|Home | About | Journals | Submit | Contact Us | Français|
Pneumocystis jirovecii pneumonia (PCP) remains the leading cause of opportunistic infection among human immunodeficiency virus (HIV)–infected persons. Previous studies of PCP that identified case-fatality risk factors involved small numbers of patients, were performed over few years, and often focused on patients who were admitted to the intensive care unit.
The objective of this study was to identify case-fatality risk factors present at or soon after hospitalization among adult HIV-infected patients admitted to University College London Hospitals (London, United Kingdom) from June 1985 through June 2006.
We performed a review of case notes for 494 consecutive patients with 547 episodes of laboratory-confirmed PCP.
Overall mortality was 13.5%. Mortality was 10.1% for the period from 1985 through 1989, 16.9% for the period from 1990 through June 1996, and 9.7% for the period from July 1996 through 2006 (P = .142). Multivariate analysis identified factors associated with risk of death, including increasing patient age (adjusted odds ratio [AOR], 1.54; 95% confidence interval [CI], 1.11–2.23; P = .011), subsequent episode of PCP (AOR, 2.27; 95% CI, 1.14–4.52; P = .019), low hemoglobin level at hospital admission (AOR, 0.70; 95% CI, 0.60–0.83; P < .001), low partial pressure of oxygen breathing room air at hospital admission (AOR, 0.70; 95% CI, 0.60–0.81; P < .001), presence of medical comorbidity (AOR, 3.93; 95% CI, 1.77–8.72; P = .001), and pulmonary Kaposi sarcoma (AOR, 6.95; 95% CI, 2.26–21.37; P =.001). Patients with a first episode of PCP were sicker (mean partial pressure of oxygen at admission ± standard deviation, 9.3 ± 2.0 kPa) than those with a second or third episode of PCP (mean partial pressure of oxygen at admission ± standard deviation, 9.9 ± 1.9 kPa; P =.008), but mortality among patients with a first episode of PCP (12.5%) was lower than mortality among patients with subsequent episodes of PCP (22.5%) (P = .019). No patient was receiving highly active antiretroviral therapy before presentation with PCP, and none began highly active antiretroviral therapy during treatment of PCP.
Mortality risk factors for PCP were identifiable at or soon after hospitalization. The trend towards improved outcome after June 1996 occurred in the absence of highly active antiretroviral therapy.
Pneumocystis jirovecii is a major cause of life-threatening pneumonia in the immunocompromised host . In industrialized countries, P. jirovecii pneumonia (PCP) is the leading opportunistic infection among HIV-infected patients . The incidence of PCP has decreased with widespread use of chemoprophylaxis, beginning in 1990, and the introduction of HAART in 1996, but PCP remains a serious clinical problem [3, 4]. PCP also illustrates societal difficulties, because many patients from underserved populations do not receive a diagnosis of HIV infection until they present with PCP.
It is unclear why some patients have mild PCP that responds to antimicrobial drugs, whereas others experience severe PCP that is fatal despite treatment. These data suggest that some isolates of P. jirovecii may have increased virulence or may have developed antimicrobial resistance. Alternatively, host defenses against Pneumocystis species may be compromised in ways too subtle to be detected by available clinical laboratory tests. In the absence of a reliable cultivation method for P. jirovecii, these issues have been addressed mainly by clinical and epidemiological studies.
Several large multicenter cohort studies have identified non-availability of a CD4+ cell count , a CD4+ cell count ≥50 cells/µL , earlier calendar year of diagnosis , increasing patient age [6, 7], risk factor for HIV infection , presence of a comorbid condition , and a previous episode of PCP  as being associated with poor outcome from PCP. These studies had differing definitions for diagnosis (laboratory-confirmed PCP or “empirical” PCP) and survival (1 or 3 months); in addition, each study examined only a small number of risk factors. By contrast, several single-center studies, each of which describe <150 patients, have identified increasing patient age [8, 9], prior receipt of PCP prophylaxis , poor oxygenation at hospitalization [11, 12], elevated lactate dehydrogenase enzyme levels , low hemoglobin  and serum albumin  levels, bacterial infection , cytomegalovirus (CMV) infection , neutrophilia in bronchoalveolar lavage (BAL) fluid [16, 17], comorbidity [18, 19], pneumothorax [20–22], and the need for mechanical ventilation [20–22] as being associated with death. These studies are inconsistent, because not all of these studies show these factors to be associated with mortality.
