Early hydrocortisone in sepsis stopped for futility, and the timing hypothesis is still untested

What was tried

Siriraj Hospital in Bangkok opened NCT06217939 in June 2024 with a narrow question. Current practice, following the Surviving Sepsis Campaign, reserves low-dose hydrocortisone for septic shock that stays hypotensive despite vasopressor support. The trial asked whether giving it much earlier changes survival.

Adults with sepsis by Sepsis-3 criteria and a mean arterial pressure below 65 mmHg were randomized 1:1 to early hydrocortisone or a saline sham. The early arm received a 50 mg intravenous bolus then 200 mg by continuous infusion over 24 hours for two days, without taper. Both arms could receive open-label hydrocortisone on the guideline trigger, failure to reach the hemodynamic goal despite norepinephrine or epinephrine at 0.25 mcg/kg/min or more for at least four hours after vasopressor initiation, at which point study drug was stopped.

Randomization and study drug had to be delivered within three hours of the onset of hypotension. Immunocompromised, pregnant, post-cardiac-arrest and corticosteroid-treated patients were excluded, as were those given etomidate. Masking covered participants, care providers and investigators. Baseline interleukin-6, interleukin-10, tumor necrosis factor, procalcitonin, C-reactive protein and cortisol were drawn at randomization. The primary endpoint was 28-day all-cause mortality, with time to shock control, in-hospital mortality and length of stay secondary.

Enrollment reached 116. The registry records termination on 25 December 2025 on a data and safety monitoring board recommendation citing preliminary evidence of futility. No results have been posted.

The biological hypothesis

The target is the glucocorticoid receptor, NR3C1. Open Targets scores the NR3C1 association with septic shock at 0.733, with a clinical evidence component of 0.994 and a genetic association component of 0.783. Seventy-one drugs or clinical candidates act at this receptor, so tractability was never the constraint.

The case for early dosing rests on the idea that critical illness-related corticosteroid insufficiency develops during the shock state rather than after it. If impaired cortisol production, impaired tissue sensitivity and unrestrained inflammatory signaling begin at hemodynamic decompensation, waiting for vasopressor refractoriness means intervening after the window in which receptor occupancy could restrain the cytokine response. A 2024 systematic review of seven trials totaling 3,063 patients reported lower ICU and hospital mortality with earlier rather than later initiation of low-dose hydrocortisone (Medicine 2024).

What actually happened

No mortality figures, effect estimates or confidence intervals have been posted, so the direction and magnitude of any observed difference are not public.

The literature this trial entered has never converged. Annane and colleagues reported in 2002 that hydrocortisone plus fludrocortisone reduced 28-day mortality among corticotropin nonresponders, 53 versus 63 percent, hazard ratio 0.67 (95 percent CI 0.47 to 0.95, JAMA 2002). CORTICUS found no 28-day mortality difference in 499 patients, 34.3 versus 31.5 percent, and more superinfection (N Engl J Med 2008). ADRENAL randomized 3,800 patients and found 90-day mortality of 27.9 versus 28.8 percent, odds ratio 0.95 (95 percent CI 0.82 to 1.10), alongside faster shock resolution (N Engl J Med 2018). APROCCHSS, published the same year, found 90-day mortality of 43.0 versus 49.1 percent, relative risk 0.88 (95 percent CI 0.78 to 0.99, N Engl J Med 2018). The 2024 Society of Critical Care Medicine focused update gave only a conditional recommendation for corticosteroids in septic shock.

Failure mechanism, best guess

This reads as an efficacy failure of trial size rather than of pharmacology, compounded by a design feature that blunts the comparison. The two arms were not early hydrocortisone versus no hydrocortisone. They were early hydrocortisone versus guideline-timed hydrocortisone, since control patients received open-label drug once they hit the vasopressor threshold. The trial therefore tests a timing difference measured in hours within a population where a substantial fraction of controls eventually receive the same drug. That contrast is smaller than steroid versus placebo, and the two largest placebo-controlled trials could not agree on whether even the larger contrast moves 90-day mortality.

Against that background, a 28-day mortality endpoint in 116 patients has almost no chance of separating the arms. Control-arm mortality across the major trials ranges from 28 to 49 percent, and detecting a reduction of a few percentage points requires thousands of patients. The futility signal is most likely a statement about the trial's information content rather than a refutation of early dosing.

