Prognostic factors for severe Pneumocystis jiroveci pneumonia of non-HIV patients in intensive care unit: a bicentric retrospective study

With the widespread use of Pneumocystis jirovecii pneumonia (PJP) prophylaxis and highly active antiretroviral therapy (HAART), the incidence and mortality of PJP in HIV patients have declined substantially in Europe and the United States [1, 2]. Current estimates of hospital mortality for PJP in HIV patients range from 7 % to 11 % [3]. Most importantly, respiratory failure due to PJP requiring ICU admission was less common in HIV patients [4]. In contrast, PJP rate is increasing in non-HIV patient [5] and the reported mortality of PJP in immunocompromised non-HIV patients ranges from 48 % to 67 % [3]. The clinical course, and inflammatory response might contribute to the different mortality rates between the two groups. In addition, the sample sizes of previous studies exploring the prognostic factor were small [6‐8]. The prognostic factor for severe PJP in the intensive care unit (ICU) setting has not been well described.

In order to determine the prognostic factors for PJP in ICU setting, we retrospectively collected data for a consecutive series of PJP patients requiring ICU admission from October 2012 to October 2015.

Overall, 348 patients were discharged with a diagnosis of PJP during the study period, of whom 266 met exclusion criteria. Consequently, 82 patients were available for the final analysis, including 72 patients PUMCH and 10 patients in CJFH (Fig. 1).

Table 1 shows characteristics and outcomes of confirmed PJP in non-HIV patients. The entire 82 patient cohort had a mean age 53 ± 17 years of and APACHE II score of19 ± 5, with 41.4 % male. Most of the patients had an underlying disease of connective tissue disease (79.3 %) and history of corticosteroid therapy (84.1 %). All patients were treated with trimethoprim-sulfamethoxazole (TMP-SMX). Not a single patient received PJP prophylaxis. PJP was diagnosed by methenamine silver stain, PCR, or both in 13, 45, and 24 patients respectively, without any difference between 2 hospitals (Additional file 1: Table S2). PJP-PCR positive samples included sputum (n = 5), trachea aspirate (n = 17), and BALF (n = 49). Methenamine silver stain positive samples included sputum (n = 4), trachea aspirate (n = 12), and BALF (n = 19). All pulmonary aspergillosis was diagnosed as EORTC probable invasive aspergillosis with the presence of a host factor, a clinical criterion, and a mycological criterion. For the 38 patients receiving caspofungin, total days on caspofungin were 8 ± 6. Twenty-five patients received empirical caspofungin therapy for less than 7 days for suspected invasive fungal infection. Another 6 patients received caspofungin as a combination therapy with amphotericin B or voriconazole for aspergillosis. Although the clinical efficacy of caspofungin as salvage therapy for PJP remained controversial [12], it was administered in 7 patients in our cohort study as a salvage regimen of whom 5 patients died during study period.

During their hospital stay, 62 (75.6 %) of the 82 patients died. Compared with survivors, the non-survivors had older age (55 ± 16 vs. 45 ± 17, p = 0.014), higher APACHE II score (20 ± 5 vs. 17 ± 5, p = 0.01), lower WBC (7.68 ± 3.44 vs. 10.48 ± 4.62, p = 0.005), less fever (80.6%vs. 100 %, p = 0.033), more hypotension (58.1 % vs. 20 %, p = 0.003), and more pneumomediastinum (29 % vs. 5 %, p = 0.027), while the difference was not statistically significant for lymphocyte counts, CD4 cell count, type of respiratory support on ICU admission. Four patients received high-frequency oscillatory ventilation and one patient received extracorporeal membrane oxygenation. All these five patients died during the hospital stay.

To investigate the role of potential confounding prognostic factors, a multivariate analysis was performed for hospital mortality (Table 2). The Hosmer and Lemeshow goodness-of-fit test were not rejected (p = 0.640), indicating adequate model fit. No interaction terms were found to be significant in this model, and there was no collinearity between any of the independent variables. The multivariate model indicated that age [odds ratio (OR)1.051; 95 % CI 1.007-1.097; p = 0.022], WBC [OR 0.802; 95 % CI 0.670-0.960; p = 0.016], and pneumomediastinum [OR 16.514; 95 % CI 1.330-205.027; p = 0.029] were independently significantly associated with hospital mortality.

