INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Infection with human immunodeficiency virus (HIV) results in a generalized immunosuppression that is predominantly due to the cytopathic effects of HIV on helper/inducer T lymphocytes (1). Since CD4⁺ T cells are responsible for inducing and regulating a broad spectrum of immunologic responses, including immunoglobulin synthesis, lymphokine production, and the regulation of cytotoxic and suppressor cell activity, the impairment of this cell population renders patients with AIDS susceptible to opportunistic infections and malignancies that involve various tissues, including the respiratory tract (2).

Respiratory failure is a common finding in patients with AIDS-related interstitial lung disease. In view of the frequent, fatal pulmonary complications, several groups have recently attempted to identify pulmonary abnormalities that predict the clinical outcome of pulmonary involvement during HIV infection. The presence of gas exchange abnormalities and the accumulation of neutrophils in the bronchoalveolar lavage (BAL) have been suggested to strongly correlate with imminent mortality in HIV-seropositive patients (3). Nevertheless, these studies were based on the analysis of data obtained from selected groups of infected individuals. Because the number of patients was small, it was impossible to evaluate how the changes in the individual BAL parameters alter the patients' prognosis in HIV infection. These goals can only be achieved by a larger multicentric co-operative study.

In this study, we obtained a detailed analysis of BAL cell populations from 161 HIV-seropositive patients who underwent bronchoscopy while being evaluated for the presence of pulmonary opportunistic infections. Morphologic and phenotypic evaluation of BAL cell populations was therefore related to clinical features and immunologic data in peripheral blood (PB) and retrospectively analyzed to determine the most powerful prognostic factors.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients and Criteria for Admission

This multicentric study was designed to determine retrospectively whether immunologic analysis of BAL has prognostic value for patients with HIV infection. To avoid variability in results, clinical and laboratory data have been collected from five institutions located in four countries (Italy, England, France, and Portugal); all institutions followed technical recommendations and guidelines for the standardization of BAL procedures previously reported (7). The participating groups were asked to provide cytoimmunologic data of BAL analysis from HIV-seropositive patients who underwent investigation for respiratory symptoms and/or abnormal chest radiograph findings from January 1991 through December 1994. Our analytic approach was restricted to 161 HIV-seropositive selected patients in whom a complete morphologic and immunologic analysis of BAL cellular components was available, including: cell recovery; differential count of macrophages, lymphocytes, neutrophils, and eosinophils; and flow cytometry analysis of CD3, CD4, and CD8 BAL T-cell populations. To minimize the risk factor for the development of adverse effects with BAL, patients who underwent morphologic and immunologic analysis of BAL should have had minimal prerequisite lung function status and arterial blood gases (Pa_O2 > 60 mm Hg, FEV₁ > 60% predicted, Sa_O2 > 90%), platelet count > 20,000/ml, and prothrombin time > 50%. A volume of 100 to 200 ml and a sample of 50% of the instilled volume with a minimum of 50 ml were considered acceptable. Subjects undergoing BAL analysis to obtain a specific diagnosis of opportunistic infections but in whom the underlying deterioration of pulmonary function limited the size of instilled fluid were excluded from this retrospective study because they lacked an appropriate quantitative and qualitative analysis of BAL cell populations. At the time of the initial BAL evaluation, each patient underwent history and physical examination, routine blood studies, and complete immunophenotyping in the PB; HIV seropositivity was confirmed by both enzyme-linked immuno-assay and Western blot analysis. BAL samples were submitted for cytologic studies, stained for Pneumocystis carinii, and stained and cultured for mycobacteria and fungi. Culture for cytomegalovirus (CMV), IgM assays, or immunofluorescence for early CMV antigen were not routinely performed. The diagnosis of CMV pneumonia was performed upon demonstration of the characteristic cytologic abnormalities (intranuclear or cytoplasmic inclusion bodies) in pulmonary cells. Quantitative BAL cultures for the diagnosis of bacterial pneumonia were not recommended as a routine practice, being warranted only in patients with compatible clinical illness.

A detailed questionnaire with special attention to BAL data was completed by the different centers. Information required included: (1) date of BAL execution, (2) results of immunologic and microbiologic analysis of the BAL, (3) the following demographic variables present at the time of the initial BAL: age, gender, HIV risk behavior, and stage of HIV infection (including absolute number of PB CD4⁺ T cells), (4) results of 6- to 8-mo interval follow-up; in particular, institutions were asked to censor patients who were still alive at their last recorded date of contact from the medical files. The follow-up period ended on June 30, 1995. A total of 142 of the 161 patients underwent zidovudine therapy for at least 6 mo during the follow-up. All patients were clinically evaluated at 6-mo intervals.

