INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: Mycobacterium tuberculosis; HIV infection; contacts

Tuberculosis (TB) and the human immunodeficiency virus (HIV) infection are currently the two leading infectious causes of death worldwide, responsible for almost four million death cases annually (1). The HIV pandemic increased TB prevalence globally, but its impact has been highest in resource-poor areas of the world such as Africa, Asia, and some countries of Latin America (2).

HIV infection can directly cause an increase in TB cases because coinfected patients are more likely to experience endogenous reactivation of a latent TB infection (LTBI) or rapidly progress from recent TB infection to disease (3). Indirectly, an increase in the number of infectious TB-HIV coinfected patients represents a source of potential transmission of Mycobacterium tuberculosis to their contacts, with the possibility of secondary cases of TB infection and disease (4). Several studies during the past decade evaluated the infectiousness of pulmonary TB patients coinfected by HIV, with varied results. Some detected a higher risk of TB transmission by HIV-seropositive index cases to their close contacts (5), whereas others reported a reduced risk or no difference in the risk of TB infection and disease among contacts of HIV-seropositive TB patients (8).

There is still controversy about the relative infectiousness of TB among HIV-infected persons, and the relative infectiousness may be influenced by the local prevalence of the two diseases. To address this, we prospectively studied the close contacts of HIV-1 seropositive and HIV-1-seronegative individuals with pulmonary TB in Rio de Janeiro City, Brazil. The incidence of TB in Rio de Janeiro City was 120 per 100,000 population in 1997; it was more than twice the national average of 54.7/100,000 (14). By late 1998, the cumulative AIDS case rate in Rio de Janeiro City was 240.8/100,000 (15). Our main objective was to assess the impact of HIV on the transmission dynamics of TB in an area with a high prevalence of TB and HIV coinfection.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Evaluation of Index Cases

Patients who sought care at the Clementino Fraga Filho University Hospital of the Federal University of Rio de Janeiro (HUCFF/UFRJ) any time from January 1995 through August 1997 were eligible to participate in the study if they were at least 18 yr of age and had signs, symptoms, and chest radiographs suggestive of pulmonary TB. Because bacteriologic confirmation of M. tuberculosis by culture would not be available for all eligible study participants, having at least one respiratory specimen (sputum or bronchoalveolar lavage, [BAL], or both) that was positive for acid-fast bacilli (AFB) was the main criterion for entry into the study. Eligible patients who provided informed written consent were enrolled as index cases. For a detailed description of the methods used in this study, see online data supplement.

Evaluation of Contacts

We defined a close contact as someone who lived and slept in the same household as an index case or someone who lived elsewhere but who reported contact with the index case at the case's household for at least 20 h per week during the previous 3 mo. Index cases were asked to invite all of their close contacts to participate in the study.

If the contact provided informed written consent, we administered a standardized questionnaire, a review of clinical symptoms characteristic of TB (fever, cough, bloody sputum/hemoptysis, and weight loss), a tuberculin skin test (TST), and a chest radiograph. The TST was performed by trained, experienced nurses and consisted of 2 tuberculin units (TU) of purified protein derivative (PPD) RT 23 (Statens Serum Institut, Copenhagen, Denmark) applied by the Mantoux method (16). The result was considered positive for an induration  10 mm ( 5 mm for HIV-seropositive individuals). All contacts enrolled in the study had at least one TST applied and read.

When the TST reading was performed, a 10-ml blood sample was obtained for HIV serology. HIV testing was performed on contacts younger than 18 yr old only if they were the child of an HIV-seropositive woman or there were risk factors for HIV infection (history of blood transfusion, intravenous drug use, unprotected sexual intercourse, or multiple sex partners) reported in the responses to their questionnaire.

Tuberculin-negative contacts returned 4 mo later for a second TST. A tuberculin conversion was defined as an increase of at least 10 mm in the diameter of the induration (17). Contacts were followed for at least 12 mo.

Statistical Analysis

Data collection and management were performed using dBase 3.0 software. We used SPSS 7.0 for Windows (SPSS, Inc., Chicago, IL) for statistical analyses.

