Field performance and cost-effectiveness of a point-of-care triage test for HIV virological failure in Southern Africa

Abstract

Introduction: Antiretroviral therapy (ART) monitoring using viral load (VL) testing is challenging in high-burden, limited-resources settings. Chemokine IP-10 (interferon gamma-induced protein 10) strongly correlates with human immunodeficiency virus (HIV) VL. Its determination could serve to predict virological failure (VF) and to triage patients requiring VL testing. We assessed the field performance of a semi-quantitative IP-10 lateral flow assay (LFA) for VF screening in South Africa, and the cost-effectiveness of its implementation in Mozambique.

Methods: A cross-sectional study was conducted between June and December 2021 in three primary health clinics in the Western Cape. Finger prick capillary blood was collected from adults on ART for ≥1 year for direct application onto the IP-10 LFA (index test) and compared with a plasma VL result ≤1 month prior (reference test). We estimated the area under the receiver operating characteristic curves (AUC), sensitivity and specificity, to evaluate IP-10 LFA prediction of VF (VL>1000 copies/ml). A decision tree model was used to investigate the cost-effectiveness of integrating IP-10 LFA combined with VL testing into the current Mozambican ART monitoring strategy. Averted disability-adjusted life years (DALYs) and HIV acquisitions, and incremental cost-effectiveness ratios were estimated.

Results: Among 209 participants (median age 38 years and 84% female), 18% had VF. Median IP-10 LFA values were higher among individuals with VF compared to those without (24.0 vs. 14.6; p<0.001). The IP-10 LFA predicted VF with an AUC = 0.76 (95% confidence interval (CI) 0.67-0.85), 91.9% sensitivity (95% CI 78.1-98.3) and 35.1% specificity (95% CI 28.0-42.7). Integrating the IP-10 LFA in a setting with 20% VF prevalence and 61% VL testing coverage could save 13.0% of costs and avert 14.9% of DALYs and 55.7% new HIV acquisitions. Furthermore, its introduction was estimated to reduce the total number of routine VL tests required for ART monitoring by up to 68%.

Conclusions: The IP-10 LFA is an effective VF triage test for routine ART monitoring. Combining a highly sensitive, low-cost IP-10 LFA-based screening with targeted VL confirmatory testing could result in significant healthcare quality improvements and cost savings in settings with limited access to VL testing.

Keywords: CXC chemokine IP-10; Southern Africa; cost-effectiveness analysis; human immunodeficiency virus; point-of-care test; therapeutic drug monitoring.

Figure 1

(a) A decision tree model for antiretroviral therapy (ART) monitoring comparing the standard of care of using viral load (VL) testing only in a two‐step algorithm recommended by the World Health Organization with three proposed algorithms where the IP‐10 lateral flow assay (LFA) is integrated combined with VL confirmatory testing: (b) immediately after performing the IP‐10 LFA; (c) three months later after receiving enhanced adherence counselling (EAC); (d) three months later after receiving EAC and after a second positive IP‐10 LFA result. Each branch depicts individual steps of ART monitoring algorithm and is conditional on the previous step. The grey square node represents a decision node, grey circular nodes represent chance nodes, and grey triangles illustrate terminal nodes. PLHIV, people living with HIV; VF, virological failure.

Figure 2

Performance of the IP‐10 lateral flow assay (LFA) reading values (arbitrary units, Cube Reader) models in predicting virological failure (VF). (a) Comparison between the area under the curve (AUC) for univariable and multivariable models. (b) Univariable model cut‐off IP‐10 LFA reading values with their respective sensitivity and specificity values.

Figure 3

Tornado diagram illustrating fluctuations on the incremental cost‐effectiveness ratios (ICERs) between the minimum (low) and maximum (high) values of individual model parameters in univariate sensitivity analysis in a high transmission scenario (1:4). ICERs were estimated comparing Strategy 2c against Strategy 2b. Value for each parameter is substituted one by one. Dark grey colour represents the low value (lower 95% CI bound or minimum), while light grey colour represents the high value (upper bound of the 95% CI or maximum) of each of the parameters. The black vertical solid line represents the baseline value of the ICER (1995 US$/DALY averted). The area shaded dotted grey represents a dominant scenario by Strategy 2c (lower costs and improved outcomes). Parameters that have the highest impact on the model are shown at the top, while the least impact is displayed at the bottom. *Strategy 2b dominated Strategy 2c (lower costs and improved outcomes), so the ICER was not estimated. DALY, disability‐adjusted live years; EAC, enhanced adherence counselling; LFA, lateral flow assay; VL, viral load.