We performed a retrospective study of a well-described cohort of adult patients with HIV infection who had confirmed PCP and presented to a major HIV treatment center over a 21-year period to identify factors present at or soon after hospitalization that were associated with mortality. Complications occurring subsequent to hospitalization, including treatment failure, pneumothorax, need for intensive care unit (ICU) admission, and need for mechanical ventilation, were excluded from analysis.
Consecutive HIV-infected adults with PCP who were admitted to University College London Hospital (London, UK) from June 1985 through June 2006 were identified. All patients had demonstration of cysts of P. jirovecii in BAL fluid, induced sputum, or necropsy lung tissue; patients with empirically diagnosed PCP were excluded. Patients were identified from a manual and electronic search of hospital discharge summaries and were cross-referenced with the electronic hospital database and with pathology department records. University College London Hospital is a 936-bed university-affiliated hospital providing inpatient care for adult HIV-infected patients. The study was performed with University College London Hospital Research Ethics Committee approval.
Information recorded from case note review included patient age, sex, risk factor for HIV infection, awareness of HIV serostatus at hospital admission, previous history of PCP, and receipt of PCP prophylaxis. HAART became available in July 1996, but no patient received HAART in the 2 months before presentation with PCP; 3 patients had previously received HAART but had discontinued it >3 months prior to presentation. No patient began HAART during treatment of PCP . The season and year in which PCP was diagnosed were noted. Partial pressure of oxygen (PaO2) breathing room air at hospital admission, hemoglobin level, CD4+ cell count (≥1 month before hospital admission), presence of comorbid condition, presence of pulmonary Kaposi sarcoma (KS) , and presence of copathogens in BAL fluid (bacteria or CMV) [24, 25] were recorded. Serum albumin level, lactate dehydrogenase level, and plasma HIV load are not measured for patients with PCP at our center. CD4+ cell counts were not available until January 1990. Before December 1989, adjunctive corticosteroids were used sporadically for the treatment of severe PCP; subsequently, they were protocolized for patients with PaO2 <9.3 kPa breathing room air at admission. Data recorded on each patient’s hospital record included PCP treatment regimen, receipt of adjunctive corticosteroids, admission to the ICU, need for mechanical ventilation, development of pneumothorax, and outcome. Outcome was described either as death or as survival at 4 weeks after diagnosis of PCP.
Summary statistics are presented for episodes of PCP. Univariate analysis was performed using logistic regression to explore factors associated with death and to compare outcomes from first and subsequent episodes of PCP. Outcomes from the period before the routine use of adjunctive steroids (1985–1989), from the period after their introduction but before HAART was available (from 1990 through June 1996), and from the subsequent period (from July 1996 through 2006) were compared using multinomial logistic regression. Nonnormal variables were either normalized through a log10 transformation (for age and neutrophil count) or categorized (CD4+ cell count, hemoglobin level, and PaO2 at admission). Seasonal trends in mortality were explored by calendar month and by season, as previously described . Associations with mortality are presented only for factors determined at or soon after admission to the hospital, whereas associations with time of admission and associations between first and subsequent episodes of PCP are also presented for factors determined later in the course of hospitalization, including need for ICU admission, need for mechanical ventilation, and development of pneumothorax.
To provide a parsimonious multiple logistic regression model for mortality risk factors, a stepwise backward model selection procedure was performed using variables that showed evidence of association with death in the univariate analysis. For variable selection, a P value of .1 was used as the cutoff value. Among factors included in the selection procedure, the prevalence of missing data was highest for CD4+ cell count. To allow this factor to be included in the model-selection process, missing CD4+ cell count data were imputed as either <50 or ≥50 cells/µL, as predicted by an imputation model based on hemoglobin and neutrophil counts . To assess whether mortality risk factors changed over time, we assessed the interaction between each factor selected in the final model (excluding KS because of small numbers) and our time factor. In univariate and multivariate analysis, standard errors were adjusted for clustering if a patient had ≥1 episode. Stata software, version 9.0 (Stata), was used for statistical analysis. A P value <.05 was regarded as statistically significant.
