Editorials

Immunotherapies for sepsis and the impact of study design

John H Powers III, Charles Natanson

BMJ 2025; 388 doi: https://doi.org/10.1136/bmj.r48 (Published 15 January 2025)Cite this as: BMJ 2025;388:r48

Enrolling the right number and type of patient is crucial

Sepsis is a life threatening syndrome initiated by microorganisms.1 Severe and often lethal injuries (eg, shock, multiorgan failure, and metabolic derangements) are the manifestation of the syndrome, with systemic inflammatory cascade activation and circulating mediators thought to cause host injury.

Scores of trials evaluating host-immune modifiers, many of which were the initial ventures of a new biotech industry 30 years ago, did not improve patient outcomes in sepsis.2 Undaunted by this history, investigators in the linked trial(doi:10.1136/bmj-2024-082583) evaluated another immune therapy, thymosin-α-1, in patients with sepsis, enrolling 1106 participants at 22 sites in China.3 The trial compared thymosin-α-1 with placebo plus usual care in a blinded randomised trial evaluating all cause mortality at 28 days. The results showed 24% mortality with thymosin versus 23% in the control group (hazard ratio 0.99, 95% confidence interval 0.77 to 1.77), which indicate no improvement with thymosin. Several positive aspects of the trial’s design include a superiority hypothesis with direct patient-outcomes of all cause mortality using inferential statistics, rather than non-inferiority hypotheses on surrogate endpoints using descriptive statistics commonly used in infection trials.456789

Yet, questions remain regarding the trial’s methods. Were the types and numbers of patients enrolled requisite for studying this syndrome? In sepsis, clinical and preclinical evidence show that potential benefits of immunotherapies may be limited to patients with a higher baseline risk of sepsis mortality—ie, people with immune dysregulation that could cause death.10 In patients with lower baseline mortality risk, suppressing appropriately functioning immune responses may be harmful and worsen outcomes.11 Study investigators hypothesised 35% mortality in the control group, yet results showed 23% mortality, indicating a less sick population with most participants having appropriate immune responses. A systematic review of mortality rates in randomised trials of sepsis between 1991 and 2013 showed that while mortality has continuously declined, when controlling for baseline severity of illness, mortality has not changed over time. These data suggest that more recent sepsis trials enrolled patients who may not benefit.12 One reason for discordance between mortality in prior studies versus current trials may be exclusion criteria that eliminate sicker patients. While sicker populations likely include patients who are beyond help including deaths unattributable to sepsis, this population also potentially includes sicker patients with septic shock who seem most likely to benefit from immunotherapies.

Because of this potential paradoxical situation of immunotherapy, even in higher baseline risk populations, to detect statistically significant effects greater than chance may require enrolling larger numbers of participants than included in the linked study. Numbers of participants required is based partly on the amount of benefit of interventions. Small effects often are not worth detecting clinically with an exception for all cause mortality in life threatening syndromes such as sepsis.

Surrogate outcomes instead of survival are not informative in sepsis trials given the present understanding of the syndrome. Sequential organ failure assessment (SOFA) scores are composites of biomarkers not on the causal pathway of disease.13 Changes in these scores have previously been shown not to reflect most treatment effects on patient survival.14 Despite patient level correlations with mortality, these sequential organ failure assessment scores are not valid trial level surrogates and are not useful endpoints for studies in serious lethal syndromes.

Despite the lack of success with host directed treatments in sepsis, the effort is worth pursuing. Paul Ehrlich studied several hundred compounds before finding salvarsan for syphilis.15 Host directed therapies, such as steroids, have shown consistent patient benefits in improving septic shock and potentially for survival in people with severe pneumonia and covid-19.1617

The focus of resources and policy in infectious disease on small molecule drugs with in vitro biological activity against antimicrobial resistant pathogens in pathogen focused development1819 misses the point. Seventeen of 18 deaths (94%) occurred in patients with organisms susceptible in vitro to current drugs.20 Poor outcomes may be due to an inaccurate early diagnosis but also due to severe dysregulated immune responses, not an inability to inhibit bacterial growth. Therefore, drugs with improved in vitro potency at growth inhibition are worth developing but will not benefit all types of patients who are seriously ill with an infection if immune dysregulation is the problem. Host directed therapies may benefit the greater number of patients with susceptible as well as antimicrobial resistant disease.

This study emphasises again that research methods matter when studying medical interventions. Per Einstein, the definition of insanity is doing the same thing over and over and expecting different results. Beneficial interventions can be discarded if not studied appropriately. Improving patient outcomes in sepsis may require studying patients with specific highly lethal infections by site of infection (pneumonia v urinary tract infections) instead of a broad array of populations pooled together into a poorly understood syndrome. These methods may allow future studies to expand the types of interventions evaluated, including host directed therapies, enrolling patients sick enough to benefit in site specific infections. Hypotheses about immune therapies have not been adequately evaluated given the methods used to date. Broadly anti-inflammatory interventions with promising data in preclinical and phase two studies need testing in patients at high risk of sepsis attributable mortality. Enough participants should be enrolled in randomised confirmatory trials using superiority hypotheses about direct patient outcomes analysed by inferential statistics. The results will be worth the effort both for patients with sepsis and for justifying the costs of treatments.