key: cord-0946604-zftw2osv authors: Björk, Jonas; Bonander, Carl; Moghaddassi, Mahnaz; Rasmussen, Magnus; Malmqvist, Ulf; Kahn, Fredrik; Inghammar, Malin title: Surveillance of COVID-19 vaccine effectiveness: a real-time case–control study in southern Sweden date: 2022-03-02 journal: Epidemiol Infect DOI: 10.1017/s0950268822000425 sha: 648ba81886c08c5442305a3e4b6f4f803e722099 doc_id: 946604 cord_uid: zftw2osv The extensive register infrastructure available for coronavirus disease 2019 surveillance in Scania county, Sweden, makes it possible to classify individual cases with respect to hospitalisation and disease severity, stratify on time since last dose and demographic factors, account for prior infection and extract data for population controls automatically. In the present study, we developed a case–control sampling design to surveil vaccine effectiveness (VE) in this ethnically and socioeconomically diverse population with more than 1.3 million inhabitants. The first surveillance results show that estimated VE against hospitalisation and severe disease 0–3 months after the last dose remained stable during the study period, but waned markedly 6 months after the last dose in persons aged 65 years or over. Vaccines are since their introduction in December 2020 the primary mitigation strategy to combat coronavirus disease 2019 . The emergence of new severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variants with higher risk of transmission (i.e. variants of concern, VOC) has stressed the importance of continuously surveil COVID-19 vaccine effectiveness (VE) and waning immunity [1] [2] [3] [4] . The population and health care registers in Sweden and the other Nordic countries contain extensive individual-level data for all residents that can be cross-referenced [5] . Such register infrastructures offer excellent opportunities for detailed epidemiologic surveillance but are currently underused. The present study aimed to investigate how the Swedish register infrastructure can be used to surveil COVID-19 VE in real time. To this end, we developed an automatic case-control sampling design using data with complete population coverage from Scania (Skåne), an ethnically and socioeconomically diverse region in southern Sweden exceeding 1.3 million inhabitants. wThe study cohort included all persons residing in Skåne, Sweden, on 27 December 2020 (baseline; n = 1 384 530) when vaccinations started [6] , and was followed for 315 days until 7 November 2021. Individuals who died or moved out from the region were censored on the date of death or relocation, yielding 1 353 488 persons remaining in the cohort at the end of the follow-up. The first to be vaccinated in Sweden were nursing home residents, their caregivers and frontline health care workers, followed by the general population in age groups in descending order, currently down to age 12. Three different vaccines have been used in Sweden: BNT16b2 mRNA (Comirnaty, Pfizer-BioNTech), mRNA-1273 (Spikevax, Moderna) and ChAdOx1-SARS-CoV-2 (Vaxzevria, AstraZeneca). The timing of the second dose depended on vaccine type and schedule but was given in median 42 days after the first. From 1 September 2021, a third booster dose was offered, starting with nursing home residents, older people and immunocompromised. The different data sources were linked using the personal identification number assigned to all Swedish residents at birth or immigration [7] . Individual-level data on country of birth, civil status, residency and vital status were obtained from the Swedish Total Population Register. Weekly updates on vaccination date, type of vaccine and dose were obtained from the National Vaccination Register, and data on positive SARS-CoV-2 test results from the electronic system SMINet, both kept at the Public Health Agency of Sweden. Data from regional registers and electronic health records were accessed continuously. Hospitalisations 5 days before until 14 days after a positive SARS-CoV-2 test result and U07.1 (ICD-10) among the diagnoses were regarded as caused by COVID-19. Severe disease among the hospitalised was defined as a need of oxygen supply ≥5 l/min or admittance to an intensive care unit (ICU). To avoid conflation of varying infection pressure over time in the population with waning VE, we used continuous density casecontrol sampling [8] nested within the study cohort described above. A case was defined as a person with a first-time positive test or any positive test more than 12 weeks after a prior positive test. For each case, 10 controls without a positive test the same week as the case or 12 weeks prior were randomly selected from the complete study cohort, matched with respect to sex and age (5-year groups: 0-4, 5-9, 10-14, …, 95-99 and 100+). We used conditional logistic regression (Stata SE 14.2, Stata Corp, command clogit) for the 1:10 case:control matched sets to estimate the odds ratio (OR) and VE = 1 − OR together with 95% confidence interval (CI, accounting for individual clustering in the full-period estimates) for the association between vaccination status (at least two vs. zero doses) and risk of infection (positive test), being hospitalised or developing severe COVID-19. Only doses received at least 7 days before the case date were counted within each matched set when vaccination status was assessed. Vaccination status among those with at least two doses was grouped according to time since last dose (0-3, 3-6 or >6 months) and vaccine type. Results were (besides the matching) presented unadjusted, but prior infection was included in a sensitivity analysis. VE was monitored weekly, from week 10 (8 March, 71 days after vaccination start) to week 44 in 2021 (7 November, 315 days after vaccination start). In total, 52 125 COVID-19 cases occurred during this period, of whom 1492 (2.9%) were hospitalised and 611 (1.2%) classified as severe (oxygen supply ≥5 l/min or ICU admittance). The weekly infection rate varied between 297 (week 18) and down to 13 (week 27) per 100 000 persons. Routine sequencing of samples of infected cases in the study region showed that the SARS-CoV-2 alpha variant (B.1.1.7) was the dominating VOC until week 26 after which the majority of cases were delta (B.1.617.2) [9] . The identified cases were considerably younger on average towards the end vs. at the start of the follow-up (Table 1) . BNT16b2 mRNA (Pfizer-BioNTech) was the dominant vaccine type with 79% of all administrated doses during the follow-up. The estimated effectiveness