CT scans were performed for all those 82 patients on ICU admission. Pneumomediastinum was confirmed by CT in 14 patients and chest x-ray plus subcutaneous emphysema in 5 patients. Six patients developed pneumomediastinum before hospital admission. Other 13 patients had time intervals between ICU admission and pneumomediastinum with a mean of 9 days (range 2–30 days). The potential risk factors contributed to pneumomediastinum were listed in Table 3. There was statistically significant difference in the percentage of patients treated with non-rebreathing mask (NRM) on ICU admission between pneumomediastinum and non-pneumomediastinum (47.4 % vs. 23.8 %, p = 0.048). Tidal volume, plateau pressure, and PEEP was similar between those two groups.

In this bicentric retrospective observational study across 4 Chinese ICUs, we found that age, WBC, and pneumomediastinum were significantly associated with hospital mortality in non-HIV immunocompromised patients with severe PJP who had been admitted to ICU. Use of non-rebreathing mask might contribute to the development of pneumomediastinum.

The most interesting finding of our study was that pneumomediastinum was associated with increased hospital mortality. Pneumomediastinum is the presence of extra-alveolar air in the mediastinum, which is believed to arise from free air leaking from ruptured alveoli. It was described as an uncommon complication of opportunistic infections in HIV-infected patients [13, 14]. However, 24.4 % patients in our study developed pneumomediastinum. The incidence rate discrepancies may be due to different underlying disease (HIV vs. Non-HIV) and few reported incidence rate in previous studies. A retrospective radiographic analysis reported an incidence rate of 11.1 % (4 of 36) in a cohort of moderate non-HIV PJP patients with a mortality rate of 33.3 % [15]. We also found more pneumomediastinum developed in the NRM group. This might be explained by higher trans-pulmonary pressure and tidal volume during spontaneous breath resulting in air leak, which was consistent with a previous report of HIV patients [13]. Delayed intubation was considered as a risk factor for worse outcome [6]. Although there was no difference in the time interval from symptom onset to intubation in our study, use of non-rebreathing mask instead of positive pressure support on ICU admission suggested delay intubation which was very difficult to define.

Development of pneumothorax was independently associated with increased mortality in previous studies [6, 16]. We did not find any difference in mortality between patients with and without pneumothorax. Protective lung ventilation strategies might account for the different findings. The tidal volumes in our study were smaller than Festic and colleagues (7 ml/kg vs. 10 ml/kg). Despite the application of protective lung ventilation, one fourth patient in this cohort developed pneumomediastinum, which suggested that pneumomediastinum was not a complication of intervention. As Cho et al. [17] reported, the development of the pulmonary cysts and bronchiectasis that were noted in follow up CT but were not visible on CT at admission could be risk factors for development of pneumomediastinum.

We also found lower WBC was related to increased mortality. Although WBC has never been reported as risk factors, previous study [18] suggested a trend of higher WBC in non-HIV patients and survivors, which was consistent with our findings.

Overall mortality of the patients in our study was 75.6 %. Although the reported mortality rates of ICU non-HIV patients with PJP in previous studies were 38.9-84.2 % [6‐8, 18‐28] (Additional file 2: Table S1), most of the mortality rates were less than 70 %, which were lower than that of our study. The high hospital mortality rate in our study possibly was related to different underlying diseases and no prophylaxis of PJP for those patients. However, although prophylaxis for PJP was recommended for HIV patients, the efficacy of prophylaxis for immunocompromised non–HIV patients has not been well established [29‐31], especially for the patients with underlying disease of connective tissue disease. In a recent study of ICU patients with PJP, adjunctive steroid was associated with increased mortality [16]. This might be the cause of high hospital mortality in our study. Considering that most of the patients in our study received steroid therapy before ICU admission, the use of steroid was not avoidable. Moreover, there was no difference in steroid therapy between survivors and non-survivors, and the effects of other covariates remained significant.

In a retrospective study, Chen and colleagues reported the characteristics and prognostic factors of 69 HIV-negative patients with PJP from PUCMH during 10-year study period [25]. In comparison, we had enrolled 72 patients with confirmed PJP during a 3-year period in our cohort (Additional file 1: Table S2, Additional file 3: Table S3). Increasing awareness of the disease in immunocompromised patients among clinicians and widespread implementation of PCR technique for PJP diagnosis might account for the discrepancy between Chen’s study and ours.

Our finding suggested that PJP in non-HIV patients requiring ICU admission remains a challenge for clinician. Poor prognostic factors included older age, lower WBC, and development of pneumomediastinum.