Scientific Papers

Effectiveness of the booster dose of inactivated COVID-19 vaccine against Omicron BA.5 infection: a matched cohort study of adult close contacts | Respiratory Research


This was a retrospective cohort study including all adult close contacts of COVID-19 in Urumqi from August 1 to September 7, 2022. The study followed the Strengthening Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines. The collection of specimens, epidemiological and clinical data for SARS-CoV-2 infected individuals and their close contacts is part of a continuing public health investigation of COVID-19 outbreaks, ruled in the Protocol on the Prevention and Control of COVID-19 by the National Health Commission of the People’s Republic of China, which was exempt from ethical approval (i.e., institutional review board assessment). This study was approved by the institutional ethics committee of Xinjiang Medical University. Individual verbal consent was obtained when collecting personal information and human samples by governmental healthcare professionals in the field. All study data were completely anonymized. This study used secondary data without personal identity or human sample provided by the Urumqi Center for Disease Control and Prevention.

Study setting

Mainland China implemented the “zero COVID-19” policy from 2020 to October 2022 (after the end of our study period), and thus, no large-scale COVID-19 outbreak occurred in the context of “zero COVID-19” control measures in Urumqi before August 2022, which means that the population, with a size of 3.8 million, was (largely) infection-naive. The COVID-19 vaccines received by almost all vaccinees in mainland China were Sinopharm vaccines (BBIBP-CorV) and Sinovac vaccines (CoronaVac), which were inactivated COVID-19 vaccines developed and administered under the supervision of Chinese authorities. Among all vaccinated subjects included in this study, BBIBP-CorV and CoronaVac were administered, and the majority were BBIBP-CorV. Since both BBIBP-CorV and CoronaVac were inactivated vaccines with similar contents, vaccine types were not further compared in this study. By the end of July 2022, before the start of the study period, the coverage of 2-dose and booster inactivated vaccines was 90% and > 72% [37], respectively, for the general population of mainland China, which was similar to that in Urumqi city. Most of the non-vaccinees (i.e., those who received 0 dose) in mainland China were those with existing conditions making them unsuitable for receiving vaccines due to medical concerns.

During the period from August 1 to September 7, 2022, the first group of COVID-19 cases in Urumqi, which is an epicenter of the outbreak, was detected. The outbreak was seeded by Omicron BA.5.2 variants (classified using PANGO lineage designation [38]), and the confirmation of these genetic variants was conducted through whole-genome sequencing of 11 randomly selected COVID-19 cases in the initial days of the outbreak. The outbreak in the Xinjiang Uygur Autonomous Region, China, started on August 7 and reached its peak on August 13. According to the “zero COVID-19” policy, a series of intensive control measures were then swiftly implemented by the local government on August 10, including city-wide lockdown, travel ban, mass case detection, symptom-based surveillance, contact tracing, case isolation and contact quarantine. Since the start of the outbreak, mandatory reverse transcription polymerase chain reaction (RT-PCR) tests were administered by the local authority on a daily basis for all citizens in Urumqi (city-wide mass testing). Test-positive individuals and his/her close contacts were immediately quarantined.

All individuals who had an epidemiological link to a laboratory-confirmed COVID-19 case were classified as close contacts of COVID-19. Information on exposure history was collected and documented through interviews with individuals with confirmed COVID-19 cases as well as their digital records of travel history through an online platform (i.e., China’s COVID-tracking QR code downloaded on individual mobile phones). As a major part of the contact tracing program conducted by the city-level Centers for Disease Control and Prevention, the contact history of each individual who was suspected to have exposure risks was linked to COVID-19 cases on a pairwise basis. The epidemiological link was identified for individuals who had unprotected contact [e.g., without sufficient personal protective equipment (PPE)] with a COVID-19 case within 4 days before his or her test-positive date, because a considerable amount of transmission could occur at an early stage after infection.

With large efforts to actively detect SARS-CoV-2 infections in Urumqi, as well as other places in mainland China (before November 2022), mandatory reverse transcription polymerase chain reaction (RT-PCR) tests were administered by the local authority on a daily basis for all close contacts. SARS-CoV-2 infections were laboratory-confirmed by performing RT-PCR tests (cycle threshold [Ct] value < 40) on specimens collected from nasopharyngeal or oropharyngeal swabs.