Student's t test was used to compare the means of continuous variables. Univariate analyses of the association of clinical, laboratory, and sociodemographic categorical variables with TST positivity were performed using the chi-square test and estimating the crude odds ratios (OR) with their respective 95% confidence intervals (CI). We calculated two-tailed p values, at a significance level of 5%.

Multivariate logistic regression models were constructed to identify the factors independently associated with TB infection among contacts, starting with all of the variables that were statistically significant in the univariate analysis (p < 0.05), plus the sex and age of both index cases and contacts. We defined three different groups of contacts with different probabilities of TB infection before their contact with the index case, and repeated analyses with each subgroup. The groups were defined as follows: Group 1all of the contacts; Group 2household contacts only; and Group 3household contacts younger than 15 yr of age. We calculated the adjusted OR and 95% CI of the independent variables in the final model.

The aforementioned multivariate logistic regression analysis assumes that the risk of TB infection is independent among the contacts of an index case. However, the risk of TB infection among the contacts of the same index case is not independent; there is a statistical dependence of TB infection among the contacts of the same index case, and a higher probability that all contacts or no contact in the same household are infected (18, 19). To reduce this possible bias, we analyzed the set of contacts of each index case as a single unit. The unit was defined as positive if at least one of its contacts was TST-positive. In this complementary analysis, Set 1 included all 86 index cases; Set 2 was composed of 84 index cases with household contacts; and Set 3 contained 50 index cases with household contacts younger than 15 yr of age. We excluded the independent variables based on single contacts and fit the previously selected logistic model to the new subgroups.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Index Cases

Of the 637 cases of TB diagnosed during the study period, 389 (61%) had pulmonary TB and 248 (39%) had extrapulmonary TB only. Among the pulmonary TB cases, 385 (99%) submitted one or more sputum specimens for microscopic examination. A positive AFB smear from a respiratory specimen was detected in 191 patients with pulmonary TB (50%). We were able to contact 109 (28%) of the 389 pulmonary TB cases, of whom eight refused to take part in the study and 15 were persistently sputum smear- negative. Eighty-six persons with smear-positive pulmonary TB agreed to participate in the study and were enrolled; they represented 45% of all sputum smear-positive cases diagnosed during the study period. There were no statistically significant differences in the age or sex distribution of participating versus non-participating patients with TB (data not shown). For further explanation of the results, see online data supplement.

HIV serology was not routinely offered to patients with TB outside of this study population, and the prevalence of HIV infection among all smear-positive patients with pulmonary TB in our hospital was not available.

Twenty-three patients with pulmonary TB (27%) tested positive for HIV-1 antibodies. The majority of them (74%) had another AIDS-defining condition when TB was diagnosed, atypical chest radiographic findings (64%), and CD4+ cell counts values  200/mm³ (65%). There were no statistically significant differences in sex, race, or the frequency and duration of clinical symptoms characteristic of TB between HIV-seropositive and HIV-seronegative index cases. The mean age of HIV-seronegative index cases was 40.5 yr (SD ± 16.0) versus 37.9 yr (SD ± 10.9) among HIV-seropositive patients (p = 0.400), but HIV-seropositive index cases were more frequently between 25 and 45 yr of age when compared with HIV-seronegative patients (p = 0.020).

Contacts

We identified a total of 371 close contacts of the 86 index cases with smear-positive pulmonary TB. Of these, 11 (3%) were not enrolled in the study: 7 (2%) refused to provide written, informed consent; 2 (1%) did not return for a TST reading after the PPD was placed; one 6-yr-old boy had an open scar from a second dose bacillus Calmette-Guérin (BCG) vaccination and the TST was not applied; and one close contact reported a prior history of contact with a different TB patient. All 360 contacts who were enrolled in the study (or their parents or a legal guardian, in the case of minors) provided informed written consent for all aspects of the protocol.

We evaluated 104 (29%) close contacts from 23 HIV-seropositive index cases (mean, 4.5 contacts/index case) and 256 (71%) from 63 HIV-seronegative index cases (mean, 4.1 contacts/index case) (p = 0.441). The distribution of contacts by sex, race/ethnicity, age category, relationship to the index case, BCG vaccination, predisposing clinical conditions, and intimacy of contact, by HIV serostatus, is shown (Table 1). The median age among contacts was 23 yr (range, 1 mo to 95 yr) and the average age was 27.3 yr (SD ± 19.7); 28.7 yr among contacts of HIV-seronegative index cases and 23.7 among contacts of HIV-seropositive index cases (p = 0.020).

