The Pulmonary Complications of HIV Infection Study (PCHIS) [17] prospectively followed HIV-seropositive patients who had demographic variables similar to those of patients with AIDS in the United States. Participants with asymptomatic or symptomatic HIV infection were recruited from sites in the eastern, midwestern, and western United States. A previous report on this cohort [18] identified determinants of delayed-type hypersensitivity response and risk factors for tuberculin reactivity. We examined the incidence of tuberculosis among patients enrolled in the PCHIS for a median observation period of approximately 4.5 years.

Methods

Top Methods Results Discussion Author & Article Info References

Patients and Study Design

The PCHIS was a multicenter, prospective study of the frequency and spectrum of pulmonary disorders in persons infected with HIV. From November 1988 through February 1990, 1171 HIV-seropositive persons and 182 HIV-seronegative persons were enrolled at centers in six U.S. cities: New York; Newark, New Jersey; Detroit; Chicago; San Francisco; and Los Angeles. Participants were followed through 31 March 1994. We report on 1130 HIV-infected persons from the PCHIS who were followed past baseline.

Participants were recruited to represent a range of severity of HIV disease. Approximately half of the participants at each center had CD4 lymphocyte counts of 400 cells/mm³ or more and no HIV-related symptoms, and half had CD4 lymphocyte counts of less than 400 cells/mm³ or symptomatic HIV infection. Both groups included persons from one of three HIV-transmission categories: homosexual men, male and female injection drug users, and women who had acquired HIV through heterosexual contact. Exclusion criteria were AIDS, as defined by the Centers for Disease Control and Prevention (CDC) [19]; acute pulmonary processes; use of immunosuppressive therapy in the past 6 months; and treatment of tuberculosis in the past 12 months. The study was approved by the institutional review board at each site, and participants gave informed consent.

At baseline and at regular intervals, clinical monitoring (including T-lymphocyte subset analysis and chest roentgenography) was done, and participants were acutely evaluated if new pulmonary symptoms occurred. Centers used the same predetermined diagnostic algorithms that were initiated if specified criteria were met. Complete details of the study design have been described elsewhere [17].

Delayed hypersensitivity was tested at baseline and then annually using purified protein derivative (PPD) tuberculin at a strength of 5 TU per 0.1-mL dose and mumps antigen (Connaught Laboratories, Inc., Swiftwater, Pennsylvania). Tests were administered by intradermal injection of 0.1 mL of antigen by the Mantoux method and read by experienced nurses 48 to 72 hours later in most cases (the interval exceeded 4 days in 18 persons). A positive response to PPD was defined as an induration at least 5 mm in diameter. Any person who was not positive when first tested but who became positive on any subsequent test was considered to be a tuberculin converter. A positive response to mumps was defined as any degree of induration (>0 mm). The criteria for anergy was nonreactivity (0 mm) to both PPD and mumps antigen.

Pulmonary tuberculosis was defined by the isolation of M. tuberculosis from a respiratory tract specimen or by improvement on chest radiography in response to specific multidrug antituberculous therapy. Patients who were diagnosed with extrapulmonary tuberculosis had clinically compatible disease and response to specific therapy, with or without the isolation of the organism from a site outside the lung. Patients who were considered to have both pulmonary and extrapulmonary tuberculosis met each set of criteria. These requirements are consistent with those of the CDC for reporting cases of tuberculosis [20], except that we did not require patients to be PPD positive if they had disease that was not mycobacteriologically confirmed.

Statistical Analysis

Tuberculosis rates were calculated as the number of cases divided by the number of years that patients were followed multiplied by 100. Except as noted below, the length of time that patients were followed was calculated for each person starting from enrollment and continuing until one of the following occurred: diagnosis of tuberculosis, death from any cause, the last study visit, or 31 March 1994. Statistical significance for comparison of rates was determined by tests for person-time data done on the basis of binomial distribution [21]. P values were determined by an exact test when sample sizes were insufficient and by an asymptotic test when sizes were sufficient. All tests were two-sided, and a P value of 0.05 was considered significant. Exact 95% CIs were calculated for rates by assuming the numerator to be a Poisson variable [22] and for rate ratios using a modified binomial model [21].

