Direct comparison of three residual risk models for hepatitis B virus window period infections using updated input parameters. Issue 3 (20th January 2020)

• Record Type: Journal Article
• Authors: Lelie, Nico; Vermeulen, Marion; van Drimmelen, Harry; Coleman, Charl; Bruhn, Roberta; Reddy, Ravi; Busch, Michael; Kleinman, Steve
• Is Part Of: Vox sanguinis. Volume 115:Issue 3(2020)
• Page Start: 133
• Page End: 145
• Publication Date: 2020-01-20
• Subjects: NAT testing -- residual risk estimation -- serological testing; Blood -- Periodicals; Blood -- Transfusion -- Periodicals; Immunohematology -- Periodicals; Immunopathology -- Periodicals; 615.39
• DOI: 10.1111/vox.12889
• Languages: English
• ISSNs: 0042-9007

Abstract

Abstract: Background and Objectives: Comparison of two models for estimating residual transfusion transmission risk by NAT screened window period (WP) donations in South African repeat donors gave identical results for HIV but not for HBV. In order to understand discrepant HBV modelling outcomes, the values of input parameters in three HBV WP risk models were reviewed and subsequently applied to the same South African screening data generated by HBsAg PRISM and two NAT assays (Ultrio and Ultrio Plus). Two of the models were also compared using individual donation (ID)‐NAT screening data from different geographical regions. Methods: Values of input parameters were derived from two published data sources and used in three risk models [(1) the incidence rate–WP risk day equivalent model, (2) the NAT yield WP ratio model and (3) the anti‐HBc‐negative HBsAg yield period ratio model] and subsequently applied to the same ID‐NAT screening data. Results: The HBV WP transmission risk in South African repeat donations during a one‐year Ultrio Plus NAT screening period was estimated as 22, 43 and 17 per million, respectively, for the three models, as compared to 56, 117 and 48 per million for HBsAg PRISM screening. The approximate two‐fold higher estimate calculated with the NAT yield WP ratio model was corroborated in repeat donations from three of four regions in a multi‐regional study. When another set of model input values (with shorter viraemia periods and a higher proportion of acute occult infections) was applied to the South African screening data, the relative difference in risk estimates between the three models became smaller. Conclusions: Window period risk modelling for HBV is more complex than for HIV. Multiple factors affect the modelling outcomes. These include the values used for the length of transient HBsAg and HBV‐DNA‐positive phases, the proportion of acute occult and vaccine breakthrough infections and the assumption of random appearance of donors throughout the entire acute resolving infection phase. A substantial proportion of HBV WP NAT yields have very low viral load and lack donor follow‐up data calling into question their definitive classification into the early acute (infectious) replication stage. Since these possible WP NAT yields most highly impact the NAT yield WP ratio model, we recommend relying on the more conservative estimates of the incidence rate–WP risk day equivalent model.