Pneumocystis jirovecii pneumonia in intensive care units: a multicenter study by ESGCIP and EFISG

Pneumocystis jirovecii pneumonia (PJP) was most commonly described as an opportunistic, life-threatening disease in patients with human immunodeficiency virus (HIV) infection [1, 2]. However, due to the increased use of immunosuppressants, biologic agents, and corticosteroids for treating diseases such as inflammatory diseases, hematological malignancies, and solid neoplasms, novel populations at risk of PJP have emerged over the last decades, with reported mortality ranging from 33 to 69% [3‐7].

Patients with severe PJP often require intensive care management. Polymerase chain reaction (PCR) targeting Pneumocystis DNA from respiratory samples and serum (1,3)-β-D-glucan (BDG) are widely used to support the diagnosis of PJP in these patients, despite the heterogeneity in assay design and positivity threshold [8‐12]. The distribution of predisposing diseases/conditions in critically ill patients admitted to intensive care unit (ICU) and subjected to diagnostic workup for PJP, or diagnosed with PJP, has seldom been explored, and usually in small, single center cohorts limiting confidence in findings [13].

Overall, 600 patients were included in the study (Fig. 1). Of them, 115 were classified as presumptive/proven PJP (19.2%), and 444/600 were classified as no PJP (74%). The remaining 41/600 patients (6.8%) were classified as “PJP diagnosis inconclusive”. Among patients with PJP, 31/115 (27.0%) and 84/115 (73.0%) were classified as proven PJP and presumptive PJP, respectively. The baseline demographic and clinic characteristics were similar between patients with proven PJP and presumptive PJP (Additional file 1: Table S2). As shown in Table 1, only 8.8% of patients subjected to diagnostic workup for PJP had HIV infection, whereas hematological malignancy, solid tumors, inflammatory diseases, and solid organ transplants were present in 23.2%, 16.2%, 15.5%, and 10.0% of patients, respectively. Concomitant coronavirus disease 2019 was present in 75 patients (12.6%). Only 3.8% of patients were under PJP prophylaxis (in all cases with trimethoprim/sulfamethoxazole) at the time of BDG/PCR testing. The median SOFA score at the time of the BDG/PCR testing was 6 (interquartile range 4–9), and as many as 28.9% of patients had septic shock. The frequency of presumptive/proven PJP by baseline condition/disease was 42.3% (22/52) in patients with HIV infection, and 55.9% (19/34) among HIV infected patients with AIDS, while among patients with hematological malignancy, solid tumor, inflammatory diseases, and solid organ transplants it was 25.9% (36/139), 28.9% (28/97), 18.3% (17/93), and 10.0% (6/60), respectively. Of note, the frequency of presumptive/proven PJP in our study population was as high as 47.1% (16/34) and 40.0% (14/35) in patients with non-Hodgkin lymphoma and metastatic solid neoplasms, respectively.

The results of univariable and multivariable analyses of factors associated with presumptive/proven PJP, conducted in 559 patients (after the exclusion of 41/600 patients with inconclusive PJP diagnosis) are available in Tables 2 and 3, respectively. As shown in Table 2, HIV infection, AIDS, hematological malignancy, non-Hodgkin lymphoma, solid tumor, metastatic solid tumor, Charlson Comorbidity Index, previous chemotherapy, previous radiotherapy, increasing serum CRP values, and bilateral ground-glass opacities were associated with PJP in univariable analysis. Conversely, an inverse association with PJP was observed for increasing length of hospital and ICU stay before PJP testing, invasive mechanical ventilation, increasing lymphocyte cell count, and increasing neutrophil cell count. In multivariable model A (Table 3), AIDS (odds ratio [OR] 3.31; 95% CI 1.13–9.64, p = 0.029), non-Hodgkin lymphoma (OR 3.71; 95% CI 1.23–11.18, p = 0.020), vasculitis (OR 5.95; 95% CI 1.07–33.22, p = 0.042) metastatic solid tumor (OR 4.31; 95% CI 1.76–10.53, p = 0.001) and bilateral ground glass on CT scan (OR 2.19; 95% CI 1.01–4.78, p = 0.048) retained an association with PJP, whereas an inverse association with PJP was retained for invasive mechanical ventilation (OR 0.43; 95% CI 0.24–0.80, p = 0.007), and increasing lymphocyte cell count (OR 0.64; 95% CI 0.42–1.00, p = 0.049). The results of the additional multivariable model B including center as a random effect were in line with the directions of effects observed in model A (Table 3).