Two hundred ninety-two (81%) were household contacts. The other 68 contacts (19%) lived in another dwelling but had contact with the index case in his or her home: 51 (75%) of these persons reported daily contact and 17 (25%) reported weekly contacts.

HIV Serostatus of Contacts

The majority of contacts (298/360; 83%) denied behavioral risk factors for HIV infection. Among those who answered positively to the questions (n = 62; 17%), being a homosexual or bisexual man (n = 7; 11%), reporting sexual intercourse with an HIV-seropositive person (n = 8; 13%), and having a history of blood transfusion within the last 10 yr (n = 34; 55%) were the most frequently reported risk factors for HIV infection.

Serologic testing for the detection of HIV antibodies was performed in 254 (71%) contacts, 72% (75/104) of the contacts of HIV-seropositive index cases, and 70% (179/256) of the contacts of HIV-seronegative index cases (p = 0.700). Eleven contacts were HIV-seropositive, representing an overall HIV infection seroprevalence of 4%; 11% (8/75) among contacts of HIV-seropositive TB patients, and 2% (3/179) among contacts of HIV-seronegative index cases (OR 7.0; CI 95% 1.80 to 27.2; p = 0.003). Of the 11 HIV-seropositive contacts, six (54%) were women and sexual partners of HIV-seropositive index cases.

The 106 contacts who did not receive HIV counseling and testing were more likely to be male (61% versus 37%, p < 0.005) and younger than 15 yr of age (91% versus 12%, p < 0.005), relative to contacts who did receive HIV counseling and testing.

TB Infection Among Contacts

The mean time from the beginning of the index case's TB treatment to the contact's first PPD application was 50.0 d for contacts of HIV-seropositive index cases and 19.6 d for the contacts of HIV-seronegative index cases (p = 0.004). The initial TST was positive in 33% of contacts (Table 2). Among 242 contacts with an initial negative TST response, 38 did not return for the second TST (11% of overall sample); 13 were contacts of HIV-seropositive index cases, and 25 were contacts of HIV-seronegative index cases (p = 0.710).

A tuberculin conversion was detected in 21% (42/204) of contacts after the second TST; 8% (5/63) among contacts of HIV-seropositive index cases and 26% (37/141) among contacts of HIV-seronegative index cases (OR 0.24; 95% CI 0.09 to 0.65; p = 0.003). Considering the results of two TST, the overall rate of TB infection among contacts was 44% (160/ 360). Table 2 summarizes the tuberculin response of contacts relative to the HIV serostatus of index cases.

The characteristics of index cases, their households, and TB infection among their contacts were determined for the univariate analysis (Tables 3 and 4). The severity of disease in the index case and his or her capacity for disseminating bacilli in the air were associated with a higher risk of TB infection in the contacts. Living in a crowded household increased the risk of TB infection among the contacts. Contacts of female index cases were more likely to be TST-positive than were contacts of male index cases, and black race/ethnicity was associated with an almost threefold increase in the risk of TB infection relative to white race/ethnicity. Contacts who were 5 yr of age or older were more likely to have a positive tuberculin response, than contacts younger than 5 yr of age. Contacts who were the spouses or offspring of an index case tended to have a higher risk of a positive TB response (p = 0.070). Tobacco smokers had an increased risk of a positive TST response (p = 0.030).

The tuberculin response was positive ( 5 mm) in four HIV-seropositive contacts. Among the other seven HIV-seropositive contacts, two were tuberculin converters and five were tuberculin-negative.

Contacts as Sets

We constructed multivariate logistic regression models to identify those variables that were independently associated with TB infection among three different groups of contacts: Group 1 with all of the contacts (n = 360); Group 2 with only household contacts (n = 292); and Group 3 (n = 107) with household contacts younger than 15 yr of age (Table 5). The HIV serostatus and grade of sputum smear positivity in the index case, having 4 or more persons living in the same household, and the contact's age remained in the final model for the three different groups of contacts. There was a strong, protective (OR < 1) association between HIV infection in the index case and the risk of TB infection in the contacts. This association was particularly significant when the analysis was restricted to household contacts younger than 15 yr of age (p = 0.006). The odds of TB infection among close contacts of HIV-seropositive index cases were almost 50% lower (OR = 0.54), compared with contacts of HIV-seronegative index cases in Group 2 and approximately 80% lower (OR = 0.23) in Group 3. 

