Invited Commentary 

March 27, 2023

Should We Add Fludrocortisone to Hydrocortisone for Treatment of Septic Shock?

Michael A. Matthay, Issa J. Dahabreh, B. Taylor Thompson

JAMA Intern Med. 2023;183(5):460-461. doi:10.1001/jamainternmed.2023.0257

Mortality for patients with septic shock is greater than 30% in most studies. The mechanisms that contribute to poor survival are primarily explained by the development of multiorgan failure from activation of both proinflammatory and procoagulant pathways that lead to acute kidney failure, acute respiratory failure, and often central nervous systemic dysfunction and hematologic abnormalities.¹ The primary treatment for septic shock includes antibiotics, intravenous fluids, vasopressors, mechanical ventilation, dialysis, and surgical source control if indicated, but effective pharmacologic therapies are limited. The current Surviving Sepsis guidelines give a weak recommendation for treatment with intravenous hydrocortisone 200 mg/d based on moderate-quality evidence.² However, 2 randomized, placebo-controlled clinical trials reported a survival benefit with glucocorticoid therapy in septic shock using a combination of both hydrocortisone and the mineralocorticoid fludrocortisone compared with placebo.^3,4 One other small trial (n = 509 patients) compared hydrocortisone alone with hydrocortisone plus fludrocortisone.⁵ The results did not demonstrate a significant reduction in mortality with the addition of fludrocortisone, but it did show a numerically lower mortality by 3% with combination therapy (relative risk, 0.94; 95% CI, 0.77-1.14; P = .50). Thus, the potential value of adding fludrocortisone to hydrocortisone for pharmacologic treatment of septic shock has not been resolved.

In this issue of JAMA Internal Medicine, Bosch and colleagues⁶ report the results of a retrospective cohort study of routinely collected data that aimed to emulate a target clinical trial to compare adding vs not adding fludrocortisone to hydrocortisone for reducing mortality and morbidity from septic shock. Among 384 394 patients with septic shock who received norepinephrine identified in claims data, 88 275 eligible patients were treated with corticosteroids within 3 days of admission; of those, 85 995 received hydrocortisone alone and 2280 received both hydrocortisone and fludrocortisone on the same day. The primary end point of hospital mortality or discharge to hospice was significantly lower by 3.7% in patients treated with both hydrocortisone and fludrocortisone (47.2%) compared with patients treated with hydrocortisone alone (50.8%; number needed to treat = 28). The secondary outcomes of vasopressor-free days and hospital-free days also showed significant benefit with the addition of fludrocortisone to hydrocortisone.

There are several mechanisms by which mineralocorticoid therapy might be effective in treating septic shock. The biological activity of mineralocorticoids is mediated by interaction with the intracellular mineralocorticoid receptor, which is expressed in vascular smooth muscle, endothelial cells, monocytes, macrophages, and neutrophils, and activation leads to increased leukocyte adhesion and production with increased production of tumor necrosis factor-α and reactive oxygen species that may aid in the control of bacterial infection. In experimental sepsis, mineralocorticoids increase histamine, serotonin, bradykinin, and catecholamines and hasten the reversal of shock and facilitate lung edema clearance. Plasma aldosterone levels are low in some patients with septic shock, and some evidence suggests that the mineralocorticoid pathway is more frequently impaired than the glucocorticoid pathway in early sepsis. Taken together, these data suggest that mineralocorticoid deficiency in sepsis may contribute to the immunoparalysis described in some patients with sepsis and that mineralocorticoid therapy may be important in restoring homeostasis and enhancing innate immunity.⁷

In view of the plausible biological mechanism and prior trial evidence, the results of the study by Bosch and colleagues⁶ are not entirely surprising. The study included a large number of patients captured in a database that included approximately 25% of all US hospital admissions over the study period of 2016 to 2020. Demographics, comorbidities, and acute organ failure were well balanced between the 2 groups. Patients were only included if they were admitted to an intermediate or intensive care unit and treated with norepinephrine, the most common vasopressor for therapy of septic shock, and treatment had been initiated within 3 days of hospital admission. The number of days from admission to corticosteroid treatment was similar between the 2 groups.

An important limitation in giving the results of these observational analyses a causal interpretation is the potential for confounding by unmeasured variables, especially because the analyses compared the addition of an active treatment (fludrocortisone) with a different route of administration than the baseline intravenous hydrocortisone treatment, against no additional treatment.⁶ Furthermore, by restricting all analyses to hydrocortisone initiators, the population of patients to whom the results apply becomes harder to define and is not necessarily the same as the population of all individuals with septic shock who are candidates for corticosteroid therapy. To address this issue, while avoiding misalignments between the time of eligibility, treatment initiation, and the start of follow-up, the investigators could have used analyses that allow for delayed initiation of treatment after meeting eligibility criteria (eg, within a grace period of 3 days after developing shock requiring norepinephrine treatment).⁸Such analyses, as well as those reported by Bosch and colleagues,⁶ could benefit from large-scale, rich, finely time-stamped clinical data. The claims data available for this study did not contain information on vasopressor dosage to estimate the severity of shock, and there was no information on how many patients in each group required mechanical ventilation, and therefore the study lacks data on these important end points. Finally, the study did not have biological data to help compare the severity of sepsis in terms of vascular injury or inflammatory responses that could identify heterogeneity among the patients with septic shock.

We now know that septic shock is not homogeneous, and there is evidence for subphenotypes of sepsis that may respond differently to early goal-directed resuscitation or even steroids, with evidence for benefit in some subgroups and harm in others.^9,10 It is likely that progress in identifying more effective therapies for septic shock will require a focus on specific subgroups that may respond more favorably to a therapeutic approach. Thus, lumping septic shock into a single clinical entity may not be optimal, and it will be more important to identify prospectively treatment responsive subsets of patients.¹⁰

To answer the question of whether adding fludrocortisone to hydrocortisone would have benefit in treatment of septic shock, a very large randomized and prospective trial would be needed to detect a mortality reduction of 3.7% as reported in the current retrospective study.⁶ A trial of this kind would also need to be designed to test for heterogeneity of treatment effect in prespecified sepsis subtypes, potentially using real-time biologic classifiers.

Pending additional data from randomized clinical trials, how should the results of the study by Bosch and colleagues⁶ affect the practice of clinicians who currently treat patients with septic shock with hydrocortisone? For such patients, adding fludrocortisone to hydrocortisone is reasonable given consistent findings across 2 placebo-controlled randomized trials and this large retrospective observational analysis, the low cost of this enterally administered drug, and the excellent safety profile. Nevertheless, we strongly endorse the conduct of additional randomized trials designed to quantify heterogeneity of treatment effects and support the personalization of treatment for septic shock.