INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Pneumocystis carinii is a well recognized cause of pneumonia in patients who are immunosuppressed by human immunodeficiency virus (HIV) infection and other causes, but the epidemiologic features of the infection are poorly understood. Serologic surveys have suggested that most people are exposed to P. carinii early in life (1). Animal studies demonstrated that P. carinii is transmitted by the airborne route, but the infective form of the organism is unknown (2). The communicability of P. carinii is also supported by the occurrence of hospital outbreaks and cases of pneumocystosis among patients who had prolonged contact with each other (5). Investigation of these events has been hampered by the lack of a continuous in vitro culture system, sensitive markers to characterize P. carinii isolates, and reagents to distinguish present from past infection. Studies have been performed analyzing the frequency of P. carinii pneumonia during different seasons of the year and different parts of the world (5); however, there is little published information about the distribution of cases of P. carinii pneumonia in specific geographic areas.

Cincinnati is a medium-sized city located in southwestern Ohio along the Ohio River. The University of Cincinnati Medical Center is a major teaching and referral center for the adjacent Ohio, Kentucky, and Indiana areas. In 1986, the University of Cincinnati established the AIDS Treatment Center (ATC) and became a participant in the AIDS Clinical Trials Group program sponsored by the National Institutes of Health. Through these programs, the ATC provided care for more than 90% of the known HIV-infected patients in the greater Cincinnati metropolitan area for the time period when these data were gathered; the number of new registrations at the ATC was equal to approximately 90% of the number of new HIV-positive test results reported. Databases with clinical and demographic information were established and a standardized diagnostic protocol was established to evaluate patients with respiratory problems; these tools have been used to conduct a variety of studies of HIV patients with pneumocystosis (9).

We undertook the present study to analyze the place of residents of HIV patients who developed proven P. carinii pneumonia during the late 1980s. This was an era when cases of pneumocystosis were plentiful, but also saw changes in the treatment of HIV and prophylaxis of P. carinii infection. Our data suggest that patients with pneumocystosis were clustered in specific zip code areas.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Sample

From January 1, 1986 through December 31, 1989, the records of all adult HIV patients suspected of having P. carinii pneumonia were analyzed. Sources of information included University Hospital Medical Records International Classification of Diseases (ICD) code listings, pathology reports, ATC records, and the ongoing Pulmonary Medicine registry of immunosuppressed patients evaluated by the standard diagnostic protocol. This protocol involved the aggressive use of bronchoscopy with bronchoalveolar lavage (BAL) and close cooperation among clinicians and laboratories (9). Only HIV patients experiencing their first episode of histologically or cytologically proven pneumocystosis were included as P. carinii cases; people with recurrent episodes of pneumocystosis or other immunosuppressed patients who developed P. carinii pneumonia were excluded. The P. carinii patients were compared with HIV patients who underwent the same diagnostic evaluation for respiratory symptoms over the same time but in whom the organism was not found; these individuals have been termed "non-P. carinii patients."

Data Analysis

The data were entered into DataEase electronic forms and transferred to PC-SAS (Version 6.04; SAS Institute, Cary, NC) data set for analysis. Chi-square test for trend or contingency tables were used as appropriate to compare rates of occurrence. To evaluate the seasonal incidence of P. carinii pneumonia, the rate was calculated as the number of cases per number of HIV patients attending ATC for each month. The denominator included the total of number of patients enrolled in the ATC adjusted for deaths in the previous month; patients from previous episodes of P. carinii pneumonia were excluded from the denominator of the at-risk population.

The general geographic distribution of P. carinii patients was readily available from the database. To determine whether there was spatial clustering of cases, the statistical technique of Grimson and coworkers (15) was used and attention focused on zip code zones in Cincinnati. Metropolitan Cincinnati contains 45 zip code zones; the incorporated City of Cincinnati contains 31 of these zip code zones. Rates were calculated as the number of cases of P. carinii pneumonia per 100 HIV patients attending ATC for each zip code over the 4-yr period. Two zip codes were considered adjacent if they shared a common boundary of nonzero length (15). For the analysis, the number of possible adjacencies must be calculated for a specific geographic map. The statistical test then generates a table of critical values for evaluating the number of adjacencies required for concluding that clustering exists.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Analysis of the P. carinii Patients

During the 4-yr study period, 960 patients with HIV disease were treated at the ATC. Of these, 118 patients had a first episode of P. carinii pneumonia; the diagnosis was established by bronchoscopy with BAL in 113 (96%) cases, by autopsy in three cases, and by lung biopsy in two cases. The racial composition of the P. carinii patients (82% white, 18% nonwhite) reflected that of the HIV population at the ATC (79% white, 21% nonwhite) (² = 0.13, p = 1.21); of the 21 nonwhite P. carinii patients, 20 were African American and one was Asian. The gender make-up of the P. carinii patients (97% male, 3% female) and HIV population (93% male, 7% female) was also similar (Fisher exact test, ² = 0.95, p = 0.13).

