Factors associated with tuberculosis by HIV status in the Brazilian national surveillance system: a cross sectional study

Tuberculosis (TB) remains a serious public health problem, especially in developing countries like Brazil, which is one of the 22 high burden countries of the disease worldwide [1]. Over the last decade TB incidence and mortality in Brazil have been steadily declining. However, this downward trend has not been observed among HIV-infected patients [2, 3].

Since the 1980s, HIV has been one of the main factors contributing to the resurgence of TB in developed and developing countries alike [4]. The virus has changed the natural history of active tuberculosis as well as having a marked impact on the epidemiology and clinical outcomes of TB [5, 6]. HIV infected patients have an annual risk of reactivating latent TB infection between 3 and 15% compared to 0.01 to 0.1% for the general population [7].

The implementation of collaborative TB-HIV activities is still modest in Brazil. These activities include coordination between the HIV/AIDS and TB programs for delivering integrated TB and HIV services including testing TB patients for HIV, providing ART to TB co-infected patients, providing HIV prevention services for TB patients, intensifying TB case finding among people living with HIV, and offering isoniazid preventive therapy. Tuberculosis in HIV-infected patients is often not identified until death, highlighting a failure of the health system to detect both diseases earlier [8]. A recent study of the six Brazilian states with the highest reported levels of HIV data in the Brazilian national TB reporting system, reported that 40% of TB patients had no HIV status provided [9].

Studies conducted in Brazil and other high burden TB countries have reported several socio-demographic and clinical features significantly associated with TB-HIV co-infection [10–12]. However, to the authors’ knowledge, there has not been such a study done in Brazil using the Epidemiological Surveillance System. In addition, none of these studies used models of analysis using the hierarchical multivariable analysis, which can detect the factors associated with TB-HIV co-infection relative to TB only in Brazil.

In this study, we describe the epidemiological and clinical profile of TB patients by HIV status using the Brazilian National Surveillance System.

Between 2007 and 2011, 429,567 adult cases of TB were reported in SINAN, of which 185,891 (43.27%) were missing data for HIV status. Among 243,676 TB patients with known HIV status, 46,466 (19%) were TB-HIV co-infected (Figure 1).

The histopathologic examination had a higher proportion of AFB positive in the TB-HIV group compared with the TB only group (46.91% vs 31.93%, p < 0.001). DOT was greater among the TB only group (53.45% vs 38.32%, p < 0.001) and cure was markedly better as well (84.95% vs 55.75%; p < 0.001) (Table 2).

Patients diagnosed with TB after returning after default were more likely to be co-infected (OR = 2.65, 95% CI 1.13-6.25). Both pulmonary and extrapulmonary TB in the same individual were more likely to be reported as TB – HIV (OR = 2.80, 95% CI 1.56-5.02). TB–HIV co-infected patients were less likely to have a positive smear result (OR = 0.55, 95% CI 0.32-0.95), but more likely to have a suggestive histopathologic examination (OR = 2.15, 95% CI 1.13-4.07).

Finally, TB–HIV patients were more likely to get unfavorable results of tuberculosis treatment including default (OR = 2.79, 95% CI 2.45-3.18), death from TB (OR = 3.64, 95% CI 2.92-4.53) and death from other causes (OR = 7.93, 95% CI 6.66-9.43).

The prevalence of TB – HIV co-infection reported by SINAN was 19% among TB patients ≥ 15 years of age with known HIV status. Co-infection was associated with being male, black, low level of education, living in an urban area and being between the ages of 20–59 [4, 10, 18, 19]. These characteristics demonstrate the social characteristics of HIV-TB co-infection. The same epidemiological profile has been described in several studies in Brazilian cities [4, 20, 21]. These findings have been associated with the lifestyle of young adults often associated with lack of awareness of their vulnerability, exposing them to the HIV virus and tuberculosis [22].

Those socio-demographic characteristics strongly influenced the proportion of TB–HIV co-infection. Since the beginning of the new century, the AIDS epidemic in Brazil has reached new population groups and cities where it had not been previously reported, affecting the less privileged social segments in the country [3, 21, 23]. Such populations have historically been plagued with high TB rates; therefore, the introduction of the AIDS epidemic has worsened the TB problem.

