Innovative use of self-controlled methods for the evaluation of waning effectiveness of the COVID-19 monovalent third dose: comparison with a test-negative design.

Background: Waning vaccine effectiveness (VE) evaluations are affected by multiple biases, many of which could be implicitly adjusted by using self-controlled methods.

Methods: We used a self-controlled risk interval (SCRI) design to evaluate waning COVID-19 monovalent third dose effectiveness against laboratory-confirmed SARS-CoV-2 medically-attended infections (cases) and hospitalizations with COVID-19 discharge diagnosis among Veterans Health Administration enrollees who received a third dose 15 December 2021-31 August 2022. We used weeks 3-8 post-third dose as reference interval, representing the period of peak VE. We classified cases/hospitalizations occurring after this window into separate weekly risk intervals until 34 weeks post-third dose, using logistic regression to compare outcome risks in each interval with reference interval, adding an offset term to adjust for time-varying viral circulation. For comparison, we conducted a test-negative design (TND) analysis with similar criteria.

Results: For the SCRI analysis, among 609,612 qualifying veterans who received a third dose, we identified 15,224 SARS-CoV-2 eligible cases. Short-term third dose VE relative to primary series was 47 % (95 %CI 41, 50). Waning of protection was evident by week 12 (incidence rate ratio [IRR] 1.5, 95 %CI 1.3, 1.7). By week 20, the IRR (1.9; 95 %CI 1.7, 2.1) was similar to that immediately after the third dose. We identified 3714 hospitalizations with a COVID-19 discharge diagnosis. At weeks 15-16 post-third dose and thereafter, the IRR was consistently above the null. IRR reached 2.3 (95 %CI 2.0, 2.7) by weeks 33-34 post-third dose. Similarly, the TND analysis showed short-term VE of 53 % (95 %CI 51, 55) and waning by week 13 (odds ratio 1.2, 95 %CI 1.0, 1.3).

Conclusion: Protection against medically-attended infection waned within four months, while protection against hospitalization lasted longer. This demonstrates value of self-controlled methods as tools to evaluate waning of VE with observational data.