Immunologic Studies

In all centers, BAL was performed following local anesthesia as previously reported (7). Briefly, a total of 150 to 200 ml of saline solution was injected via fiberoptic bronchoscopy in 25-ml aliquots, with immediate vacuum aspiration after each aliquot. The fluid was filtered through gauze. Cells recovered from the BAL were washed three times with phosphate-buffered saline, resuspended, and then counted. Differential counts of macrophages, lymphocytes, neutrophils, and eosinophils (made from total counts of 300 cells) were determined by morphology in cytocentrifuged smears. BAL and PB lymphocytes were characterized by CD3, CD4, and CD8 monoclonal antibodies following standard flow cytometry methods (7).

To select laboratory variables for the Cox proportional hazards model, we calculated cut-off limits of BAL data obtained from 35 healthy control subjects; the values used were twice the SD from the mean of the corresponding control values. Concerning PB data, we evaluated the absolute numbers of CD4⁺ T cells (with cutoff limits as follows: > 250 cells/mm³, between 250 and 50 cells/mm³, and < 50 cells/mm³) and CD8⁺ T cells (cutoff limits as follows: > 800 cells/mm³, between 800 and 400 cells/mm³, and < 400 cells/mm³). The set of BAL variables and the corresponding limits analyzed in the study were the presence of alveolitis (total cell recovery > 250 × 10³ cells/ml) and the type of alveolitis, such as macrophage (alveolar macrophage [AMs] > 200 × 10³ cells/ml), lymphocytic (BAL lymphocytes > 30 × 10³ cells/ml), and neutrophil (BAL neutrophils > 15 × 10³ cells/ml) alveolitis. The increase in the percentage of BAL lymphocytes and neutrophils (> 20% and > 10%, respectively) and the absolute number of BAL CD3⁺ (> 24 × 10³ cells/ml), CD4⁺ (< 10 × 10³ cells/ml), and CD8⁺ (> 12 × 10³ cells/ml) lymphocytes were also taken into account.

Statistical Analysis

Univariate statistical analysis was performed using the life-table estimator and log-rank test (8). To analyze prognostic factors, we adopted the standard version of the Cox proportional hazards model (9). The variables that were significant in the univariate analysis (p < 0.05) were used to develop a multiple regression model. The estimated regression coefficients () express the link between covariates and survival. A positive coefficient is associated with an increase in the death rate, whereas a negative coefficient has the opposite meaning. The regression coefficients are given together with standard measures of significance of the covariates (prognostic factors).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Clinical and Immunologic Findings

The clinical and BAL findings of our study population are summarized in Table 1. On the basis of BAL analysis, a pulmonary infection was demonstrated in 87 of the 161 patients (54.0%) and eight patients showed more than one pathogen in the BAL specimens (in particular, P. carinii and pyogenic bacteria in two patients; P. carinii and CMV in two patients; CMV and Mycobacterium tuberculosis in one patient; P. carinii and herpes simplex virus [HSV] in one patient; P. carinii, CMV, and pyogenic bacteria in one patient; and P. carinii, CMV, HSV, and Candida albicans in one patient). It should be noted that some of the organisms we identified in BAL specimens might not have been true pulmonary pathogens (including C. albicans, CMV, and HSV).

As shown in Table 1, the T-cell subset analysis in the PB showed that the majority of patients (94 cases) had a CD4 count of < 50 cells/mm³. BAL analysis revealed the presence of an alveolitis in 61 patients (37.8%). This alveolitis was sustained by AMs in 49 patients (30.4%), by lymphocytes in 64 patients (39.7%), and by neutrophils in 42 patients (26.1%).

Univariate Analysis

The retrospective analysis was carried out during a 48-mo follow-up. During this 4-yr period, 155 of the 161 subjects died (mean follow-up time: 14.8 mo; range: 1 to 48 mo; overall mortality: 96.3%). Concerning the follow-up time, 32 patients were observed for less than 6 mo (19.8%), 36 patients for at least 6 mo (22.3%), 29 patients for 12 mo (18.0%), 18 patients for 18 mo (11.1%), 15 patients for 24 mo (9.3%), nine patients for 30 mo (5.6%), 16 patients for 36 mo (9.9%), and six patients for 48 mo (3.7%).

Demographic factors present at the time of the first evaluation that in univariate analysis were predictive of an increased risk of death within 48 mo are shown in Figure 1. They were as follows: age of > 30 yr, HIV disease status at the time of the BAL analysis (patients presenting with a diagnosis of AIDS being at higher risk), and risk groups for HIV infection (polytransfused and homosexual categories being at higher risk). While neither the presence of P. carinii (Figure 1) nor of other opportunistic infections in the BAL were predictors of mortality in univariate analysis, the concomitant detection of more than one opportunist pathogen in the same patient was associated with a decreased survival (Figure 2). Nevertheless, it should be noted that 40.9% of patients had a Pneumocystis carinii pneumonia (PCP) (Table 1); the small number of patients with other types of opportunistic infections could have prevented a proper statistical evaluation of the relationship between the presence of the different pathogens in BAL samples and survival.

