Cosyntropin testing does not predict response to glucocorticoids in community-acquired pneumonia in a randomized controlled trial.

Abstract

Glucocorticoids have been shown to improve outcome in community-acquired pneumonia (CAP). However, glucocorticoids have potential side-effects, and treatment response may vary. It is thus crucial to select patients with high likelihood to respond favourably. In critical illness, cosyntropin testing is recommended to identify patients in need for glucocorticoids. We investigated whether cosyntropin testing predicts treatment response to glucocorticoids in CAP. Predefined secondary analysis of a randomized controlled trial. Hospitalized patients with CAP. We performed 1µg cosyntropin tests in a randomized trial comparing prednisone 50mg for 7days to placebo. We investigated whether subgroups based on baseline and stimulated cortisol levels responded differently to glucocorticoids with regard to time to clinical stability (TTCS) and other outcomes by inclusion of interaction terms into statistical models. A total of 326 patients in the prednisone and 309 patients in the placebo group were evaluated. Neither basal cortisol nor a Δcortisol <250nmol/L after stimulation nor the combination of basal cortisol and Δcortisol predicted treatment response as measured by TTCS (all P for interaction >0.05). Similarly, we found no effect modification with respect to mortality, rehospitalization, antibiotic treatment duration or CAP-related complications (all P for interaction >0.05). However, glucocorticoids had a stronger effect on shortening length of hospital stay in patients with a baseline cortisol of ≥938nmol/L (P for interaction=0.015). Neither baseline nor stimulated cortisol after low-dose cosyntropin testing at a dose of 1µg predicted glucocorticoid responsiveness in mild to moderate CAP. A treatment decision for or against adjunct glucocorticoids in CAP should not be made depending on cortisol values or cosyntropin testing results.