From June 1985 through June 2006, 494 adult patients with HIV infection were admitted to the hospital with 547 episodes of PCP. PCP was diagnosed at bronchoscopic examination in 512 episodes, by induced sputum in 21 episodes, and at necropsy in 14 episodes. Four hundred forty-seven patients had a single episode of PCP, 41 patients had 2 episodes, and 6 patients had 3 episodes. Of those patients with a single episode of PCP, 429 experienced a first episode, and 18 experienced a second episode (having experienced their first episode of PCP at another treatment center).
Patients were predominantly men (94.3%), the main risk factor for HIV infection was sex with other men (87.4%), and the median patient age was 36 years (range, 21–72 years). Hospital admission with PCP represented the initial HIV diagnosis for 376 episodes (68.7%). In 423 episodes (77.3%), patients were not receiving PCP prophylaxis prior to hospital admission. Of 124 patients receiving prophylaxis, the regimen was sulfa-based in 55 patients (trimethoprim-sulfamethoxazole in 28 and dapsone-pryimethamine in 27), nebulized pentamidine in 65 patients, and other in 4 patients. The median CD4+ cell count was 40 cells/µL (range, 0–480 cells/µL); CD4+ cell counts were unavailable for 175 episodes. In 355 (95.4%) of 372 episodes, the CD4+ cell count was ≥200 cells/µL. Primary therapy of PCP was trimethoprim-sulfamethoxazole in 356 (65.1%) of 547 episodes, clindamycin with primaquine in 57 (10.4%), nebulized pentamidine in 55 (10.0%), intravenous pentamidine in 42 (7.7%), and dapsone with trimethoprim in 31 (5.7%), and atovaquone in 6 (1.1%); in 222 (40.2%) of 547 episodes, the patient received adjuvant corticosteroids.
Bacterial copathogens were identified in BAL fluid specimens in 38 (7.4%) of 512 episodes, as follows: Staphylococcus aureus in 16, Haemophilus influenzae in 9, Streptococcus pneumoniae in 5, Pseudomonas aeruginosa in 4, Klebsiella pneumoniae in 2, and Enterobacter cloacae in 2. CMV was identified in BAL fluid specimens in 18 episodes (3.5%). Pulmonary KS was identified at bronchoscopic examination in 19 patients (3.7%). In 41 episodes (7.6%), a comorbidity was detected, as follows: non-Hodgkin lymphoma in 5, acute psychosis in 5, acute neurologic diseases in 5, disseminated cryptococcal and/or histoplasma infection in 5, chronic alcohol abuse in 4, intercurrent bacterial septicemia in 4, heart failure and/or cardiomyopathy in 3, chronic respiratory disease in 3, thromboembolic disease in 2, and hypothyroidism, acute renal failure requiring renal support, decompensated hepatitis B and/or hepatitis C, cachexia, and pregnancy in 1 patient each. In 61 episodes (11.1%), patients were admitted to the ICU, and in 44 episodes (8%), the patient developed a pneumothorax. Overall, in 74 (13.5%) of 547 episodes, the patient did not survive. The majority of deaths occurred during hospitalization and were due to progressive respiratory failure.
Factors significantly associated with mortality were patient age, first episode of PCP, hemoglobin level, CD4+ cell count, PaO2 at admission, presence of a medical comorbidity, and detection of pulmonary KS (table 1). The mortality rate was highest in the autumn (19.7%) and lowest in the spring (10.6%); evidence for an association between season and mortality was weak (P = .16) (table 1).
In a multivariate model, older age, second or third episode of PCP, low hemoglobin level, low PaO2 breathing room air at admission, pulmonary KS, and medical comorbidity were independently associated with risk of death (table 1). This model was the result of a model selection process based on all episodes with complete data for all factors initially considered (including CD4+ cell count after imputation of missing values). Mortality factors did not vary by period.
