INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Bacterial pneumonia is a frequent cause of morbidity in human immunodeficiency virus (HIV)-infected individuals (1- 3). The rate of bacterial pneumonia is 5.0 to 9.3 episodes/100 person-years (py) in HIV-seropositive patients compared with 1.5 to 2.1 episodes/100 (py) in the general adult population (3- 6). HIV-infected intravenous drug users (IDUs) are at an even greater risk of developing bacterial pneumonia compared with the general population, with an incidence of 6.7 to 12.3 episodes per 100 py (7). Advancing immunosuppression is a known risk factor for bacterial pneumonia in HIV-infected patients. Patients with CD4 cell counts below 50 cells/mm³ have a risk of developing bacterial pneumonia between twofold and fivefold greater than patients with CD4 cell counts between 200-500 cells/mm (3, 6, 8, 9). Prior analyses of community-acquired pulmonary infections in HIV-infected cohorts showed an increasing incidence of bacterial pneumonia through 1995, despite the increased use of antibiotics for primary prophylaxis of Mycobacterium avium complex (MAC) and Pneumocystis carinii pneumonia (PCP) in HIV-infected persons (4, 6).

The advent of highly active antiretroviral therapy (HAART) with combination regimens including protease inhibitors has led to dramatic reductions in acquired immunodeficiency syndrome (AIDS)-related morbidity and mortality as a result of improved cellular immune function and suppression of HIV replication (10, 11, 12). Several studies have demonstrated a decline in the incidence of specific AIDS-defining opportunistic infections in the HAART era, but reductions in bacterial pneumonia have not been reported (13, 14). In vitro studies suggest improved B-cell function in persons on antiretroviral regimens including a protease inhibitor (15, 16). Thus, a decrease in bacterial pneumonia might be anticipated given the augmentation in host immunity associated with HAART.

We used a cohort of HIV-infected patients with nadir CD4 cell counts below 200 cells/mm³ to determine the impact of the use of HAART on the incidence of bacterial pneumonia.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

The study population includes 1,898 persons with HIV infection and at least one CD4 cell count of < 200 cells/mm³ who were followed for at least 6 mo in the Johns Hopkins Hospital HIV Clinic between January 1993 and June 1998. For this analysis, all follow-up time after the first CD4 count < 200 cells/mm³ was included. The procedures and methods for monitoring this cohort have been previously described (17).

Participant Evaluations

A detailed baseline medical and psychosocial evaluation was performed on all participants. Data were abstracted on standardized forms by trained abstractors at baseline and every 6 mo. In addition to charts from the HIV clinic, data were obtained from electronic hospital databases, and other institutions where patients receive care. Information considered in this analysis included laboratory variables such as CD4 cell count, and complete blood count; comorbid diseases including PCP, MAC, tuberculosis; medication use including trimethoprim-sulfamethoxazole for primary PCP prophylaxis, azithromycin or clarithromycin for MAC prophylaxis, and use of antiretroviral drug therapy with dates of initiation and termination. Demographic data including age, race, sex, and HIV transmission category were also collected. HIV-1 RNA determinations were not routinely available in the cohort until July 1996 and are not included in the analysis.

Definitions

Cases were defined as individuals with positive sputum and/or blood cultures for typical respiratory pathogens and a clinical course consistent with bacterial pneumonia. This definition excluded persons with microbiologically confirmed PCP or tuberculosis. In cases that lacked microbiologic or histologic diagnosis, bacterial pneumonia was defined as the presence of at least three clinical signs and three symptoms consistent with lower respiratory tract infection. Confirmatory symptoms included productive cough, fever, shaking chills, pleuritic chest pain, and shortness of breath. Confirmatory signs included focal radiographic infiltrates, temperature > 37.8° C, heart rate > 100 beats/ min, respiratory rate > 24 breaths/min, cyanosis, consolidation on lung examination, and an appropriate response to antibiotic therapy.

We defined injection drug use as positive if the respondent reported using injection drugs at the baseline assessment.

