Two-step tuberculin testing and standardized interviews of 793 current and former drug users were performed to determine the risk factors for tuberculin positivity. The prevalence of tuberculin positivity was 25%. Factors independently associated with tuberculin positivity among participants seronegative for human immunodeficiency virus (HIV) included crack cocaine use (adjusted odds ratio [OR], 1.6; 95% confidence interval [CI], 1.0–2.5), employment as a home health aide (adjusted OR, 2.1; 95% CI, 1.0–4.1), birth in Puerto Rico (adjusted OR, 2.2; 95% CI, 1.3–3.6), foreign birthplace (adjusted OR, 4.7; 95% CI, 1.6–13.6), African American race (adjusted OR, 2.5; 95% CI, 1.2–5.0), reported tuberculosis exposure (adjusted OR, 2.3; 95% CI, 1.2–4.4), and older age (adjusted OR, 2.9; 95% CI, 1.2–6.7). Additional risk factors among HIV-infected participants included alcoholism (adjusted OR, 2.4; 95% CI, 1.0–5.8) and high CD4+ lymphocyte count. Identification of and administration of appropriate chemoprophylaxis to drug users with these risk factors should be given high priority.
Drug users are at increased risk for Mycobacterium tuberculosis infection and tuberculosis disease [1, 2]. Evidence that drug use itself increases susceptibility to tuberculosis infection is lacking, although tuberculosis transmission has been associated with the use of crack cocaine [3, 4]. The risk of tuberculosis infection in drug users is also associated with other factors prevalent in this population, such as incarceration , homelessness , alcoholism , and foreign birthplace .
HIV infection, which is common among drug users, is a major risk factor for the development of tuberculosis disease among persons who are already infected with M. tuberculosis. However, HIV-infected individuals have lower rates of reactivity to tuberculin skin testing than do demographically similar HIV-seronegative persons, primarily as a result of a diminished delayed-type hypersensitivity response [8–11]. Whether risk factors associated with tuberculin positivity differ among HIV-seropositive versus HIV-seronegative drug users has not been determined.
Detection and treatment of latent tuberculosis infection among persons at high risk for infection and disease is a public health priority . Although typically included in such designations of high-risk groups, drug users are a heterogeneous population. Little is known about the epidemiology of tuberculosis infection among persons who use illicit drugs. Identification of specific drug use behaviors and sociodemographic characteristics associated with an increased risk of tuberculin positivity among current and former drug users would enable more-targeted allocation of prevention resources to those at highest risk for infection. We conducted a cross-sectional study to determine the prevalence of and risk factors associated with tuberculin positivity among drug users with and without HIV infection in the Bronx, New York.
Patients and Methods
Subjects were recruited from among participants in the Bronx HIV Epidemiologic Research on Outcomes (HERO) study, a longitudinal study of HIV infection among current and former drug users with and without HIV infection that is now in its 18th year . Participants in the HERO study were recruited from the Montefiore methadone maintenance treatment program and were observed during semiannual research visits, regardless of whether they remained enrolled in substance abuse treatment programs. At these visits, a standardized interview was conducted that elicited sociodemographic information, medical history, and drug use behaviors, and blood samples were obtained for performance of HIV enzyme immunoassay (with Western blot confirmation) and T cell lymphocyte studies.
From May 1995 through December 1998, consenting HERO study participants were coenrolled in a substudy to examine M. tuberculosis infection in drug users. Substudy visits coincided with semiannual HERO study visits, and data obtained for both studies were incorporated into analyses.
At enrollment in the tuberculosis substudy, an additional standardized interview was performed that elicited detailed information regarding housing status, employment history, and prior incarceration. Tuberculosis exposure was assessed, and “exposure to tuberculosis” was defined as either living in the same room or apartment as, working in the same room or area as, or spending ⩾2 h with a person believed to have active tuberculosis. Participants were also screened for alcoholism status by use of the CAGE questionnaire , with a positive response to ⩾2 items indicating alcoholism . All participants who did not have a history of severe local reaction to tuberculin skin tests (TSTs) underwent 2-step tuberculin testing as well as anergy testing, regardless of whether they reported a prior positive tuberculin reaction. A dose of 0.1 mL of 5-tuberculin unit purified protein derivative (PPD; Tubersol; Connaught Laboratories) was injected intradermally in the volar aspect of each participant's forearm with a 26-gauge beveled syringe. Anergy testing was performed concomitantly with the Multitest CMI device (Merieux Institute), which delivered a panel of 7 antigens and 1 control percutaneously to the contralateral forearm.
