Impact of time zero designation on estimated COVID-19 antiviral effectiveness in observational studies

In a well-designed clinical trial, time zero is when eligibility is determined, treatment is assigned, follow-up time begins, and each of these elements is aligned. Attaining this alignment can be challenging in observational studies, risking potential bias. We compared the impact of different time zero designations on the estimated effectiveness of nirmatrelvir-ritonavir for COVID-19. We identified US veterans who tested positive for SARS-CoV-2 from April 2022 to March 2023 and compared nirmatrelvir-ritonavir vs no treatment using 5 time zero approaches: (1a) test-date (treated) vs test-date (untreated) allowing treatment on days 0-5 with matching, (1b) day 0 only with matching, or (1c) days 0-5 with a clone-censor-weight method; (2) treatment date vs test-date with matching; or (3) treatment date vs matched index date. Thirty-day incidence of hospitalization or death was lower in the nirmatrelvir-ritonavir group than the no treatment group for all time zero approaches. Estimated risk differences (95% CI) were larger for approaches 1a (-2.10% [-2.35 to -1.86]), 1b (-2.03% [-2.40 to -1.84]), and 2 (-2.26% [-2.47 to -2.02]); -1.80% (-1.89 to -1.45) for approach 3; and lowest for approach 1c (-0.95% [-1.11 to -0.75]). Different time zero designations can influence effect estimates and should be carefully considered when designing pharmacoepidemiology studies.