assessment including ventilation, shock resolution, kidney replacement therapy, antimicrobial duration, survival, health-related quality of life, and costs.
  • Retrospective expert-panel assessment provided mechanistic insight into diagnostic plausibility and therapeutic relevance.
• Limitations:
  • Open-label clinician management increases risk of performance bias for secondary bedside outcomes.
  • 20 patients were excluded after randomisation/enrolment, so the final analysis was not a pure all-randomised ITT analysis.
  • The mNGS result was often too delayed for the earliest antimicrobial decision window: time-to-result 2.61 ± 0.97 days.
  • Only 14.3% of evaluable intervention patients had documented treatment changes based on mNGS, indicating weak clinical implementation.
  • Expert consultation was optional and almost unused: 2 patients in the intervention group and none in the control group.
  • Blood-culture time-to-result was not systematically captured, limiting direct comparison of real-world diagnostic speed.
  • Blood-culture plausibility was not systematically adjudicated in the same way as mNGS plausibility.
  • Secondary outcomes were not multiplicity-adjusted and should be treated as exploratory.
  • Healthcare cost data were available for only 129 patients, 33.2% of the cohort.
  • Some investigators had relationships with the diagnostic manufacturer, although the main funder was the German Innovation Fund and the manuscript reports no funder role in conduct or analysis.

Interpretation & Why It Matters

• Main interpretation

  Adding mNGS to usual microbiology substantially increased pathogen detection and expert-rated therapeutic relevance, but did not significantly improve the primary DOOR/RADAR endpoint.
• Clinical signal

  The trial generated important signals for shorter ventilation, faster shock resolution, and better 90-day EQ-5D-5L, but these were secondary and unadjusted findings.
• Mechanistic lesson

  DigiSep demonstrates that diagnostic information alone is an incomplete intervention; the causal chain requires rapid reporting, bedside receipt, clinician confidence, antimicrobial expertise, and actual therapy modification.
• Practice implication

  The results do not support routine mNGS as a stand-alone add-on for all ICU sepsis patients, but they strongly justify trials of faster mNGS workflows coupled to mandatory infectious diseases or antimicrobial-stewardship decision support.

Controversies & Other Evidence

• Diagnostic yield versus clinical benefit:
  • DigiSep shows a major increase in organism detection, but the primary patient-centred endpoint was not statistically significantly improved.
  • This distinction matters because more detected DNA does not automatically mean more treatable infection.
• Pathogen DNA is not synonymous with active infection:
  • Implausible positive mNGS results were commonly explained by pathogens not fitting the clinical presentation; many such results were viral-only detections.
  • Microbial cell-free DNA can persist longer than conventional cultures, complicating interpretation in disseminated or recently treated infection.¹⁰
• Stewardship integration is probably the intervention:
  • Evidence from molecular rapid diagnostics in bloodstream infection indicates that outcomes improve most consistently when rapid testing is paired with antimicrobial stewardship or infectious diseases input.⁸
  • A more recent ICU study also supports combining rapid diagnostics with immediate infectious diseases consultation to increase targeted antibiotic therapy.⁹
  • DigiSep’s optional consultation model, used in only 2 intervention patients and 0 controls, likely under-tested the full “diagnostic plus decision-support” concept.
• Expert-panel findings are informative but not definitive:
  • Retrospective experts judged treatment modification appropriate in 32.3% of control patients, and those patients had worse outcomes; however, this is post-randomisation, retrospective, non-causal, and vulnerable to confounding by severity and diagnostic complexity.
  • Interrater agreement was mixed: Fleiss’ kappa was 0.2697 for baseline plausibility, 0.3407 for baseline therapeutic relevance, 0.6134 for day-3 plausibility, 0.2173 for day-3 therapeutic relevance, and 0.2871 for therapy-change recommendation.
• Current guidelines:
  • Recent sepsis guidelines support prompt antimicrobials for septic shock and antimicrobial de-escalation when data allow, but DigiSep does not yet establish mNGS as a routine standard-care diagnostic for all sepsis patients.³

