Considerable efforts have been made to spare excessive catecholamine exposure and avoid side effects when treating patients with refractory septic shock. Until recently, physiological replacement of arginine-vasopressin (AVP) and corticosteroid (CS) deficits was considered promising but equivocal to improve end-organ function [1], reduce new-onset tachyarrhythmias [2], shorten duration of shock [3], and improve survival [4]. However, two large multicenter randomized controlled trials failed to demonstrate mortality reduction when using physiological doses of AVP [5] and CS [6] and hampered the initial enthusiasm for their widespread use.In this issue of Intensive Care Medicine, Torgersen et al. [7] evaluate the effect on mortality of adding CS to AVP in 159 patients with severe septic shock. Patients included in this single-center retrospective study were relatively ill, as outlined by high baseline sepsis-related organ failure assessment (SOFA) score (median score 15), frequent requirement of continuous renal replacement therapy (99/159, 62.3%), and overall intensive care unit (ICU) mortality (97/159, 61.0%). This sample is not fully representative of the severe sepsis population currently encountered in ICUs [8]. Because treatment assignment was not randomized, the authors used three different statistical analyses to account for baseline characteristics imbalance: multivariate logistic regression, Cox proportional hazard model, and Cox model stratified to matched pairs based on propensity score. Both adjusted analyses and propensity score determination were based on covariables such as age, year of admission, baseline SOFA score, and initial AVP dose, and ignored potentially significant confounders or cointerventions like fluid resuscitation or early effective antimicrobial therapy. Improved survival at 28 days was demonstrated with the Cox proportional hazard model, but not with other statistical analyses, maybe due to lack of power. These results should be cautiously interpreted and must not lead to hasty conclusions regarding the benefits of combining AVP and CS.Two previous studies using retrospective design and prone to similar biases reported consistent results [9,10]. Bauer et al. [9] compared 21 patients receiving AVP and CS with 21 patients infused with AVP without CS. Patients were matched for age, sex, acute physiology score component of the acute physiology and chronic health evaluation (APACHE), number of vasopressors, and primary ICU service [9]. Patients receiving CS were more likely to be alive and weaned off vasopressors at day 7 (80.9 vs. 47.6%, p = 0.02), although the 28-day and hospital mortality were not statistically different. In another retrospective analysis of 778 patients with septic shock enrolled in a multicenter randomized controlled