Adjusted rate ratios for comparisons among groups defined by demographic variables were calculated by using a Mantel-Haenszel type estimator for incidence-rate data with approximate 95% confidence limits based on the tests [21]. Seventy-three participants, including women who had acquired HIV through heterosexual contact and persons were not black, white, or Hispanic, were excluded from some calculations of adjusted rates because of small sample sizes. Distributions of time to death among patients with tuberculosis were estimated using the Kaplan-Meier method, and comparisons were made using the log-rank test.

To calculate tuberculosis rates by immunologic status, we divided the time each participant was followed into the number of years during which CD4 lymphocyte counts were 200 cells/mm³ or greater and the number of years after CD4 lymphocyte counts were less than 200 cells/mm³. Time for these CD4 groups was then summed for all participants. Rates were calculated as the number of tuberculosis cases in each group divided by the number of years followed. One participant who did not have CD4 measurements was omitted from these calculations.

Twenty-three participants who were never tested for PPD response were excluded from PPD conversion rates and calculations of tuberculosis rates by PPD status. Participants who were tested for PPD response at least once were classified into one of three groups: positive at entry, newly positive (converted), or negative. One hundred seventy-two of 1107 participants were assigned to groups on the basis of only one test. Baseline PPD status was assigned for 66 participants who were not tested at study entry by using the first reported test and time followed calculated from the date of that test. For participants who developed tuberculosis, only the results of PPD tests done before diagnosis were considered. Tuberculin converters were considered to be part of the negative group before becoming PPD positive and to be part of the newly positive group after converting.

Persons who reported a history of isoniazid use or tuberculosis before the study or who received isoniazid for at least 6 months during follow-up were considered to have completed prophylactic therapy. All others were considered to have not been given prophylaxis. Time before completion of isoniazid therapy was included with time followed in the untreated group. Restricted tuberculosis rates were calculated by PPD status for those considered to have not received prophylaxis.

Results

Top Methods Results Discussion Author & Article Info References

Patient Characteristics at Baseline

Overall, 1171 HIV-seropositive persons entered the study. Follow-up was completed for 1130 persons (96%), whose baseline characteristics are shown in Table 1. Approximately 1% of patients reported a history of tuberculosis, 8% reported a previous positive result on a PPD test, and 4% reported previous use of isoniazid. The median CD4 T-lymphocyte count among HIV-seropositive patients was 410 cells/mm³: Thirty-six percent of patients had counts of at least 500 cells/mm³, 44% had counts between 200 and 499 cells/mm³, and 19% had counts of less than 200 cells/mm³. At study entry, 6% of patients were PPD positive, 42% were reactive to mumps antigen, and 54% were anergic. A cross-sectional analysis of skin-test results at baseline in this cohort has been described in detail elsewhere [18].

Patient Follow-up

The median duration of follow-up was 53 months. By the end of the study, 655 persons had survived after a median follow-up of 57 months (range, 31 to 64 months), 354 had died after a median follow-up of 31 months (range, 1 to 63 months), and 121 withdrew or were lost to follow-up after a median of 25 months (range, 1 to 61 months).

Participants received PPD skin tests a median of three times, and 1107 (98%) participants were evaluated at least once. Sixty-six (6%) participants were PPD positive when first tested. Among the 1041 patients who were PPD negative at first testing, 29 subsequently became PPD positive (0.8 conversions per 100 person-years). During follow-up, isoniazid prophylaxis was prescribed for 110 persons (10%), but only 53 (5%) received therapy for 6 months or more.