Timing may also not be the operative variable. The VANISH post hoc analysis found that transcriptomic sepsis response signature endotype modified the corticosteroid effect, with hydrocortisone associated with higher mortality in the immunocompetent SRS2 group, odds ratio 7.9 (95 percent CI 1.6 to 39.9), interaction p value 0.02 (Am J Respir Crit Care Med 2019). If the population contains responders and harmed subgroups in unknown proportions, an unstratified average effect will sit near zero regardless of when the drug is given.

How to prevent this next time

No endpoint-level data were posted for this trial, so a Bayesian update on its own result is not available and none is offered here. The levers that apply are design levers.

Enrich rather than average. This trial already drew interleukin-6, interleukin-10, tumor necrosis factor, procalcitonin and cortisol at randomization, which is what a stratified analysis needs. A prospective design could have used an endotype or cortisol-response classifier as a randomization stratum.

Adjust the base rate before sizing. Two adequately powered placebo-controlled trials disagreed on the steroid effect, so the prior on a timing effect within a smaller contrast should have been shifted toward the null and the sample size set accordingly.

Choose an endpoint the sample can move. Time to shock control and vasopressor-free days respond to hydrocortisone consistently across ADRENAL and Annane 2002, and a physiological primary endpoint would have been interpretable at this sample size. Any design in which controls receive the drug on a clinical trigger should also carry a pre-specified crossover estimate.

The single highest leverage change would have been to make a validated responder classifier, cortisol-based or transcriptomic, a randomization stratum with the primary analysis pre-specified inside the enriched stratum, rather than testing a timing shift as an average effect in an unselected septic shock population.

What this means for similar programs

NR3C1 now carries two Claidex failure records and a mechanism risk score of 50, in the orange band. The score is not driven by weak biology, since the genetic and clinical evidence deficit contributes only 4.0 of a possible 15 points. It is driven by archetype severity, at 15.8 of 25, and by saturation at the maximum 15 of 15, reflecting 72 distinct programs against this receptor. A target with strong evidence and seven decades of use can still generate repeated failures when the unanswered question is patient selection.

For groups planning corticosteroid trials in critical illness, the unselected mortality trial has been run repeatedly and has not settled the question. The tractable remaining designs are enrichment and physiological-endpoint designs.

Open questions

What were the 28-day mortality rates in each arm at the point of stopping? What proportion of control patients crossed over to open-label hydrocortisone? Were the stored baseline cytokine and cortisol samples analyzed, and do they show endotype separation? And will the investigators publish the 116-patient dataset, which would feed the individual patient data meta-analyses this field needs?

Sources

1. • ClinicalTrials.gov, NCT06217939, Early Intravenous Hydrocortisone in Sepsis. https://clinicaltrials.gov/study/NCT06217939 - Annane D, et al. Effect of treatment with low doses of hydrocortisone and fludrocortisone on mortality in patients with septic shock. JAMA.

2. - Sprung CL, et al. Hydrocortisone therapy for patients with septic shock. N Engl J Med.

3. - Venkatesh B, et al. Adjunctive Glucocorticoid Therapy in Patients with Septic Shock (ADRENAL). N Engl J Med.

4. - Annane D, et al. Hydrocortisone plus Fludrocortisone for Adults with Septic Shock (APROCCHSS). N Engl J Med.

5. - Antcliffe DB, et al. Transcriptomic Signatures in Sepsis and a Differential Response to Steroids. From the VANISH Randomized Trial. Am J Respir Crit Care Med.

6. - Chaudhuri D, et al. 2024 Focused Update: Guidelines on Use of Corticosteroids in Sepsis, Acute Respiratory Distress Syndrome, and Community-Acquired Pneumonia. Crit Care Med.

7. - Early use of low-dose hydrocortisone can reduce in-hospital mortality in patients with septic shock: a systematic review and meta-analysis. Medicine (Baltimore).

8. - Annane D, et al. Corticosteroids for treating sepsis in children and adults. Cochrane Database Syst Rev.

9. - Open Targets Platform, NR3C1 (ENSG00000113580) and septic shock (EFO_0006834) association record. https://platform.opentargets.org/target/ENSG00000113580 - ChEMBL, hydrocortisone CHEMBL389621. https://www.ebi.ac.uk/chembl/ - openFDA FAERS drug event endpoint, hydrocortisone reaction term counts. https://api.fda.gov/drug/event.json.