INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

At the beginning of the AIDS epidemic, Pneumocystis carinii pneumonia (PCP) was rapidly recognized as the most frequent and severe respiratory infection in HIV-positive patients. Studies involving univariate analysis to look for high risk factors present at admission found that severe hypoxemia, high lactate dehydrogenase (LDH) levels, and marked neutrophilia in bronchoalveolar lavage (BAL) indicated an increased risk of death (1). The development of pneumothorax and/or a need for mechanical ventilation during the course of the disease were also associated with mortality in some studies (11).

Between 1990 and 1997, despite the development of effective antiretroviral therapy and PCP prophylaxis, PCP remained indicative of AIDS in more than 15% of cases. Most HIV-positive patients admitted for PCP were either unaware that they were HIV-positive or were not receiving follow-up (14, 15). Even in randomized trials, PCP continued to carry a 90-d mortality rate of 13 to 20% despite trimethoprime-sulfamethoxazole therapy and, in patients with a Pa_O2 < 70 mm Hg on room air, of adjunctive steroid therapy (16). Mortality rates of as much as 90% were reported in patients who required mechanical ventilation (15, 20, 21). However, the risk factors for mortality in HIV-associated PCP have not been reevaluated since the introduction of adjunctive steroids.

The purpose of this retrospective study was to conduct a multivariate analysis of the risk factors for mortality in 144 consecutive patients with AIDS and PCP admitted to a university hospital between 1990 and 1995, with particular attention to BAL neutrophilia.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Selection

All consecutive HIV-infected patients (two positive ELISAs and one positive Western Blot) with BAL-proven PCP managed between October 1, 1990 and September 15, 1995, were identified by retrospective review of the records of the cytological and parasitological laboratories of our institution.

Because this retrospective study required the availability of reliable clinical and laboratory data collected at admission and during the course of the disease, only patients hospitalized in our institution were included.

To restrict our study population to patients with new episodes of PCP as opposed to slowly-resolving PCP, we excluded patients who had been admitted for PCP to any institution during the previous 3 mo.

PCP Diagnosis and BAL Evaluation

Bronchoscopy with BAL was carried out as previously described (22) and performed by wedging the bronchoscope into a distal airway of the relevant bronchus (middle lobe or lingula in patients with diffuse radiologic changes), then the BAL was performed. The fluid was immediately divided into aliquots for examinations and cultures and dispatched to different laboratories.

Routine evaluation for Pneumocystis carinii included two stains (Toluidine blue O and May-Grunwald-Giemsa) and a specific immunofluorescence test. Results of the stains were used to establish the presence of P. carinii cysts and the diagnosis of PCP. None of the patients with negative stains had a positive immunofluorescence test.

BAL fluid was also evaluated for opportunistic bacteria, viruses, fungi, other protozoa, and acid-fast bacteria.

In the cytology laboratory, BAL fluid was cytocentrifuged (10³ cells per slide), and a staining by May-Grunwald-Giemsa permitted cellular count and typing as well as identification and quantification of PC elements. Cytomegalic inclusions were noted when present.

Protected bronchial brushing was routinely performed during bronchoscopy, before BAL.

Data Collection

For each patient with AIDS-related PCP, the following information was collected retrospectively by medical record review using a standardized procedure.

(1) Epidemiologic and clinical data at admission: sex, age, risk factors for HIV infection, history of PCP more than 3 mo before the current hospitalization, PCP prophylaxis (trimethoprime-sulfamethoxazole, pentamidine aerosols); antiretroviral therapy (azidothymidine, 2',3'-dideoxycytidine or 2',3'-dideoxyinosine); and weight loss in the previous 2 yr. If the HIV infection was revealed by the current episode, PCP was considered inaugural.

(2) Initial site of hospitalization: intensive care unit (ICU) or medical department.

(3) Established prognostic factors in AIDS-related PCP: Pa_O2 on room air, lactate dehydrogenase (LDH) level, CD4 lymphocyte count, and albuminemia.

(4) BAL data: cell counts and type, concomitant opportunistic infection, concomitant bacterial infection (defined as a positive culture of a protected specimen brush with  10³ CFU/ml).

(5) Treatment for PCP: trimethoprime (20 mg per kilogram per day)-sulfamethoxazole (100 mg per kilogram per day), or pentamidine (4 mg per kg) in patients with documented allergy to trimethoprime-sulfamethoxazole; this treatment was administered at admission when the diagnosis of PCP was suspected.

(6) Adjunctive steroids: for patients with Pa_O2 on room air at admission between 50 and 70 mm Hg, steroids were started at a dose of 40 mg of oral prednisone (or 30 mg of intravenous methylprednisolone) twice a day and continued for 21 d (16); patients with severe hypoxemia received 240 mg of intravenous methylprednisolone at tapering doses every 3 d for 21 d (17). This treatment was administered only immediately after P. carinii recovery from BAL fluid analysis.

