Setting and data
We used data collected between January 3 and February 18, 2022, when the omicron variant was predominant in Israel,13 emulate a target trial evaluating the efficacy of a fourth dose of vaccine versus three doses of vaccine. We analyzed data from Clalit Health Services (CHS), the largest integrated payer-provider healthcare organization in Israel. With over 4.7 million members, CHS covers more than half of Israel’s population. The CHS population is broadly representative of the general Israeli population.14.15 CHS health records have been fully digitized since 2000, and its data repositories include demographic, diagnostic, pharmacological, laboratory, procedural, imaging, and hospitalization data. Data on SARS-CoV-2 (polymerase-chain-reaction) infections [PCR] and antigen tests) and Covid-19 results (including hospitalizations, serious illnesses and deaths) are stored centrally by the Israeli Ministry of Health and delivered daily to the four national health organizations.
This study was approved by the CHS Institutional Review Board. A waiver of the informed consent requirement was granted. The authors guarantee the accuracy and completeness of the data contained in this report.
We included people who, at baseline (defined below), were 60 years of age or older, had been a member of the CHS for at least 1 year, and were eligible to receive the fourth dose of vaccine at any time during the period. study (i.e., had been vaccinated with a third dose of BNT162b2 at least 4 months earlier16) and had no prior PCR-confirmed SARS-CoV-2 infection. As in previous studies,17-19 we also excluded healthcare workers, people in long-term care facilities, people who are housebound, and people who had interacted with the healthcare system (for example, having seen a doctor or had blood tests ) in the previous 3 days. This latter exclusion criterion reduces the likelihood that people who chose to delay receiving a fourth dose of vaccine because they were feeling unwell (possibly with symptoms of Covid-19) would be included in the control group. . Given the rarity of missing data in the ECS dataset (<1%), we also excluded individuals with missing data on body mass index (BMI), area of population or area of residence. A detailed description of all study variables is provided in Supplementary Appendix Table S1, available with the full text of this article on NEJM.org.
We looked at five outcomes: PCR-confirmed SARS-CoV-2 infection, symptomatic Covid-19, Covid-19-related hospitalization, severe Covid-19 (defined by National Institutes of Health criteria), and Covid-related death -19. All outcomes were assessed over two follow-up periods of interest: days 7-30 after the fourth dose and days 14-30 after the fourth dose. In addition, to estimate the progressive accumulation of immunity and to assess the similarity of the study groups during the first days following vaccination (the negative control period20), PCR-confirmed infection was also assessed separately during each follow-up day.
The study design of the primary analysis was similar to that used in our previous vaccine efficacy studies,17.19 who examined the same population in a similar environment. On each day of the study period, eligible individuals who received the fourth dose of the BNT162b2 mRNA vaccine that day (four-dose group) were exactly matched to eligible individuals who had not yet received a fourth dose. to date (control group) according to a set of potential confounders: age (classified in 1-year bins), gender, area of residence, sector of the population (three categories: Arab, generalist Jew and ultra-Orthodox Jew) , calendar month in which each person received the third dose of vaccine, number of pre-existing chronic conditions defined by the CDC (on December 20, 202021) as risk factors for severe Covid-19 (classified into four classes: 0, 1, 2 and ≥3), and number of hospitalizations over the past 3 years (classified into 5 classes: 0, 1, 2, 3 or 4, and ≥5). These last two variables together were designed to capture the burden and stability of chronic disease.
Each matched pair was tracked from the date of matching until the earliest of the following events: the outcome of interest; the death; 30 days of follow-up; February 18, 2022 (last day of data collection); or the fourth vaccination dose of the control member of the matched pair (at which point the data for both members of the matched pair were censored). Controls who received a fourth dose of vaccine after being matched as controls became eligible to be recruited into the four-dose group and matched to a new control.
Cumulative incidence curves were constructed using the Kaplan-Meier estimator. For each follow-up period, only matched pairs in which data from both members had not been censored at the start of the follow-up period were included. Risk was defined as the likelihood that a given outcome will develop over the follow-up period. The estimated risks in each group were compared both as relative risks and as risk differences. Vaccine efficacy was estimated as 1 minus the relative risk. We calculated 95% confidence intervals using the nonparametric bootstrap method with 500 replicates. The widths of the confidence intervals have not been adjusted for multiplicity and should not be used to infer statistical significance.
We performed two sensitivity analyzes to explore the robustness of our estimates. First, our estimates of the observational analog of the per-protocol effect, in which matched-pair data were censored when the control received a fourth dose, would have been biased if the probability of vaccination had changed at the time of the infection (v. censorship). We therefore performed an analysis identical to the primary analysis except that when the control received a fourth dose of vaccine, the censoring of the matched pair’s data was delayed by 7 days,17 a period during which the extra dose was not yet expected to take effect. In this sensitivity analysis, controls were not subsequently recruited to the four-dose group.
Second, as an alternative to our nonparametric Kaplan-Meier approach, we also fit three parametric Poisson regression models with a log-link function22 across all eligible individuals, with each model incorporating a different definition of time-varying exposure: no fourth dose of vaccine, days 1-4 after fourth dose of vaccine, days 5 and 6, and day 7 and beyond ; no fourth dose of vaccine, days 1 to 4, days 5 and 6, days 7 to 13 and days 14 and following; and no fourth dose of vaccine and each follow-up day treated as a separate category. People were able to provide follow-up data to each of these four-dose groups (i.e. the groups based on time since receiving the fourth dose) and the control group dynamically and independently of interactions with the health care system. The outcome of interest was documented and PCR-confirmed SARS-CoV-2 infection. All models included, as covariates, the calendar date of each follow-up day and the matching factors described above, with area of residence (a covariate with hundreds of categories) replaced by a measure of burden of Covid-19 (the proportion of PCR Tests in the area of residence the day before) (Methods section S1). In this analysis, vaccine efficacy was defined as 1 minus the incidence rate ratio estimated from the model.
Analyzes were performed using R software, version 4.1.0, and additional freely available R software packages “tidyverse”, version 1.3.1 and “survminer”, version 0.4.9.