Background
Pneumocystis pneumonia (PcP) is a severe disease with high morbidity and mortality, which almost exclusively affects immunocompromised patients. PcP has long been known for its high prevalence among human immunodeficiency virus (HIV)-positive patients [
1]. Since the implementation of combination antiretroviral therapies (cART) and chemoprophylaxis its incidence has been continuously decreasing [
2]. Nowadays, most patients with HIV-associated PcP are treatment-naïve with very low CD4 cell counts; some of these patients do not know that they are HIV positive until they attend hospital [
3]. On the other hand, PcP is also frequently diagnosed in non-HIV-positive patients as immunosuppressive regimens are being increasingly used in a wide range of patient populations. Consistently, the incidence of PcP in non-HIV-positive patients is increasing [
2].
Pneumocystis infections have first been described in preterm infants following World War II [
4] and in patients with malignancies in the late 1960s [
5], more than a decade before the HIV epidemic emerged. Defects in cell-mediated immunity and use of glucocorticoids are among the strongest risk factors for the development of PcP [
6,
7].
The epidemiology of PcP has been described in several retrospective studies [
2,
8]. A study from the Mayo Clinic in 116 non-HIV-positive patients found that most frequent underlying diseases were hematological malignancies (30%), organ transplantation (25%), autoimmune disease (22%), solid tumors (13%) and other reasons [
6]. Although, most of these reports were published more than a decade ago and might not reflect the current epidemiological situation, a report from France was published a few years ago [
9]. Roux et al. reported overall mortality of 17.4%, which was significantly higher in non-HIV-positive patients (27%) than in HIV-positive patients (4%) [
9]. A few parameters such as age, prior episode of PcP, low CD4 cell count, lactate dehydrogenase (LDH), and coinfections have been reported to predict unfavorable outcome in patients with HIV infection [
10‐
12]. Reports on outcome predictors in non-HIV-positive patients are sparse.
Based on the high burden of PcP and the likelihood of unfavorable outcome particularly in non-HIV-positive patients, chemoprophylaxis with trimethoprim-sulfamethoxazole (TMP-SMX) is recommended in high-risk populations [
13]. TMP-SMX is also the treatment of choice for PcP. Adjunctive corticosteroid therapy is recommended in HIV-positive patients with severe respiratory failure [
14,
15]. However, while beneficial outcomes of higher dosage of corticosteroids in HIV-positive patients during PcP are reported [
16], the outcome of using corticosteroids in non-HIV-positive patients is not clear [
17].
We here report comprehensive epidemiological, clinical, laboratory, therapeutic and outcome data on 240 cases of PcP, including a high percentage of non-HIV-positive patients, in a tertiary care center over the last 17 years. Additionally, we aimed at identifying predictors of outcome.
Methods
Study design and population
We performed a retrospective, single-center cross-sectional analysis of all patients with a positive finding of Pneumocystis jirovecii on direct immunofluorescence testing or detection by Diff-Quick staining in the bronchial washing fluid or broncho-alveolar lavage (BAL) fluid, from January 2000 to June 2017. Our hospital is a university tertiary care center with approximately 1500 beds. Written informed consent was waived by the ethics committee due to the anonymized retrospective nature of the analysis.
All bronchial washing fluids or BAL samples from patients clinically suspected to have P. jirovecii pneumonia were evaluated within the study period. For every patient, clinical data on demographic characteristics, underlying disease, status of immune competence, treatment regimens of immunosuppression, PcP therapy regimen and mortality, were gathered in the study database. Date of diagnosis was defined as the date of microbiological confirmation. P. jirovecii direct immunofluorescence was performed using the Monofluo kit P. carinii (BioRad Laboratories, years 2000 to 2016) or the Detect IF Pneumocystis carinii kit (Axis-Shield Diagnostics Ltd., year 2017). Diff Quick staining was performed using the stain sets provided by Dade Behring or Siemens AG, respectively.
Statistical analysis
Standard descriptive statistics were used to summarize the data (e.g. continuous: mean ± standard deviation/count: absolute and relative frequencies/time-to-event: Kaplan–Meier estimator). To identify predictors of in-hospital mortality or survival we applied logistic and Cox proportional hazards models. Consequently, both odds ratio (OR) and hazard ratio (HR) estimates were reported. To obtain multivariable adjusted estimates, all main effects with univariate p values less than or equal to 0.15 were investigated simultaneously. In addition, confidence intervals (CI) were calculated with coverage of 95%. Finally, we also created receiver operating characteristics (ROC) curves for in-hospital death, which were summarized by area under the curve (AUC) estimates. ROC analyses were based on an increasingly complex (leave one out) logistic regression model with in-hospital mortality as the predicted outcome and the following predictors, which were all entered linearly: LDH alone, LDH + age, LDH + age + body mass index (BMI), and for LDH + age + BMI + estimated glomerular filtration rate (eGFR). All reported p values are nominal and two-sided. In this explorative study, we applied a significance level of α = 0.05 (two-sided). All statistical analyses were done using GraphPad Prism 5.0 (La Jolla, CA, USA) or R 3.4.2.
