INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Fewer opportunistic infections are occurring in persons infected with human immunodeficiency virus (HIV), presumably because of the effectiveness of current antiretroviral treatment (1). Nevertheless, Pneumocystis carinii pneumonia (PCP) remains the most common acquired immune deficiency syndrome (AIDS)-defining opportunistic infection in the United States, and recurrent bacterial pneumonia (BP) is a frequent AIDS-defining condition as well (2). In addition, BP can occur throughout the course of HIV infection, even affecting those persons with CD4 cell counts above 500 cells/µl (2, 3). Despite the high frequency of pneumonia in HIV-infected persons, little is known about the effects of opportunistic respiratory infections on lung function.

Most studies of the effects of HIV-associated pneumonias on pulmonary function have focused on the acute effects of pneumonia. PCP produces a well-documented decrease in the diffusing capacity of carbon monoxide (DL_CO) (4). The diffusing capacity may improve after acute P. carinii infection, but the data are not conclusive (4, 6, 7). BP and pulmonary tuberculosis also produce a decrease in DL_CO; a similar decline can be seen in patients with advanced AIDS in the absence of acute pneumonia and in patients (with or without HIV infection) who inject drugs intravenously (6, 8). FEV₁ declines during almost any acute HIV-associated disease, and FEV₁, FVC, and TLC commonly decrease during acute P. carinii infection (6, 8). Additionally, small airways dysfunction and airway hyperresponsiveness (AHR) have also been documented in persons with HIV infection, and may be more common after pulmonary infection (14, 15).

Short follow-up periods and small numbers of patients limited these previous studies of the effects of HIV-associated pneumonia on pulmonary function. Most of the studies followed patients for less than 3 mo and examined fewer than 40 patients (4, 6, 7, 13). Thus, it is unknown whether HIV-associated pneumonias cause permanent changes in pulmonary function. The goal of the study described herein was to determine whether there were lasting changes in pulmonary function in a large cohort of HIV-infected patients with PCP or BP.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Cohort and Study Design

Subjects from the Pulmonary Complications of HIV Infection Study consisted of 1,183 HIV-seropositive persons and 170 HIV-seronegative control subjects from the same HIV risk groups. Subjects were recruited from six centers across the country. Details of this cohort and the study methods have been described previously (16). Subjects were enrolled between November 1988 and February 1990, and were followed from time of enrollment until death or March 1994 (median duration of follow-up = 3.7 yr). The HIV-infected group included homosexual/bisexual men, injection drug users, and female sexual partners of HIV-infected men. Subjects with a previous AIDS-defining illness or a preexisting pulmonary diagnosis (e.g., asthma, chronic obstructive pulmonary disease) were excluded. HIV-infected subjects were randomly assigned to have interviews, physical examinations, laboratory studies, chest radiographs, and pulmonary function tests (PFTs) at 3- or 12-mo intervals. Subjects were evaluated 1 mo after an episode of pulmonary infection, and subsequently at 6-mo intervals.

For our analysis of pulmonary function, the HIV-negative control subjects were excluded. In addition, subjects older than 60 yr of age were excluded because of the possibility of accelerated decline in pulmonary function after this age and the limited numbers of older patients available to accurately model this decline.

For our analysis of the effects of opportunistic respiratory infections on pulmonary function, subjects with a first episode of PCP or BP were included. This subject group with PCP or BP was compared with the overall cohort of HIV-infected patients in order to determine the effects of pneumonia on lung function. If a subject had a second episode of pneumonia, subsequent pulmonary function tests were excluded from analysis. These exclusions were necessary because of the very limited data available for modeling repeated pulmonary infections; however, limited exploratory analysis was performed on a subset of patients (n = 10) who had episodes of both PCP and BP.

