Vaccine effectiveness of two and three doses of BNT162b2 and CoronaVac against COVID-19 in Hong Kong

Background: Hong Kong maintained extremely low circulation of SARS-CoV-2 until a major community epidemic of Omicron BA.2 starting in January 2022. Both mRNA BNT162b2 (BioNTech/Fosun Pharma) and inactivated CoronaVac (Sinovac) vaccines are widely available, however coverage has remained low in older adults. Vaccine effectiveness in this predominantly infection-naive population is unknown. Methods: We used individual-level case data on mild/moderate, severe/fatal and fatal hospitalized COVID-19 from December 31, 2021 to March 8, 2022, along with census information and coverage data of BNT162b2 and CoronaVac. We used a negative binomial model, adjusting for age and calendar day to estimate vaccine effectiveness of one, two and three dose schedules of both vaccines, and relative effectiveness by number of doses and vaccine type. Findings: A total of 12.7 million vaccine doses were administered in the 7.3 million population of Hong Kong, and we analyzed data from confirmed cases with mild/moderate (N=5,474), severe/fatal (N=5,294) and fatal (N=4,093) COVID-19. Two doses of either vaccine protected against severe disease and death, with higher effectiveness among adults [≥]60 years with BNT162b2 (VE: 88.2%, 95% confidence interval, CI: 84.4%, 91.1%) compared to CoronaVac (VE: 74.1%, 95% CI: 67.8%, 79.2%). Three doses of either vaccine offered very high levels of protection against severe outcomes (VE: 98.1%, 95% CI: 97.1%, 98.8%). Interpretation: Third doses of either BNT162b2 or CoronaVac provide substantial additional protection against severe COVID-19 and should be prioritized, particularly in older adults who received CoronaVac primary schedules. Longer follow-up is needed to assess persistence of different vaccine platforms and schedules.

