• S A Urwyler, C A Blum, M Coslovsky, B Mueller, P Schuetz, and M Christ-Crain.
• Department of Endocrinology, Diabetology and Metabolism, University Hospital Basel, Basel, Switzerland.
• J. Intern. Med. 2019 Jul 1; 286 (1): 75-87.

BackgroundA previous study found community-acquired pneumonia (CAP) patients with imbalance of high inflammation and discordantly low cortisol levels to benefit most from adjunctive corticosteroid treatment. Our aim was to validate this hypothesis in a preplanned secondary analysis of the randomized controlled STEP trial.MethodsPatients included in the STEP trial receiving 50 mg prednisone or placebo for 5 days were categorized based on pro-inflammatory cytokines (Interleukin-6/8/MCP-1), CRP and cortisol levels on admission into four groups (high/low inflammation and high/low cortisol). The primary combined end-point was mortality or ICU admission within 30 days.ResultsIn total, 632 patients (315 prednisone, 317 placebo) were included in this analysis. Prednisone did not significantly reduce the risk for the primary end-point in patients with high cytokines/low cortisol and in any other subgroups. However, we noted some differences in the strength of corticosteroid effect in the different subgroups with stronger effects in patients with high cytokines [OR 0.44 (0.10,1.72)] compared to patients with low cytokines [OR 0.68 (0.30,1.5)] (P-interaction = 0.600). The effects did not differ according to cortisol levels.ConclusionThe imbalance of high inflammation state and low cortisol levels did not predict treatment response to corticosteroids in patients with CAP. However, in line to previous research, inflammation as measured by cytokine levels irrespective of cortisol tended to predict treatment response to corticosteroids in CAP. Whether this concept may help to personalize corticosteroids to patients most likely benefitting from this treatment needs to be tested in future intervention trials.© 2019 The Association for the Publication of the Journal of Internal Medicine.

27 keywords...

hide…