Validation of the Viral Load Testing Criteria - an algorithm for targeted viral load testing in HIV-positive adults receiving antiretroviral therapy.

OBJECTIVES: Restricted capacity for viral load (VL) testing is a major obstacle for antiretroviral therapy (ART) programmes in high-burden regions. Algorithms for targeted VL testing could help allocate laboratory resources rationally. We validated the performance of the Viral Load Testing Criteria (VLTC), an algorithm with satisfactory performance in derivation (sensitivity 91%, specificity 43%). METHODS: HIV-positive adults who had been receiving first-line ART for ≥12 months at three Ethiopian public ART clinics were included. Healthcare providers collected data on variables of the VLTC: current CD4 count, mid-upper arm circumference (MUAC) and self-reported treatment interruption. VL testing was performed in parallel. Performance of the algorithm for identification of patients with VL ≥ 1000 copies/ml was evaluated. RESULTS: Of 562 patients (female 62%, median ART duration 92 months), 33 (6%) had VL ≥ 1000 copies/ml. Sensitivity for the VLTC was 85% (95% CI, 68-95), specificity 60% (95% CI, 55-64), positive predictive value 12% (95% CI, 10-14) and negative predictive value 98% (95% CI, 97-99). Use of the algorithm would reduce the number of VL tests required by 57%. Misclassification occurred in 5/33 (15%) of subjects with VL ≥ 1000 copies/ml. CONCLUSION: In validation, the VLTC performed similarly well as derivation. Use of the VLTC may be considered for targeted VL testing for ART monitoring in high-burden regions.

OBJECTIFS: La capacité restreinte de mesure de la charge virale (CV) constitue un obstacle majeur pour les programmes de traitement antirétroviral (ART) dans les régions à prévalence élevée. Des algorithmes pour des tests ciblés de la CV pourraient aider à allouer les ressources de laboratoire de manière rationnelle. Nous avons validé la performance des critères de mesure de la charge virale (VLTC), un algorithme dont la performance de dérivation est satisfaisante (sensibilité de 91%, spécificité de 43%). MÉTHODES: Des adultes VIH positifs qui recevaient un ART de première ligne depuis au moins 12 mois dans trois cliniques ART publiques éthiopiennes ont été inclus. Les prestataires de soins de santé ont collecté des données sur les variables des VLTC: nombre actuel de CD4, périmètre brachial et interruption de traitement auto-déclarée. La mesure de la CV a été réalisée en parallèle. La performance de l'algorithme pour l'identification des patients avec une CV≥1000 copies/mL a été évaluée. RÉSULTATS: Sur 562 patients (femmes 62%, durée médiane de l'ART 92 mois), 33 (6%) avaient une CV ≥1000 copies/mL. La sensibilité des VLTC était de 85% (IC95%: 68-95), sa spécificité de 60% (IC95%: 55-64), sa valeur prédictive positive de 12% (IC95%: 10-14) et sa valeur prédictive négative de 98% (IC95%: 97-99). L'utilisation de l'algorithme réduirait le nombre de tests de CV requis de 57%. Une mauvaise classification est survenue chez 5/33 (15%) des sujets avec CV ≥1000 copies/ml. CONCLUSION: En validation, les VLTC ont obtenu une performance aussi bonne comme dérivation. L'utilisation des VLTC peut être envisagée pour des mesures ciblées de la CV dans le suivi des ART dans les régions à forte charge de morbidité.

Performance of Galectin-9 for Identification of HIV Viremia in Adults Receiving Antiretroviral Therapy in a Resource-Limited Setting.