(7) Events during the first 90 d: concomitant opportunistic infection, nosocomial infection (urinary, pulmonary, catheter-related, septicemia of unknown origin), pneumothorax, need for mechanical ventilation, intolerance to trimethoprime-sulfamethoxazole (rash, fever  38.5, white blood cell < 2 g/L, platelets < 50 g/L, renal failure), and intolerance to pentamidine (hypoglycemia, diabetes, pancreatitis).

End-point Determination

The end point in this study was 90-d mortality (from the day of initial bronchoscopic diagnosis) as determined based on the patient's medical record; on information obtained by telephone from the patient, a family member, a caregiver, or the managing physician; or on examination of death certificates.

Statistical Analysis

Descriptive statistics for continuous variables are expressed as means ± standard deviations. Continuous variables were dichotomized as categorical variable by using median values as cut points.

Comparisons of patients were performed using the chi-square test for categorical variables and Wilcoxon's test for continuous variables. We used a nonconditional logistic regression model to identify determinants of 90-d mortality, a follow-up period available for all included patients and likely to reflect the acute effects of PCP. Calculations were performed on the statistical library BMDP LR (23).

To highlight the impact of parameters available at admission, two multivariate models were constructed, one with all parameters available at admission and the other with these parameters plus parameters collected during the course; p values  0.05 by two-tailed tests were considered statistically significant. Goodness of fit for the nonconditional logistic regression model was evaluated using the Hosmer-Lemeshow chi-square test (p value > 0.05) (24).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

From October 1, 1990 to September 15, 1995, 2,340 BAL fluids from HIV-infected patients were examined at the cytology and parasitology laboratories of our institution.

P. carinii was present in 228 BAL fluids from 208 HIV- infected patients. Among these 208 episodes of BAL-proven HIV-related PCP, 44 patients were not hospitalized in our institution and 20 patients had a documented PCP within 3 mo; therefore, 144 patients met our selection criteria. Ninety-day follow-up data were available for all patients.

Patient Characteristics

As shown in Table 1, there were 117 men and 27 women, with a mean age (± SD) of 37 ± 9 yr. Seventeen patients (12%) were intravenous drug users. Mean weight loss during the previous 2 yr was 7.32 ± 8 kg.

PCP was inaugural in 60 patients (41.5%). Eleven patients (7.6%) had been admitted for PCP more than 3 mo earlier; in these patients, the previous episode was managed either at our institution before October 1, 1990 or at another institution.

Forty patients (28%) were receiving PCP prophylaxis and 36 (25%) were receiving antiretroviral therapy.

At admission, Pa_O2 on room air was 61 ± 20 mm Hg, the lactate dehydrogenase level was 1,066 ± 568 IU, the CD4 lymphocyte count was 57 ± 70/mm³, and serum albumin was 30 ± 6 g/L.

BAL Data at Admission

BAL fluid examination showed increased cellularity, to 256 ± 187 10³/mm³, with increases in the proportion of lymphocytes (20.5 ± 15%), the absolute number of lymphocytes, the proportion of neutrophils (16.5 ± 21%), and the absolute number of neutrophils (Table 1). Cytomegalic inclusions were observed in 10 patients, and a bacterial infection with a positive protected specimen brush was seen in 25 (17.5%).

Treatment of PCP, Follow-up, and Mortality

As first-line treatment, 121 patients received TMP-SMZ, 20 received pentamidine, and three received another anti-PCP drug. Initially or secondarily, 85 patients (59%) also received steroids because of severe hypoxemia (Table 2). Although 36% of patients (n = 52) developed side effects from the treatments (Table 2), only 21% had to be switched from their first-line treatment to another treatment because of side effects.

A concomitant opportunistic infection was diagnosed during hospitalization in 41 patients (28.5%). Nosocomial infection occurred in 19 patients (13%). Initially or secondarily, 17 patients (12%) developed pneumothorax and 15 needed mechanical ventilation.

Pneumothorax were observed in three situations: six patients were receiving pentamidine aerosols for PCP prevention and may have had active subpleural pneumocystosis; five others received mechanical ventilation, with its possible traumatic effect; one received both pentamidine aerosols and mechanical ventilation; and five did not present these two conditions.

Vital status at 90 d was determined for all patients. Overall 90-d mortality was 21.5%; 16 patients died during the first 30 d and 15 died between Days 31 and 90. The cause of death, established on the basis of data from medical records, death certificates, or autopsy reports, was known for 21 of the 31 patients who died. PCP was the primary cause of death in 17 of these 21 patients (81%); the remaining four patients (19%) died from another opportunistic infection.

