Multisite evaluation of point of care CD4 testing in Papua New Guinea

Laboratory-based CD4 monitoring of HIV patients presents challenges in resource limited settings (RLS) including frequent machine breakdown, poor engineering support and limited cold chain and specimen transport logistics. This study assessed the performance of two CD4 tests designed for use in RLS; the Dynal assay and the Alere PIMA test (PIMA). Accuracy of Dynal and PIMA using venous blood was assessed in a centralised laboratory by comparison to BD FACSCount (BD FACS). Dynal had a mean bias of -50.35 cells/µl (r(2) = 0.973, p<0.0001, n = 101) and PIMA -22.43 cells/µl (r(2)= 0.964, p<0.0001, n = 139) compared to BD FACS. Similar results were observed for PIMA operated by clinicians in one urban (n = 117) and two rural clinics (n = 98). Using internal control beads, PIMA precision was 10.34% CV (low bead mean 214.24 cells/µl) and 8.29% (high bead mean 920.73 cells/µl) and similar %CV results were observed external quality assurance (EQA) and replicate patient samples. Dynal did not perform using EQA and no internal controls are supplied by the manufacturer, however duplicate testing of samples resulted in r(2) = 0.961, p<0.0001, mean bias =  -1.44 cells/µl. Using the cut-off of 350 cells/µl compared to BD FACS, PIMA had a sensitivity of 88.85% and specificity of 98.71% and Dynal 88.61% and 100%. A total of 0.44% (2/452) of patient samples were misclassified as "no treat" and 7.30% (33/452) "treat" using PIMA whereas with Dynal 8.91% (9/101) as "treat" and 0% as "no treat". In our setting PIMA was found to be accurate, precise and user-friendly in both laboratory and clinic settings. Dynal performed well in initial centralized laboratory evaluation, however lacks requisite quality control measures, and was technically more difficult to use, making it less suitable for use at lower tiered laboratories.