Background: Although vaccines have proved effective to prevent severe COVID-19, their effect on preventing long-term symptoms is not yet fully understood. We aimed to evaluate the overall effect of vaccination to prevent long COVID symptoms and assess comparative effectiveness of the most used vaccines (ChAdOx1 and BNT162b2).

Methods: We conducted a staggered cohort study using primary care records from the UK (Clinical Practice Research Datalink [CPRD] GOLD and AURUM), Catalonia, Spain (Information System for Research in Primary Care [SIDIAP]), and national health insurance claims from Estonia (CORIVA database). All adults who were registered for at least 180 days as of Jan 4, 2021 (the UK), Feb 20, 2021 (Spain), and Jan 28, 2021 (Estonia) comprised the source population. Vaccination status was used as a time-varying exposure, staggered by vaccine rollout period. Vaccinated people were further classified by vaccine brand according to their first dose received. The primary outcome definition of long COVID was defined as having at least one of 25 WHO-listed symptoms between 90 and 365 days after the date of a PCR-positive test or clinical diagnosis of COVID-19, with no history of that symptom 180 days before SARS-Cov-2 infection. Propensity score overlap weighting was applied separately for each cohort to minimise confounding. Sub-distribution hazard ratios (sHRs) were calculated to estimate vaccine effectiveness against long COVID, and empirically calibrated using negative control outcomes. Random effects meta-analyses across staggered cohorts were conducted to pool overall effect estimates.

Findings: A total of 1 618 395 (CPRD GOLD), 5 729 800 (CPRD AURUM), 2 744 821 (SIDIAP), and 77 603 (CORIVA) vaccinated people and 1 640 371 (CPRD GOLD), 5 860 564 (CPRD AURUM), 2 588 518 (SIDIAP), and 302 267 (CORIVA) unvaccinated people were included. Compared with unvaccinated people, overall HRs for long COVID symptoms in people vaccinated with a first dose of any COVID-19 vaccine were 0·54 (95% CI 0·44-0·67) in CPRD GOLD, 0·48 (0·34-0·68) in CPRD AURUM, 0·71 (0·55-0·91) in SIDIAP, and 0·59 (0·40-0·87) in CORIVA. A slightly stronger preventative effect was seen for the first dose of BNT162b2 than for ChAdOx1 (sHR 0·85 [0·60-1·20] in CPRD GOLD and 0·84 [0·74-0·94] in CPRD AURUM).

Interpretation: Vaccination against COVID-19 consistently reduced the risk of long COVID symptoms, which highlights the importance of vaccination to prevent persistent COVID-19 symptoms, particularly in adults.

Funding: National Institute for Health and Care Research.

Declaration of interests DP-A's department has received grants from Amgen, Chiesi-Taylor, Lilly, Janssen, Novartis, UCB Biopharma, the European Medicines Agency, and the Innovative Medicines Initiative. DP-A's research group has received consultancy fees from Astra Zeneca and UCB Biopharma. DP-A's department has organised training programmes funded or supported by Amgen, Astellas, Janssen, Synapse Management Partners, and UCB Biopharma. RK's research group has received consultancy fees from AstraZeneca and the Estonian Ministry of Social Affairs through RITA CORIVA and RITA MAITT (Machine learning and AI powered public services) projects. AU reports funding from the European Regional Development Fund (RITA 1/02–120) for her institution (Department of Family Medicine and Public Health, University of Tartu). HMEN reports support from the European Health Data and Evidence Network project grant, Innovative Medicines Initiative 2 Joint Undertaking (grant agreement 806968) for harmonisation of the Norwegian registry data into the Observational Medical Outcomes Partnership (OMOP) common data model (CDM). TD-S reports funding from the Innovative Medicines Initiative 2 Joint Undertaking (grant agreement 806968) for the institute to map the SIDIAP data to the OMOP CDM. This initiative received support from the EU Horizon 2020 research and innovation programme and European Federation of Pharmaceutical Industries and Associations. All other authors declare no competing interests.