against infection after at least two doses of any of three vaccines was 76% in median, and the weekly estimates varied between 51% and 87% with no evident time trend (Fig. 1) . At least two doses of any of the three vaccines offered high protection against hospitalisation (median effectiveness 87%, range 70-94%) and severe disease (median effectiveness 93%, range 66-96%), here aggregated monthly due to small numbers (Fig. 2) . Estimated effectiveness against infection (Supplementary Fig. S1A ) and hospitalisation ( Supplementary Fig. S1B ) 0-3 months after the last dose remained stable during the study period. Waning effectiveness against hospitalisation was consistently noted 6 months after the last dose, but with substantial fluctuations due to the small numbers. The mRNA vaccines (Pfizer-BioNTech and Moderna) exhibited higher effectiveness on average against infection than the vector vaccine (AstraZeneca; Fig. 3 ). All three vaccines offered strong protection against hospitalisation and severe disease both overall and specifically in individuals ≥65 years ( Fig. 3 and Supplementary Fig. S2 ). VE waned considerably with time since last dose, especially in individuals ≥65 years. The protection against hospitalisation and severe disease remained more satisfactory in younger individuals but was also more statistically uncertain ( Supplementary Fig. S2 ). Prior SARS-CoV-2 infection offered strong protection against new infection (average effectiveness 87%, 95% CI 86-88%) and hospitalisation (average effectiveness 90%, 95% CI 80-95%, not in tables), but did not confound the estimates of VE. The extensive register data available in our real-time surveillance system allow us to classify hospitalised cases further concerning disease severity, stratify by time since last dose and demographic factors, account for prior SARS-CoV-2 infection and extract data for population controls automatically. The density case-control sampling is a convenient approach to avoid bias from underlying time trends in the population [8] . Such bias could occur if, e.g. associations between age or time since last dose among the vaccinated and the infection pressure on society are not considered during design or analysis. Selective COVID-19 testing depending on the vaccination status could still bias the case-control sample but is unlikely to impact estimated effectiveness against hospitalisation and severe disease. The surveillance was based on data with complete population coverage, but was nevertheless hampered by limited population size and a low infection rate during parts of the study period. As a comparison, the population in Scania county is only 1/7 of the Israeli population, a well-known example where continuous surveillance of COVID-19 VE is conducted [3] . However, after adaptions, our extensive surveillance system should be possible to implement at the national level in Sweden, with a population exceeding 10 million. Our initial surveillance result presented here show more marked waning VE among persons aged 65 years or over 6 months after the last dose. This finding is not consistent with results from the Israeli population where a decrease in effectiveness of similar magnitude in all age groups has been observed [3] . Differences in design and how the varying infection pressure in society over time is handled may be one explanation for the diverging results. Consistent with our finding, recent immunological investigations observed a more decreased immune humoral response among older people and nursing home residents and also among males and persons with immunosuppression [10] [11] [12] . It is for this reason important to continue the surveillance in order to study the effects of additional booster doses in specific groups as well as in the population more broadly, and to monitor protection against VOC [13] . The present investigation demonstrates the strength of combining individual-level population and health care register data to monitor VE in real time. A natural extension of the surveillance system would be to add further individual-level data on socioeconomic conditions, disease histories and care needs among older people, all available in the Swedish register infrastructure. Such additions would make it possible to monitor protection in Epidemiology and Infection 5 Effectiveness of COVID-19 vaccines against the B.1.617.2 (delta) variant Vaccination and non-pharmaceutical interventions for COVID-19: a mathematical modelling study. The Lancet Infectious Diseases Waning immunity after the BNT162b2 vaccine in Israel Effectiveness of mRNA BNT162b2 COVID-19 vaccine up to 6 months in a large integrated health system in the USA: a retrospective cohort study Nordic health registry-based research: a review of health care systems and key registries High level of protection against COVID-19 after two doses of BNT162b2 vaccine in the working age population -first results from a cohort study in southern Sweden The Swedish personal identity number: possibilities and pitfalls in healthcare and medical research Re: Measurement of vaccine direct effects under the testnegative design SARS-CoV-2 sequencing in Scania, southern Sweden (in Swedish): Scania County Council Waning immune humoral response to BNT162b2 COVID-19 vaccine over 6 months Significant reduction in vaccine-induced antibody levels and neutralization activity among healthcare workers and nursing home residents 6 months following coronavirus disease BNT162b2 mRNA vaccination Effectiveness of Pfizer-BioNTech and Moderna vaccines in preventing SARS-CoV-2 infection among nursing home residents before and during widespread circulation of the SARS-CoV-2 B.1.617.2 (Delta) variant -National Healthcare Safety Network COVID-19: do vaccines work against omicron-and other questions answered Conflict of interest. All authors declare no conflicts of interest, no support or financial relationship with any organisation or other activities with any influence on the submitted work.Ethical standards. Ethical approval for the study was obtained from the Swedish Ethical Review Authority (2021-00059). All personal data were handled and analysed pseudonymised, which means that the key that allows for identification of individual persons were stored technically and organisationally separated from the research data.Data availability statement. Aggregated surveillance data from the present study are publicly available at https://sodrasjukvardsregionen.se/kliniskastudier/covid-vacciner-skyddseffekt/.