Bias in control selection associated with the use of rapid tests in influenza vaccine effectiveness studies

• Published in: International journal of epidemiology Vol. 54; no. 4
• Main Authors: Poukka, Eero, Murphy, Caitriona, Mak, Loretta, Cheng, Samuel M S, Peiris, Malik, Tsang, Tim K, Sullivan, Sheena G, Cowling, Benjamin J
• Format: Journal Article
• Language: English
• Published: England 11.06.2025
• Subjects: Adolescent; Adult; Aged; Bias; Case-Control Studies; Child; Child, Preschool; COVID-19 - prevention & control; Female; Hong Kong - epidemiology; Humans; Infant; Influenza B virus; Influenza Vaccines - administration & dosage; Influenza Vaccines - immunology; Influenza, Human - diagnosis; Influenza, Human - prevention & control; Male; Middle Aged; SARS-CoV-2; Selection Bias; Vaccine Efficacy; Young Adult; Hong Kong; influenza; vaccine effectiveness; test-negative design; rapid test; PCR test; bias
• ISSN: 1464-3685; 1464-3685
• DOI: 10.1093/ije/dyaf089

In test-negative design (TND) studies that use rapid tests to estimate influenza vaccine effectiveness (VE), a common concern is case-control misclassification due to imperfect diagnostic accuracy. An imperfect test can also fail to exclude from the control group people infected with other influenza types or other vaccine-preventable respiratory viruses. We investigated these biases while evaluating VE for the 2023/24 seasonal influenza vaccination. A TND study was conducted among outpatients aged ≥6 months of age who visited an outpatient clinic in Hong Kong between 15 December 2023 and 13 August 2024. VE was estimated for polymerase chain reaction (PCR)- and rapid-test-confirmed influenza A and B with exclusions of other types of influenza and SARS-CoV-2 based on either PCR or rapid test. Alternatively, for the exclusion of SARS-CoV-2, we adjusted the analysis with COVID-19 vaccination. VE was estimated by using logistic regression adjusted for confounders. In a study population of 1691 participants, VE against influenza A by PCR was 49% [95% confidence interval (CI) 26%-65%] after the exclusion of PCR-confirmed influenza B and SARS-CoV-2 controls. The corresponding VE against influenza B was 65% (95% CI 35%-81%). VE estimated by adjusting for COVID-19 vaccination status yielded similar estimates to these. When case-control status and exclusions from test-negative controls were determined by using the rapid test, VE was reduced by 5%age-15%age points. In TND studies using rapid tests, reduced sensitivity affects both case-control classification and control group exclusions, potentially causing bias. New methods for these biases could help to adapt participatory cohorts for the monitoring of VE for influenza, COVID-19, and respiratory syncytial virus.