We aimed to determine whether primary vaccination with extended dosing intervals provides higher protection than regular intervals for adolescents aged 12–17 receiving mRNA vaccines. Specifically, we examined 1) whether primary vaccination with an extended dosing interval provides higher protection against infections after receiving the second dose compared to primary vaccination with a regular dosing interval; and 2) whether the increased risk during the inter-dose period due to an extended dosing interval may counterbalance the additional protection gained from extending the dosing interval. Analysis was restricted to the age group because > 99% of children aged 0–11 receiving mRNA vaccines with the interval between first dose and second dose ≥ 56 days.
Study population and vaccination eligibility
Hong Kong had a population of 7.4 million people, including 389,400 adolescents aged 12–17 years at the end of 2021 (30). Adolescents became eligible for COVID-19 vaccination on 4 April 2021, and were able to receive CoronaVac (Sinovac Biotech) and Comirnaty® (BNT162b2, Pfizer-BioNTech) vaccines. Initially, the recommended dosing interval was 21 days. To reduce the myocarditis risk associated with mRNA COVID-19 vaccine, the recommended dosing interval was extended to 84 days from 23 December 2021. The recommended dosing interval was reduced to 56 days from 17 June 2022 (11).
Data sources
We obtained a COVID-19 database from the Center of Health Protection in Hong Kong, which included vaccination records and case details linked by unique identifiers. Both datasets contained demographic and relevant medical information, including date of birth, sex, and underlying conditions. We excluded adolescents with underlying conditions to avoid potential confounding associated with preferential vaccination for people with underlying conditions.
The vaccination dataset included individuals with any recorded vaccination, detailing the date and type of vaccine for each dose. We assumed that each vaccine dose required 14 days to become effective. Those vaccinated within 14 days of SARS-CoV-2 notification were considered unvaccinated; i.e. observation time commenced 14 days after the second dose. Individuals infected within 14 days after dose 2 vaccination were excluded in the analysis. Adolescents were considered to have received a regular dosing schedule if the interval between their first and second dose was 21–27 days, while an extended interval was assigned if doses were received more than or equal to 28 days apart.
The case dataset included all confirmed COVID-19 cases (PCR or RAT) in Hong Kong, with their notification date, hospitalization outcome, and mortality outcome (all-cause or COVID-related). Cases detected prior to 1 January 2022 were excluded from the analysis because all were infected by ancestral strains in 2020–2021 and the incidence rate was very low (12,631 cases among 7 million residents) (31, 32).
Observed relative vaccine effectiveness of extended versus regular dosing interval since second dose
Statistical analyses were performed using R version 4.0.5 (R Foundation for Statistical Computing, Vienna, Austria). Individuals infected prior to receipt of their second dose during the Omicron outbreaks were excluded due to potential hybrid immunity (33). Data were analyzed to estimate the relative hazard of infection among adolescents receiving vaccination according to an extended versus regular dosing schedule. Analyses were restricted to the period 1 January to April 30, 2022, marking the first Omicron wave and prior to the introduction of new circulating Omicron variants in May 2022. During this period, all RAT-confirmed cases required a mandatory confirmatory PCR test. We assumed that each individual could only be infected once during the fifth wave between January and April 2022 (17, 18). In all analysis, we assumed each vaccine dose required 14 days to be effective.
We employed a Cox proportional hazard model to estimate the hazard ratio (HR) of infection for extended versus regular dosing intervals. To account for the varying infection risk during an epidemic the time-to-event was based on calendar time, with observation time commencing 1 January 2022 and ending on the date of case notification or 30 April 2022. Individuals started to contribute person-time to the analysis 14 days after receiving their second dose and after 1 January 2022 (34). The exposure was the vaccination dosing interval, dichotomized to regular (21–27 days) and extended (28 + days). In sensitivity analyses the threshold was varied to 56 days (see below). To account for waning VE, time since vaccination was included as a time-varying term, calculated as the number of days since 14 days post-second-dose. Other variables included were age and sex, which were treated as time-independent variables.