BACKGROUND: Targeted viral load (VL) testing has been proposed for antiretroviral treatment (ART) monitoring in resource-limited settings. In this study, we have investigated the performance of the host biomarker galectin-9 (Gal-9), alone and in combination with interferon-γ-inducible protein 10 (IP-10), in identifying individuals at increased likelihood of viremia during ART. SETTING: Cohort of HIV-positive adults receiving ART at Ethiopian health centers. METHODS: We included participants with detectable viremia (VL ≥150 copies/mL) 12 months after starting ART and sex-matched nonviremic controls. Performance to identify individuals with VL ≥1000 copies/mL was determined for Gal-9 and the Gal-9/IP-10 combination, respectively, using receiver operating characteristic (ROC) analysis. RESULTS: Among 191 participants (50.3% women), 46 (24.1%) had VL ≥1000 copies/mL, 23 (12.0%) had 150-999 copies/mL, and 122 (63.9%) had <150 copies/mL. Gal-9 and VL were positively correlated (r s = 0.451, P < 0.001). Sensitivity and specificity for Gal-9 to identify individuals with VL ≥1000 copies/mL were 91.3% (95% CI: 79.2-97.6) and 54.5% (95% CI: 46.0-62.8), respectively. The area under the ROC curve for Gal-9 was 0.810 (95% CI: 0.745-0.875), which was similar to that of the combination of Gal-9 and IP-10 [0.849 (95% CI: 0.792-0.905)]. Assuming 10% prevalence of VL ≥1000 copies/mL, using Gal-9 for targeted VL testing instead of universal VL testing would reduce the number of VL tests from 10 to 5 to identify 1 viremic individual, with misclassification of 1 in 10 viremic individuals. CONCLUSIONS: Gal-9 is a potential screening marker for targeted VL monitoring in ART recipients. Further studies are needed to determine optimal threshold levels.

Interferon-γ-Inducible Protein 10 (IP-10) Kinetics after Antiretroviral Treatment Initiation in Ethiopian Adults with HIV.

Interferon-γ-inducible protein 10 (IP-10) has been suggested as a marker for targeted viral load (VL) monitoring during antiretroviral treatment (ART). We aimed to determine the kinetics of IP-10 during the initial year of ART, with particular regard to the impact of tuberculosis (TB) co-infection on IP-10 secretion. Longitudinal plasma IP-10 levels were quantified in 112 treatment-naive HIV-positive adults at Ethiopian health centers, through enzyme-linked immunosorbent assay (ELISA) using samples obtained before and during the initial 12 months of ART. All participants underwent bacteriological TB investigation before starting ART. In virological responders (VRs; defined as VL < 150 copies/ml with no subsequent VL ≥ 1,000 copies/ml), IP-10 kinetics were analyzed using linear regression models. Among 91/112 (81.3%) participants classified as VRs, 17 (18.7%) had concomitant TB. Median baseline IP-10 was 650 pg/ml (interquartile range [IQR], 428-1,002) in VRs. IP-10 decline was more rapid during the first month of ART (median 306 pg/ml/month) compared with later time intervals (median 7-48 pg/ml/month, P < 0.001 in each comparison). Although VRs with TB had higher IP-10 levels at baseline (median 1106 pg/ml [IQR, 627-1,704]), compared with individuals without TB (median 628 pg/ml [IQR, 391-885]; P = 0.003), the rate of IP-10 decline during ART was similar, regardless of TB-status. During the initial year of ART, IP-10 kinetics followed a biphasic pattern in VRs, with a more rapid decline in the first month of ART compared with later time intervals. Baseline IP-10 was higher in individuals with TB versus individuals without TB, but the kinetics during ART were similar. IMPORTANCE To reach the goal of elimination of HIV as public health threat, access to antiretroviral treatment (ART) has to be further scaled up. To ensure viral suppression in individuals receiving ART, novel and robust systems for treatment monitoring are required. Targeting viral load monitoring to identify individuals at increased likelihood of treatment failure, using screening tools, could be an effective use of limited resources for viral load testing. Interferon-γ-inducible protein 10 (IP-10), a host inflammation mediator, has shown potential for this purpose. Here, we have investigated IP-10 kinetics in Ethiopian adults with HIV during the initial year after ART initiation. IP-10 levels decreased in parallel with viral load during ART, and prevalent tuberculosis at ART initiation did not influence IP-10 kinetics. This study shows satisfactory performance for IP-10 as a surrogate marker for viral load in persons starting ART, with no influence of concomitant tuberculosis.