Incidence of Tuberculosis

Cases of Tuberculosis

Tuberculosis was diagnosed during the study period in 31 HIV-seropositive patients (0.7 cases per 100 person-years [95% CI, 0.5 to 1.0 cases]). Sixteen participants (52%) had pulmonary involvement only, 7 (23%) had extrapulmonary disease only, and 8 (26%) had both pulmonary involvement and extrapulmonary disease. Of the 23 patients with cases that were considered confirmed, 21 had pulmonary involvement. Twenty-one of the 24 cases that affected the lungs and only 1 of 7 cases without pulmonary involvement were confirmed by culture. Mycobacterium tuberculosis was isolated from respiratory tract specimens from 19 patients (6 of 13 patients had a positive smear). In only 4 of 15 patients with extrapulmonary disease, M. tuberculosis was recovered from extrapulmonary sources (urine in 1 patient, blood in 1 patient, bone marrow in 1 patient, and lymph nodes in 1 patient). Among the remaining patients with extrapulmonary disease, 1 had involvement of the pleura, 1 had involvement of the meninges, 1 had involvement of the uvea, and 3 had involvement of the lymph nodes. No specific site was identified in 5 persons.

Twenty-seven participants had CD4 cells measured within 6 months (median, 2.3 months) before the diagnosis of tuberculosis. The median CD4 cell count in these persons was 144 cells/mm³ (range, 2 to 543 cells/mm³). Eleven of 28 participants who were tested had a positive PPD reaction before diagnosis. Of these 11 participants, 8 were PPD positive at baseline, and 3 converted from 0 mm to at least 5 mm. Among 18 patients who were tested with both PPD and mumps antigen within 6 months (median, 2.6 months) before diagnosis, all 9 patients with a CD4 count of less than 200 cells/mm³ were anergic, and 7 of 9 patients with a count greater than 200 cells/mm³ were PPD positive.

By the end of the study, an AIDS-defining event (other than tuberculosis) had been reported in 12 of the 31 patients (39%). The median survival time after diagnosis of tuberculosis was 23.6 months. Seventeen of the 31 patients (55%) died during follow-up. Tuberculosis was reported as a primary or secondary cause of death for 7 patients. However, in 6 patients, diagnosis was established after death and treatment was never given. Death was inversely proportional to the last CD4 count before diagnosis of tuberculosis (Figure 1). Ten of 16 persons with a CD4 count of less than 200 cells/mm³ died; the median time from diagnosis to death was 6.6 months. Seven of 15 participants with CD4 counts of 200 cells/mm³ or more also died; their median time to death was 45 months (P < 0.02).

View larger version (13K): [in this window] [in a new window]   Figure 1. Estimated time to death for HIV-seropositive persons who received a diagnosis of tuberculosis. The solid line represents patients whose last CD4 count before the diagnosis of tuberculosis was 200 cells/mm3 or more; the broken line represents patients whose count was less than 200 cells/mm3. Sample sizes at baseline and at 12 and 24 months are shown in parentheses. Among patients with CD4 counts of 200 cells/mm3 or more, 7 persons died and the 12-month survival rate was 92%. Among patients with CD4 counts of less than 200 cells/mm3, 10 persons died and the 12-month survival rate was 34%. This difference in mortality was significant (P = 0.02 by log-rank test).

Predictors of Tuberculosis at Study Entry

The effects of transmission category, ethnic group, and study center location were first examined individually (Table 2). To examine the independent effects of injection drug use, black ethnic group, and residence in the eastern United States, we considered the confounding induced by the demographic composition of the cohort. Transmission category and ethnic group varied by study center region (in the eastern United States, 50% of persons in the cohort were injection drug users and 46% were black; in the midwestern and western United States, 12% of persons in the cohort were injection drug users and 14% were black). In addition, ethnic group varied by transmission group (68% of injection drug users were black and 25% were white; 9% of homosexual men were black and 83% were white). After we adjusted for ethnic group and location, the rate of tuberculosis among injection drug users compared with that among homosexual men decreased (risk ratio, 1.3 [CI, 0.5 to 3.7]). The rate among black persons compared with that among white persons was similarly reduced by adjustment for transmission category and location (risk ratio, 2.1 [CI, 0.5 to 8.8]). After adjustment for transmission group and ethnic group, eastern residence remained the strongest demographic predictor for tuberculosis (risk ratio, 4.1 [CI, 2.0 to 8.4]). The incidence of tuberculosis was similar in the eastern United States among injection drug users (2.2 cases per 100 person-years [CI, 1.1 to 3.9]) and homosexual men (1.9 cases per 100 person-years [CI, 0.9 to 3.6]). No differences in rates were seen between sexes.

