Statistical considerations for cross‐sectional HIV incidence estimation based on recency test

Abstract

Longitudinal cohorts to determine the incidence of HIV infection are logistically challenging, so researchers have sought alternative strategies. Recency test methods use biomarker profiles of HIV-infected subjects in a cross-sectional sample to infer whether they are "recently" infected and to estimate incidence in the population. Two main estimators have been used in practice: one that assumes a recency test is perfectly specific, and another that allows for false-recent results. To date, these commonly used estimators have not been rigorously studied with respect to their assumptions and statistical properties. In this article, we present a theoretical framework with which to understand these estimators and interrogate their assumptions, and perform a simulation study and data analysis to assess the performance of these estimators under realistic HIV epidemiological dynamics. We find that the snapshot estimator and the adjusted estimator perform well when their corresponding assumptions hold. When assumptions on constant incidence and recency test characteristics fail to hold, the adjusted estimator is more robust than the snapshot estimator. We conclude with recommendations for the use of these estimators in practice and a discussion of future methodological developments to improve HIV incidence estimation via recency test.

Recent HIV Infection: Diagnosis and Public Health Implications

“…Incidence can be calculated using the following simple estimator (snapshot): I = R / (μ × N), where R is the number of HIV-positive and recently infected people (based on a recency test), N represents HIV-negative people, and μ is the average duration of infection among recently infected people [ 124 , 125 ]. It is suggested that the snapshot estimator is unbiased when the incidence does not change over the period of time that precedes a cross-sectionally collected sample.…”

“…It is suggested that the snapshot estimator is unbiased when the incidence does not change over the period of time that precedes a cross-sectionally collected sample. Moreover, it assumes that the duration-specific test-recent probability function goes to zero, indicating zero test-recent probability for a subject infected long enough [ 125 ]. On most occasions, however, FRR is not 0, showing variability across demographic and epidemiological contexts [ 23 ].…”

“…In other words, the false-recent test probability is not associated with the duration of the infection. The assumptions and statistical properties of the two above-mentioned estimators have recently been described in detail [ 125 ]; when the assumptions on the constant incidence and recency test characteristics do not hold, the adjusted estimator is more robust than the snapshot estimator.…”

Fusing trial data for treatment comparisons: Single vs multi‐span bridging

“…With placebo counterfactual methods, a standardized estimate of the outcome of interest in the placebo arm of a trial is obtained for comparison with an active control in a different trial. [4][5][6][7][8] While this method can account for some differences in trial populations, covariate adjustments are limited to a few discrete covariates. Bayesian methods have also been proposed to estimate placebo counterfactuals but require strong assumptions from expert knowledge or prior data.…”

“…More recently, researchers have estimated placebo counterfactuals, where a counterfactual refers to what an individual's outcome would have been had they been assigned to a placebo rather than active treatment. With placebo counterfactual methods, a standardized estimate of the outcome of interest in the placebo arm of a trial is obtained for comparison with an active control in a different trial 4‐8 . While this method can account for some differences in trial populations, covariate adjustments are limited to a few discrete covariates.…”