The diagnosis of pneumonia was based on previously described criteria (16). Patients were included in the PCP group if they had evidence of P. carinii on microscopic examination of lung-derived specimens or if they had a compatible clinical and chest radiographic presentation with an appropriate response to anti-Pneumocystis therapy. BP was similarly defined by the presence of a compatible clinical and radiographic presentation, with the isolation of a likely pathogen from blood and/or an adequate sputum/bronchoalveolar lavage fluid (BALF) sample, or with an appropriate response to antibacterial treatment. Patients with a productive cough but no radiographic findings were assumed to have acute bronchitis rather than BP.

Measurements of FEV₁, FVC, TLC, and DL_CO were made at each center according to American Thoracic Society guidelines (17). Quality control of the testing was assessed by site visits and by regular testing of one person with stable lung function at each center. DL_CO was corrected for hemoglobin concentration (18).

Modeling of Pulmonary Function

Multivariate models of FEV₁, FVC, TLC, and DL_CO were constructed with the mixed procedure of the Statistical Analysis System (SAS) (19). By including random slope and intercept terms, this procedure accounted for the lack of independence of repeated measures in a single individual. Only data from baseline and routine follow-up visits were used for analysis. Data that were collected within the 30 d preceding or the 90 d following a diagnosis of PCP or BP were excluded so that acute effects of these diseases would not contaminate measurements. The model controlled for a large number of predictor variables known or thought to influence pulmonary function. These variables included age, age squared (to allow for accelerating decline in pulmonary function), height, height squared, sex, race, cigarette smoking (ever versus never), HIV risk group, injection drug use, CD4 cell count, time in the study (a surrogate for time HIV-infected), study site (versus Site 1), and number of previous visits (to account for practice effects). For simplicity, all variables, even those that did not reach statistical significance at p < 0.05, were included in the models. Exclusion of such variables from the models did not alter the results presented. The estimates for time in the study are interpreted as the degree to which HIV infection accelerated a decline in pulmonary function with time beyond the decline seen with normal aging. Age varied widely among subjects in the study, and mainly reflected age before HIV infection, whereas time in the study varied more widely within subjects, and included only HIV-infected time. Thus, the effect of time in the study, once controlled for age, is to provide an estimate of how much faster the decline in pulmonary function was while patients were in the study (and HIV-infected) than it was over most of the patients' lives (mainly before HIV infection). Use of pack-years of smoking as a predictor, instead of ever- versus never-smoked, had very little impact on the estimate, but resulted in fewer available observations because of missing data; therefore, ever-smoked was used as the predictor variable. For CD4 cell count, the average value at the previous, current, and subsequent visit was used when available, in order to reduce variability and missing values. A practice effect was modeled as a linear improvement until the seventh trial, after which exploratory analysis showed little improvement.

To assess the impact of infection on pulmonary function, the predictor variables of infection with PCP or BP were included in the model. The fitted effects of these variables estimated the permanent change in pulmonary function produced by each infection. Time since PCP was also included, to estimate any acceleration in rate of decline in pulmonary function over time following PCP as compared with the rate before PCP. All visits were analyzed until the subject's death, end of study, or subsequent pulmonary infection. Interactions of PCP with the variables of BP, sex, race, HIV risk group, and smoking were examined for the outcome of FEV₁.

Residuals from the models were not normally distributed, with more large residuals than would be expected under normality. Because the lack of normality would make standard calculation of p values and confidence intervals (CIs) invalid, the bias-corrected, accelerated bootstrap method, with individual histories being the resampling units, was used to obtain valid CIs for estimated effects (20). Values of p were obtained as one minus the highest confidence level that would exclude zero, and are reported as p < 0.001 if all 2,000 bootstrap estimates were positive or all were negative.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Demographic Characteristics

A total of 1,149 HIV-infected persons were included in the overall cohort used to model pulmonary function (Table 1). Among these persons, 141 subjects developed either PCP (n = 71) or BP (n = 70) during the study period. Most of the subjects in the overall cohort and in the subgroup that developed PCP or BP were male, white, and homosexual/bisexual, and had a history of current or prior cigarette use.

