Classification of HIV virological failure using whole blood versus plasma viral load

Introduction: HIV viral load testing is the preferred monitoring approach for HIV infected patients on combination antiretroviral therapy (cART) as it is more sensitive than CD4 count and clinical monitoring. In resource limited settings, timely plasma separation and transportation to testing laboratories is a major barrier to the access of HIV viral load testing. The 2015 World Health Organisation guidelines recommend that cART should be initiated in all adults and children living with HIV regardless of disease stage or CD4 count, thereby escalating the demand for HIV viral load testing. Potential solutions to expand implementation and scale up of viral load testing in low and middle income countries are whole blood testing through point of care (POC) viral load assays or dried blood spots (DBS) collected at the health facility. Utilization of whole blood instead of plasma would simplify sample collection, storage and transportation requirements and be cost effective. However, the paucity of studies comparing whole blood HIV viral load across different test platforms, especially in the correct classification of virological failure, has resulted in the lack of a standardised programmatic approach to whole blood viral load testing. Methods: We evaluated four HIV whole blood viral load test methods namely Alere q HIV-1/2 POC, Abbott RealTime HIV-1 DBS original and updated protocols, and Roche CAP/CTM DBS free virus elution (FVE) protocol, against the standard of care, plasma viral load, on 299 samples across the viral load spectrum from South African patients on cART. Virological failure was defined at >1000 copies/ml. Proportions of correct classification of virological failure and overall correlation with plasma were used for evaluating each method's performance. Results: Alere q, Abbott original and updated, and Roche FVE correctly classified virological failure in 61%, 89%, 87% and 76% of all samples tested respectively. The performance varied across plasma viral load categories. Alere q showed good correlation above plasma viral load of 1000 copies/ml, with correct classification of virological failure in 100% of samples. However, below the plasma threshold of 1000 copies/ml, Alere q demonstrated significant over-quantification, resulting in reduced specificity and upward misclassification of virological failure in 39% of all samples tested. Abbott original and updated also had good sensitivity of 98% and 91% respectively and the best overall correlation with plasma (r² = 0.76 and 0.72 respectively), but there was upward misclassification in 10% and 8% of samples tested respectively. Roche FVE had the best specificity of 99% but with significantly reduced sensitivity of 53%, especially between 1000–10,000 copies/ml of plasma, resulting in downward misclassification in 24% of all samples tested. Greatest variability between the different testing methods was seen when plasma viral load was 40-1000 copies/ml. Correlation was best for all whole blood viral load assays at >10,000 copies/ml. Conclusion: The key finding highlighted by this study is the great variability between the different whole blood test methods. Various factors influence the ability to quantify whole blood HIV viral load such as input volume used in each assay vary, sample treatment/processing (DBS versus fresh blood samples versus FVE), extraction (RNA selective, total nucleic acid extraction), amplification target and detection methods are different for each of the platforms tested. Based on our study, Alere q and Abbott DBS need to raise their whole blood threshold for virological failure in order to reduce upward misclassification and Roche FVE needs to achieve better sensitivity around its limit of detection. Receiver operating characteristic curve analysis can be used to determine the optimum threshold of virological failure for each assay