Hydrocortisone Outcomes in Hospital Bloodstream Infection

This single-study briefing examines an ancillary observational analysis of the prospective EUROBACT-2 cohort. It assessed whether hydrocortisone exposure was associated with outcomes among adult ICU patients with hospital-acquired bloodstream infection (HA-BSI) and sepsis. HA-BSI was defined as a positive blood culture collected more than 48 hours after hospital admission, with the infection managed in the ICU. [pmid:42743132]

Evidence

The analysis included 2,401 patients; 608 (25%) received hydrocortisone during the HA-BSI episode. Investigators compared recipients with patients not receiving hydrocortisone, using 1:1 nearest-neighbour propensity-score matching without replacement, a caliper of 0.2 standard deviation of the logit propensity score, and multivariable logistic regression with cluster-robust standard errors by matched pair. Matching yielded 521 pairs, or 1,042 patients. [pmid:42743132]

In the matched cohort, day-28 mortality was 50% (260/521) among hydrocortisone recipients and 42% (220/521) among controls. In the multivariable analysis, hydrocortisone use was associated with day-28 mortality, with an odds ratio of 1.448 (95% CI 1.096–1.914; p = 0.009). This is an association from an adjusted observational comparison, not a randomized treatment effect. [pmid:42743132]

The reported association was observed in the subgroup with sepsis without shock and in the female subgroup, but not in patients with septic shock or those receiving norepinephrine-equivalent dose at least 0.25 µg/kg/min. Exploratory secondary analyses associated hydrocortisone exposure with fewer vasopressor-free days, fewer mechanical ventilation-free days at day 28, and a less favorable SOFA trajectory from HA-BSI diagnosis to day 7. [pmid:42743132]

Analysis — Association and Applicability

The central signal is directionally unfavorable: after matching and multivariable adjustment, hydrocortisone exposure was associated with higher odds of death by day 28 in this ICU HA-BSI population. The observed mortality proportions and adjusted odds ratio describe the study’s matched comparison; they do not show that hydrocortisone caused the difference. Propensity-score matching can align measured characteristics between comparison groups, but it cannot substitute for random allocation or remove bias from factors that were not measured or not fully modeled. [pmid:42743132]

The population is narrower than sepsis broadly. Participants had microbiologically documented, late-onset hospital-acquired bloodstream infection managed in the ICU, a setting the investigators describe as distinct from many randomized corticosteroid trials. Therefore, the result is most directly informative for the study population and exposure comparison rather than for all corticosteroid use in all forms of sepsis. The paper’s subgroup findings add context but should not be treated as definitive evidence that effects differ by shock status, vasopressor exposure, or sex. [pmid:42743132]

The secondary-outcome pattern is consistent with the primary association in that it was less favorable among exposed patients, but these outcomes were exploratory and post-baseline. Statistical significance in an adjusted association does not by itself establish clinical significance, treatment causation, or a change in the balance of benefits and harms. The study is best read as a hypothesis-generating signal that identifies an important question for designs better able to address treatment selection and timing. [pmid:42743132]

Limitations

Hydrocortisone was not randomized, and the rationale for its administration was not recorded. Consequently, residual confounding by indication may remain despite matching and multivariable adjustment: patients selected for hydrocortisone may have differed in clinically important ways that the analysis could not fully capture. This limitation is central because the reported effect estimate is associative. [pmid:42743132]

The exact date and time of hydrocortisone initiation relative to blood-culture sampling or HA-BSI diagnosis were not available, so time-dependent bias cannot be excluded. The investigators also characterize the female-subgroup association as exploratory and hypothesis-generating rather than evidence of a sex-specific treatment effect. Secondary outcomes likewise warrant caution because of their observational, post-baseline nature. [pmid:42743132]

This briefing reports what the study compared and found; it does not provide a treatment recommendation, a patient-specific conclusion, or a prediction of outcomes. [pmid:42743132]