Modeling of Pulmonary Function in the HIV-Infected Cohort

The multivariate models of FEV₁, FVC, TLC, and DL_CO for the overall cohort showed that age, height, sex, and race were among the significant predictors of pulmonary function (Table 2). Injection drug use was associated with lower values of FEV₁ (149 ml, p = 0.009) and DL_CO (2.0 ml/min/mm Hg, p = 0.001), whereas a history of cigarette use was associated with a lower DL_CO (2.0 ml/min/mm Hg, p < 0.001) and a trend toward a lower FEV₁ (67 ml, p = 0.058). Increasing CD4 cell count did not predict changes in FEV₁, FVC, or TLC, but did predict a small change in DL_CO (0.2 ml/min/mm Hg, p < 0.001). Time in the study (used as a surrogate for time with HIV infection) showed a significantly faster decline in FEV₁ and DL_CO (FEV₁: 27.4 ml/yr, p = 0.015; DL_CO: 0.4 ml/min/mm Hg/yr, p = 0.004) than would be predicted from aging alone. Both study site (data not shown) and practice effect (previous PFT) influenced pulmonary function; controls for these effects were applied in subsequent analyses.

Impact of PCP or BP on Pulmonary Function

The multivariate model showed that pulmonary infection had a negative impact on lung function (Figure 1). FEV₁ declined significantly after PCP (264 ml, p = 0.001) and after BP (109 ml, p = 0.005). This change is equivalent in magnitude to the estimated effect of aging from 35 to 43 yr of age. Similar results were found when examining FVC. PCP predicted a 254-ml decline in FVC (p = 0.004), and BP predicted a 117-ml decline (p = 0.007). Furthermore, the ratio of FEV₁ to FVC decreased significantly after PCP (2.1% decline, p = 0.003) and after BP (1.3% decline, p = 0.031). These declines were present at all times after an episode of PCP or BP. They did not appear to resolve with increasing time since PCP; in fact, declines in FEV₁ and FVC after PCP appeared to continue to worsen over time as compared with what would have been expected without a history of PCP, although this acceleration did not reach statistical significance (39.4 ml/yr for FEV₁, p = 0.49; 71.4 ml/yr for FVC, p = 0.33).

View larger version (9K): [in this window] [in a new window]   Figure 1.   Changes in pulmonary function after infection with PCP or BP. *p < 0.05, p < 0.01, and p  0.001.

There were no statistically significant lasting changes in TLC after PCP (195 ml, 95% CI: 466 to +56 ml, p = 0.13) or BP (83 ml, 95% CI: 226 to +18 ml, p = 0.12). Although the model showed a significant decline in DL_CO after pulmonary infection, the absolute changes were small (PCP: 1.6 ml/min/mm Hg, p = 0.047; BP: 1.4 ml/min/mm Hg, p < 0.001). As with its effect on FEV₁ and FVC over time, PCP also seemed to accelerate the rate of decline in DL_CO per year (2.1 ml/min/mm Hg/yr, p = 0.006).

Interaction between PCP and BP

Ten patients who developed both PCP and BP were analyzed in order to examine possible interactions between the two infections. A history of both PCP and BP in the same patient was associated with an FEV₁ that was 177 ml lower than would be expected from the sum of the two infections' separate effects, but this interaction did not reach statistical significance (data not shown). Interactions of PCP with other predictors were also examined for FEV₁. The predictors of sex, race, smoking, and HIV risk group did not show any statistically significant interaction with the history of PCP (data not shown).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Our study showed that pulmonary infection in HIV-infected individuals is associated with a permanent decline in lung function. We found that both PCP and BP were associated with permanent decreases in FEV₁, FVC, and FEV₁/FVC, and with small decreases in DL_CO. Also, these effects did not appear to resolve over time; indeed, there was some suggestion that the decline in lung function observed after PCP was greater than that expected from aging or from duration of HIV infection. Although other studies have suggested that lung function may not return to baseline after infection (6, 8, 10, 13), the present study is the first in which repeated measures were made in individuals over long periods of time after infection. The study is also unique in its use of a model based on a large number of HIV-infected patients for examining longitudinal changes in pulmonary function.

