Pneumocystis jirovecii pneumonia mortality risk associated with preceding long-term steroid use for the underlying disease: A multicenter, retrospective cohort study.

Abstract

Long-term steroid use increases the risk of developing Pneumocystis pneumonia (PcP), but there are limited reports on the relation of long-term steroid and PcP mortality. Retrospective multicenter study to identify risk factors for PcP mortality, including average steroid dose before the first visit for PcP in non-human immunodeficiency virus (HIV)-PcP patients. We generated receiver operating characteristic (ROC) curves for 90-day all-cause mortality and the mean daily steroid dose per unit body weight in the preceding 10 to 90 days in 10-day increments. Patients were dichotomized by 90-day mortality and propensity score-based stabilized inverse probability of treatment weighting (IPTW) adjusted covariates of age, sex, and underlying disease. Multivariate analysis with logistic regression assessed whether long-term corticosteroid use affected outcome. Of 133 patients with non-HIV-PcP, 37 died within 90 days of initial diagnosis. The area under the ROC curve for 1-40 days was highest, and the optimal cutoff point of median adjunctive corticosteroid dosage was 0.34 mg/kg/day. Past steroid dose, underlying interstitial lung disease and emphysema, lower serum albumin and lower lymphocyte count, higher lactate dehydrogenase, use of therapeutic pentamidine and therapeutic high-dose steroids were all significantly associated with mortality. Underlying autoimmune disease, past immunosuppressant use, and a longer time from onset to start of treatment, were associated lower mortality. Logistic regression analysis after adjusting for age, sex, and underlying disease with IPTW revealed that steroid dose 1-40 days before the first visit for PcP (per 0.1 mg/kg/day increment, odds ratio 1.36 [95% confidence interval = 1.16-1.66], P<0.001), low lymphocyte counts, and high lactate dehydrogenase revel were independent mortality risk factor, while respiratory failure, early steroid, and sulfamethoxazole/trimethoprim for PcP treatment did not. A steroid dose before PcP onset was strongly associated with 90-day mortality in non-HIV-PcP patients, emphasizing the importance of appropriate prophylaxis especially in this population.