4f - Adjusting for seroconversion and migration in the CD4 back-calculation method for HIV incidence estimation

Background : Back-calculation is used to estimate unobserved incidence and undiagnosed HIV prevalence by linking a model of infection and disease progression to observed HIV diagnosis and CD4 count data. The model is implemented annually for gay and bisexual men (GBM) in the United Kingdom (UK) to assess progress towards HIV elimination, and was adapted for use in the ECDC HIV Modelling Tool. Two current limitations of the model are: 1) individuals are increasingly likely to be diagnosed during seroconversion, when a transient dip in CD4 count may occur, so an individual’s progression towards AIDS may be incorrectly classified; and 2) estimates do not account for population migration, so the model is ill-suited to groups with large migrant populations. We aim to address these limitations by incorporating both a recent infection biomarker and migration information into the model.

Methods : The existing (CD4-only) model was extended to a ‘dual-biomarker’ model by the inclusion of recent infection states, informed by serological tests for recent infection (RITA). Both models were applied to diagnosis data for UK GBM between 2011-2019 and compared.

Results : Between 2011-2019, RITA information was available for 59% of HIV diagnoses among GBM. Estimated HIV incidence in 2019 was 1020 (95% credible interval (CrI) 540-1790) using the CD4-only model, and 960 (95% CrI 630-1570) using the dual-biomarker model. Estimates of undiagnosed HIV prevalence from the dual-biomarker model were significantly lower than those from the CD4-only model, particularly during earlier years.

Conclusions Incorporation of additional biomarker information in the back-calculation model improved estimate precision, with undiagnosed prevalence estimates closer to those from other methods. Work is ongoing to include migration probabilities, informed by routine migration statistics, relevant to UK heterosexual populations.