The maximal diameter of induration to the PPD test and each of the Multitest antigens and control was measured by a research assistant in accordance with accepted (ballpoint) technique  48–72 h after placement. Participants with a cutaneous reaction to the baseline PPD test of <10 mm induration underwent placement of a second PPD test 1 week later. Before initiation of this study, each research assistant successfully completed a course in skin testing technique taught by the New York City Department of Health Bureau of Tuberculosis Control.
Participants received monetary compensation for each visit. The study protocol was approved by the institutional review board for the protection of human subjects of Montefiore Medical Center, and all participants provided written informed consent.
Tuberculin positivity was defined as a ⩾5-mm induration in response to either the baseline PPD test or the booster test in HIV-seropositive individuals and as a ⩾10-mm induration in response to either PPD test in HIV-seronegative individuals . Cutaneous anergy was defined as the absence of reaction to each of 7 Multitest CMI antigens of ⩾2 mm more than that produced by the control prong of the testing device and by less than a 2-mm response to the PPD test . All participants with a positive TST result were referred to a specialist for further medical evaluation.
The associations of tuberculin positivity with sociodemographic characteristics, drug use behaviors, and clinical variables, including HIV serostatus and CD4+ lymphocyte count, were determined. Given the substantial effect modification by HIV serostatus, all analyses were conducted for the HIV-infected and -uninfected populations separately. Comparisons were performed using the χ2 test or Fisher's exact test, as appropriate. Multivariate logistic regression analysis was performed to assess independent predictors of tuberculin positivity and included all variables with bivariate associations of P < .10 for either the HIV-infected or -uninfected subgroup, with the following exceptions: (1) crowding was not included in either multivariate model because data were only available for a subset of the population (420 subjects), (2) foreign birthplace was not included in the multivariate model for HIV-seropositive participants because of the small sample size, and (3) CD4+ lymphocyte count was not included in the multivariate model for HIV-seronegative participants. Corresponding ORs and 95% CIs were computed. SPSS software, version 10.0 (SPSS), was used for all analyses.
Participant characteristics. A total of 806 HERO study participants enrolled in the tuberculosis substudy. This analysis included 793 individuals (98%) for whom a TST was performed and the results interpreted. The characteristics of the participants at the time they underwent the TST are listed in table 1. All participants had a history of heroin use, and 716 (90%) of 793 reported that they were currently receiving methadone maintenance treatment. Two hundred fifty-six subjects (32%) were HIV seropositive; the median CD4+ lymphocyte count among these individuals was 338 cells/mm3 (range, 1–1732 cells/mm3). Compared with the remainder of the cohort, the HIV-infected participants were significantly more likely to be African American, to receive public assistance, and to report a history of injection drug use, incarceration, and homelessness (P < .05).
Prevalence of tuberculin positivity. Overall, the prevalence of tuberculin positivity was 25% (198 of 793 subjects). The majority (174 [88%] of 198) of positive responses were detected with the initial TST. Of the 446 initially tuberculin-negative participants who received a second TST, 24 (5%) had a booster response. There was no difference in the prevalence of booster response on the basis of HIV serostatus or CD4+ lymphocyte count.
Anergy testing was performed for 532 (99%) of 537 HIV-seronegative individuals and 253 (99%) of 256 seropositive individuals. Anergy was less prevalent among HIV-seronegative participants (36 [7%]) than it was among HIV-seropositive participants (90 [35%]; P < .0005). Conversely, tuberculin positivity was more common among HIV-seronegative participants (155 [29%]) than it was among HIV-seropositive participants (43 [17%]; P < .0005). However, among nonanergic participants only, this difference was no longer significant (31% vs. 25%, for HIV-seronegative vs. HIV-seropositive participants; P = .16). Among 236 HIV-infected individuals with T cell lymphocyte data, tuberculin positivity was more common in those with a CD4+ lymphocyte count >500 cells/mm3 (19 [31%] of 61 subjects) than in those with a CD4+ lymphocyte count of 201–500 cells/mm3 (16 [16%] of 103) or those with a CD4+ lymphocyte count ⩽200 cells/mm3 (6 [8%] of 72; P = .002).
Risk factors associated with tuberculin positivity. Univariate associations between the participants' characteristics and positive TST results are presented in table 2. Among HIV-seronegative individuals, the factors significantly associated with tuberculin positivity were African American race, age of ⩾35 years, birth in Puerto Rico or a foreign country, crack cocaine use, reported tuberculosis exposure, and employment in a shelter or as a home health aide. Among 278 HIV-uninfected participants with living space data, tuberculin positivity was more common among those living in crowded conditions, which were defined as >2 people per sleeping room (P = .07). A positive CAGE test result for alcoholism was also of borderline significance.