Co-infected patients were less likely to have prevalent diabetes in our study. It is important to highlight that TB patients with diabetes have demonstrated worse outcomes of TB treatment. A study carried out in Brazil using the Brazilian national surveillance system (SINAN) demonstrated that TB patients with diabetes were more likely to die of TB [24].

TB/HIV co-infected patients were less likely to be AFB positive than TB patients only which can be attributed to immunodeficiency. This result corroborate with other published studies which should indicate the need to develop more sensitive diagnostic techniques to confirm TB among HIV co-infected such culture. Another tool is a scoring system used to diagnose smear-negative pulmonary TB in children and adolescents, in HIV-infected adults suspected of having smear negative pulmonary TB. A study conducted in Brazil from a cohort of 2,382 HIV-infected adults [25], 1276 were investigated and 128 were diagnosed with pulmonary TB. The scoring system of the Brazilian Ministry of Health for the diagnosis of pulmonary TB in children and adolescents was adapted by the authors of the present study for HIV-infected adults and presented a good capacity for discriminating patients who did not have pulmonary TB, in the studied population.

The treatment of co-infected TB – HIV patients is difficult. Patients with TB require long-term treatment with various medications. For patients with TB-AIDS, adherence with the treatment regimen is difficult due to the extra burden of drug taking, resulting in higher default rate as showed in our study [26–28]. Because of this difficulty and poor adherence to tuberculosis treatment, new strategies for monitoring treatment should be devised, taking into account the characteristics of the co-infection as recommended by Brazilian National Tuberculosis Control. However, in our study the TB – HIV subjects were less likely to be covered under the DOT, although it was not statistically significant. This retention in DOT for TB/HIV co-infected patients was poor in relation to available AIDS and TB treatment models in Brazil. In addition, in some Brazilian cities there is no encouragement for patients on HAART to receive treatment supervision in the community such food vouchers and transportation subsidies.

We also found similar results as other studies [3, 9, 26], indicating that co-infected patients were more likely to have unfavorable treatment outcomes (default and death from TB). The explanation of this unfavorable outcome in our study is not easy. First, CD4 count is related to the severity and clinical presentation of tuberculosis in co-infection, extrapulmonary and disseminated as observed in our study. Second, among co-infected patients, mortality is commonly related to delayed diagnosis of TB because some HIV-infected individuals postpone seeking health care in order to avoid receiving an AIDS diagnosis [10]. As we can see HIV status plays an important role in TB treatment outcome. However, for 43% of patients reported in SINAN during the study period, HIV status was unknown. This information is similar to what was found by a study [9] carried out in Brazil, from 2003 to 2008. That study showed that the group with unknown HIV status showed intermediate outcomes between the groups above (TB only and TB-HIV), suggesting that this group includes some with HIV infection.

We analyzed the epidemiological profile of tuberculosis patients with unknown HIV status with patients with known status in our study period, and we observed that the group with unknown HIV status shared more similar socio-demographic characteristics with the TB-HIV group, although some clinical characteristics of disease were more similar with TB only group as TB form (pulmonary form was more prevalent, data not shown). These facts, coupled with what the study [9] cited above points out, leads us to believe that the group of patients for which we have no known HIV status is a mix of HIV positive and negative individuals. We cannot quantify or identify the direction of the bias our results may have suffered, but as already mentioned, this is a limitation that is inherent to operational research using secondary data such as national disease reporting systems.

Our study has some other limitations. Firstly, the proportion of missing data for some variables was quite significant. Nevertheless, our large sample size still allowed us to maintain high statistical power for all analyses, but not accounting for potential biases. Secondly, we did not have access to information about antiretroviral therapy (ART) history or CD4 count among our group of HIV co-infected patients. Therefore, we did not know if the patients were diagnosed with TB prior to, concomitant with, or after the diagnosis of HIV infection.

Our analysis provides better understanding of the socio-demographic and clinical differences between HIV and non-HIV related TB in Brazil, providing evidence for developing targeted interventions directed towards reducing both infections in high risk groups. We recommend that the Brazilian National Tuberculosis Control Plan, which currently recommends all patients diagnosed with TB be tested for HIV, be strengthened through better integration and communication between AIDS and TB programs.