Immunologic Status

Tuberculosis rates were determined for groups stratified by CD4-lymphocyte count. During the study period, 645 participants (57%) had at least one CD4 count of less than 200 cells/mm³; 17 cases of tuberculosis were diagnosed in this group (1.2 cases per 100 person-years) compared with 14 cases among persons with counts of 200 cells/mm³ or more (0.5 cases per 100 person-years; risk ratio, 2.4 [CI, 1.1 to 5.2]). Results were similar after adjustment for ethnic group and location. A similar trend was seen among homosexual men (1.1 cases per 100 person-years for CD4 counts of less than 200 cells/mm³ compared with 0.2 cases per 100 person-years for counts of 200 cells/mm³ or more; risk ratio, 7.3 [CI, 2.0 to 41.9]). However, the incidence of tuberculosis among injection drug users was the same in both strata (1.7 cases per 100 person-years; risk ratio, 1.0 [CI, 0.3 to 3.2]).

Tuberculin Status

Table 3 shows the incidence of tuberculosis by PPD status and CD4 strata. Three persons who had not received tuberculin testing before diagnosis were excluded. Of the 1107 participants tested, 80 completed at least 6 months of isoniazid prophylaxis or therapy and were excluded from the calculation of restricted rates. However, 59 persons who received isoniazid for a mean of 2 months (range, 1 week to 5 months) were included.

Compared with patients who remained PPD negative, patients who were PPD positive were more likely to develop tuberculosis, whether they were positive at first testing (restricted risk ratio, 10.9 [CI, 3.5 to 30.0]) or newly positive (restricted risk ratio, 13.1 [CI, 2.5 to 46.6]). Among PPD-negative patients, the incidence of tuberculosis was greater for persons with a CD4 count of less than 200 cells/mm³ at study entry than for those with higher counts (restricted risk ratio, 2.9 [CI, 0.8 to 9.2]). When patients were grouped by region, the incidence of tuberculosis was greater at sites in the eastern United States than at all other sites for patients who were PPD positive at baseline (4.6 cases per 100 person-years [CI, 1.7 to 10.1] in the eastern United States compared with 1.7 cases per 100 person-years [CI, 0.2 to 6.0] elsewhere) and those who were PPD negative (1.3 cases per 100 person-years [CI, 0.6 to 2.3] in the eastern United States compared with 0.2 cases per 100 person-years [CI, 0.08 to 0.4] elsewhere). Among the 29 tuberculin converters, 3 of 15 from the eastern United States developed tuberculosis (10.3 cases per 100 person-years [CI, 2.1 to 30.0]) compared with none from other regions.

The incidence of tuberculosis varied according to the magnitude of PPD response (Figure 2). Tuberculosis occurred in 9 of 70 patients with indurations of at least 10 mm (3.9 cases per 100 person-years [CI, 1.8 to 7.4]), in 2 of 24 patients with indurations of 5 to 9 mm (2.4 cases per 100 person-years [CI, 0.3 to 8.8]), and in 0 of 56 patients (24 from the eastern centers) with indurations of 1 to 4 mm. In contrast, tuberculosis was diagnosed in 17 of 985 participants who had no response to PPD testing (0.5 cases per 100 person-years [CI, 0.3 to 0.7]). If participants who were considered to have received prophylaxis or treatment before the study were excluded from analysis, the differences became more pronounced. The best criterion for positive results on a PPD test appeared to be an induration of 5 mm or more (3.5 cases per 100 person-years [CI, 1.8 to 6.3]); the best criterion for a negative result was an induration of less than 5 mm (0.5 cases per 100 person-years [CI, 0.3 to 0.7]).

