© 2002 American Thoracic Society
Treatment, Tuberculosis, and Human Leukocyte AntigenEast London Tuberculosis Services Department of Respiratory Medicine Homerton University Hospital London, United Kingdom Good studies of disease associations with human leukocyte antigens (HLAs) are difficult to execute. There must be a good reason to suspect an association, that is, an immunologic basis to the disease. Both subjects and control subjects must be exposed to the same etiologic factor. The number of subjects must be sufficient to allow for adequate power to detect a difference and to adjust the probability of no association with HLA by the number of tests performed. The number of different HLA alleles is vast and correction of the p values by the Bonferroni method [pcorr = 1 - (1 - p)n] means that finding an association with 13 probes for the DRB1 locus, for instance, requires p < 0.004 to attain significance. An association between tuberculosis and HLA is expected. The epidemic due to human immunodeficiency viruses has conclusively demonstrated the importance of CD4+ T cells, which recognize antigen in the context of major histocompatibility complex Class II molecules, for example, HLA-DR, HLA-DQ, and HLA-DP molecules. Studies also suggest a role for CD8+ T cells in tuberculosis, where peptides are presented by Class I molecules (HLA-A, HLA-B, and HLA-C in human populations). Exposure to tuberculosis in the control population cannot be guaranteed by a positive tuberculin skin test, as mycobacterial antigens are notoriously cross-reactive. Studies, which have demonstrated an association between tuberculosis and HLA, have used control subjects from a population in which the incidence of tuberculosis is very high, or hospital staff working in a tuberculosis ward (1–3). The form of tuberculosis is likely determined by the human host and therefore failure to use a group of patients with a well-defined clinical form of tuberculosis will necessarily obscure any association with immune response genes. In this issue of AJRCCM (pp. 916–919), Sharma and coworkers (4) note that many different groups using different methods have found a convincing association between HLA-DR2, and especially the subtype HLA-DR15, with the development of smear-positive pulmonary tuberculosis. The attributable risk (36%, with 95% confidence intervals of 17–48% in one study [3]) is consistent with the general view that factors other than the presentation of antigenic peptides by Class II HLA molecules to CD4+ T cells are also important in the pathogenesis of tuberculosis.
Sharma and coworkers suggest that the absence of HLA-DQA1*0102 and the presence of HLA-DQB1*0201 are associated with drug-induced hepatitis during the treatment of tuberculosis. It is reasonable to suspect an immunologic basis for isoniazid-induced hepatitis and to test for an HLA association. Isoniazid is associated with drug-induced systemic lupus erythematosus (5). Liver biopsies have confirmed an immunologic basis to recurrent hepatitis occurring after rechallenge with the drug (6). Rifampin, taken intermittently, causes a "flu"-like syndrome, which is thought to be due to the deposition of immune complexes of the drug with circulating antibody, but its hepatotoxic effect is likely to be direct or mediated by an effect on the metabolism of isoniazid (7). In the present study (4), the patient population was well selected to exclude other causes of hepatitis and to ensure exposure to both tuberculosis and isoniazid. However, the clinical form of tuberculosis in these patients was heterogeneous. The definition of drug-induced hepatitis was rigorous, except for those whose only criterion for inclusion in that group was "normalization of liver function after withdrawal of all antituberculosis drugs" with "absence of serologic evidence of infection with hepatitis virus A, B, C, or E." Some physicians take a minor abnormality in liver function (i.e., less than indicated in the other criteria) as confirmation that the patient is adhering to treatment. All the patients in this study were able to tolerate isoniazid after rechallenge. Only 3 of 47 patients with drug-induced hepatitis had the DQA1*0102 allele compared with 68 of 251 control subjects ( Haplotype frequencies can be helpful in estimating whether genes close to or within the major histocompatibility complex, other than the HLA molecules themselves, are associated with disease. The classic example of this has been the association of B8-DR3 with autoimmune disease and polymorphic complement genes. This haplotype has been associated with autoimmune hepatitis, perhaps by its linkage with the complement allele C4A-QO (8). In this study (4), the haplotype with an increased frequency in those with drug-induced hepatitis did not include any of the previously associated alleles, was not related to the autoimmune haplotype, and when corrected for the number of haplotypes examined, gave p = 0.085. Older age is a key risk factor in the development of isoniazid-induced hepatitis (9) and this study confirms this association. Low serum albumin and high alcohol intake were also associated with drug-induced hepatitis. A poor immune response accompanies malnutrition, alcohol, and increasing age (10–12). Whether these have obscured the relationship between HLA and hepatitis due to treatment for tuberculosis, or whether they negate the original hypothesis, is debatable. As drug-resistant tuberculosis becomes more of a global problem, efforts to use the most effective treatments available will increase. The poor association between HLA and hepatitis owing to treatment for tuberculosis makes it unlikely that the use of steroids or other immunosuppressive agents will allow the continued administration of first-line drugs for tuberculosis to those who are unable to tolerate treatment because of drug-induced liver disease. REFERENCES
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
2 = 9.35), but even at this level of significance correction for the 8 alleles tested gives a p value of about 0.07. The positive association of DQB1*0201 with drug-induced hepatitis (