The diagnostic performance of respiratory Pneumocystis PCR for the diagnosis of presumptive/proven PJP and for proven PJP only is shown in Table 4. Overall, respiratory Pneumocystis PCR was performed in 561/600 patients (93.5%), with results available for 550/561 patients (in the remaining 11 patients, Pneumocystis PCR was performed but the results were not provided). Using any positive results for the molecular detection of Pneumocystis DNA as the criterion for positivity, respiratory Pneumocystis PCR showed 100% sensitivity and 100% NPV both for diagnosis of presumptive/proven PJP and for proven PJP (in this latter case the analysis was conducted in the subgroup of patients who underwent microscopy, 190/550, 34.5%). Specificity was 85% and 61% for the diagnosis of presumptive/proven PJP and proven PJP, respectively. In subgroups according to the type of respiratory specimen (sputum, tracheal aspirate, or bronchoalveolar lavage fluid) specificity for the diagnosis of presumptive/proven PJP was lower for Pneumocystis PCR performed on sputum (68%) compared with deeper samples (85% for both tracheal aspirate and bronchoalveolar lavage fluid [BALF]), with Pneumocystis PCR on BALF showing overall the best diagnostic performance (Table 4). The diagnostic performance of serum BDG for the diagnosis of presumptive/proven PJP and for proven PJP only is shown in Table 5. Overall, serum BDG testing was performed in 327/600 patients (54.5%), mostly employing the Fungitell assay (293/327, 89.6%), whereas the Wako assay was employed only in 34/327 cases (10.4%), as shown in Additional file 1: Table S1. The distribution of BDG values according to presumptive/proven PJP diagnosis is shown in Additional file 1: Figure S1 (results reported for the Fungitell assay). Based on Youden index, the best compromise between sensitivity and specificity for the diagnosis of presumptive/proven PJP using the Fungitell assay was a cut-off level of 230 pg/mL, showing 82% sensitivity and 82% specificity (figure S2). The diagnostic performance of respiratory Pneumocystis PCR and serum BDG in subgroups according to baseline disease/condition is available in Additional file 1: Tables S3 and S4, respectively. Although with moderate strength, correlations were identified between increasing serum BDG values and decreasing Ct values of Pneumocystis PCR on BALF (63 pairs, p < 0.001, Spearman r − 0.49) and between increasing serum BDG values and decreasing Ct values of Pneumocystis PCR on tracheal aspirate (10 pairs, p = 0.041, Spearman r -0.67). Finally, an increased PPV was observed when both respiratory Pneumocystis PCR and serum BDG were positive in comparison with the performances of the two markers separately, whereas when both were negative their NPV did not substantially vary compared with the NPV obtained using only respiratory Pneumocystis PCR or only serum BDG (Fig. 2 and Additional file 1: Table S5).

Overall, 98.3% of patients with presumptive/proven PJP received PJP therapy (113/115, 98.3%), mostly with trimethoprim/sulfamethoxazole (96.5%, 109/113). Information on concomitant/alternative infections due to pathogens other than Pneumocystis jirovecii are available in Additional file 1: Table S6. As shown in Additional file 1: figures S3 and S4, both cumulative 30-day mortality and cumulative 90-day mortality were high and similar in patients with and without PJP (i.e., with other causes of pneumonia, see Additional file 1: Table S6) in our study population. More in detail, cumulative 30-day mortality was 52% and 47% in patients with presumptive/proven PJP and without PJP, respectively (log-rank test, p = 0.74), and cumulative 90-day mortality was 67% and 68% in patients with presumptive/proven PJP and without PJP, respectively (log-rank test, p = 0.76). Cumulative 30-day and cumulative 90-day mortality stratified by baseline condition/disease (HIV infection vs. other baseline diseases or conditions) in patients with presumptive/proven PJP are also available in Additional file 1: Figures S5 and S6.

In the present large, multinational cohort study, we showed that: (i) more patients admitted to ICU are subjected to PJP diagnostic workup compared with the past, with a reduction in the relative frequency of patients with HIV infection and an increase in patients with hematological and neoplastic diseases, solid organ transplant, and inflammatory diseases; (ii) combining serum BDG and respiratory Pneumocystis PCR improved the PPV for the diagnosis of PJP in critically ill patients admitted to ICU.