Statistical Analysis

The data were analyzed using the statistical package SAS (version 7.0; SAS Institute, Cary, NC). The dependent variable in the analysis was an episode of bacterial pneumonia. Incidence rates were calculated by summing the number of episodes of bacterial pneumonia diagnosed in each 6-mo period and dividing by total follow-up time during the half-year and multiplying by 100. The incidence rate for 1993-1994 was compared with the incidence in 1996-1998 by Poisson regression analysis. Poisson regression analysis was also used to assess the univariate association between the rate of bacterial pneumonia and demographic and clinical variables. The covariates of greatest interest were known risk factors for developing bacterial pneumonia (prior PCP, CD4 cell count, race, and age dichotimized into > 40 or < 40 yr), prophylactic treatments (trimethoprim-sulfamethoxazole and macrolides), and antiretroviral regimens (monotherapy or dual therapy with nucleoside analogues, or combination therapy including a protease inhibitor). CD4 counts and medication usage were modeled as time- dependent covariates and required at least 30 d of use. Covariates that were significant in the univariate analysis or known to be biologically important were entered into a multivariate Poisson regression model using a generalized estimating equation methodology (PROC GENMOD). This model had a dependent variable of bacterial pneumonia, used a Poisson distribution and log-link function, and assumed an exhangeable correlation structure. The model was used to estimate the risk of bacterial pneumonia relative to type of antiretroviral therapy controlling for the covariates CD4 cell count (categorized as  50, 51 to 100,  101), trimethoprim-sulfamethoxazole prophylaxis, macrolide prophylaxis, HIV transmission category, and prior episodes of PCP. The Wald chi-square test was used to test significance.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

There were 1,898 HIV-infected patients with nadir CD4 cell counts < 200 cells/mm³ included in this analysis. The median duration of follow-up was 721 d, with total follow-up time of 27,720 person-months. The population was 74% male, 78% African American, and had a median age of 36 yr (range, 17 to 76 yr). The major risk factor for HIV transmission was injection drug use (IDU, Table 1). Overall, 11% of patients were lost to follow-up for more than 1 yr.

Age, race, sex, proportion from each HIV transmission category, and the proportion of persons in each category of CD4 cell count remained stable over the 5 yr of observation (chi-square, p > 0.05).

During the study, the pattern of antiretroviral drug use changed dramatically among patients with CD4 cell counts < 200. The proportion of patients with any antiretroviral therapy remained relatively constant from 1993 to June 1998 at 79 to 86%. However, there was a change from the routine use of nucleoside analogue monotherapy to the use of protease inhibitor-containing regimens in the first half of 1996 (Fig-ure 1). The use of protease inhibitor-containing regimens rose from 0% before 1995, to 12% in 1995, to 53% in 1996, and to > 60% in 1998. Between 1996 and 1998 there was a significant difference in use of protease inhibitor-containing regimens between homosexual males (67%) and IDUs (50%). However, the use of protease inhibitor-containing regimens was similar between individuals in categories of CD4 cell count ( 50, 51 to 100,  101).

The rate of use of trimethoprim-sulfamethoxazole prophylaxis for PCP prevention was constant during the 66 mo of observation, ranging from 85 to 92%. The median CD4 cell count at initiation of PCP prophylaxis ranged between 121- 126 cells/mm³ and did not differ by study year (rank sum test p > 0.2).

The proportion of persons receiving MAC prophylaxis with macrolides changed dramatically over the study interval among persons with CD4 cell counts < 50 cells/mm³. The proportion of eligible patients receiving macrolides was less than 20 % in 1993 but increased to 63% by the first half of 1996. The use of macrolides increased steadily to 87%of patients with a nadir CD4 cell count < 50 cells/mm³ by end of the first semester of 1998.

During the study interval, there were 352 episodes of bacterial pneumonia in 172 members of the clinic cohort. The 352 episodes of pneumonia represent 221 microbiologically confirmed cases and 131 cases defined by clinical criteria. After remaining relatively constant from 1993 to 1995, the incidence of bacterial pneumonia declined significantly in 1996 and continued to decline through 1997. Bacterial pneumonia decreased from 22.7 episodes/100 py in the first half of 1993 to 12.3 episodes in the first half of 1996 and reached a nadir of 9.1 episodes/100 py in the second half of 1997 (Figure 1). The incidence in 1996-1998 was significantly lower than in 1993-1994 (p < 0.05). An analysis of only microbiologically confirmed episodes of bacterial pneumonia showed a similar reduction in incidence of bacterial pneumonia, with 14.3 episodes/100 py in the first half of 1993 compared with 8.4 episodes/100 py in the first half of 1996 (p = 0.07) and reaching a nadir of 5.7 episodes/100 py in the second half of 1997 (p = 0.04, Figure 1).