Evaluation of the CD4 counts in the 98 P. carinii patients who had them performed revealed that the disease occurred when HIV infection had reached an advanced stage. Almost all (96%) of the patients had CD4 counts < 200/µl and 55% had counts < 50/µl (Table 1).

Data analysis was complicated by two major recommendations issued by the Centers for Disease Control (CDC) during the study period: the use of zidovudine (AZT) for the treatment of HIV infection in May 1987; the use of drugs for the prophylaxis of P. carinii pneumonia in June 1989 (16, 17). Because accurate information about the use of AZT or anti- P. carinii drugs was not always available, several steps were taken to evaluate the impact of the CDC recommendations on our study. First, the P. carinii patients were divided into the following cohorts: "pre-AZT," diagnosed from January 1, 1986 to May 31, 1987; "AZT," diagnosed from June 1, 1987 to May 31, 1989; "AZT plus P. carinii prophylaxis," diagnosed from June 1, 1989 to December 31, 1989 (Table 2). The cohorts were then compared for the severity of P. carinii pneumonia based on the alveolar-arterial oxygen gradient: mild (gradient of < 35 mm Hg), moderate (35 to 45 mm Hg), or severe (> 45 mm Hg) (9). Patients with severe P. carinii pneumonia constituted 52% of the cases in the pre-AZT cohort compared with only 25% of the cases in the AZT plus P. carinii prophylaxis cohort (² = 10.0, p = 0.04).

An attempt was made to determine if the trend toward milder P. carinii pneumonia over time was due to the more widespread use of AZT and P. carinii chemoprophylaxis or earlier diagnosis and treatment of the disease. We reasoned that if people knew they were infected with HIV and thus at risk for P. carinii pneumonia, they might seek medical attention when they first developed respiratory symptoms. However, the frequency of P. carinii patients (70%) who knew they were infected with HIV at the time they were diagnosed with P. carinii pneumonia was the same in each cohort. This information suggested that a large number (30%) of P. carinii patients remained outside the healthcare system despite advances in the care of HIV patients that occurred over the study period.

Because P. carinii pneumonia occurred at an advanced stage of HIV and was diagnosed by BAL, factors related to these variables might influence interpretation of the data. We sought to identify a comparable group of HIV patients who had causes other than P. carinii of their respiratory illness. This effort was helped by the uniform diagnostic protocol which is based on the use of BAL (9). The protocol discouraged the use of empiric therapy of P. carinii pneumonia; a review of 229 patients treated for pneumocystosis from 1986 to 1990 at our institution found that only two (0.9%) patients were treated without a specific diagnosis (9).

Analysis of Non-P. carinii Patients

During the 4-yr period of the study, 52 HIV patients were identified who presented with a clinical picture compatible with pneumocystosis but whose diagnostic work-up revealed causes other than P. carinii for their illness. These individuals did not have a history of pneumocystosis and did not develop the disease when followed for 1 yr after their diagnostic work-up. The non-P. carinii patients has similar demographic characteristics (96% male, 4% female, 69% white, 27% black, 4% Hispanic) but differed in CD4 counts. In contrast to persons with P. carinii pneumonia, 57% of the non-P. carinii patients had CD4 counts < 200/µl and only 31% had counts < 50/µl (Table 1) (² = 42.5, p < 0.001). Thus, the non-HIV patients had less severe HIV infection than the P. carinii patients.

Our efforts to find patients with advanced HIV infection and low CD4 counts who might serve as a comparison group for P. carinii patients were unsuccessful. The experience with Mycobacterium avium-intracellulare is illustrative: when we examined the records of M. avium-intracellulare patients, we learned that virtually all of them had also experienced P. carinii pneumonia.

Analysis of Geographic Clustering

Most of the P. carinii pneumonia patients (81%) and the non- P. carinii patients (75%) resided within the metropolitan Cincinnati area. The difference between the groups was not significant (² = 0.34, p = 0.54). Analysis of Cincinnati zip code data revealed that P. carinii cases lived in 32 of the 45 zip code zones in the Cincinnati metropolitan area. ATC patients resided in all Cincinnati zip code zones.