PCP predicted the greatest changes in lung function. As might be expected, PCP produced a lasting decrease in DL_CO, although this decline was small and has uncertain clinical significance. More surprisingly, PCP was associated with airflow obstruction. Both FEV₁ and FVC were impaired after infection, and the ratio of FEV₁ to FVC declined significantly. These results suggest that in addition to the effects expected from the alveolar involvement seen in PCP, infection with Pneumocystis may cause airways obstruction. Our findings are consistent with the finding in previous work that PCP leads to small airways dysfunction and the suggestion that AHR may be related to opportunistic infections (14, 15). The mechanism of this obstruction is a subject for speculation, but it may involve a reduction in lung elastic recoil, since PCP produces an effect on the airways similar to that of emphysema. This hypothesis is consistent with the finding in a recent study that patients with HIV infection have evidence of focal air trapping and emphysema on high-resolution computed tomography (HRCT) of the chest (21). The observed declines in DL_CO are also consistent with such a hypothesis.

BP also produced lasting changes in lung function. BP was associated with lower values of FEV₁ and FVC, and with a lower FEV₁/FVC. Since pneumonia may be accompanied by bronchitis, the findings of airways obstruction are not surprising. Although transient bronchial reactivity after viral infections is a common clinical problem, similar permanent changes have not been documented in the general population after bacterial or viral infections (22, 23). Studies of normal hosts with bacterial or viral pneumonias have been unable to document decrements in FEV₁ or FVC as early as 2 wk after resolution of symptoms, although declines in DL_CO may persist (22, 23). Because the presence of HIV infection was associated with worsening lung function in our study and others (10, 12) and because the virus itself can be recovered from BALF (24- 26), it is possible that HIV infection interacts with lung infections to produce these declines. In fact, acute tuberculosis and other infections have been shown to increase the replication of HIV in the lungs (27, 28). It is possible that a similar mechanism leads to permanent changes in pulmonary function after BP.

Previous studies have shown increased bronchial reactivity and small airways dysfunction in persons with HIV (29). Although anatomic evidence of obstruction has been shown in HIV-infected patients, these changes have not been correlated with a history of opportunistic infection (21). Our data build on these observations by demonstrating permanent obstructive changes in pulmonary function after PCP and BP.

Our study also builds on past studies of DL_CO in HIV infection. Previous authors have reported decreases in DL_CO with advanced AIDS (4, 5, 9, 10, 12, 13), and our results also show that DL_CO declines with a decreasing CD4 cell count. A decrease in DL_CO in symptomatic patients predicts the presence of PCP (30, 31), and our data suggest that some of this decrease may be permanent. A recent study (32) also correlated decreases in DL_CO with evidence of emphysema on HRCT. Unlike our findings, these changes did not correlate with a history of previous lung disease.

One limitation of our study was the inability to examine the effects of multiple pulmonary infections. If each bout of infection predicts the same decline in FEV₁ found in our study, then significant airways obstruction might develop in patients with repeated pulmonary infections. Knowledge of the respiratory effects of pulmonary infections is important in the care of HIV-infected patients. Anecdotally reported increases in respiratory complaints after episodes of pneumonia may have their source in these alterations in pulmonary function. Study site substantially influenced pulmonary function values. However, because the effects of PCP and BP are based on changes within subjects over time, and subjects also did not change their study sites, systematic differences between sites would not provide spurious results.

In summary, we have shown that PCP or BP in HIV-infected individuals leads to lasting decreases in lung function, most notably to airflow obstruction. The changes are greatest in patients who have had PCP. The clinical implications of these changes are unknown, but repeated infections may lead to significant pulmonary symptoms in this group of patients.