Among HIV-infected subjects, the factors significantly associated with tuberculin positivity were: African American race, a CAGE test result positive for alcoholism, reported tuberculosis exposure, and a CD4+ lymphocyte count >500 cells/mm3. Sex, receipt of public assistance, homelessness, any history of staying in a shelter, previous incarceration, a history of injection drug use, and any history of work in a hospital were not significantly associated with tuberculin positivity in either subgroup.
To determine the relative importance of these factors, separate multivariate logistic regression analyses were performed for the HIV-seronegative and HIV-seropositive participants, with tuberculin positivity used as the dependent variable (table 3). Factors independently associated with tuberculin positivity among the HIV-seronegative participants included: birth in Puerto Rico (adjusted OR, 2.2; 95% CI, 1.3–3.6), birth in a foreign country (adjusted OR, 4.7; 95% CI, 1.6–13.6), African American race (adjusted OR, 2.5; 95% CI, 1.2–5.0), reported tuberculosis exposure (adjusted OR, 2.3; 95% CI, 1.2–4.4), employment as a home health aide (adjusted OR, 2.1; 95% CI, 1.0–4.1), age of ⩾35 years (adjusted OR, 2.9; 95% CI, 1.2–6.7), and crack cocaine use (adjusted OR, 1.6; 95% CI, 1.0–2.5). Employment in a shelter (adjusted OR, 2.3; 95% CI, 1.0–5.6) was also associated with tuberculin positivity but was of borderline significance. These estimates were essentially unchanged when the analysis was limited to seronegative participants who were not anergic.
Among HIV-infected participants, factors independently associated with tuberculin positivity included the following: birth in Puerto Rico (adjusted OR, 3.4; 95% CI, 1.1–10.1), reported tuberculosis exposure (adjusted OR, 4.6; 95% CI, 1.5–13.6), a CAGE test result positive for alcoholism (adjusted OR, 2.4; 95% CI, 1.0–5.8), and less immunodeficiency (for a CD4+ lymphocyte count of 201–500 cells/mm3, the adjusted OR was 3.4 [95% CI, 1.0–11.8]; for a CD4+ lymphocyte count >500 cells/mm3, the adjusted OR was 8.5 [95% CI, 2.4–29.7]). African American race (adjusted OR, 8.1; 95% CI, 1.0–69.8) was also associated with tuberculin positivity but was of borderline significance. These estimates were not substantially different when the analysis was limited to HIV-infected participants who were not anergic.
One in 4 study participants demonstrated reactivity to the TST when Centers for Disease Control and Prevention (CDC) criteria were used, reflecting a high prevalence of M. tuberculosis infection among drug users receiving methadone maintenance treatment in the Bronx. In this heterogeneous group, among HIV-seronegative participants, the risk factors for tuberculin positivity were varied and included use of crack cocaine, employment as a home health aide, birth in Puerto Rico or a foreign country, African American race, reported tuberculosis exposure, and older age; among HIV-seropositive participants, the risk factors were alcoholism, high CD4+ lymphocyte count, birth in Puerto Rico, and reported tuberculosis exposure.
To our knowledge, this is the first study that has shown an association between crack cocaine use and tuberculin positivity in a community-based cohort of drug users. Previously, crack cocaine use has been implicated in the transmission of M. tuberculosis during 2 outbreaks in California [3, 4]. Recent crack cocaine use has also been reported as a risk factor for tuberculin reactivity among hospitalized patients with mental illness . It has been hypothesized that crack cocaine use facilitates tuberculosis transmission by inducing coughing or because it may be used in poorly ventilated crack houses . Alternatively, crack cocaine users with tuberculosis may delay seeking medical care and thus remain infectious for longer periods of time than do nonusers, promoting transmission within crack cocaine—using networks .
By use of the CAGE questionnaire, we found an independent association between alcoholism and tuberculin positivity among HIV-infected participants that persisted after controlling for numerous other factors. Although heavy alcohol use has been associated with active tuberculosis [19, 20], a relationship between alcohol use and tuberculin positivity is less well established. In a study of welfare clients in New York City, Friedman et al.  found that the rate of tuberculin reactivity in alcoholics far exceeded that noted in other drug users. In this study, we demonstrated that the CAGE questionnaire, an easy-to-use screening test for alcoholism, is a sensitive tool to detect drug users at high risk for infection with M. tuberculosis.
Employment as a home health aide was strongly associated with tuberculin positivity among HIV-seronegative participants, whereas any history of work in a hospital was not. Home care workers are estimated to be at greater risk for acquiring tuberculosis than are hospital employees, presumably because of the lack of engineering controls and intermittent respirator use in their work environment . The Bronx was hard hit by the tuberculosis epidemic of the 1980s and 1990s, and many of the affected patients received care through the home health agency system. We believe that ours is the first study to demonstrate this occupational hazard of work as a home health aide, particularly among current and former drug users, who are not often considered to be employed by the health care industry. Employment in a homeless shelter was also weakly associated with tuberculin positivity in HIV-seronegative individuals, presumably because of occupational exposure to residents with active disease. We did not detect an association between either having lived in a shelter or having been incarcerated and tuberculin positivity in this population, although congregate living has been associated with tuberculin reactivity in other studies [22, 23].