Study design, participants, and variables

This was a matched cohort study including all adult close contacts of COVID-19 between August 1 and September 7, 2022, in Urumqi, China. For participant selection before matching, we excluded contacts who had received fewer than 2 doses of vaccines because we aimed to study booster vaccination versus 2-dose vaccination among adults. Note that only a small proportion (1.0%) of participants were partially vaccinated (i.e., received 1 dose). Those contacts who had missing information on the date of the last vaccine dose (before exposure) were excluded. Those exposed within 14 days since the last dose were also excluded [31], which was to account for the time lag for vaccines to develop protective effects within human hosts [39]. The participants’ selection procedures are visualized in the flowchart shown in Fig. 1.

Fig. 1
figure 1

Flowchart of samples selection, and subsequent propensity score matching before statistical analyses

For the eligible contacts, we extracted individual-level information, including the age and sex of both the contacts and their linked source cases, contact settings (i.e., household, community, workplace, and unknown settings), timeline-list data of vaccination and exposure history, vaccination status of source cases, and RT-PCR test result for SARS-CoV-2 infection. The vaccination status of close contacts was considered the variable of interest, and we considered 2-dose vaccination as the reference level against booster vaccination. For the outcome, we considered RT-PCR test-positive status for SARS-CoV-2 infection as the primary outcome variable for both asymptomatic and symptomatic infections. As most of the Omicron infections were asymptomatic or mildly symptomatic [32], we also observed that < 10% of adult infections in our cohort were symptomatic.

Propensity score matching

The propensity score was estimated by a multivariate logistic regression model. In this study, close contacts who received (only) 2 doses of vaccine were matched to close contacts who received the booster (i.e., third) dose of vaccine at a ratio of 1:5 using the nearest-neighbor approach with discard, where participants were matched for sex, the age of both index (source) cases and close contacts, the calendar date of contact (in the form of the epidemiological week of 2022), and contact setting strata. These variables were chosen based on possible or known associations with transmission risks or contact patterns to reduce the likelihood of selection bias in vaccination status and transmission risks, and facilitated the comparison between the testing outcomes of SARS-CoV-2 based on vaccination status.

For each variable, the after-matching standardized mean differences (SMD) between 2-dose and 3-dose vaccinees were calculated, and SMD < 0.1 was considered a satisfactory balance of the baseline conditions between the two cohorts [40, 41]. The matched cohort was used to assess the vaccine effectiveness.

Statistical analyses

We stratified the cohort by vaccination status and testing outcome for SARS-CoV-2 infections. The characteristics of the eligible (i.e., before-match) and after-matching cohorts were described with the use of frequency distributions and measures of central tendency.

Using the matched cohort, multivariate conditional logistic regression models were adopted to explore the association between vaccination status and SARS-CoV-2 infection risk among close contacts with COVID-19 in terms of the odds ratio (OR). The vaccine effectiveness (VE) was calculated based on the OR, such that VE = (1 − OR) × 100% when OR < 1; or VE = − (1 − 1/OR) × 100% when OR > 1 [42,43,44]. We controlled for potential confounding variables, including the sex and age of both source cases and contacts, the epidemiological week of contact history, the vaccination status of source cases, and contact settings. We assessed the statistical uncertainty by using the 95% confidence interval (CI). Our reference group was the two-dose group, so the vaccine effectiveness we studied was marginal VE. Although survival analysis with time-varying risk can be applied to the estimation of VE in situations that proportional hazard (PH) assumption was violated [35, 45], we adopted conditional logistic regression models [46, 47], which was conservative for the retrospective data after matching.

Subgroup analyses were performed in which we assessed VE by sex (male and female), age (18–39, 49–60, and ≥ 61 years), time lag from last vaccine dose to contact with COVID-19 cases (15–180, and ≥ 181 days), and the vaccination status of source case (0–1 dose, 2, and 3 doses). For data visualization, we employed the Kaplan–Meier estimator to construct cumulative incidence curves [48]. The hazards of Omicron infection were stratified by the vaccine dose of contacts, and compared using log-rank tests for a statistically significant difference.

All data processing and matching procedures were performed in R statistical software (version 4.1.1) [49], and specifically, propensity score matching was conducted using the package “MatchIt” [50].

Role of the funding sources

The funding sources had no role in the design, conduct, and reporting of the study or in the decision to submit the manuscript for publication.



Source link