Laboratório de Biossegurança Nível 3
Ministerio da Saúde- SVS - CGLab
Fundação Oswaldo Cruz - Bahia.
terça-feira, 10 de maio de 2011
tuberculose e diabetes
Cross-sectional assessment reveals high diabetes prevalence among newly-diagnosed tuberculosis cases
Blanca I Restrepo a, Aulasa J Camerlin a, Mohammad H Rahbar b, Weiwei Wang b, Mary A Restrepo a, Izelda Zarate a, Francisco Mora-Guzmán c, Jesus G Crespo-Solis c, Jessica Briggs d, Joseph B McCormick a & Susan P Fisher-Hoch a
a. School of Public Health in Brownsville, University of Texas Health Science Center at Houston, 80 Fort Brown (SPH Bldg.), Brownsville, TX 78520, United States of America (USA).
b. The Center for Clinical and Translational Sciences, University of Texas Health Science Center at Houston, Houston, USA.
c. Secretaría de Salud de Tamaulipas, Ciudad Victoria and Matamoros, Mexico.
d. School of Biological Sciences, University of Texas at Austin, Austin, USA.
(Submitted: 31 December 2010 – Revised version received: 22 March 2011 – Accepted: 23 March 2011 – Published online: 30 March 2011.)
Bulletin of the World Health Organization 2011;89:352-359. doi: 10.2471/BLT.10.085738
Tuberculosis (TB) continues to be the leading killer among bacterial diseases worldwide. In 2009, more than 9 million new cases were diagnosed and 1.7 million people died from the disease.1 The World Health Organization (WHO) suspects that TB control is being undermined by the growing number of patients with diabetes mellitus in the world, which currently stands at an estimated 285 million but is anticipated to reach 438 million by 2030.2,3 Prior to the 1950s reports of an association between diabetes (primarily type 1) and TB were frequent in the literature, but they waned as insulin and drugs against TB became available.4,5This association (now with type 2 diabetes) was recognized again in the 1990s6–9and is currently supported by a growing body of literature.6–17 According to a recent meta-analysis, diabetes patients have three times the risk of contracting TB as non-diabetics (95% confidence interval, CI: 2.3–4.3)18 and studies report the fraction of TB cases attributable to diabetes to be between 15% and 25%.9,13,16The biological basis for the association between both diseases is not fully understood but studies suggest that diabetes depresses the immune response, which in turn facilitates infection with Mycobacterium tuberculosis and/or progression to symptomatic disease. This is corroborated by the fact that diabetes is generally diagnosed before TB develops.4,19–22
Despite the suggested importance of diabetes as a risk factor for TB, most contemporary studies are based on secondary data or self-reported diabetes status.6–17 The contribution of diabetes to the burden of TB may be more conspicuous in countries where both diseases are highly prevalent: Bangladesh, Brazil, China, India, Indonesia, Pakistan, and the Russian Federation are high-burden countries and rank among the 10 countries with the highest numbers of diabetes patients2,23 and also classified as high-burden for TB. However, the risk of TB attributable to diabetes has only been reported for India and it was estimated from secondary data.13 We need to gain a deeper understanding of the differences between TB patients with and without diabetes to assist in the development of guidelines to prevent co-morbidity.
Our research team is strategically located on the Texas–Mexico border, where TB is endemic. In 2007 the overall incidence of TB was 10.5 cases per 100 000 in south Texas and 38 per 100 000 in north-eastern Mexico (personal communications, JL Robles, Secretaría de Salud de Tamaulipas [SSA-Tamaulipas]; Brian Smith and Nita Ngo, Texas Department of State and Health Services [DSHS]). Diabetes prevalence among adults over the age of 20 years is 19.4% in South Texas and 15.1% in north-eastern Mexico.24,25 Our previous studies were conducted with data extracted from existing surveillance databases for TB control and were based on self-reported diabetes.17 In this study our objective was to estimate the risk of TB attributable to diabetes in this population along the Mexican border using primary data from patients newly diagnosed with TB and tested for diabetes.
