Variability in HIV viral load quantification in real-world service delivery settings: findings from an observational cohort study in Rwanda

While developments in HIV viral load (VL) testing technologies have improved detection of low-level viremia, there is substantial variation in the quantitative results of available polymerase chain reaction tests, particularly around assay lower limits of quantification (LLOQ). We aimed to describe testing results for paired specimens from the same participants reported by two laboratories in Rwanda using different assays, characterize discordancy in VL quantification, and identify factors associated with quantifiable VL results at a threshold of 40 copies/mL. In an observational cohort study of people living with HIV (PWH), aged ≥ 40 years enrolled in HIV care, we used two laboratories to process paired research specimens. We used Kappa statistics and plots to examine between-lab agreement at the LLOQs for assays used by the two labs (Lab A: 20 copies/mL for COBAS AmpliPrep/Taqman and Abbott Alinity-m HIV1 assays; Lab B: 40 copies/mL for Abbott RealTime assay and at thresholds of 200, 1000 and 2000 copies/mL. We used Poisson regression to examine participant characteristics independently associated with unsuppressed viral loads (≥ 200 copies/mL). 593 results were reported by both laboratories for 572 unique participants. The mean age of study participants was 54.4 years, and 58% were female. Median time on antiretroviral therapy was 16 years, 93.7% were on dolutegravir-based regimens, and 96.9% had CD4 cell counts of > 200 cells/mm 3 . For Lab A, 29.2% of specimens had quantifiable VLs at assay LLOQs of 20 copies/mL and 22.1% had quantifiable VLs at a common threshold of 40 copies/mL, compared with 4.7% of specimens processed by Lab B (LLOQ 40 copies/mL). Kappa statistics showed poor to moderate between-lab agreement below a threshold of 200 copies/mL, with high agreement at thresholds of 1000 and 2000 copies/mL and differences by assay type. CD4 cell counts < 200 cells/mm 3 were associated with unsuppressed VLs reported by each laboratory. While VL quantification differed substantially between assays and laboratories used in a real-world service delivery setting, there was high agreement at cut-offs generally used in clinical decision-making. Our findings support the use of higher viral suppression cut-offs used in UNAIDS’ 95-95-95 targets because lower thresholds are vulnerable to misclassification.