View larger version (23K): [in this window] [in a new window]   Figure 1.   Survival curves at 48 mo. Lack of statistically significant differences in the univariate analysis, according to detection of P. carinii in the BAL.

View larger version (33K): [in this window] [in a new window]   Figure 2.   Survival curves at 48 mo. Statistically significant differences in the univariate analysis, according to different demographic parameters: age, CDC stage at the time of the BAL execution, risk category for HIV infection, and detection of one or more opportunist pathogens in the BAL.

Laboratory features associated with poor prognosis were the demonstration of a BAL neutrophilia and/or a low number of CD4 T BAL cells (Figures 3 and 4). In contrast, the presence of an alveolitis and an increase in the absolute numbers of AMs and CD3⁺ T cells was associated with a slight but significant improvement in survival (Figure 3). It can be seen in Figure 4 that while patients with a low number of BAL CD4 T cells have poorer prognosis, when the study groups were divided according to the absolute number of circulating CD4 T cells the CD4 T-cell count in the PB had no significant influence on predicting survival.

View larger version (35K): [in this window] [in a new window]   Figure 3.   Survival curves at 48 mo. Statistically significant differences in the univariate analysis, according to different immunologic parameters: presence of an alveolitis, BAL neutrophilia, and an increase in the absolute numbers of alveolar macrophages and CD3+ T cells.

View larger version (20K): [in this window] [in a new window]   Figure 4.   Survival curves at 48 mo. Statistically significant differences in the univariate analysis, according to the numbers of peripheral blood and BAL CD4+ T cells. HIV-infected patients with pulmonary involvement and low BAL CD4 cell counts have a poorer prognosis. In contrast, the CD4+ T cell count in the peripheral blood has no effect on predicting survival.

The following other data were not associated with an increased mortality: gender (p = 0.307), type of opportunistic pathogen isolated from the BAL (p = 0.409), absolute number of PB CD8 T cells (p = 0.903), absolute number of PB CD4 T cells (p = 0.145), absolute number of BAL lymphocytes (p = 0.403) and polymorphonuclear cells (p = 0.374), percentage of BAL lymphocytes (p = 0.435), and absolute number of BAL CD8 T cells (p = 0.060).

Multivariate Analysis

Results of the multivariate Cox's analysis are shown in Table 2. Significant independent predictors of short survival were as follows: older age, risk group for HIV infection, cell recovery (alveolitis), and low absolute number of BAL CD4 T cells. While the HIV disease status at the time of BAL, the presence of more than one pathogen in the sample, the demonstration of neutrophilia, and an increase in the absolute numbers of AMs and CD3 T cells were all significant in the univariate analysis, these parameters were no longer significant when analyzed by multivariate technique, suggesting the influence of other variables.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In this multicenter study on the prognostic value of pulmonary analysis in HIV-seropositive patients, we tested for prognostic factors that predict mortality within 48 mo from the first BAL evaluation. In the univariate analysis, an age of > 30 yr, HIV disease status, risk category for HIV infection, number of pathogens isolated, BAL neutrophilia, and low number of pulmonary CD4 T cells were predictors of more rapid mortality. The presence of an alveolitis or of an increased number of AMs and CD3 BAL T cells suggested a better prognosis for the patient. In the multivariate analysis, significant independent predictors were an age of > 30 yr, type of HIV risk factor, the degree of alveolitis, and the low number of pulmonary CD4 T cells.

We have demonstrated that an older age and HIV transmission category may significantly influence the risk of death in patients with HIV infection. These demographic data are reflective of the worsening clinical course that can be observed in older and polytransfused HIV-infected subjects and are in agreement with previous reports of the relationship between the risk of respiratory illnesses in different groups of HIV-infected patients with pulmonary complications (10).

It is surprising that the type of infection identified did not register as a prognostically important modifying factor in our series. Two possible explanations can be taken into account. Since an optimal host response against pulmonary opportunists depends on the local accumulation of effector immunocompetent cells, it is likely that alterations at any level of the cell network leading to the development of alveolitis represent a negative prognostic factor. This first hypothesis is supported by the evidence that: (1) the presence of an alveolitis rather than the type of infection correlates with a better prognosis; (2) while the presence of more than one pathogen in the sample was a predictor of survival in univariate analysis, this parameter was no longer significant when analyzed by multivariate techniques, suggesting the influence of other variables (for instance, the degree of the local immunosuppression or other immunologic variables).