Comparison between ICU Patients and Patients from Other Departments

Seventy-one patients (49.3%) were hospitalized at admission in the respiratory ICU. As compared to the 73 patients admitted to other departments, the 71 patients admitted to the ICU were more likely to have inaugural PCP (p = 0.02) and had lower Pa_O2 on room air (p < 0.0001), a higher lactate dehydrogenase level (p < 0.0001), a lower CD4 lymphocyte count (p = 0.04), and a higher percentage of neutrophils in BAL fluid (p < 0.05). Cytomegalic inclusions were observed in nine of the 71 patients admitted to the ICU versus only one of the 73 other patients. Patients admitted to the ICU had a significantly (p = 0.007) higher mortality rate (29.5 versus 11%).

Predictors of 90-d Mortality

Univariate analysis (Table 2). The same predictors of mortality were found with 30-d or 90-d outcome as end point (Table 1).

Among clinical and laboratory variables (including BAL) at admission, four were significantly associated with 90-d mortality: Pa_O2 < 60 mm Hg, lactate dehydrogenase > 1,000 IU, albuminemia < 30 g/L, and BAL neutrophilia > 10%. Weight loss of more than 7 kg during the previous 2 yr, a positive protected specimen brush, a CD4 lymphocyte count less than 50 per mm³, or cytomegalic inclusions in BAL fluid did not significantly affect 90-d mortality.

Among events recorded during hospitalization, pneumothorax, a need for mechanical ventilation, and nosocomial infection were each associated with 90-d mortality. In contrast, a concomitant opportunistic infection did not significantly influence this parameter.

At admission, BAL neutrophilia, but not Pa_O2 on room air, was related to the occurrence of pneumothorax and to a need for mechanical ventilation during the course of the disease (Figure 1).

View larger version (30K): [in this window] [in a new window]   Figure 1.   BAL neutrophilia and PaO2 on room air at admission in patients with pneumothorax (PNO) or with the need for mechanical ventilation (MV).

Multivariate analysis (Tables 3 and 4). In the first model including potential determinants of 90-d mortality at admission, Pa_O2 < 60 mm Hg on room air and BAL neutrophilia > 10% were independent predictive factors of 90-d mortality (Table 3 and Figure 2).

View larger version (15K): [in this window] [in a new window]   Figure 2.   Survival (Kaplan-Meier curves) in patients with AIDS- related PCP. (A) Survival according to PaO2 on room air at admission. (B) Survival according to BAL neutrophilia.

Ninety-day mortality was 31% in the 64 patients with Pa_O2 < 60 mm Hg and 33.3% in the 54 patients with BAL neutrophilia > 10%; in the 33 patients with both parameters, 90-d mortality was as high as 40% (Figure 3).

View larger version (14K): [in this window] [in a new window]   Figure 3.   BAL neutrophilia and PaO2 on room air at admission in all patients.

In a second model including potential determinants of 90-d mortality at admission and during hospitalization, none of the parameters evaluable at admission was significant; the only risk factors for 90-d mortality were development of pneumothorax and a need for mechanical ventilation (Table 4).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We evaluated clinical, laboratory, and BAL findings to identify poor prognosis factors in 144 patients with AIDS-related P. carinii pneumonia (PCP) managed after a consensus conference recommended adjunctive steroids to treat selected cases of PCP. Our results confirm earlier reports that 90-d mortality was influenced by four factors present at admission, namely, hypoxemia (1), high lactate dehydrogenase (1, 3- 6), low serum albumin (3, 5, 6), and BAL neutrophilia (4, 7), as well as by the occurrence during therapy of pneumothorax or a need for mechanical ventilation (13, 15, 21). Thus, the influence of these factors on 90-d mortality persisted despite the use of adjunctive steroids (9, 13, 15, 16). In contradiction with data reported by Kumar and Krieger (25), we found that parameters reflecting AIDS activity (e.g., CD4+ counts) failed to predict 90-d mortality in the univariate or multivariate analyses. However, Kumar and Krieger studied only ventilated patients; in our study, this subgroup of patients was too small to allow a meaningful analysis.

The main finding from our study was that 90-d mortality was influenced by four independent factors, of which two are available at admission, namely, Pa_O2 and BAL neutrophilia, and two emerged during hospitalization, namely, pneumothorax and a need for mechanical ventilation. Moreover, BAL neutrophil count at admission was significantly correlated with the occurrence of pneumothorax and a need for mechanical ventilation during the course of PCP.

