<p indent="0mm">The clinical trial of vaccine efficacy is designed to compare the difference of incidence rate of disease or infections between vaccine group and placebo group, and to calculate the vaccine protection rate. Up to now, 22 COVID-19 vaccine candidates were under evaluation in phase III clinical trials, with the first two mRNA vaccines Moderna, and Pfizer/BioNTech announced efficacy of 94.1% and 95.0% for symptomatic COVID-19. However, several doubts concerning these results were raised, particularly by Peter Doshi, who speculated that a real efficacy of Pfizer/BioNTech’s mRNA vaccine could be as low as 29%. The debates around the efficacy of the COVID-19 vaccines became hot than ever in China, which might lead to increased vaccine skepticism, vaccination hesitation and undermine public confidence. Therefore, we need to have a better understanding on the efficacy data and interpret the results correctly. The selection of appropriate clinical endpoints with specific definition is critical. Hierarchical endpoints are often used in vaccine clinical trials to comprehensively evaluate the protective efficacy. Definition of hierarchical endpoints on basis of WHO clinical progression scale for COVID-19 could help to evaluate vaccine efficacy against COVID-19 with different severity, and compare the data from different trials. In general, vaccine efficacy is expected to be higher for the clinical endpoints with higher severity. In addition, both sensitivity and specificity of the disease surveillance system could have impacts on the vaccine efficacy. Specificity is more important than sensitivity, especially in the case of low incidence rate of the endpoint, which has a great impact on the protective efficacy of the vaccine, and often leads to the underestimation of vaccine efficacy. We compared phase III efficacy trial protocols of COVID-19 vaccine candidates sponsored by Moderna, Pfizer/BioNTech, AstraZeneca/University of Oxford and Johnson & Johnson, according to these impact factors mentioned above. Comparing with Moderna, AstraZeneca/University of Oxford and Johnson & Johnson’s trials to adopt a sensitive definition of suspected COVID-19 cases, Pfizer/BioNTech’s surveillance for diseases was mostly based on a passive surveillance involving self-report and self-sampling for case screening, which could compromise the sensitivity of the surveillance system capturing suspected COVID-19 cases. In addition, diagnostic testing to confirm the causative pathogen is important. However, the sensitivity of quantitative RT-PCR, the current goldstandard assay for diagnosis of SARS-CoV-2, is imperfect and influenced by viral load, sample type, and timing. Thus, collecting more than one specimen from suspected cases at more than one time points during the disease course, could results in a better sensitivity as well as specificity. Importantly, the serologic tests of antibody targeting at the N protein of SARS-CoV-2 could be used to identify whether it is naturally infected by SARS-CoV-2. At last, sharing and opening of trial protocols and the disclosure of trial data will contribute to better and more scientific evaluation of the efficacy of COVID-19 vaccine candidates. Extending the period of case surveillance and accumulating data on the protective efficacy of COVID-19 vaccines can contribute to understand the true protective efficacy of the vaccine.