Interpretation: Third doses of either BNT162b2 or CoronaVac provide substantial 41 additional protection against severe COVID-19 and should be prioritized, particularly in older 42 adults who received CoronaVac primary schedules. Longer follow-up is needed to assess 43 persistence of different vaccine platforms and schedules. 44 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 22, 2022. ;https://doi.org/10.1101https://doi.org/10. /2022 Funding:  Vaccines Evaluation Program, Chinese Center for Disease Control and 45 Prevention 46 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Study design and population 87
We assessed vaccine effectiveness of the BNT162b2 and CoronaVac vaccines using an 88 ecological study design, which has been previously employed to provide estimates of vaccine 89 effectiveness in Israel. 26 The study population consisted of residents of Hong Kong aged 20 90 years and over, where the population with zero, one, two or three doses of either vaccine at 91 risk at a given time was derived using detailed data from the vaccination programme and 92 population census. Information on all laboratory-confirmed SARS-CoV-2 cases in Hong 93 Kong from December 31, 2021 to March 8, 2022 was obtained from nationwide individual 94 level surveillance data provided by the Centre for Health Protection and linked to clinical 95 outcome data provided by the Hong Kong Hospital Authority. 96 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) confirmed COVID-19 cases were admitted to hospitals for isolation and standardized clinical 120 management, regardless of symptom presentation, with their hospitalization records stored in 121 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 22, 2022. ; https://doi.org/10.1101/2022.03.22.22272769 doi: medRxiv preprint the data system managed by Hospital Authority of Hong Kong. After mid-February 2022, 122 due to the large number of incident cases, hospitalisation was reserved for patients with more 123 severe disease, and milder cases were required to isolate at dedicated government quarantine 124 facilities or at home. In the Hospital Authority data system, records of patients' test results, 125 medication and condition changes were documented and integrated into a centralized 126 database from which we extracted relevant information on those experiencing mild/moderate 127 disease prior to February 16, 2022 and severe disease and death at any time. We excluded 128 those with conflicting information in the database, i.e. persons with a worst recorded 129 condition of 'mild' but also experiencing a fatal outcome within hospital. Severe disease was 130 defined as any severe, critical or fatal COVID-19 case (definitions for each in Appendix). Those who received vaccines other than BNT162b2 or CoronaVac, or who received a mixed 144 primary series of one dose of BNT162b2 and one dose of CoronaVac, were excluded from the 145 analysis. In addition, for the purposes of this analysis we also exclude those who switched 146 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted  protection (VE >90%) against severe disease and death. We found no effect of two doses of 240 CoronaVac and a limited effect of BNT162b2 against mild/moderate disease, with the caveat 241 that many individuals had received their second dose several months before exposure to the 242 SARS-CoV-2 virus. Limited protection against mild/moderate disease was restored with third 243 doses for both vaccines, but we were only able to estimate VE for the short period since Hong 244 Kong broadly recommended administration of third vaccine doses, and it is unclear how long 245 this protection will last. 246 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 22, 2022. ;https://doi.org/10.1101https://doi.org/10. /2022 12 247 Although improved effectiveness of a third dose of vaccination was observed against severe 248 outcomes in younger age groups, absolute VE of two doses of vaccination remains high in 249 this age group for both vaccines and the relative effects should be interpreted accordingly. 28  Despite the overall consistency between our results and those presented in other studies, it is 276 possible that VE, particularly against severe outcomes, has been overestimated in our study. 277 Vaccine hesitancy in Hong Kong is highest among the elderly and in individuals with 278 underlying health conditions. 36 In this scenario so-called 'healthy vaccinee bias', by which 279 vaccine recipients are healthier than their unvaccinated peers, may inflate the estimates. 35 280 Although we have accounted for age in the current estimates, a lack of individual-level data 281 on controls mean that this cannot be formally assessed with currently available data. 282 However, our estimates for BNT162b2 and CoronaVac are similar to other studies using 283 alternative designs, and we anticipate the magnitude of overestimation is unlikely to be 284 substantial. 18,33 Employing alternative study designs using unvaccinated cohorts as a 285 comparator group to estimate VE may offer additional problems, as unvaccinated individuals 286 are a small proportion of some age cohorts, in particular younger age groups in Hong Kong, 287 and the characteristics of those individuals are likely to differ substantially to those 288 vaccinated. This bias, inherent to observational studies, is present in much of the existing VE 289 literature at this stage of the pandemic. To address this concern, we also estimated a relative 290 VE of three versus two doses of each vaccine type, as these cohorts are likely to be more 291 comparable (Table 3). We find a third dose of either vaccine provides additional protection, 292 reiterating the public health value of a third dose for minimizing severe disease and death but 293 also for reducing health system congestion, public concern and indirect costs stemming from 294 milder episodes during a COVID-19 epidemic. 295 296 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 22, 2022. ;https://doi.org/10.1101https://doi.org/10. /2022 We compared performance of the mRNA BNT162b2 and inactivated CoronaVac vaccines 297 and found higher VE for BNT162b2 following one and two doses, but similar performance 298 after three doses (Table 2). Our estimates are likely to be affected by time since vaccination, 299 where typically more time has passed since administration of second than third doses which 300 have only been widely available in Hong Kong since the beginning of January 2022 (Table  301 1). Improved effectiveness may partially reflect a recent, rather than a third, vaccine dose. 302 This hypothesis is supported by data from an observational study in Malaysia which 303 compared the duration of protection of the BNT162b2 and CoronaVac vaccines. They find 304 more rapid waning of CoronaVac, in particular for mild/moderate and severe outcomes, but 305 to a lesser extent for COVID-19 related mortality. 23 Moreover, a recent study of humoral and 306 cellular responses among Hong Kong vaccinees over time found that neutralising antibodies 307 against variants of concern dropped to detection limit only three months after vaccinations, 308 along with diminishing memory T cell responses, primarily among CoronaVac recipients. 36 309 310 Our study has a number of limitations arising from available data and the nature of the 311 epidemic within Hong Kong. Firstly, we used census data from the correct time period to 312 construct the source population, and any differential population movement by vaccine status 313 over the duration of the vaccination program could affect the validity of our estimates. As we 314 are estimating vaccine effectiveness in real-time, there are large amounts of missingness in 315 clinical data, which is especially problematic when assuming a population level denominator, 316 as the assumed number of people still at risk will be overestimated. However, this is mostly 317 an issue for mild/moderate outcomes, as we used complete records on COVID-19 mortality 318 to derive estimates and we expect severe cases are fully documented. Secondly, there are 319 some differences in testing requirements by vaccine status, particularly for those required to 320 regularly test because of occupation. However, we expect that VE estimates against severe 321 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 22, 2022. ;https://doi.org/10.1101https://doi.org/10. /2022 15 outcomes will be only marginally susceptible to biases related to testing requirements. 322 Finally, in Hong Kong there was a clear preference for the BNT162b2 vaccine in younger age 323 groups and for CoronaVac in older adults. We have addressed this confounding in estimates 324 presented by stratifying by age categories and adjusting estimates by 10-year age categories 325 and calendar day, however some residual confounding by age is possible in the vaccine 326 platform-specific estimates and other factors may confound the relationship between vaccine 327 status, type and risk of infection that cannot be accounted for in this design. 328

329
Our findings indicate that two dose schedules of both BNT162b2 and CoronaVac vaccines 330 offer strong protection against severe disease and death outcomes, however higher levels of 331 protection were observed among those who received two doses of BNT162b2 compared to 332 those receiving two doses of CoronaVac, particularly in older age groups. Three recent doses 333 of both vaccines offer very high levels of protection for older adults against severe outcomes, 334 with no differences observed across vaccine types. It will be important to increase uptake of 335 third vaccine doses, particularly in older adults who have so far received two doses of 336 CoronaVac. Further investigation of the durability of protection provided by both vaccines is 337 warranted and planned. 338 339 340 341 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2022. ;https://doi.org/10.1101https://doi.org/10. /2022  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2022. ;https://doi.org/10.1101https://doi.org/10. /2022  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2022. ;https://doi.org/10.1101https://doi.org/10. /2022

Fatal cases
Vaccine type BNT162b2 CoronaVac Unvaccinated . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2022. ;https://doi.org/10.1101https://doi.org/10. /2022