Referral patients with probable TB were enrolled between March 2006 and September 2008 at the TB clinics in Hidalgo and Cameron County Health Departments, which are the TB reference centres for South Texas counties (South Texas), and the SSA-Tamaulipas in Matamoros, Mexico (north-eastern Mexico). Jail inmates and individuals < 20 years of age were excluded. Individuals fulfilling the inclusion criteria but refusing to participate (n = 80; 95% from South Texas) did not differ with respect to age, gender, race or ethnicity from those enrolled (data not shown). Participants signed informed consent. The study was approved by the institutional review boards of the participating institutions in Mexico and the United States.
For TB diagnosis we used standard WHO definitions: culture positive forMycobacterium tuberculosis (confirmed case), sputum smear positive for acid-fast bacilli when culture data were not available (smear-positive case), or clinical diagnosis only when microbiological test results were negative or not available (clinical case: symptoms compatible with TB and documentation of anti-TB treatment for at least 6 months).26 Participants in whom TB was ruled out or whose test results were inconclusive were excluded (Fig. 1). We used the American Diabetes Association classification guidelines for epidemiological studies: hyperglycaemia and/or self-reported diabetes.27 To measure blood glucose we used blood anticoagulated with ethylenediaminetetraacetic acid (EDTA) and a hand-held glucometer (AccuCheck Advantage, Roche Indianapolis, United States of America). We defined hyperglycaemia as fasting blood glucose ≥ 126 mg/dl or a random blood glucose ≥ 200 mg/dl. We froze whole EDTA-treated blood and batch tested it for glycosylated haemoglobin (HbA1c) (GLYCO-Tek Affinity Column, Beaumont, USA) and we defined chronic hyperglycaemia as HbA1c ≥ 6.5%.28 HIV antibody tests were performed on 222 (95%) of the TB patients at the health departments or in our laboratory (Mulitspot HIV-1/HIV-2 rapid test; Biorad Laboratories, Redmond, USA). In all other cases (11, or 5%) HIV status was based on self-reporting. Using the validated quantity frequency questions described previously, we defined alcohol abuse on the basis of consumption frequency and on the average number of alcoholic drinks consumed in a typical day.29 We defined alcohol abuse as at least one weekly episode of drinking seven drinks or more or binge drinking (at least two monthly episodes of drinking 10 drinks or more). Past or present drug abuse was self-reported. Household income was self-reported and we have reported the data in dollars (average exchange rate of 11 Mexican pesos to one United States dollar (US$) during the study period). We recorded height and weight to calculate the body mass index [BMI, (weight in pounds × 703)/(height in inches)2] and we classified patients as underweight (BMI < 18·5), normal (18·5 ≤ BMI < 25), overweight (25 ≤ BMI < 30) or obese (BMI ≥ 30).
Fig. 1. Classification of participants in study of the contribution of diabetes to tuberculosis rates, South Texas and north-eastern (NE) Mexico, 2006–2008
MOTT, mycobacteria other than Mycobacterium tuberculosis; TB, tuberculosis.a Inconclusive TB, smear and culture results unavailable or tuberculosis was diagnosed by physician but there was no documentation of TB treatment for at least 6 months.b Data required for classification as a diabetic included self-reported diabetes or blood glucose information.