View larger version (18K): [in this window] [in a new window]   Figure 2. Incidence of tuberculosis in HIV-seropositive patients by size of purified protein derivative tuberculin induration. Height of bars represents the rate of tuberculosis (in cases per 100 person-years) for each group. The numbers above the bars are the rates (in cases per 100 person-years), 95% CIs, and numbers of persons who received a diagnosis of tuberculosis/total number of persons.

Control Antigen Testing

Response to mumps antigen was investigated as a predictor of tuberculosis. Table 4 shows tuberculosis rates by paired PPD and mumps responses, stratified by CD4 count measured at the time of delayed-type hypersensitivity testing. Among PPD-negative participants, the rate of tuberculosis was higher in those who did not react to mumps antigen than in those who did (risk ratio, 3.9 [CI, 1.1 to 21.2]). This effect was seen for both high and low CD4 strata. The rate was also higher among participants who were PPD positive at study entry and did not react to mumps antigen (5.8 cases per 100 person-years [CI, 2.3 to 11.9]) than among those who did (0.8 cases per 100 person-years [CI, 0.02 to 4.3]).

Discussion

Top Methods Results Discussion Author & Article Info References

The salient observations of our study include an incidence of tuberculosis in HIV-infected persons that is in the middle of those previously reported [6-14], a strong relation between geographic location and risk for tuberculosis, support for current PPD tuberculin interpretative criteria, and an association between risk for tuberculosis and response to mumps antigen that was independent of PPD status.

Tuberculosis occurred at a rate of 0.7 cases per 100 person-years. Eastern location was the strongest demographic risk factor. Because CD4 counts at study entry were not lower in the eastern United States than they were elsewhere, the elevated risk probably did not arise from increased susceptibility to tuberculosis on immunologic grounds. In the east, both the higher prevalence at baseline of a positive result on a PPD test (12% compared with 3.5% elsewhere) and the higher incidence of PPD conversion (1.7 conversions per 100 person-years [CI, 1.0 to 2.8] compared with 0.5 conversions per 100 person-years elsewhere [CI, 0.3 to 0.8]) implicate a greater degree of exposure to M. tuberculosis as the decisive factor. In this context, the increased incidence of tuberculosis seen among PPD-positive persons in New York and Newark, New Jersey, compared with that in other regions suggests that in the eastern United States, a positive tuberculin test is more likely to represent recent infection.

The incidence of tuberculosis appeared to vary directly in relation to the diameter of the PPD induration and was eightfold greater among persons who had reactions of 5 mm or more than it was for persons with 0- to 4-mm responses. Participants with indurations of 1 to 4 mm did not appear to have a greater risk for tuberculosis than nonresponders (that is, persons with indurations of 0 mm). Because most participants with positive PPD test results also had relatively elevated CD4 counts, the influence of CD4 value on the rate of tuberculosis could only be ascertained for those persons who were PPD negative. Among such persons, the incidence of tuberculosis increased threefold at CD4 counts of less than 200 cells/mm³.

In addition, response to mumps antigen predicted risk for tuberculosis, regardless of tuberculin status. Among participants who reacted to mumps antigen, risk for tuberculosis increased approximately sevenfold in PPD-positive persons and fourfold in PPD-negative persons. Moreover, among PPD-negative persons, reaction to mumps antigen predicted a risk for tuberculosis that was independent of CD4 count.

Preventive therapy was not used to the extent possible in this cohort. Only 55% of PPD-positive participants who were eligible for isoniazid received it, and only 59% of those who received isoniazid complied with therapy for 6 months or more. The rate of tuberculosis in PPD-positive participants was 1.6 cases per 100 person-years for those who received isoniazid compared with 4.7 cases per 100 person-years for those who did not. Prophylaxis showed little effect among PPD-negative persons, including those with CD4 counts of less than 200 cells/mm³; however, persons who were perceived to have a higher risk for tuberculosis may have been preferentially selected to receive isoniazid.

