Reply to Black et al. High HIV incidence or poor test performance?

Abstract

We note the concern of Black et al. [1] in the implications of our findings on mother-to-child transmission (MTCT) of HIV and our recommendations related to prevention of MTCT programmes [1,2]. They also acknowledge that despite the availability of antiretroviral prophylaxis a significant proportion of perinatal HIV infections could be attributed to ‘incident maternal HIV infections’ (recent HIV infections during pregnancy). However, their main concern with our findings lies in the possible effect of an unreliable testing methodology on the reported seroincidence. We agree with Black et al. [1] that rapid tests with varying performance and more importantly poor sensitivity at the initial visit could contribute to false ‘seroconversions’ at the subsequent visit. Hence, Black et al. [1] imply that the HIV incidence in the study population is considerably lower than that reported. We confirm that the Abbott Determine HIV Rapid test (Abbott Laboratories, Abbott Park, Illinois, USA), known for its high sensitivity, and the serial testing algorithms were used at both time points, with user variability at the different times. It is likely that the sensitivity of tests was compromised at the first visit by unsupervised nonstudy staff, although this could not be determined in the absence of laboratory confirmations of these initial results [3]. Direct virological testing was not performed in this study. Hence, if HIV infections were missed at the initial visit and subsequently incorrectly interpreted as a new infection at the repeat visit by the same test, we would not have been able to confirm whether this was due to the infection being missed as a result of a ‘window period’ or poor technical performance of the rapid test. If we are to implicate the lack of sensitivity of the rapid tests as a contributory factor to the potentially large proportion of 72 cases of ‘false seroconversions’, this could most certainly contribute to an overestimate of HIV incidence. However, a similar application of the above explanation regarding the lack of sensitivity at a subsequent testing opportunity would logically provide us with an underestimate of incident HIV infections (false-negative results) at the repeat visit. We also agree with Black et al. [1] that varying testing algorithms do make interstudy comparisons difficult, and for the purpose of measuring true HIV incidence in a population, one needs to use more superior tests and testing algorithms. We would like to add that varying study designs, target populations and measures of seroincidence could also make interstudy comparisons difficult. Our study [2] reported both an incidence risk of 3% and an incidence rate of 10.7 per 100 woman-years. Our incidence risk is lower than that reported by Rehle et al. [4] in their cross-sectional survey; however, the 5.2% could also be attributed to a different testing method used. Furthermore, the antenatal HIV prevalence in Rehle et al.'s [4] study population and our study [2] population was 37 and 28%, respectively. The much lower HIV incidence (0%) reported in the Western Cape population is probably expected with the associated lowest antenatal HIV prevalence (12.6%) in South Africa [5,6]. The variable HIV incidence risks reported in the different studies are therefore plausible if one considers the related antenatal HIV prevalence in these populations. We strongly disagree with Black et al. [1] that retesting of pregnant women should be postponed until a more reliable testing algorithm is used in studies. We concur with the opinion of Karim [7] that the availability of incidence data, however inaccurate they may be in the absence of more reliable testing algorithms, remains useful information in understanding of the HIV epidemic in settings with high antenatal HIV prevalences. Furthermore, the multiple prenatal clinic visits by pregnant women provide greater opportunities for HIV testing using rapid tests in a resource-limited setting, hence it would be prudent to offer a retest at a subsequent visit if the likelihood of false-negative results range between 1.4 and 5% at the initial visit. Nonetheless, repeat testing has been demonstrated to be cost-effective and provides added opportunities to prevent perinatal HIV infection through the use of PMTCT prophylaxis, modified obstetric practice and safe infant-feeding practice [8,9].