Sources of data from the general population
For South Texas we estimated the prevalence of diabetes among adults 20 years of age or older from our community-based cohort, an ongoing population-based study in Cameron county in which the interview instrument used and the definition of diabetes are the same as in the present TB study.24,27 The incidence of TB and the prevalence of HIV infection were provided by Nita Ngo (personal communication, Texas DSHS) and we obtained the estimated population for 2005–2007 from the United States Census Bureau.30 For north-eastern Mexico we took the prevalence of diabetes from a study conducted in the United States-Mexico border population.25 Statistics for HIV infection and TB and the population size for the city of Matamoros were provided by JL Robles and JS Hernández, SSA-Tamaulipas (personal communication) and from published data.25,31
The study design followed STROBE guidelines.32 We used two-sample t-tests to compare continuous variables and χ2 or Fisher’s exact tests for categorical variables. Given the low incidence of TB and HIV infection in our study populations and because our estimation of odds ratios and relative risks (RRs) yielded similar numbers, we report RRs in this study. For each specific exposure variable (diabetes or HIV infection), the association with TB is expressed as an RR using the following formula:
We calculated the prevalence of exposure to diabetes among new TB cases,p(exposed|new TB cases) directly from our study data, and we obtained the prevalence of exposure to diabetes among the general population, p(exposed), from the aforementioned surveillance data. To calculate the 95% CI of the RR, we first calculated the exact binomial CI of p(exposed|new TB cases), denoted by (r1,r2). P(exposed) is considered a constant because it is estimated with a much larger sample size. The 95% CI of the RR is then given by (r1/(1−r1) × (1−p(exposed))/p(exposed), r2/(1−r2) × (1−p(exposed))/p(exposed)). The attributable risk among general population, ARpopulation, is defined as the proportion of incidence rate of TB due to diabetes or HIV infection in the general population that would be reduced if the exposure were eliminated and is estimated by ARpopluation = p(exposed)(RR−1)/[1 + p(exposed)(RR−1)]. The attributable risk among the exposed (to diabetes or HIV infection in our study), ARexposed, is defined as the reduction in the incidence of TB among the exposed population that would be experienced if the exposure were eliminated. It is estimated by ARexposed = 1−1/RR. Data were analysed with SAS version 9.1 (SAS Institute, Cary, USA). Significance was set at P ≤ 0.05 and marginal significance atP ≤ 0.10.
Characteristics of enrolled patients
A total of 333 TB suspects were enrolled in the study, and 233 fulfilled the criteria for analysis: 61 in South Texas and 172 in north-eastern Mexico (Fig. 1). All participants had pulmonary TB: 109 (46.8%) were confirmed by culture (55 in South Texas and 54 in north-eastern Mexico), 118 (50.6%) had a positive sputum smear (3 in South Texas and 115 in north-eastern Mexico) and 6 (2.6%) had clinical TB (3 in South Texas and 3 in north-eastern Mexico). Most TB patients were born in Mexico. Nearly all were Hispanics and more than 60% were male (Table 1). Patients from north-eastern Mexico were younger, more likely to be employed and less educated than those from South Texas. Mexican patients were more likely to abuse alcohol or consume illicit drugs than South Texas patients. Cigarette smoking was reported by one third of the 170 participants (34% Texas; 29% Mexico; data not shown). HIV infection rates were low. South Texas patients had a higher average BMI. Diabetes was common in patients from both countries: 39.% in South Texas and 36% in north-eastern Mexico (Table 1). Among diabetes patients, those from north-eastern Mexico were less likely to be aware that they had diabetes before this study [1/25 (4.2%) in South Texas, 12/62 (19.4%) in north-eastern Mexico)], a finding more common among males (12 males versus 1 female, P = 0·03). More than two thirds of the diabetes patients had chronic hyperglycaemia.
The prevalence of diabetes among South Texas adults was estimated at 19.5%.24Cameron county cohort participants resembled TB patients in their age distribution, ethnicity, Mexican origin (66%), low education and socioeconomic status. The self-reported average annual household incomes, by quartiles, were $7470, $12 000, $20 000 and $300 000. In contrast, the cohort had a higher proportion of females (67%) and obese individuals (60%).
There were three possible sources of diabetes prevalence estimates from the general population in north-eastern Mexico: (i) a nationwide study conducted in 2000 (9.5%);33 (ii) a report from a diabetes prevention programme on the United States-Mexico border conducted from 2001–2002 data (15.1%);25,31 and (iii) unpublished data from the SSA-Tamaulipas (15%). Although they did not provide detailed sociodemographic data for the Mexican participants, we used the last two sources because their figures were similar and more recent and because they would yield a more conservative estimate of attributable risk. For the United States-Mexico border study, combined statistics for the 4027 participants from both sides of the border (2122 from Mexican states, among these 331 from Tamaulipas) showed a relatively homogeneous age distribution (nearly 20% in each age group), a predominance of females (71%) and high rates of overweight (37%) and obesity (38%).