View larger version (52K): [in this window] [in a new window]   Figure 1.   Relation between the change in the incidence of bacterial pneumonia and increased use of protease inhibitor-containing regimens.

The reduction in bacterial pneumonia was similar among men and women, persons  40 and < 40 yr old, and among persons receiving and not receiving chemoprophylaxis for opportunistic infections. Although IDUs had a slight decline in the incidence of bacterial pneumonia during the period of study, it was significantly less than that of gay males. Between 1995 and 1998 bacterial pneumonia rates declined nonsignificantly in IDU from 19.2 episodes/100 py to 14.1 episodes/100 py (p = 0.31). This contrasts with a reduction in gay males from 15 episodes/100 py in 1995 to 1.9 episodes/100 py in 1998 (p = 0.002).

By univariate analysis, race, sex, age, and prophylactic antibiotic use were not associated with the risk of bacterial pneumonia (Table 2). The risk for bacterial pneumonia was associated with the intensity of antiretroviral therapy as a time-dependent covariate. Compared with the use of no therapy, nucleoside analogue monotherapy was associated with an increased risk of bacterial pneumonia (risk ratio [RR] 1.46, 95% confidence interval [95% CI] 1.02 to 2.07]), dual nucleoside analogue therapy was associated with a reduction in the incidence (RR 0.65, 95% CI 0.38 to 1.13), and combination therapy containing a protease inhibitor was associated with an even greater reduction in risk of bacterial pneumonia (RR 0.49, 95% CI 0.28 to 0.84). Patients receiving nucleoside analogue monotherapy had a threefold greater risk of developing bacterial pneumonia than patients receiving a regimen containing a protease inhibitor, whereas patients on dual nucleoside analogue therapy had a 1.3 times greater risk of developing bacterial pneumonia relative to those on protease inhibitor therapy.

Injection drug users had double the risk of bacterial pneumonia relative to the gay males (RR 2.04, 95% CI 1.46 to 2.87). Similar to prior studies, patients with a prior episode of PCP had more than four times the risk of developing bacterial pneumonia relative to those without prior episodes of PCP (RR 4.62, 95% CI 1.86 to 11.48).

The relative risk of developing bacterial pneumonia was strongly associated with the most recent CD4 cell count. Of the 352 episodes of bacterial pneumonia, 234 episodes occurred in persons with a CD4 cell count between 0-50 cells/ mm³, 55 episodes occurred at 51 to 100 cells/mm³ and 63 episodes occurred  101 cells/mm³. Furthermore, the median CD4 cell count at diagnosis of pneumonia did not differ significantly between year of developing bacterial pneumonia. Compared with those with CD4 cell counts  101 cells/mm³, patients with CD4 cell counts of 51 to 100 cells/mm³ had an increased risk of pneumonia (RR 1.64, 95% CI 1.05 to 2.56) as did patients with CD4 cell counts of 0-50 cells/mm³ (RR 2.10, 95% CI 1.50 to 2.95).

Our multivariate Poisson model demonstrated that there were differences in risk of bacterial pneumonia based on the type of antiretroviral regimen prescribed after controlling for CD4 cell category, HIV transmission category, and prior episode of PCP (Table 3). Only those persons on combination antiretroviral regimens containing a protease inhibitor had a significant reduction in risk of bacterial pneumonia (RR 0.55, 95% CI 0.31 to 0.94). Nucleoside analogue monotherapy continued to be associated with an increased risk for bacterial pneumonia compared with no therapy (RR 1.49, 95% CI 1.0 to 2.13); dual therapy was associated with a nonsignificant decrease relative to patients not receiving any antiretroviral therapy.

