Overestimation of the South African HIV incidence using the BED IgG assay

Abstract

We thank Rehle et al. for their important study of HIV incidence in South Africa which we read with great interest. We agree with the authors that the incidence of HIV in South Africa is probably extremely high particularly among young women and believe that the study will help us focus HIV prevention efforts on appropriate subgroups. We have serious concerns however about the applicability of the BED IgG assay to the South African HIV epidemic. In light of recent evidence we are concerned that Rehle et al. have overstated the true absolute incidence of HIV in South Africa. As the name implies the BED assay was developed using sequences from HIV subtypes B D and E. To compensate for imperfect sensitivity and specificity Rehle et al. use a correction factor based on McDougal et al.s study of subtype B virus. Given that the majority of HIV infections considered by Rehle et al. were (apparently) of subtype C the applicability of the McDougal correction and indeed of the BED assay itself to these samples is problematic. More questions arise in light of a recent report by Karita et al. that the BED assay does not perform well in subtype C virus infections; investigators found a specificity of 71% (95% confidence interval (CI) 54 - 84%) substantially different from one estimate of specificity used in the McDougal correction (94% for infections more than 360 days in the past). In addition Karita et al. found that using the BED assay with the McDougal correction resulted in overestimation of incidence in prospective Ugandan samples (subtype not available but probably A and D) reporting a corrected BED incidence of 6.4% and a true incidence of 1.3 - 1.7%.4 We are therefore concerned that the incidence figures reported by Rehle et al. may be overestimates. If indeed these figures are incorrect this will make future comparisons with more accurate measures of incidence difficult and could lead to spurious conclusions with regard to the course of the epidemic. Given these concerns and the currentUNAIDS recommendation against using the BED assay for incidence estimation it would be helpful if the authors clarified their findings with a quantitative sensitivity analysis of their estimates. Until the BED assay has been further validated we believe that BED-derived estimates of HIV incidence must be interpreted with caution. (full-text)