Although patients with HIV infection and AIDS were still those eventually showing a higher frequency of PJP in our cohort (55.9%), the absolute number of HIV patients with PJP was much lower than the absolute number of patients with PJP and other predisposing diseases or conditions, which is in line with results from previous smaller cohorts [5, 6]. Overall, while HIV infection remains one of the factors conferring the highest risk of PJP, the expanding denominator of ICU patients with other predisposing conditions overall makes PJP more commonly encountered in non-HIV critically ill patients nowadays, a fact that should be considered by ICU physicians caring for patients with pneumonia and forms of immunosuppression other than HIV infection. Besides AIDS, other predisposing conditions showing an independent association with PJP in multivariable models were non-Hodgkin lymphoma, metastatic solid tumor, and vasculitis. Among factors inversely associated with PJP diagnosis, the apparently reduced risk conferred by invasive mechanical ventilation at the time of PJP diagnostic workup may reflect the fact that PJP was most frequently the reason for ICU admission and for requirement of invasive mechanical ventilation rather than a complication developing during ICU stay. The increased risk conferred by low lymphocyte counts is conversely more likely testifying a true risk factor, in line with what observed in HIV patients with low CD4 + lymphocyte cell count [21]. Of note, although a few associations did not retain statistical significance in the additional multivariable model including center as a random effect (model B), the directions of effects were in line with those observed in the primary model (model A), suggesting limited influence of between-center variability on results. This is important considering the possible heterogeneity in the diagnostic workup for PJP in the different participating centers (e.g., testing in all patients with interstitial pneumonia vs. testing only in patients with risk factors), and strengthens the generalizability of the detected associations.

Regarding the secondary objective of assessing the diagnostic performance for PJP of respiratory Pneumocystis PCR and serum BDG, in our study respiratory Pneumocystis PCR showed high sensitivity (100%) and relatively low specificity (up to 85% for BALF, and lower for sputum samples) for the diagnosis of presumptive/proven PJP. Notably, although the majority of patients in our cohort underwent PCR testing on BALF, some did not; this could represent an important limitation for the diagnostic assessment of PJP in real life setting, considering the overall better diagnostic performance of BALF testing compared with other respiratory samples [22‐25]. In our opinion, some important considerations may stem from our results: (i) in presence of a negative respiratory Pneumocystis PCR, PJP seems to be very unlikely (i.e., PCR retains a high NPV), even in presence of a positive serum BDG; (ii) concomitant positivity of respiratory Pneumocystis PCR and serum BDG increases PPV in comparison with PCR alone, potentially helping to distinguish between Pneumocystis colonization and infection in the absence of established cut-offs for quantitative molecular tests (and/or when qualitative molecular tests are employed) [25, 26].

A first important limitation, related to our primary descriptive analysis, is the lack of precise information about previous/ongoing corticosteroid treatment (e.g., dosage, length of therapy), a recognized predisposing condition to PJP, that was unfortunately unavailable retrospectively. However, in our opinion, this limitation eventually did not undermine the identification of the spectrum of baseline diseases nowadays encountered in critically ill patients with PJP. Future studies with a more detailed information on steroid treatments could help to define the possible reasons for the low prevalence of patients on PJP prophylaxis in our cohort despite the high frequency of patients with at least one predisposing disease/condition. Notably, a low frequency of PJP prophylaxis in non-HIV patients at risk of PJP was already reported by Dunbar and colleagues and by Roux and colleagues [27, 28]. The lack of detailed information about corticosteroid treatments also precluded the evaluation of the diagnostic performance of respiratory Pneumocystis PCR and serum BDG using the EORTC/MSGERC definition of probable PJP, although it should be highlighted that this would have resulted in a significant incorporation bias [14]. Of course, our definition of presumptive PJP, although more suitable for the aims of the present study, is not standardized and also still hampered by a nonnegligible risk of incorporation bias, that should be necessarily acknowledged. This also considering the relatively low frequency of patients who underwent computerized tomography (56.5%), the absence of which could have further hampered the identification of PJP and its differential diagnosis in some cases. Second, in the assessment of the diagnostic performance of respiratory Pneumocystis PCR we defined any positive result as the criterion for positivity, to preserve homogeneity of interpretation despite the wide heterogeneity of employed molecular tests across centers (Additional file 1: Table S1), whereas higher fungal loads are typically more indicative of active infection than low fungal loads [29]. In this regard, the highly fragmented data in terms of Ct values available for this study prevented us from reliably assessing potential thresholds to distinguish between colonization and infection. A third limitation arises from the recognized suboptimal sensitivity of the diagnostic gold standard (microscopy) [30] and from the low number of microscopy tests performed in our real life cohort, that forced us to strongly rely on a definition of presumptive disease as diagnostic reference. Finally, it should be noted that mortality rates reported in this study, either for PJP or for pneumonia caused by other pathogens, are only presented descriptively and unadjusted for other prognostic factors.