In this study, both birth in Puerto Rico and birth in a foreign country were independent predictors of tuberculin positivity. The importance of foreign birthplace as a risk factor for both M. tuberculosis infection and disease is well recognized [24, 25], with most cases of tuberculosis among foreign-born persons residing in the United States resulting from infection in the person's country of birth . Among the immigrants in our study, foreign acquisition of M. tuberculosis infection was likely: most individuals came from Caribbean countries where tuberculosis is endemic. In contrast, the incidence of tuberculosis disease in Puerto Rico is comparable to that in the United States , and thus the source of infection among Puerto Rican participants with a positive TST result is less certain. The possibility that some of the positive tests among the foreign-born subjects resulted from vaccination with bacille Calmette-Guérin (BCG) cannot be excluded, because we did not have reliable data regarding BCG vaccination status in our study sample. However, the effect of BCG vaccination is likely to be small, because all of the study subjects were adults and the TST response due to prior BCG vaccination is unlikely to persist for >10 years after vaccination .
A higher prevalence of tuberculin positivity among African American persons, as occurred in our study, has also been reported by others [11, 29, 30]. However, Cantwell et al.  demonstrated that much of the increased risk of tuberculosis associated with race may be attributed to socioeconomic status. Crowding in particular has been associated with a higher rate of tuberculosis infection, because a greater degree of shared airspace increases exposure to droplet nuclei that contain M. tuberculosis bacilli. In this study, crowding was weakly associated with tuberculin positivity among a subset of the HIV-seronegative study population with living space data. Our failure to detect a significant effect of socioeconomic status on tuberculin response may be because both of our indicators (crowding and receiving public assistance) measured only current socioeconomic status, whereas an individual's risk of infection is dependent upon a lifetime history of exposure. In addition, because 81% of our study population reported receiving public assistance, we may have lacked the power to detect statistically significant differences by using this indicator of socioeconomic status.
By analyzing the factors associated with tuberculin positivity among HIV-infected and HIV-uninfected drug users separately, we attempted to isolate the powerful effect of HIV-related immunosuppression on tuberculin reactivity. Our finding that alcoholism was associated with tuberculin positivity among HIV-infected subjects only suggests that the dynamics of tuberculosis transmission may be different among HIV-seropositive and HIV-seronegative drug users. The association we describe between tuberculin positivity and CD4+ lymphocyte count among HIV-infected persons is consistent with our earlier report .
Our study has several limitations. Although the positive predictive value of the TST is increased by the high prevalence of tuberculosis infection in our study population, it remains imperfect for the detection of prevalent infection. False-negative reactions secondary to anergy would have resulted in misclassification, especially among HIV-seropositive participants. However, the effect of such misclassification appears to have been small, because our results were unchanged by the exclusion of anergic subjects from our analyses. In addition, use of self-reported data may have resulted in an underascertainment of some potential risk factors, including incarceration and injection drug use. However, the validity of self-reported data in our study is strongly supported by such findings as the high correlation between reported exposure to tuberculosis and tuberculin positivity. An additional issue is that we report only the association of specific risk factors for tuberculin positivity among drug users receiving methadone treatment. Further research is warranted to determine whether the presence of some of these factors should merit targeted screening efforts in other populations. Finally, as a result of the cross-sectional nature of this study, we could only identify factors that are associated with tuberculin positivity and could not demonstrate causality.
In summary, in this heterogeneous population of current and former drug users receiving methadone treatment, we found substantial variation in the magnitude of risk of tuberculin positivity. Abuse of specific substances (crack cocaine and alcohol) was associated with an elevated risk, whereas a history of injection drug use was not. The risk from crack cocaine use was particularly evident among HIV-seronegative drug users, whereas alcoholism was associated with increased risk among seropositive drug users. The risk of tuberculin positivity was also greater among drug users who worked as home health aides, who were born in Puerto Rico or a foreign country, who were older, who were African American, or who reported having had contact with someone with active tuberculosis. In ongoing efforts to eliminate tuberculosis in the United States, identification of and, when appropriate, provision of chemoprophylaxis for drug users and perhaps others with these risk factors should be given high priority. However, given the high prevalence of tuberculosis infection among drug users, and given the fact that infected drug users are also at increased risk for progression to tuberculosis disease , we recommend that all drug users continue to be targeted for tuberculin testing in accordance with the current guidelines from the CDC .