Ron Brookmeyer1. 1. Woodrow Wilson School of Public and International Affairs, Princeton University, Princeton, New Jersey, USA. rbrook@jhsph.edu
Abstract
OBJECTIVE: To evaluate adjustment procedures that have been proposed to correct HIV incidence rates derived from cross-sectional surveys of biomarkers (e.g. BED capture enzyme immunoassay). These procedures were motivated by some reports that the biomarker BED approach overestimates incidence when compared to cohort studies. DESIGN: Consideration of the Hargrove and McDougal adjustment procedures that adjust biomarker estimates of HIV incidence rates for misclassification with respect to the timing of infections. METHODS: : Performed mathematical and statistical analysis of the adjustment formulas. Evaluated sources of error in cohort studies of incidence that could also explain discrepancies between cohort and biomarker estimates. RESULTS: The McDougal adjustment has no net effect on the estimate of HIV incidence because false positives exactly counterbalance false negatives. The Hargrove adjustment has a mathematical error that can cause significant underestimation of HIV incidence rates, especially if there is a large pool of prevalent long-standing infections. CONCLUSION: The two adjustment procedures of biomarker incidence estimates evaluated here that purport to correct for misclassification do not increase accuracy and in some situations can introduce significant bias. Instead, the accuracy of biomarker estimates can be increased through improvements in the estimates of the mean window period of the populations under study and the representativeness of the cross-sectional samples. Cohort estimates of incidence are also subject to important sources of error and should not blindly be considered the gold standard for assessing the validity of biomarker estimates.
OBJECTIVE: To evaluate adjustment procedures that have been proposed to correct HIV incidence rates derived from cross-sectional surveys of biomarkers (e.g. BED capture enzyme immunoassay). These procedures were motivated by some reports that the biomarker BED approach overestimates incidence when compared to cohort studies. DESIGN: Consideration of the Hargrove and McDougal adjustment procedures that adjust biomarker estimates of HIV incidence rates for misclassification with respect to the timing of infections. METHODS: : Performed mathematical and statistical analysis of the adjustment formulas. Evaluated sources of error in cohort studies of incidence that could also explain discrepancies between cohort and biomarker estimates. RESULTS: The McDougal adjustment has no net effect on the estimate of HIV incidence because false positives exactly counterbalance false negatives. The Hargrove adjustment has a mathematical error that can cause significant underestimation of HIV incidence rates, especially if there is a large pool of prevalent long-standing infections. CONCLUSION: The two adjustment procedures of biomarker incidence estimates evaluated here that purport to correct for misclassification do not increase accuracy and in some situations can introduce significant bias. Instead, the accuracy of biomarker estimates can be increased through improvements in the estimates of the mean window period of the populations under study and the representativeness of the cross-sectional samples. Cohort estimates of incidence are also subject to important sources of error and should not blindly be considered the gold standard for assessing the validity of biomarker estimates.
Authors: Song Duan; Sheng Shen; Marc Bulterys; Yujiang Jia; Yuecheng Yang; Lifeng Xiang; Fei Tian; Lin Lu; Yao Xiao; Minjie Wang; Manhong Jia; Huazhou Jiang; Sten H Vermund; Yan Jiang Journal: BMC Public Health Date: 2010-04-07 Impact factor: 3.295
Authors: Carlos A Velasco de Castro; Beatriz Grinsztejn; Valdiléa G Veloso; Francisco I Bastos; José H Pilotto; Mariza G Morgado Journal: BMC Infect Dis Date: 2010-07-28 Impact factor: 3.090