Using extracted data from Kaplan-Meier curves to compare COVID-19 vaccine efficacy

Although various COVID-19 vaccines have shown efficacy against placebo in randomized clinical trials, no head-to-head comparisons are yet available. This study aims to compare the efficacy of available COVID-19 vaccines. Vaccine trials searched in May 2021 were included. Data were extracted from Kaplan-Meier (KM) curves using the WebPlotDigitizer program for the individual participant (IP) data simulation. A mixed-effect acceleration failure model with log-logistic and Weibull distributions was used to estimate relative effects for individual vaccines as well as grouped by class: inactivated virus, mRNA, and viral vector. Primary studies were considered as the random effect in the model. Hazard ratios (HR) were estimated and compared across vaccine groups. All vaccines were efficacious in lowering symptomatic infection compared to placebo. CoronaVac, Ad26.COV2.S, ChAdOx1 nCoV-19, rAd26/rAd5, WIV04, HB02, and BNT162b2 showed 7.61 (4.50, 12.87), 6.77 (4.08, 11.24), 5.01 (2.93, 8.57), 4.50 (2.52, 8.01), 3.90 (2.04, 7.45), 3.18 (1.62, 6.21), and 2.15 (1.22, 3.78) times significantly higher risk of infection than mRNA-1273. mRNA vaccines were the most efficacious vaccine group compared to inactivated virus and viral vectors with HRs (95% CI) of 0.27 (0.20, 0.37) and 0.28 (0.21, 0.37), respectively. Although all vaccines showed significant protection compared to no vaccination. mRNA vaccines, including mRNA-1273 and BNT162b2, showed the highest efficacy in preventing symptomatic COVID-19 infection. Simulated IP data from the KM curve might allow treatment comparison when there is no primary study comparing active treatments.