The relationship between the grade of sputum smear positivity in the index case and crowding in the household was less significant when the analysis was restricted to Group 3 (p = 0.099 and p = 0.061, respectively). This result is likely a consequence of the smaller sample (n = 107), because the additive risk remained (OR > 1.95 in Groups 1 and 2). Increasing age was associated with a higher risk of TB infection among all three groups of contacts; the additive risk varied from 1% to 14% for each year of a contact's life. Cough, which was significantly associated with a positive TST response in the univariate analysis, was not important in the multivariate logistic regression model.

When we analyzed the contacts as sets to evaluate the hypothesis of statistical dependence, the HIV seropositivity of the index cases was associated with a reduced risk of TB infection among contacts, for household contacts younger than 15 yr of age (Table 5).

Tuberculosis Disease Among Contacts

TB disease was diagnosed in nine (3%) contacts. Three cases of TB were diagnosed at the initial visit, another three during the first month of evaluation, and the last two after 425 and 609 d of follow-up, respectively. Eight contacts with TB disease were contacts of HIV-seronegative index cases. The prevalence of TB disease among contacts of HIV-seronegative and HIV-seropositive index cases was 3% (8/256) and 1% (1/104), respectively (p = 0.46). In six cases, the diagnosis of active TB was based on sputum smear light microscopy. Two children with hilar adenopathy and a young woman with a pleural effusion who responded to multidrug chemotherapy for TB, were diagnosed without bacteriologic confirmation. Seven contacts received HIV counseling and testing, and one had a positive HIV test result.

Six of nine contacts with TB disease had an initial TST response  10 mm. One contact with a TST conversion abandoned treatment for LTBI in the second month of therapy and was diagnosed with TB after 1 yr.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We assessed the impact of HIV on the transmission dynamics of smear-positive pulmonary TB in an area with a high prevalence of TB and HIV coinfection. In our study population, TB-HIV coinfected patients were less likely to infect their close contacts with M. tuberculosis, relative to HIV-seronegative patients with TB. Our findings agree with previous studies reporting that the HIV seropositivity of index cases did not represent a greater risk of TB infection and disease to their contacts (8). In our analyses, we tried to elucidate and control the factors inherent to the contacts, the environment, and the index cases that could have influenced TB transmission. There were significant differences in the sex and age distribution of contacts of HIV-seropositive and HIV-seronegative index cases. However, these two variables were included in the multivariate logistic regression models, thereby controlling their impact on TB transmission and infection. There were no significant differences between the two groups of contacts based on the socioeconomic level of contacts, the index case's years of schooling, and salary level. Similarly, the reported intimacy of contact with the index case, prior BCG vaccination, and predisposing health conditions were not significantly different among the groups of contacts.

The time between the initiation of treatment for TB in the index case and the application of the first PPD in the contact was greater for contacts of HIV-seropositive index cases than for contacts of HIV-seronegative index cases. It is possible that the "window period" for the tuberculin conversion was surpassed when the contacts of HIV-seropositive patients were initially evaluated for TB infection, explaining the absence of a statistically significant difference in the TB infection rate between the two groups of contacts when the results of the first TST were compared. Repeated TST controlled this bias.

The prevalence of HIV infection among contacts was 4% overall and 11% among the contacts of HIV-seropositive index cases. An HIV test result was available for 71% of the contacts, and there was no significant difference in the HIV testing rate among contacts of HIV-seropositive index cases and contacts of HIV-seronegative patients. Most of those who were not tested for HIV were younger 15 yr of age and lacked risk factors for HIV infection. The higher rate of TB disease among contacts of HIV-seropositive patients described by other investigators (7, 10) could be explained by the HIV coinfection rates among these contacts rather than a greater infectiousness of TB in HIV-seropositive index cases.