Summary

• DigiSep randomised 389 analysed ICU patients with sepsis or septic shock to mNGS plus standard microbiology versus standard microbiology alone.
• The primary 28-day DOOR/RADAR endpoint was not statistically significantly improved: 3.21 ± 1.54 versus 3.49 ± 1.51; mean difference −0.28; 95% CI −0.58 to 0.03; P=0.149.
• mNGS greatly increased diagnostic yield: at day 3, positivity was 45.9% for mNGS versus 4.7% for blood culture.
• Secondary exploratory signals favoured mNGS for absolute ventilation duration, 6.6 ± 9.4 versus 9.3 ± 10.6 days, and shock resolution, 6.9 ± 7.4 versus 8.8 ± 8.5 days.
• The major implementation limitation was low clinical actionability: only 14.3% of evaluable intervention patients had therapy changes based on mNGS despite much higher retrospective expert-rated therapeutic relevance.

Overall Takeaway

DigiSep is an important translational trial because it moves mNGS in sepsis from diagnostic-performance studies into patient-centred randomised evaluation. Its central lesson is that mNGS detects more pathogens, but clinical benefit depends on speed, interpretive expertise, and actual antimicrobial decisions; the next trial should test an integrated rapid mNGS-plus-stewardship pathway rather than a diagnostic add-on alone.

Bibliography

• Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA. 2016;315:801-10.
• Rudd KE, Johnson SC, Agesa KM, Shackelford KA, Tsoi D, Kievlan DR, et al. Global, regional, and national sepsis incidence and mortality, 1990-2017: analysis for the Global Burden of Disease Study. Lancet. 2020;395:200-11.
• Prescott HC, Antonelli M, Alhazzani W, Møller MH, Alshamsi F, Azevedo LCP, et al. Surviving Sepsis Campaign: international guidelines for management of sepsis and septic shock 2026. Intensive Care Med. 2026;52:863-936.
• Gu W, Deng X, Lee M, Sucu YD, Arevalo S, Stryke D, et al. Rapid pathogen detection by metagenomic next-generation sequencing of infected body fluids. Nat Med. 2021;27:115-24.
• Brenner T, Decker SO, Vainshtein Y, Grumaz S, Stevens P, et al. Improved pathogen identification in sepsis or septic shock by clinical metagenomic sequencing. J Infect. 2025;91:106565.
• Brenner T, Skarabis A, Stevens P, Axnick J, Haug P, Grumaz S, et al. Optimization of sepsis therapy based on patient-specific digital precision diagnostics using next generation sequencing (DigiSep-Trial)—study protocol for a randomized, controlled, interventional, open-label, multicenter trial. Trials. 2021;22:714.
• Evans SR, Rubin D, Follmann D, Pennello G, Huskins WC, Powers JH, et al. Desirability of outcome ranking (DOOR) and response adjusted for duration of antibiotic risk (RADAR). Clin Infect Dis. 2015;61:800-6.
• Timbrook TT, Morton JB, McConeghy KW, Caffrey AR, Mylonakis E, LaPlante KL. The effect of molecular rapid diagnostic testing on clinical outcomes in bloodstream infections: a systematic review and meta-analysis. Clin Infect Dis. 2017;64:15-23.
• Kramme E, Käding N, Graf T, Junge S, Scharf C, Kluge S, et al. Rapid diagnostic testing combined with an immediate infectious disease consultation increases the rate of septic intensive care unit patients on targeted antibiotic therapy. Front Cell Infect Microbiol. 2025;14:1513408.
• Eichenberger EM, de Vries CR, Ruffin F, Sharma-Kuinkel B, Park LP, Hong D, et al. Microbial cell-free DNA identifies aetiology of bloodstream infections, persists longer than conventional blood cultures, and its duration of detection is associated with metastatic infection in patients with Staphylococcus aureus and gram-negative bacteraemia. Clin Infect Dis. 2022;74:2020-7.