Twenty-four of the 45 zip code zones had 10 or more ATC patients. Analysis was performed on the 24 zip code zones with 10 or more ATC patients and on all 45 zip code zones together.

The total number of ATC patients in a zip code zone was used as the denominator to determine the rate of pneumocystosis in that zone; because the error rate was likely to be high in zones with less than 10 patients in the denominator, the P. carinii pneumonia rates were not determined for those zones. The 24 zip code zones with 10 or more ATC patients in the denominator were ranked by pneumocystosis rates, and the six highest rates are shown in Table 3. Zip code zones 1-4 were at highest risk and were clearly distinguished from other zip code zones; the difference in the rate of P. carinii pneumonia between the fourth and fifth ranked zip codes (6 cases/100 at risk) was greater than the difference between any subsequently ranked zip code zones.

Among the four zip codes zones with the highest rates of pneumocystosis, three adjacencies were found. Based on the test statistic for the 24 zones, a trend toward clustering was shown (p = 0.0723) (Table 4). When all 45 Cincinnati zip code zones were included in the analysis, clustering of acute P. carinii cases was observed (p = 0.0211) (Table 4). The proximity of the zip code zones is shown in Figure 1.

View larger version (32K): [in this window] [in a new window]   Figure 1.   Rates of P. carinii pneumonia cases/100 ATC patients for only the central Cincinnati zip code zones are illustrated. Three of the four zip code zones with the highest rates (black dots) were adjacent to each other in west-central Cincinnati.

Non-P. carinii patients resided in 22 zip code zones; 18 of these zones had 10 or more ATC patients and the rates were calculated for these individuals in the same manner as they were for the P. carinii cases. The rate of pneumocystosis was different than the rate at which the non-P. carinii patients occurred in the zip code zones (Table 3).

There was the possibility that the geographic clustering of the P. carinii cases was due to socioeconomic issues related to advanced HIV infection (e.g., loss of income or health insurance necessitating a move to lower cost housing) rather than related to P. carinii. However, our analysis showed that the mean household incomes for three of the four zip code zones with the highest rates of P. carinii pneumonia were among the higher income areas in the city (18). Median household income among the 31 zip code zones in the incorporated limits of the City of Cincinnati was $23,368 (ranging from $4,999 to $47,311). The median household income and the rank (highest to lowest) for the four zip code zones with the highest rates of P. carinii pneumonia were: zip code 45238, $32,424, rank 5; zip code 45224, $31,101, rank 6; zip code 45239, $30,071, rank 7; and zip code 45223, $20,153, rank 21.

Analysis of Seasonal Clustering

Episodes of pneumocystosis occurred throughout the year without any seasonal clustering. Similarly, no temporal clustering was found in the episodes of pneumonia that occurred in the non-P. carinii patients.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Although the epidemiology of HIV infection has been studied extensively, the epidemiologic features of some opportunistic infections afflicting patients with HIV infection have received less attention. This study has used a statistical technique, which previously was helpful in evaluating the spatial clustering of cases of sudden infant death syndrome (15), to analyze the geographic distribution of P. carinii pneumonia cases in Cincinnati. The results suggested that the patients with an initial episode of pneumocystosis were more likely to reside in three zip code zones in western Cincinnati. The same pattern was not found in non-P. carinii patients evaluated in a similar manner or in the overall ATC HIV patient population.

Our analysis relied upon the generation of a probability table for a specific number of zip code zones. When the analysis was limited to the 24 zones with 10 or more ATC patients in the denominator for the incidence rate calculation, a statistical trend toward clustering was observed; when the analysis included all 45 zones in the metropolitan area from which ATC patients were drawn, a clustering effect was shown. As would be expected, the three adjacencies were more improbable with the higher number of zip code zones, and the lower p value reflects this lower probability. Calculation using the 24 zones has the advantage of using rates based on denominators with 10 or more patients. However, using all 45 zones in the metropolitan area may be appropriate for this model; our analysis was constrained because adjacencies were only sought among the zip code zones for which a P. carinii pneumonia rate was calculated.

A companion study by Morris and coworkers used a different city (San Francisco), era (late 1990s), and statistical technique (multivariate analysis) to demonstrate a significant association between the frequency of pneumocystosis in HIV patients and zip code (19). Although the results showed a lower rate associated with the zip code (versus an increased rate in our report), these studies are really complementary because they both establish that local geographic distribution is an important feature of the epidemiology of this infection.