A second TST applied to contacts with an initial induration less than 10 mm permitted a more reliable estimate of the TB infection rate among them. Without a second TST, approximately 12% of TB-infected contacts would not have been identified and would not have received the benefits of treatment for LTBI. We used a strict definition of a tuberculin conversion ( 10 mm increase in the diameter of the induration), which is larger than that usually recommended when evaluating close contacts (17). Although this criterion caused the TST to be less sensitive, it permitted a more specific estimation of TB infection. This was particularly important in a region with a high coverage of BCG vaccination and high rates of atypical mycobacterial infection (20). We also assessed TST conversions after only two TST. Consistent with other studies (21- 24), our analyses indicated that prior BCG vaccination was not significantly associated with TST positivity.

The multivariate logistic regression analysis applied to three different groups of contacts that were progressively more specific for TB infection, confirmed a lower risk of TB infection among contacts of HIV-seropositive, smear-positive pulmonary patients. The strength of this association was greater when only contacts younger than 15 yr of age were analyzed, despite the smaller sample. When we controlled for possible statistical nonindependence by studying the contacts as sets, the HIV seropositivity of the index case was consistently associated with a lower risk of TB transmission.

The factors associated with a reduced risk of TB transmission from a person with pulmonary TB are not fully understood. Those factors associated with the transmission of bacilli, such as duration of cough and a cavitary chest radiograph, could differ significantly between HIV-seropositive and HIV-seronegative TB patients (25, 26). However, these characteristics were not significantly associated with transmission in our study population. The stage of immunodepression of HIV- infected TB patients, based on the CD4+ cell count, was also not associated with TB infection among contacts.

A reduced bacillary load in pulmonary TB patients infected with HIV could explain their reduced infectiousness. A study in Zambia found a weak association between the HIV seropositivity of the index case and a positive TST response among the contacts, when controlling for the bacillary load of the index case (9). TB patients with HIV infection more frequently have atypical findings in their chest radiograph or normal radiographs (27, 28) that could represent a reduced burden of bacilli, rather than a reduced immunologic response to a mycobacterial infection. In our study, the presence of a smear of grade 3+ in the sputum of the index case was associated with a higher risk of TB infection among contacts. However, this statistically significant association was not present when the contacts were evaluated as sets. We do not believe that the reduced risk of M. tuberculosis transmission was associated with a shorter period of infectiousness of the HIV-infected patients because the duration of disease was comparable between the two groups of patients and the initial response of HIV-infected TB patients to anti-TB treatment is similar to that of HIV-seronegative ones (25, 27, 29, 30).

We did not determine whether there were specific virulence properties related to the bacillus that were responsible for the differences in the infectiousness between index cases. However, TB outbreaks among HIV-infected patients involving a particular restriction fragment length polymorphism (RFLP) pattern of M. tuberculosis have been described (31, 32). Some outbreaks report the predominance of one or more specific strains in defined geographical areas (31). These studies indicate that certain intrinsic characteristics of the bacilli may be important in defining local TB transmission dynamics.

The prevalence of TB disease among contacts (nine cases; 3% of total) detected in our study was less than that described by the authors of other studies: 5% in Zaire (8); 6% in Kenya (10); 6% in Barcelona, Spain (7); and 6% in the Dominican Republic (12). In a study evaluating contacts of HIV-seronegative TB patients in Rio de Janeiro City, Lapa e Silva and coworkers reported that the prevalence of TB disease was 4% (34). In a different study in Rio de Janeiro City, the proportion of TB disease among contacts of patients with drug-resistant pulmonary TB was 9% (35). Because of the strict enrollment and evaluation procedures, it is unlikely that there were undiagnosed TB cases among the contacts in our study. It is likely that the treatment of LTBI using isoniazid on TST converters reduced the number of TB cases among the contacts (30).

The contacts of HIV-seropositive persons with pulmonary TB were less likely than the contacts of HIV-seronegative persons to have a positive TST response by 1 yr of follow-up. Our data reinforce the need for contact tracing and screening the close contacts of smear-positive pulmonary TB patients, independently of the HIV serostatus of index case. In countries such as Brazil, strategies to identify and treat close contacts who were recently infected with TB could have a positive impact on TB prevention and control in the near future.