Studies of the incidence of tuberculosis among HIV-seropositive persons have been done predominantly in populations in which injection drug use is prevalent [7-912, 13]. Although these studies have found incidence rates that exceed ours by a factor of 10, comparisons of similar subgroups often yield similar results. For example, the incidence of tuberculosis in untreated PPD-positive persons in New York, Madrid, and Barcelona varied from 9.7 to 16.2 cases per 100 person-years [7-9]; in PPD-positive persons in our cohort living in the eastern United States who were not given preventive therapy, this rate was 7.7 cases per 100 person-years. On the other hand, two large multicenter studies from the United States and France [11, 14], in which fewer than half of the participants used injection drugs, showed overall rates of 1.4 cases per 100 person-years and 2.2 cases per 100 person-years, respectively-results that are similar to our overall incidence of 0.7 cases per 100 person-years. Although the use of injection drugs is clearly associated with an increased incidence of tuberculosis, injection drug users are often concentrated in areas that have a high prevalence of tuberculosis. Therefore, it is not surprising that such a location emerged as the principal demographic determinant of risk for tuberculosis.

Although some investigators have proposed a reduction in the size of induration that indicates positivity in tuberculin testing, the resulting loss of specificity could lead to the treatment of excessive numbers of persons who do not have tuberculosis [23]. Our study suggests that the best cutoff point for a positive PPD test result is an induration of 5 mm or more; this corresponds with the current CDC recommendations [24]. In addition, the utility of anergy testing has been questioned, particularly because it may not reliably distinguish true-negative from false-negative results on PPD testing [18]. Nevertheless, the incidence of tuberculosis in anergic persons has approached that in PPD-positive persons in some studies and has substantially exceeded that in PPD-negative nonanergic persons [7-9, 13]. Not only does our study show an effect for mumps antigen among PPD-negative persons that is independent of the CD4 count, it shows a similar effect among PPD-positive persons. These observations can best be understood when antigen responsiveness is viewed as an independent measure of cell-mediated immune response rather than as a means to interpret tuberculin test results [25, 26].

Recent studies that have used genotypic analyses of M. tuberculosis isolates have emphasized the importance of newly acquired infection [27-31]. Strategies that target anergic persons for preventive therapy must contend with the observation that 40% to 60% of cases may represent recent infection that can often be traced to unsuspected outbreaks from a common source. Although lifelong prophylaxis has been considered [32], resources may be better spent on efforts to interrupt transmission.

Despite our large sample size and relatively long period of observation, only 31 cases of tuberculosis were diagnosed. This rendered some multivariate analyses impossible and is the principal limitation of our study. Moreover, some subpopulations that are known to have high incidence rates of tuberculosis, such as Hispanic persons, were underrepresented in this cohort [33, 34]. This could lead to a 30% to 40% underestimation of the true overall incidence. In a study [11] of a community-based cohort with demographic characteristics that represented those of patients with AIDS in the United States more closely than did those in our study, the frequency of tuberculosis was twofold higher than ours. However, the mean CD4 count of patients in that study was 205 cells/mm³ at baseline. This is about half that of patients in our study and may have led to an overestimation of the incidence rate. Therefore, despite its shortcomings, our study should reasonably reflect the epidemiology of tuberculosis in HIV-seropositive patients in the United States.

Approximately 800 000 to 1 200 000 persons in the United States are currently infected with HIV [35]. Given our observation of 0.7 cases of tuberculosis per 100 person-years, HIV infection would account for 5840 to 8760 cases of tuberculosis annually. This is 22% to 33% of the total number of cases of tuberculosis in the United States in 1992 or 26% to 38% of the total number of cases in 1995 [5, 36]. This range is consistent with the estimated 2000 co-infected patients reported from 13 urban tuberculosis clinics, in which the overall seroprevalence of HIV was 21% [37]. In addition, our findings suggest that the risk for tuberculosis is 50 to 200 times greater in HIV-infected persons than in the general population, depending on the subgroups examined. However, because many populations with a high incidence of HIV are independently at increased risk for tuberculosis [18], the specific contribution of HIV infection to this risk may be considerably lower.