TB risk attributable to diabetes or HIV infection
The prevalence of diabetes was significantly higher among TB patients from the two study sites than among the corresponding general populations (39.3% versus 19.5% in South Texas; 36.0% versus 15.1% in north-eastern Mexico).24,25 Taking into account these statistics and the TB incidence among adults(18 per 100 000 in South Texas and 32 per 100 000 in north-eastern Mexico), we estimated that in both countries diabetes patients had three times the risk of developing TB as non-diabetics and that those aged 35 to 64 years had five times the risk (Table 2; no data for north-eastern Mexico). The estimated fraction of TB cases attributable to diabetes in adults 20 years old or older was 26% in South Texas and 24% in north-eastern Mexico (Table 2). In South Texas, this figure increased to 48% among individuals 35–64 years old. The fraction of TB cases attributable to diabetes across the population was 63% in South Texas and 68% in north-eastern Mexico.
Based on the official prevalence of HIV infection (0.19% for South Texas and 0.24% for north-eastern Mexico), similar calculations for HIV infection indicated that the contribution of this exposure to TB was ≤ 5% at the population level and 94% among HIV-positive individuals in both countries. Altogether, these data indicate that HIV infection is a major risk factor at the individual level, but that diabetes has a larger impact on TB at the population level.
TB patients with diabetes
Given the important contribution of diabetes to TB incidence, we further characterized TB patients to define the profile of those with diabetes. To do this we merged the data sets from both countries. TB patients with diabetes were more likely to be older and overweight or obese and less likely to abuse alcohol or drugs (Table 3). Smoking did not differ significantly between TB patients with diabetes or without it (24% and 34%, respectively) [(odds ratio: 0.6; 95% CI: 0.3–1.2), data not shown].
Although diabetes is a recognized risk factor for TB, whether TB induces a transient hyperglycaemia that would be classified as “diabetes” remains controversial.34 We found that patients with diabetes had been aware of having diabetes for approximately 8 years on average before their TB diagnosis, and they were more likely to report co-morbidities classically associated with diabetes than patients without diabetes (data not shown). These data suggest that TB developed in patients who already had diabetes.
Our study shows that the prevalence of diabetes among TB patients from adjacent populations in South Texas and north-eastern Mexico is very high and among the highest in the world.6–17 One quarter of the cases of TB in these medically-underserved communities along the Texas–Mexico border were attributable to diabetes. With diabetes on the rise in TB-endemic areas, our findings highlight the re-emerging impact of diabetes, now type 2, on TB control in regions of the world where both diseases are prevalent.
Our study was designed to estimate the contribution of diabetes to TB. To accurately define diabetes, we used primary data collected prospectively from newly-diagnosed TB patients. However, the study has some limitations. First, we may have overestimated the prevalence of diabetes because TB can cause transient hyperglycaemia.8,35 If so, our margin of error is ≤ 3% based on the concordance between our current definition based on hyperglycaemia as well as self-reported diabetes and the stricter definition based solely on chronic hyperglycaemia (HbA1c ≥ 6.5%; kappa 0.93; 95% CI: 0.88–0.98).
