|Home | About | Journals | Submit | Contact Us | Français|
Hypersensitivity reaction (HSR), associated with the antiretroviral abacavir, does not usually have serious outcomes unless recognised late or after rechallenge. Pharmacogenetic testing of human leucocyte antigen (HLA) B*5701 for HSR risk is being used increasingly in clinical practice backed by strong evidence among Caucasians and Hispanics.1 In the Western Australian landmark studies, updated to improve diagnostic accuracy, positive predictive values (PPVs) and negative predictive values (NPV) of B*5701 for HSR were 78.9% and 99.4%,1 respectively. When B*5701 screening was introduced into this cohort,2 and in Brighton, UK,3 no HSRs were observed in 148 and 185 B*5701 negative patients, respectively, treated with abacavir.
However, associations between B*5701 and HSR have not been adequately investigated in black Africans who represent a substantial proportion of patients infected with HIV in the UK, perhaps resulting in the test being used in these populations without appreciation of screening effectiveness. The potential problem is illustrated by estimations of “numbers needed to screen” summarised in table 11.
In Brighton, among the relatively small numbers from black (predominantly African) ethnicity, the B*5701 carriage rate was recently reported to be 5.3% (high for black Africans).3 However, as the molecular test used was unable to discriminate between HLA B*57 alleles, only one of the four black African patients designated B*5701 positive was truly positive by retrospective high‐resolution DNA sequencing, giving a true carriage rate of 1.3%. The other three patients were positive for HLA B*5703, more commonly found in Africans.4 Assuming that the strong associations between B*5701 and HSR found in Caucasians apply to black Africans (table 11),), a PPV of a B*5701‐specific test for HSR in the African Brighton cohort would be 42.2%, equating to 182 black Africans being screened to correctly identify 1 at risk of HSR, and incorrectly identifying only 1.4. This compares with only 14 Caucasians being screened to identify 1 true case, with an incorrect identification of 0.27. A substantial proportion of black African patients infected with HIV in the UK are Ugandan or Zimbabwean. The B*5701 carriage rate may be as high as 3% in Kampalan Ugandans,4 down to 0.4% among the Shona majority of Zimbabwe (http://www.allelefrequencies.net). Using identical assumptions as above, PPVs for Kampalan Ugandans and Zimbabweans would equal 63.2% and 18.2%, equating to 53 and 1371 being screened to correctly identify 1 patient at risk of HSR, and incorrectly identifying 0.6 and 4.5, respectively. Although these numbers suggest a possible utility in some Ugandans (and then only if the association between B*5701 and HSR holds), testing as a whole might still be inappropriate given the low B*5701 carriage among Nilotic and Sudanic populations in Kenya,4 ethnic groups found in large proportions among Ugandans at our clinic. More sobering estimates are arrived at when considering results of a multicentre case–control study5 where case definitions were independently evaluated by doctors treating patients with HIV, who were blinded to the results, but where diagnostic criteria were not enhanced by epicutaneous patch testing as in Western Australia.1 Using strict case definitions, a strong association between HSR and B*5701 was found in whites but not in a subset of black Americans, although the sample size could not exclude a weak association. Sensitivities and specificities for whites were 60% (lower than that found in other studies) and 96%, respectively, but for blacks 16% and 98%, respectively. Using these results for black Americans, the calculated PPVs for Ugandans and Zimbabweans, 19.8% and 3.1%, respectively, correspond to 168 and >8000 patients being screened to correctly identify 1 patient at risk, and incorrectly identifying 4 and 31, respectively. Relating genetic test accuracies in black Americans to black Africans may seem speculative but, despite racial admixture in the black Americans, HLA B loci genetic distances are often shorter between them than between various black African populations.4
Despite biological plausibility for the role of B*5701 in the pathogenesis of HSR,1,6 a lack of association in African populations may be explained by variation in genetic linkage to other critical genes involved, some known, others yet to be discovered. Indeed, HSR seems to be largely dependent on combinations of B*5701 with polymorphisms in the heat‐shock protein Hom and tumour necrosis factor α genes as part of the ancestral 57.1 haplotype.1
Given these uncertainties, how can apparent routine integration of B*5701 testing into clinical practice among black Africans be dealt with? The original Australian studies (and draft British HIV Association treatment guidelines, 2006) point out that B*5701 testing may be specific for Caucasians, and clinicians needed to maintain vigilance for HSR for all patients. Certainly, B*5701‐specific tests should confirm positive results from non‐specific ones. Cheap, accurate and rapid non‐sequence‐based typing methods able to do this are now becoming an available and attractive option.7 When the diagnosis of HSR is in serious doubt, a negative test for B*5701 with a negative epicutaneous patch test6 may be useful to guide safety of rechallenge, but more work is required to prove this. Although B*5701 screening is likely to be cost effective in Caucasians,8 our calculations suggest that cost effectiveness for many black African populations is highly questionable. Furthermore, toxicities associated with alternative antiretroviral therapies make inappropriate denial of abacavir more relevant given the rare occurrence of severe HSR‐related outcomes actually seen in current clinical practice. Therefore, in general, doctors should confidently prescribe abacavir to black African patients, provided adequate vigilance for HSR can be maintained, without testing for B*5701, until results of at least properly designed, perhaps ethnic‐specific, case–control studies become available. Adequately powered prospective studies are unlikely to be practical given the prevalence rates. Implementation of such policies needs to take account of the lack of research in mixed populations and sensitivity around perceptions of discrimination.
Related to all this is the serious need for increased training of healthcare workers in understanding general concepts in genetics,9 so that as and when pharmacogenetic tests are introduced, counselling around them is appropriate and consistent. Although challenging, pharmacogenetic research in general needs to be proactive in recruiting diverse ethnic groups, in enough numbers from the outset, before introduction of routine clinical testing to avoid systematic discrimination in accessing new personalised medicine technologies.
We thank Professor Nick Carter (Genetics, St George's University of London) and Dr Andrew Copas (Medical Statistics; Royal Free and University College Medical School) for their comments.