Four hundred thirteen episodes of PCP (75.5%) occurred before July 1996, when HAART became available in the United Kingdom. However, no patient received HAART during the study period (table 2). Before July 1996, patients were more likely to be male, to have sex with other men as their risk factor for HIV infection, and to have multiple episodes of PCP (P < .001 for each). In July 1996 and after, patients were more likely to be unaware of their HIV infection status before presenting with PCP. Before July 1996, 316 (76.5%) of 413 episodes involved patients who were aware of their HIV infection status, compared with 60 (44.8%) of 134 episodes in July 1996 and after (P < .001). In July 1996 and after, patients were less likely to be receiving PCP prophylaxis (15 [11.2%] of 134 episodes), compared with before July 1996 (109 [26.4%] of 413 episodes; P = .001). Rates of pulmonary KS, CMV in BAL fluid specimens, and medical comorbidity did not differ between patients before July 1996 and in or after July 1996. Patients presenting with PCP in July 1996 or after had lower PaO2 ( ± SD) at hospital admission (8.7 ± 2.1 kPa), compared with those presenting before this time (9.6 ± 2.0 kPa; P < .001). Mortality was lower (9.7%) in July 1996 and after than it was before July 1996 (14.8%); the difference was not statistically significant (P = .142), but the power to detect such a difference was low.
There was a significant difference between the PaO2 at hospital admission in first-episode PCP (mean PaO2 ± SD 9.3 ± 2.0 kPa, when compared with the PaO2 in second- and third-episode PCP (mean PaO2 ± SD, 9.9 ± 1.9 kPa; P = .008) (table 3). Almost twice as many patients with second- and third-episode PCP (14.1%) developed a pneumothorax during hospitalization, compared with patients with first-episode PCP (7.1%). Patients with first-episode PCP and patients with a second or third episode of PCP did not differ with respect to rates of detection of pulmonary KS, detection of CMV in BAL fluid specimens, or presence of medical co-morbidities; however, in 81 (17%) of 476 first episodes of PCP and 43 (61%) of 71 second or third episodes of PCP, the patient was receiving prophylaxis (P < .001; table 3).
This is, to our knowledge, the largest single-center study to date that has examined PCP-associated mortality risk factors; it describes patients presenting over a 21-year period and includes detailed data on several clinical risk factors. Mortality from HIV-associated PCP at this center over the 21-year study period was 13.5%. Analysis of mortality risk factors specifically excluded events occurring subsequent to patient hospitalization, including treatment failure, need for ICU admission, need for mechanical ventilation, and development of pneumothorax. These latter events may be regarded as being “on the road to death” and, as such, might bias interpretation of mortality risk factors. The difference in mortality rate in July 1996 and after (in the era of HAART), compared with before July 1996, was small and was not statistically significant; no patient in this study received HAART before presenting with PCP or during treatment of this infection.
The epidemiology of patients with PCP at our center evolved during the study period. In later years, heterosexual sex was the major HIV risk factor, and an increasing number of patients were female. After June 1996, patients were more likely to be unaware of their HIV infection status, were less likely to be receiving PCP prophylaxis before presenting with PCP, and had lower CD4+ cell counts, compared with patients presenting in or before June 1996. These observations reflect the overall evolving epidemiology of HIV infection in the United Kingdom .
This study found some evidence that survival among patients with PCP improved in the post-HAART era. These data contrast with previous reports from the early HAART era that suggest more clearly an improved survival among patients with PCP [29, 30]. One multicenter study described the outcomes of 1660 episodes of PCP in 78 hospitals (overall mortality, 11.3%). In patients who had received HAART, mortality was 9.9%; in those who had not received HAART, mortality was 12.0% . By contrast, another multicenter study conducted during the same study period described 1231 episodes of PCP from 34 hospitals. Mortality rates of 8.2%–13.5% were identified ; improved survival, when compared with survival a decade earlier, was ascribed to general improvements in medical care, rather than to a direct effect of HAART.
Improved survival has been described in the era of HAART among the subgroup of patients with severe PCP admitted to the ICU [22, 31, 32]. Whether improved outcome is attributable to the direct effects of HAART  or to general improvements in ICU care (in particular, protective ventilator strategies) remains unclear . In the present study, no patient received HAART before or during hospitalization with PCP. This observation might in part account for the lack of apparent improvement in mortality in the era after June 1996, when HAART became available.