The Morris report (19) and our study have a number of common features. Both were retrospective in nature and involved the use of HIV patients who had causes other than P. carinii of their pulmonary illness as controls. Because both P. carinii and the non-P. carinii control patients in these studies were evaluated by a standard diagnostic protocol and because treatment was based on the histologic demonstration of the organism, it seems unlikely that factors such as patient selection or variability in practice pattern among clinicians could have materially affected the results. In both reports the P. carinii patients were more immunosuppressed than the control patients, as judged by CD4 counts; the influence of this difference on the geographic distribution of cases of pneumocystosis is unknown. Cases of pneumocystosis tended to occur in the more affluent areas of Cincinnati and San Francisco; this result was somewhat surprising, given the association of poverty with infectious diseases such as tuberculosis and with people whose resources have been exhausted in attempting to deal with a fatal illness. The rate of P. carinii pneumonia among African Americans was similar to that among other racial/ethnic groups. The prospective, multicenter Pulmonary Complications of HIV Infection Study showed a lower risk of developing pneumocystosis among blacks than among whites (20). However, another study that involved analysis of the national CDC database did not find this result, and suggested that racial differences in acquired immunodeficiency syndrome (AIDS)-related opportunities infections may be more influenced by factors such as organism exposure, diagnostic evaluation, and access to health care (21).

Our study and the report of Morris and coworkers also raise broader questions about the geographic features of P. carinii. In the pre-AIDS era, analysis of published reports and data obtained by CDC for pentamidine requests revealed that P. carinii pneumonia was an uncommon disease that occurred mainly in immunosuppressed patients (22). These patients were receiving medical care at specialized, referral hospitals that were often located far from their homes. The geographic distribution of P. carinii pneumonia at that time probably reflected the location of these health care facilities rather than the residence of the patients. More recent studies have shown that P. carinii pneumonia occurs more often in HIV-infected patients in industrialized countries than in tropical or developing countries, even though the frequency of exposure to P. carinii (based on serologic testing) is similar in both populations (25). This disparity may be due to variations in medical care and the greater risk of the development of more virulent infections such as tuberculosis in patients residing in tropical and developing countries.

One possible explanation for our results is exposure to a common environmental source of P. carinii. Studies have shown that P. carinii, in the past considered a protozoan, is taxonomically more closely related to fungi (29). Point source outbreaks of histoplasmosis, blastomycosis, and coccidiomycosis have been well documented in which disruption of the soil by activities such as excavation have caused conidia to become airborne and available for inhalation (30). Analyses of outbreaks of histoplasmosis in Indianapolis, Indiana, and of cases of blastomycosis in north central Wisconsin and Rockford, Illinois, have revealed local geographic differences in frequency of infection (31). One study from Denver, Colorado, reported that over a 3-yr period pneumocystosis occurred in 10 of 73 renal transplant recipients at Denver General Hospital, but none of 34 recipients at the Veterans Affairs Hospital located one block away; the same medical and surgical physicians cared for all patients at both institutions (34). Environmental sampling is part of epidemiologic investigations, and because cultivation of fungi from soil samples is often unsuccessful, more sensitive techniques such as the polymerase chain reaction (PCR) will probably be used in the future. Reports demonstrating the ability to detect P. carinii by PCR in air (8) suggest that environmental sampling for this organism will be performed more frequently in the future.

Another possible explanation for the apparent clustering in this report is exposure to common human sources of P. carinii. In previous outbreaks of P. carinii pneumonia it has been difficult to demonstrate person-to-person transmission (5). One report described five cases of P. carinii pneumonia that occurred over a 22-mo period among renal transplant recipients who shared the same outpatient area with HIV-infected patients; renal transplant recipients who developed P. carinii pneumonia had more encounters with HIV-infected patients than did control subjects (35). Many of our ATC patients knew each other socially or through their visits to the ATC. However, it was not possible to systematically investigate these possible exposures through information in our database. Molecular markers that distinguish (at least to some degree) among P. carinii isolates are beginning to find their way into epidemiologic studies (8, 29). A recent molecular analysis of three clusters of pneumocystosis suggested that person-to-person transmission could only have accounted for a minority of the cases within the cluster (36). Further development of molecular epidemiologic approaches to study P. carinii would be of interest.

Opportunistic infections remain a leading source of morbidity and mortality among HIV-infected individuals. Epidemiologic studies of opportunistic infections are needed to learn more about their natural history and to develop better strategies for diagnosis, treatment, and prevention. Statistical techniques such as those employed in our study and in the report by Morris and coworkers (19) are valuable because they can provide new insights into the epidemiologic features of these infections and suggest fertile areas of new investigation.