Since most patients had a PCP, a second hypothesis is plausible, i.e., the small number of patients with other types of infections could have prevented a proper statistical evaluation of the relationship between the presence of the different pathogens in BAL samples and survival. A larger, co-operative study should contribute to definitively determining whether infection by itself is a risk factor for mortality. In this regard, retrospective analysis of survival in HIV-infected patients with positive CMV culture and cytologic study from BAL has demonstrated that CMV retrieved by BAL is associated with significantly higher 3- and 6-mo mortalities (16). Nonetheless, it should be noted that it is still unclear what constitutes CMV pneumonitis since BAL cells of patients with AIDS may be positive for the CMV in up to half of the cases and it is controversial as to whether CMV is a respiratory pathogen or a colonizer in these subjects (17). Further prospective CMV cultures and cytologic studies from BAL may be of benefit in resolving the role of CMV in local immunosuppression.

As reported above, previous studies have demonstrated that the morphologic and phenotypic evaluation of BAL cell populations has some predictive value in HIV-infected individuals with respiratory complications (4). The present study confirms and extends these data, thus helping to define the natural history of HIV-associated pulmonary complications. In fact, BAL factors found to be independently associated with survival represent only a fraction of BAL parameters that had previously been associated with survival in patients with AIDS (4). This distinction is best understood by examining the relationship between BAL cellularity and survival. Although an increased BAL neutrophilia was previously shown to be associated with an increased risk of death (4, 5), in the present study it was not an independent predictor. Similarly, in our univariate models, increased numbers of AMs and BAL CD3 T cells were associated with improved survival but were no longer significant in the multivariate model. Thus, it is possible that these variables have no direct association with survival, but rather they are confounders. Alternatively, the sample size could be too small to reach statistical significance in a model with more explanatory variables.

In the multivariate model, only the presence of a low number of BAL CD4 T cells was an adverse prognostic factor, while an increased BAL cellularity was linked to improved survival. These findings should be interpreted to show that the degree of alveolitis and the concentration of BAL CD4 T cells (i.e., two parameters closely interrelated to the effectiveness of host local immune mechanisms) have the greatest impact among BAL markers that have been analyzed. It is amazing that while patients with a low number of BAL CD4 T cells have a particularly poor prognosis, circulating CD4 T-cell counts were not a sensitive measure of risk of mortality in our patients. We suggest that circulating CD4 T cells do not reflect the interstitial lung process nor, in particular, do they reflect the bulk of the pulmonary CD4 T-cell pool. In the majority of our patients (58.4%), the CD4 T-cell count was < 50 cells/ mm³. The fact that the degree of immunodepression was quite similar in the PB of most subjects (only 12 cases had a PB CD4 count > 250 cells/mm³) explains the negligible effect on predicting survival of the PB CD4⁺ count among patients with matching values of circulating CD4⁺ T cells. However, lung involvement in HIV-infected patients is known to be associated with the intraalveolar accumulation of T cells that are phenotypically different from those in the PB, and sometimes numerous T cells are still seen locally while the patient is severely lymphopenic (18). That the number of BAL T cells showed great variability indicates that the immune function in the lung is regulated differently than in the blood and explains why BAL CD4 cell count evaluation is a significant predictor of survival in an apparently homogenous population of strongly immunosuppressed patients. Our observations may partially explain why previous studies have shown that the rate of PB CD4 T-cell decline can predict survival only before CD4 T-cell counts reach 100 cells/mm³ (21). Likely, in patients with a low number of circulating CD4 T cells and better prognosis (Figure 4), a normal number of tissue CD4 T cells may contribute to maintaining a relatively effective T cell-mediated immune response against opportunists.

In conclusion, this study supports the hypothesis that the evaluation of the degree of alveolitis and BAL CD4 T cells might provide additional prognostic information that could be important in monitoring patients with HIV infection. However, sample size limitation may have influenced BAL findings of this retrospective analysis. For instance, the guidelines for BAL execution may have caused potential bias in patient populations since a complete cytologic and immunologic analysis of BAL was not performed in individuals with severe deterioration of their pulmonary function. Nonetheless, although our observations cannot be extended to the general population of patients with HIV, they are certainly applicable to patients with milder respiratory symptoms or with abnormal chest radiograph findings who have a clear indication for BAL (i.e., the need to establish a diagnosis) and in whom BAL has an acceptable risk/benefit ratio. Further multicentric prospective studies should be planned to evaluate whether the rate of decline of BAL CD4 T cells and the presence of an alveolitis as well as the decline of PB CD4 T-cell counts can predict prognosis during the course of the disease.