A rapid DNA-sequencing test for sepsis missed its primary goal in a German trial

A rapid test that reads microbial DNA straight from a patient's blood did not improve the main outcome it was designed to change in people with sepsis, according to DigiSep, a randomised trial run across 24 German intensive care units [s1][s2]. The finding is a cautionary one for a technology marketed as a way to identify the responsible microbe faster and tailor antibiotics sooner: on the trial's prespecified primary measure, patients who received the sequencing test fared no better than those who got standard cultures alone [s1].

Sepsis is a dysregulated, body-wide response to infection that can tip into organ failure within hours, and speed matters — guidelines push clinicians to start broad antibiotics fast and then narrow them once the bug is known [s1]. The trouble is that blood cultures are slow and often negative, especially once antibiotics are flowing, which is the gap metagenomic sequencing promises to fill [s1].

What the trial did

DigiSep was a randomised, controlled, open-label, multicentre trial funded by Germany's public Innovation Fund [s1][s2]. It enrolled adults with sepsis or septic shock and assigned them either to standard-of-care microbiology plus metagenomic next-generation sequencing (mNGS) of circulating cell-free DNA, or to standard-of-care microbiology alone [s1]. The sequencing arm included 200 patients and the control arm 189 [s1]. Rather than waiting days for a culture to grow, mNGS reads fragments of microbial DNA in a blood sample and can in principle name the pathogen within hours [s1]. Crucially, the primary endpoint was not survival but a composite called the Desirability of Outcome Ranking / Response Adjusted for Duration of Antibiotic Risk (DOOR/RADAR), which rewards good outcomes achieved with less antibiotic exposure [s1].

What it found

On that primary endpoint, the two groups were statistically indistinguishable at 28 days. The DOOR/RADAR score was 3.21 ± 1.54 in the sequencing group and 3.49 ± 1.51 in the control group, a difference whose 95% confidence interval ran from −0.58 to 0.03 and therefore included no effect [s1]. Several secondary measures, which the authors are careful to label exploratory, did move in the test's favour. The duration of mechanical ventilation was shorter — 6.6 ± 9.4 days versus 9.3 ± 10.6 days (95% CI −5.03 to −0.34) — and shock resolved faster, at 6.9 ± 7.4 days versus 8.8 ± 8.5 days (95% CI −3.75 to −0.04) [s1]. Health-related quality of life at 90 days, measured on the EQ-5D-5L, was modestly higher in the sequencing group (0.312 ± 0.386 versus 0.208 ± 0.373; p=0.047) [s1]. In the roughly one-third of patients with available claims data, healthcare costs over 180 days did not differ between groups [s1].

How to read it

The honest summary is that DigiSep did not meet its primary endpoint, and everything encouraging in it comes from secondary analyses the trial was not powered to prove [s1]. That ordering matters. When a primary outcome is neutral, downstream findings are hypothesis-generating rather than confirmatory, because the more comparisons a study makes, the more likely one crosses a significance threshold by chance [s1]. The ventilation and shock-resolution signals are biologically plausible — knowing the organism sooner could in theory sharpen treatment — but confidence intervals that only just clear zero, in an open-label trial where clinicians knew who got the test, are exactly the kind of result that later studies often fail to reproduce [s1]. The open-label design is a genuine limitation here: expectations about a new diagnostic can nudge bedside decisions about when to wean a ventilator or stop vasopressors [s1].

It is also worth noting what the trial did not show. There was no benefit on the primary ranking, and the cost analysis — limited though it was to a subgroup — gave no sign that the extra test paid for itself over six months [s1]. For hospitals weighing whether to buy into metagenomic sequencing for sepsis, that combination is sobering, and it sits against a broader pattern in critical care where plausible diagnostics repeatedly fail to translate into better outcomes once tested properly [s1].

What to watch

The authors frame mNGS as promising but unproven, and the gap between a neutral primary endpoint and encouraging secondaries is precisely what a larger, blinded, adequately powered trial exists to settle [s1]. The open questions are whether faster pathogen identification actually changes antibiotic decisions at the bedside, whether any downstream benefit is large enough to justify the cost, and which patients — if any — gain most [s1]. Until those are answered, DigiSep is best read as a reason for caution rather than a case for adoption [s1][s2].

This article describes research and is not medical advice. Decisions about sepsis diagnosis and treatment rest with treating clinicians.

Sources

• DigiSep trial: clinical metagenomics in sepsis and septic shock — Intensive Care Medicine, 30 June 2026
• DigiSep trial registration, NCT04571801 — ClinicalTrials.gov