The multivariate analysis also demonstrated that decreased CD4 cell count, prior episode of PCP, and IDU transmission category remain significant risk factors for bacterial pneumonia after controlling for type of antiretroviral therapy (Table 3). Use of trimethoprim-sulfamethoxazole or macrolide antibiotics was not associated with the likelihood of bacterial pneumonia in this analysis. To discern whether there was an interaction between protease inhibitor use and macrolide therapy, we controlled for macrolide therapy in our preliminary model and found no interaction between antiretroviral drug use and macrolide therapy. Macrolide therapy did not confound the relation between antiretroviral therapy and bacterial pneumonia. In addition, after adjusting for the covariates in Table 3, calendar year was not associated with risk of bacterial pneumonia.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study demonstrates a reduction in the incidence of bacterial pneumonia associated with the use of HAART in a group of patients at high risk for bacterial pneumonia. Our analysis found a reduction in bacterial pneumonia that is temporally related to the introduction of combination antiretroviral therapy containing a protease inhibitor. As in other studies, injection drug use, CD4 cell level, and prior PCP were associated with bacterial pneumonia (2, 3, 8). We were unable to detect an association between chemoprophylaxis with trimethoprim-sulfamethoxazole and macrolides and bacterial pneumonia. Furthermore, there was no interaction between chemoprophylactic regimens and antiretroviral drug use.

The use of protease inhibitor-containing antiretroviral drug regimens has been associated with marked reductions in the incidence of a variety of AIDS-related opportunistic diseases. Most reports have focused on declining rates of PCP, cytomegalovirus, and MAC, infections controlled primarily by cell-mediated immunity. Our results demonstrate reductions in bacterial pneumonia, traditionally considered under the control of humoral immunity and neutrophil function. Immunity against community-acquired bacterial pneumonia also relies on T-cell function, however, as evidenced by the strong association of advanced HIV and rate of pneumonia. Clinical studies of patients receiving protease inhibitors for greater than 12 wk have documented reconstitution of both T-cell and B-cell function (15, 16). It is possible, moreover, that reduction in HIV viral load by antiretroviral therapy may independently restore B-cell function. Furthermore, protease inhibitors could improve local immunity, which could decrease nasopharyngeal colonization with pathogenic bacteria. The significant difference in risk of bacterial pneumonia with the addition of protease inhibitors after controlling for T-cell count found in this study further supports the hypothesis that protease inhibitors may restore humoral immunity independent of T-cell number.

This observational cohort study has several limitations. The use of clinical criteria to define bacterial pneumonia may have misclassified those with other types of pneumonia as having bacterial pneumonia. Our substudy of microbiologically confirmed pneumonia, however, demonstrated a similar reduction in pneumonia in those on protease inhibitor-containing regimens. We were unable to control for the potential confounders of smoking behaviors and compliance to drug regimens, because reliable data on these variables were not available. Smoking practices and medication compliance may have been significantly different in persons not on antiretroviral therapy. However, prior studies have demonstrated only a small increase in risk of bacterial pneumonia in smokers versus nonsmokers, which would be unlikely to explain the significant protective effect of protease inhibitor-containing regimens and temporal decline in bacterial pneumonia witnessed during our study interval (2, 4). Furthermore, although we lacked surrogate measures of compliance with protease inhibitor therapy, one would anticipate that decreased compliance would have biased our association of protease inhibitor use and bacterial pneumonia toward no effect. We minimized selection bias as all persons with at least one CD4 cell count < 200 followed in clinic for  6 mo were enrolled. However, clinicians may have instituted protease inhibitor therapy in persons who were more frequent and reliable clinic attendees, which could have resulted in a comparison of two groups disparate in unmeasured risk behaviors.

In conclusion, the increased use of protease inhibitor-containing regimens has been paralleled by a marked reduction in bacterial pneumonia. The principal expense associated with HIV care is in cost of hospitalization. We have documented a significant reduction in a frequent cause of hospitalization in patients with HIV infection (18). Our findings of persistent increases in risk of bacterial pneumonia with declining CD4 cell count warrant aggressive control of viral load and timely administration of HAART to persons presenting with low CD4 cell counts. Follow-up studies will be required to determine whether the reduction in the incidence of bacterial pneumonia is maintained with longer intervals of use of protease inhibitor-containing regimens.