These four independent predictors of 90-d mortality involve different pathophysiologic mechanisms. Hypoxemia and a need for mechanical ventilation merely reflect the acute respiratory failure and ventilation-perfusion alterations caused by PCP. The mortality rate in ventilated patients (86.5%) may be due to the fact that in all these patients ventilation was used because trimethoprime-sulfamethoxazole, or pentamidine, plus steroids failed to prevent further deterioration, a course known to be strongly related to a worse prognosis (26). The 12% rate of occurrence of pneumothorax in our study is within previously reported ranges (13). Pneumothorax can occur as a result of rupture of the lung parenchyma because of active subpleural pneumocystosis in patients receiving pentamidine aerosols for PCP prevention (27, 28). The volutraumatic or barotraumatic effects of mechanical ventilation can also cause pneumothorax (29). Finally, local recruitment of neutrophils resulting in a high percentage of neutrophils in BAL fluid may result from two very different mechanisms in patients with PCP, namely, a concomitant bacterial infection (30) or early-stage acute respiratory distress syndrome. A positive protected specimen brush was found in only 25 of our 144 patients and in only 14 of the 54 patients with a BAL neutrophil count greater than 10%. Fernandez and colleagues (6) reported that bacterial superinfection was a predictor of short-term mortality. We failed to replicate this finding. Neutrophils have been recovered in BAL fluid from patients with ARDS, and their percentage was correlated with severity of the syndrome (31). This suggests that severe respiratory failure in PCP may be related to a granulocyte-mediated inflammatory process rather than to the presence of P. carinii itself. Support for this hypothesis has been provided by a report that high interleukin-8 levels in BAL fluid from patients with AIDS-related PCP were associated with a poor prognosis (32). Numerous studies have reported the impact of neutrophil alveolitis in AIDS-related PCP. Mason and colleagues (8) and Smith and coworkers (9), in 60 and 19 patients, respectively, found that the proportion of PMNs in BAL fluid was related to severity but not to mortality. In a cohort of 60 patients with AIDS- related PCP, Jensen and colleagues (7) reported significantly higher mortality rates in patients with than in those without neutrophil alveolitis.

The therapeutic implications of the predictive value of hypoxemia and BAL neutrophilia are very different. Our confirmation of the independent predictive value of hypoxemia less than 60 mm Hg at admission has no influence on current treatment practices because steroids are now used in all patients with a Pa_O2 less than 70 mm Hg (16). In contrast, greater than 10% BAL neutrophilia is not included in any available treatment guidelines. In patients with a Pa_O2 greater than 70 mm Hg, for whom therapeutic recommendations do not include steroids, BAL neutrophilia greater than 10% in the absence of documented pulmonary bacterial infection may define a subset of patients likely to benefit from steroids given according to the conference consensus protocol. In our study, 10 patients met these criteria; none had a positive protected specimen brush, and 50% died. In contrast, among the 25 patients with a Pa_O2 between 50 and 70 mm Hg and a BAL neutrophilia > 10%, all of whom received steroids, only 4% (n = 1) died (Figure 3). If steroids are used in this subset of patients with Pa_O2 > 70 mm Hg and BAL neutrophilia > 10%, they should be started early, at the time of TMP-SMZ initiation, because of the inconsistent efficacy of rescue steroid therapy.

In patients with Pa_O2 < 70 mm Hg, there is no room for doubt about the need for giving steroid therapy, irrespective of the BAL neutrophil count. However, in our study, mortality was higher in patients with Pa_O2 < 50 mm Hg and BAL neutrophilia > 10% than in patients with Pa_O2 < 50 mm Hg and BAL neutrophilia < 10% (52 versus 31%); all these patients received adjunctive steroids as recommended by the consensus conference (Figure 3). This difference in mortality, not constantly found in previously published reports (8, 9), may warrant consideration of other treatment protocols in the subset of patients with Pa_O2 < 50 mm Hg and BAL neutrophilia > 10%, for instance, a long course of high-dose steroids as used by Meduri and colleagues (33) in unresolving ARDS.

These suggestions are based on findings in a cohort of patients with PCP documented by BAL fluid analysis. Moreover, this cohort included consecutive HIV-infected patients with PCP, none of whom was treated before documentation of the diagnosis by BAL.

Some institutions use either empirical treatment in suspected AIDS-related PCP in well-defined situations (34) or induced sputum examination as the first-line investigation followed by BAL if the sputum is negative. Others routinely perform BAL in suspected AIDS-related PCP because BAL has been recognized as a marker of effective management (35) and as an effective means of screening for concomitant bacterial or opportunistic infection. Whatever the strategy used, we suggest that the proportion of neutrophils in BAL fluid, which is independently correlated to severity and mortality, deserves attention.

In conclusion, PCP still occurs in patients with AIDS and remains potentially fatal despite the use of adjunctive steroids. Pa_O2 on room air at admission and BAL neutrophilia are independent predictors of 90-d mortality and are correlated with disease course parameters strongly associated with 90-d mortality in AIDS-related PCP. Our findings suggest that early therapeutic intervention, including steroid use in patients with BAL neutrophilia not caused by bacterial infection, may improve the chance for successful treatment.