Discussion
Here we present the single-center experience with regard to epidemiology, treatment and outcome of PcP in a tertiary care center over a period of 17 years. A substantial group of our PcP cases (about 50%) were not related to HIV infections and these patients had a worse clinical course with higher ICU admission and mortality rates than patients with HIV-associated PcP. Three major non-HIV-positive groups were identified: (1) solid organ transplant recipients, (2) patients with malignancies and (3) patients with rheumatic diseases. PcP occurred almost exclusively in patients who did not receive chemoprophylaxis with TMP-SMX. Overall, about one quarter of all patients suffering from PcP did not survive the disease. In those who required intensive care (approximately 40% of all patients), the in-hospital mortality increased up to 58%. Based on the - in principle - reversible nature of PcP, almost no patients were denied intensive care in the event of acute respiratory failure, as the in-hospital mortality rate in non-ICU patients was 1.6%. Moreover, we found that the extent of initial LDH elevation was a predictor of in-hospital mortality, and that in univariate comparisons, in-hospital mortality was higher in patients whose TMP-SMX dose was below 15 mg/kg BW.
Non-HIV-positive populations at risk of PcP have been described before. The distribution across these risk groups varies widely (SOT 3–31%, hematologic diseases 26–39%, rheumatic diseases 11–25%), which is likely due to the clinical emphasis in predominantly monocentric studies [
9,
18‐
20]. Several centers have reported a decrease in the percentage of HIV-infected patients among all cases of PcP in the era of cART [
19,
21]. We did not confirm this decrease, nor did Fillatre et al. [
18].
Our data showed that the clinical course and outcome of PcP differed between HIV-infected and non-HIV-infected patients. In general, HIV-infected patients require fewer ICU admissions and have better overall survival than non-HIV-infected patients, which is in line with previous reports [
9,
22]. The PcP mortality rate in HIV-positive patients was in the range of 1–15% [
9,
18,
19] compared to 30–40% in HIV-negative patients [
23‐
25]. The better outcome of HIV-positive patients might be the consequence of (1) lower age (− 3.8 years in our cohort), (2) fewer co-morbidities, (3) the reversible nature of the immune defect upon anti-viral treatment strategies and hypothetically (4) greater awareness of PcP in HIV-positive individuals, where it is the most common AIDS-defining disease.
In mixed cohorts, ICU admission and mechanical ventilation occurred in 22–40% and 13–22% of patients, respectively [
9,
18,
22]. Both events are frequently reported as negative predictors of survival. Nevertheless, newer data on ICU mortality in patients with PcP are scarce. Only one recent study reported relatively high ICU mortality in HIV-negative patients of 53% compared to 15% in HIV-positive patients [
18]. Mortality in patients with invasive mechanical ventilation was 60% and 28%, respectively, in these two groups [
22]. Compared to other reports [
21,
26], we had a relatively high ICU admission rate of 25% in HIV-positive patients, which can in part be explained by the selective transfer of more patients with more progressive disease to our university hospital as a tertiary center.
The literature is inconclusive on outcome predictors, but on multivariable analysis, a large prospective study from France demonstrated a significant survival benefit in younger patients [
9]. Although observed in HIV-positive patients before [
27], here we report for the first time that in a mixed population of HIV-positive and non-HIV-positive patients the initial levels of serum LDH at hospital admission are not only useful in the diagnostic evaluation [
28] but may also be an independent predictor of survival. This might potentially help the clinician in early identification of the sickest patients at high risk of unfavorable outcome. With regard to the specificity/sensitivity, the ideal LDH cutoff value to predict outcome has yet to be defined.
With regard to the actual treatment of PcP we found that a reduction in the recommended TMP-SMX dose below 15 mg/kg BW (for whatever reason) might be associated with higher mortality (13.1 vs. 55.8%). Given that this observation was not confirmed in the multivariable analysis, relevant confounders must be considered.
Our study has several limitations. This was a retrospective observational study based on the medical records of patients with PcP from a single institution. Consequently, causal claims cannot be made. The observation that under-dosing with TMP-SMX is associated with higher mortality might be confounded by a higher percentage of patients with renal impairment. More robust evidence from well-designed studies is needed to make firm conclusions or even to make implications about changes in standard PcP management.