Unvaccinated adolescents were not considered, therefore VE estimates of primary series were therefore projected backward to day 0 based on the estimated waning rate. Primary series VE was derived from the relative risk of infection at 14 days after vaccination (r1), versus the risk of infection at the end of VE waning (r2); i.e. VE = 1-r1/r2 (Fig. 3). The risk of infection was assumed to increase over time consistent with waning VE, such that r1 < r2. We tested the impact on different assumptions on primary series VE and VE waning, or allowed them to be estimated from data. In our main analysis, when the end day of waning was set to Y days, the assumed function of waning was f(X) = minimum(X,Y), and put into the regression, so that the infection risk increased log-linearly from r1 to r2 in Y days, and protection could reach 0% by 90 or 180 days and then stay at zero thereafter. This regression coefficient was estimated, and therefore when it was estimated to be positive, the infection risk was increasing since second dose and hence the primary series VE was positive, and vice versa. When waning was incorporated into the model as a time-varying linear term defined by the number of days since second dose with an assumed function of f(X) = X. There was no upper bound to the number of days since second dose; i.e. waning was assumed to never end and decline beyond zero to negative values (replicating the assumptions of Lai et al. (11)).
VE was estimated from the model at different assumptions about waning, where protection was assumed to wane to 0% perpetually or within 90 or 180 days, since previous studies suggest that VE wanes to negligible around 90–180 days (35–39). Waning after 90 days in the regular dosing interval group and 180 days in the extended dosing interval group was also modeled, to test the potential of reducing waning rate from extended dosing intervals. In addition, we explored the effect of waning when VE was set to 40% or 25%. We conducted sex-specific analyses and sensitivity analyses. We varied the threshold for the dosing interval to 56 days instead of 28 days and excluded participants with extreme dosing intervals (> 100 days).
We tested a case-density sampling approach, which allows cases to be selected as controls during their period at risk (i.e. prior to infection) in a matching analysis. This approach can accommodate the time-varying infection risk observed during epidemics. The case-to-control ratio was 1:4, matched by age and sex. In these simulations, we examined whether the HR could be reliably approximated by the odds ratio (OR). Two conditional logistic regression models were explored. In the first, the data were additionally matched by the date of the second vaccine dose, which more closely resembles density sampling. In the second, days since the second dose was included as an unmatched covariate and incorporated into the model as a time-varying linear term, as used by Lai et al (11).
To assess the impact of approaches or assumptions on primary series VE or duration of protection in estimating the protection of extended versus regular dosing intervals, we developed a simulation model to test different estimation approaches to determine if they could provide unbiased estimates (SI appendix). This model described infection risk since the first dose, assuming risk was proportional to community case numbers. We then excluded individuals with infections before their second dose, mimicking the construction of the real dataset. We tested the true value of HR = 1 and 0.85, corresponding to no effect and moderate effect of extended versus regular dosing interval.
Observed relative vaccine effectiveness of extended versus regular dosing interval since first dose
Restricting the comparison of extended versus regular dosing intervals to the infection risk since the second dose ignores the potential increased risk of infection during the inter-dose interval. Therefore, we used the same calendar-time proportional hazard model to evaluate the impact of increased risk of infection during the inter-dose period. In this analysis, individuals infected prior to receipt of their first dose during the Omicron outbreaks were excluded due to potential hybrid immunity (33). We compared the infection risk for adolescents who received vaccination 21–27 days since first dose (regular dosing interval) versus those were not, including those receive second dose 28 days or more after first dose (extended dosing interval), or who did not. Two ranges of intervals were examined: 1) 42–98 days, consistent with the recommendation to separate doses by 84 days from 23 December 2021 and adopted during the study period, and 2) 42–70 days after first dose, corresponding to a 28-day inter-dose interval + 14 days to allow for seroconversion, and a 56 days interval, which was the interval recommended by the Hong Kong government from 17 June 2022.
Comparison of infection risk since first dose instead of second dose
Given that almost all adolescents (98%) who received vaccination in 2022 had an extended dosing interval and the Omicron outbreak in Hong Kong also started in January 2022, comparing the risk of infection since first dose among adolescents with regular and extended dosing intervals in the real dataset may not be robust. Therefore, we further used the simulation to compare the risk of infection during the inter-dose period for the extended and regular dosing groups. In the simulation, the VEs of primary series were ranged from 0–45%, the HR of infection of extended versus regular dosing intervals ranged from 0.6 to 1, and duration of protection were set to 90 days in both vaccination groups. In a sensitivity analysis the duration of protection was extended to 180 days for the extended dosing group.