Open AccessCan visual interpretation of NucliSens graphs reduce the need for repeat viral load testing?
Author(s) -
Newten Handireketi,
Collins Timire,
Hemant Deepak Shewade,
Ellen Munemo,
Charles Nyagupe,
Sandra Chipuka,
Lucia Sisya,
Hlanai Gumbo,
Ezekiel Dhitima,
Anthony Harries
Publication year - 2019
Publication title -
plos one
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.99
H-Index - 332
ISSN - 1932-6203
DOI - 10.1371/journal.pone.0223597
Subject(s) - viral load , interpretation (philosophy) , virology , physics , computer science , computational biology , biology , human immunodeficiency virus (hiv) , programming language
Background In Zimbabwe, viral load (VL) testing for people living with HIV on antiretroviral therapy is performed at the National Microbiology Reference Laboratory using a NucliSens machine. Anecdotal evidence has shown that invalid graphs for “Target Not Detectable (TND)” will upon repeat VL testing produce a valid result for virus not detected, therefore removing the need to repeat the test. This needs formal assessment. Objectives To determine i) intra- and inter-rater agreement of the visual interpretation of NucliSens graphs (Target Detectable [TD], TND and No Line [NL]) between two laboratory scientists and ii) sensitivity, specificity and predictive values of the NucliSens graphs compared with repeat VL results. Method Cross sectional study using secondary data. Two laboratory scientists independently rated graphs one week apart for intra-rater agreement and compared final ratings with each other for inter-rater agreement. Consensus interpretations of graphs were compared with repeat VL results. Kappa coefficients were used to obtain measures of agreement. Results There were 562 patients with NucliSens graphs and repeat VL. Kappa scores were: 0.98 (Scientist A); 0.99 (Scientist B); 0.96 (Scientist A versus Scientist B); and 0.65 (NucliSens graphs versus VL). Sensitivity, specificity, positive predictive value and negative predictive value for graphs compared with VL were 71%, 92%, 79% and 89% respectively. Conclusion Intra-and inter-rater agreements were almost perfect. The negative predictive value translates to a false negative rate of 11%. If repeat VL testing is not done, the clinical consequences need to be balanced against cost savings and the risks outweigh the benefits.