The incidence of tuberculosis in a population can be specified by the risk for infection with or acquisition of M. tuberculosis and the risk for disease after infection has occurred [38, 39]. The former depends primarily on such exogenous factors as prevalence of infection and opportunity for exposure, and the latter is largely determined by such endogenous factors as competence of cell-mediated immunity. Risk may be modified, in turn, by efforts to avoid infection and the use of measures to prevent disease. The interplay of these variables probably accounts for the differences in rates that have been seen in various studies. These rates have varied to the extent that a single rate at which HIV-infected cohorts develop tuberculosis cannot be readily derived. Because strategies to prevent infection differ from those intended to prevent disease after infection has occurred, the distinction between recent and reactivated disease must be better delineated in future studies.

Appendix

The following were the participating clinical centers for The Pulmonary Complications of HIV Infection Study Group.

University of California, San Francisco, San Francisco, California: Philip C. Hopewell, John Stansell, Joan Turner, and Dennis Osmond. Northwestern University, Chicago, Illinois: Jeffrey Glassroth, Melinda Mossar, and Robert Hirschtick. Beth Israel Medical Center, New York, New York: Mark J. Rosen, Lori Meiselman, Kim K. Manghisi, Christopher Cardozo, and Thomas H. Kalb. University of Medicine and Dentistry of New Jersey, New Brunswick, New Jersey: Lee B. Reichman, Bonita T. Mangura, Claudia Hanson, Fran Dowell, and Margaret O'Toole. University of California, Loss Angeles, Los Angeles, California: Jeanne M. Wallace, Bert Shapiro, Barbara LeMaire, Barbara Richer, Janet Au, and Anne Coulson. Henry Ford Hospital, Detroit, Michigan: Paul A. Kvale, Norman Markowitz, Louis Saravolatz, Christine Johnson, Joanne Huitsing, and Annmarie Krystoforski.

The Data Coordinating Center for the group was Research Triangle Institute, Research Triangle Park, North Carolina: W. Kenneth Poole, A. Vijaya Rao, Kim Clayton, Nellie I. Hansen, Matthew C. Jordan, Jim Thompson, David Myers, Lisa LaVange, Judith Katzin, William Fulkerson, Timothy Wilcosky, Yu Lou, and Steven Game.

National Heart, Lung, and Blood Institute, Bethesda, Maryland: Anthony R. Kalica, Janet Wittes, Dean A. Follman, and Robert Wise (consultant from Johns Hopkins University, Baltimore, Maryland).

The Policy and Safety Monitoring Board for the group included Reuben Cherniack (Chair), John G. Bartlett, John E. Connett, Ronald P. Daniele, John F. French, Dixie E. Snider Jr, and Gerard M. Turino.

From Henry Ford Hospital, Detroit, Michigan; Research Triangle Institute, Research Triangle Park, North Carolina; San Francisco General Hospital, San Francisco, California; Northwestern University, Chicago, Illinois; University of Medicine and Denistry of New Jersey, Newark, New Jersey; University of California, Los Angeles, Los Angeles, California; and Mount Sinai Medical Center, New York, New York.

Ms. Hansen and Dr. Wilcosky: Research Triangle Institute, PO Box 12194, Research Triangle Park, NC 27709.

Dr. Hopewell: Chest Service, Room 5K1, San Francisco General Hospital, 1001 Potrero Avenue, San Francisco, CA 94110.

Dr. Glassroth: Medical College of Pennsylvania, 3300 Henry Avenue, Room 235H, 6th Floor, Philadelphia, PA 19129.

Drs. Mangura and Reichman: University of Medicine and Dentistry of New Jersey, University Hospital 1354, 150 Bergen Street, Newark, NJ 07103-2425.

Dr. Wallace: Olive View Medical Center, University of California, Los Angeles, Department of Medicine, 2B-182, 14445 Olive View Drive, Sylmar, CA 91342-1495.

Dr. Rosen: Beth Israel Medical Center, 7-Dazian, First Avenue at 16th Street, New York, NY 10029.