Division of Infectious Diseases, Department of Medicine, School of Medicine, Johns Hopkins University, Baltimore, Maryland, USA.
Fred Hutchinson Cancer Research Center, Seattle, Washington, USA.
J Int AIDS Soc. 2021 Dec;24(12):e25830. doi: 10.1002/jia2.25830.
Cross-sectional incidence testing is used to estimate population-level HIV incidence and measure the impact of prevention interventions. There are limited data evaluating the accuracy of estimates in settings where antiretroviral therapy coverage and levels of viral suppression are high. Understanding cross-sectional incidence estimates in these settings is important as viral suppression can lead to false recent test results. We compared the accuracy of multi-assay algorithms (MAA) for incidence estimation to that observed in the community-randomized HPTN 071 (PopART) trial, where the majority of participants with HIV infection were virally suppressed.
HIV incidence was assessed during the second year of the study, and included only individuals who were tested for HIV at visits 1 and 2 years after the start of the study (2016-2017). Incidence estimates from three MAAs were compared to the observed incidence between years 1 and 2 (MAA-C: LAg-Avidity <2.8 ODn + BioRad Avidity Index <95% + VL >400 copies/ml; LAg+VL MAA: LAg-Avidity <1.5 ODn + VL >1000 copies/ml; Rapid+VL MAA: Asanté recent rapid result + VL >1000 copies/ml). The mean duration of recent infection (MDRI) used for the three MAAs was 248, 130 and 180 days, respectively.
The study consisted of: 15,845 HIV-negative individuals; 4406 HIV positive at both visits; and 221 who seroconverted between visits. Viral load (VL) data were available for all HIV-positive participants at the 2-year visit. Sixty four (29%) of the seroconverters and 3227 (72%) prevelant positive participants were virally supressed (<400 copies/ml). Observed HIV incidence was 1.34% (95% CI: 1.17-1.53). Estimates of incidence were similar to observed incidence for MAA-C, 1.26% (95% CI: 1.02-1.51) and the LAg+VL MAA, 1.29 (95% CI: 0.97-1.62). Incidence estimated by the Rapid+VL MAA was significantly lower than observed incidence (0.92%, 95% CI: 0.69-1.15, p<0.01).
MAA-C and the LAg+VL MAA provided accurate point estimates of incidence in this cohort with high levels of viral suppression. The Rapid+VL significantly underestimated incidence, suggesting that the MDRI recommended by the manufacturer is too long or the assay is not accurately detecting enough recent infections.
横断面发病率检测用于估计人群 HIV 发病率，并衡量预防干预措施的效果。在抗逆转录病毒治疗覆盖率和病毒抑制水平较高的情况下，评估估计值的准确性的数据有限。了解这些环境中的横断面发病率估计值很重要，因为病毒抑制可能导致最近的检测结果出现假阴性。我们比较了多检测算法（MAA）对发病率的估计准确性与社区随机 HPTN 071（PopART）试验中的观察结果，该试验中大多数 HIV 感染者的病毒均受到抑制。
在研究的第二年进行了 HIV 发病率评估，仅包括在研究开始后 1 年和 2 年（2016-2017 年）接受 HIV 检测的个体。将三种 MAA 的发病率估计值与第 1 年至第 2 年的观察发病率进行比较（MAA-C：LAg-亲和度<2.8 ODn + BioRad 亲和度指数<95%+VL>400 拷贝/ml；LAg+VL MAA：LAg-亲和度<1.5 ODn + VL>1000 拷贝/ml；Rapid+VL MAA：Asanté 近期快速检测结果+VL>1000 拷贝/ml）。三种 MAA 的平均近期感染持续时间（MDRI）分别为 248、130 和 180 天。
该研究包括：15845 名 HIV 阴性个体；4406 名在两次就诊时均为 HIV 阳性的个体；以及 221 名在两次就诊之间发生血清转换的个体。所有 HIV 阳性参与者在第 2 年就诊时均可获得病毒载量（VL）数据。64 名血清转换者（29%）和 3227 名（72%）流行阳性参与者的病毒受到抑制（<400 拷贝/ml）。观察到的 HIV 发病率为 1.34%（95%CI：1.17-1.53）。MAA-C 和 LAg+VL MAA 的发病率估计值与观察到的发病率相似，分别为 1.26%（95%CI：1.02-1.51）和 1.29%（95%CI：0.97-1.62）。Rapid+VL MAA 估计的发病率明显低于观察到的发病率（0.92%，95%CI：0.69-1.15，p<0.01）。
在病毒抑制水平较高的队列中，MAA-C 和 LAg+VL MAA 提供了发病率的准确点估计值。Rapid+VL 显著低估了发病率，这表明制造商推荐的 MDRI 过长或该检测未能准确检测到足够的近期感染。