The present study identified that mortality from second- or third-episode PCP was greater than that from first-episode PCP. This finding confirms results from the Adult and Adolescent Spectrum of HIV Disease Project, which identified recurrent PCP as a risk factor for mortality from PCP , and results from a study by Mansharamani et al. , which showed higher mortality associated with second- or third-episode PCP (17% and 21%, respectively) than with first-episode PCP (11%). Details of patient disease severity are not provided in either study, limiting comparison with the present study. By contrast, a study from the pre-HAART era by Dohn et al.  showed that patients with second- or third-episode PCP had milder disease, as indicated by PaO2 at admission to the hospital, and better outcome than patients with first-episode PCP. In the present study, patients with subsequent-episode PCP had milder disease (as indicated by PaO2 at admission to the hospital) but worse outcome. There are 2 possible explanations for this discrepant observation. First, persistent abnormalities of lung function following PCP , together with the demonstration of post-PCP fibrosis on CT , suggest reduced pulmonary compliance and increased susceptibility to pneumothorax. In the present study, almost twice as many patients with second - or third-episode PCP had a pneumothorax (14.1%), compared with patients with first-episode PCP (7.1%). Second, there might be greater impairment of immune function among persons with recurrent PCP than among those with first-episode PCP. This explanation is supported by findings from this study, because patients with recurrent PCP had lower CD4+ cell counts and lower systemic inflammatory responses (as measured by peripheral blood neutrophil counts). This hypothesis is supported by demonstration of impaired antibody responses in HIV-infected patients recovering from recurrent PCP, compared with responses in those recovering from first-episode PCP .
In the present study, pulmonary KS was an independent risk factor for mortality. An explanation for this observation is not immediately apparent. Reports of reduced gas transfer factor and coefficient in HIV-infected patients with pulmonary KS and without intercurrent opportunistic infection suggest that pulmonary KS may contribute to impaired gas exchange during PCP [23, 38]. Medical comorbidity, when present, adversely impacted outcome from PCP. Likely explanations are multifactorial and include an additional inflammatory burden from intercurrent infective processes, underlying chronic respiratory compromise, and intercurrent cardiovascular disease.
Although season was not a significant risk factor for mortality, the present study reports possible seasonal changes in PCP-associated mortality. Additional details regarding overall hospital admissions and deaths and changes in staffing and in policies and procedures over the study period are needed to make further inferences. However, reports of seasonal variation in concentration of Ascomycetous species and other fungi in air  suggest that patients might have been exposed to a greater burden of Pneumocystis species in the autumn or been exposed to a more virulent genotype . Recent reports describing an association between Pneumocystis species colonization and worsening of chronic lung disease [39, 40] raise the possibility that medical problems caused by Pneumocystis species extend beyond HIV-infected patients and emphasize the need for further studies examining environmental aspects of this infection.
This study has several limitations. First, it was retrospective and included only patients with laboratory-proven PCP, precluding analysis of patients with presumptive PCP, who may have had different clinical manifestations or severity of the disease . Second, it only examined mortality risk factors present at or soon after hospital admission. Other factors that are regarded as being on the “road to death” were excluded from analysis. Lack of patient demographic and clinical data, including race and/or ethnicity, smoking history, and laboratory test results (e.g., serum albumin and lactate dehydrogenase levels) further limits comparison with other studies.
In summary, at this center, overall mortality from PCP over the 21-year period of the study was 13.5%. The difference between the mortality rate in the era of HAART (after June 1996) and the mortality rate in or before June 1996 was small and was not statistically significant. Factors identified early in the course of patient hospitalization were associated with risk of death. The continuing presentation of patients with PCP and their attendant mortality rate underscores the need for earlier diagnosis of HIV infection to identify patients before they develop PCP.
Analysis of this dataset was performed by P.D.W. in partial fulfilment of the requirements for the degree of MSc in Epidemiology at London School of Hygiene and Tropical Medicine (London, UK).
Financial support The Medical Research Service, Department of Veterans Affairs, and the National Institutes of Health (RO1 AI-06492 and F33 AI-065207).
Potential conflicts of interest All authors: no conflicts.