Second, inherent biases in data sets may have affected the comparison of TB patient data with data from the corresponding adult populations. We carefully selected age-matched data from official health department statistics or research studies in which diabetes was confirmed with blood tests. For South Texas the methods used to characterize the general population and TB patients were nearly-identical, but sociodemographic data for Mexico were less detailed. The data suggest that in both countries the proportion of females and the average BMI were higher in the general population than among TB patients, not surprisingly since TB is more prevalent among males and can cause dramatic weight loss. In any event, even if obesity in the general populations were overrepresented, this would lead to an underestimation of the TB risk attributable to diabetes because people with obesity are more prone to diabetes than people of normal weight.2,27
A third limitation comes from the assumptions made in calculating attributable risk: that a causal association exists between diabetes and TB and that other risk factors for TB are equally distributed among TB patients with and without diabetes. In our study, known risk factors for TB were less frequent among TB patients with diabetes (Table 3). Therefore, our attributable risk calculations may be underestimates of the contribution of diabetes to TB. Finally, M. tuberculosisinfection was not confirmed by culture in a substantial number of participants. Nevertheless, we are now conducting routine culture on all TB suspects and find that only 0.5% of smear-positive cases have atypical mycobacteria.
Nearly all our diabetes patients (particularly in South Texas) were aware of having diabetes for at least 6 months before being diagnosed with TB. These findings reveal that diabetic patients with TB are not new to the health-care system and highlight the fact that opportunities for preventing TB among diabetes patients are often missed. While not all diabetes patients with latent TB infection should take prophylactic treatment,36 such patients should be made aware of their risk of TB and should discuss with their physicians the potential risks and benefits of taking preventive anti-TB treatment. Our findings and what is currently known about the natural history of TB suggest that people with a history of diabetes and its complications who have had recent contact with a TB patient are prime candidates for preventive treatment. This would involve screening contacts for diabetes during contact investigations, a measure that would be most cost-efficient if only contacts 35 years or older were targeted. In our setting, diabetic individuals 35–64 years of age accounted for nearly 50% of the TB cases in this age group. These measures will also make it necessary to update the educational curriculum of health professionals to increase their awareness of the re-emerging association between TB and diabetes.
The impact of diabetes on TB control varies by country. In the United States TB rates are disproportionately higher among Hispanics, African Americans and American Indians, all of whom are also at higher risk for diabetes. Further research on the contribution of diabetes to TB in these populations is needed.7,37,38 In other regions of the world the contribution of diabetes to TB is largely unknown Of the seven countries that have the highest numbers of TB and diabetes patients in the world, India is the only one for which the risk of TB attributable to diabetes has been reported, based on secondary data. The results suggest an important contribution (20% for smear-positive TB)13 that warrants further investigation.
Our findings suggest that in Mexico the fraction of diabetics who know they have the illness is lower than in South Texas, especially among males, and the same may be true in other developing countries. Better access to health care in the United States may explain why awareness is higher, even though South Texas is one of the country’s poorest regions.30 This finding points to the potential benefits of partially integrating TB and diabetes control programmes worldwide. Males may benefit the most given the higher prevalence of TB among them39 and the fact that the acute symptoms of TB often motivate them to present to the health system. In contrast, type 2 diabetes is insidious and can persist for years without diagnosis when access to medical care is limited.40 Delayed diagnosis occurs when patients present with complications from chronic hyperglycaemia. The infrastructure for TB control could serve to improve the early detection of diabetes patients, particularly in developing countries. In light of these findings, diabetes control worldwide could be greatly enhanced by the sharing of resources, experience and infrastructure among resource-limited programmes for TB and diabetes control.41
We thank the following for their support: Diana Gomez, Perla Martinez, Caroline Mullin, Paula Pino, Nancy Rouse, and Mary Walsh from the University of Texas School of Public Health, Brownsville campus; Nita Ngo and Brian Smith from the Texas DSHS, and José Luis Robles and Jorge Sebastián Hernández from the SSA-Tamaulipas; Eduardo Olivarez, Gloria Salinas, Lydia Serna and Richard Wing from the Hidalgo County Health Departments; Herminia Fuentes, Raul Loera, Olga Ramos and the staff at the Secretaría de Salud de Matamoros; and Yvette Salinas and the staff from the Cameron County Department of Health and Human Services.
Support for this work was provided by NIH NIAID 1 R21 AI064297-01-A1, NIH NCMHD P20 MD000170-04, NIH CCTS-CTSA 1U54RR023417-01 and UL1 RR024148.