The study, published in eClinicalMedicine, drew on the SCIFI-PEARL project, a population-based research infrastructure that links multiple Swedish national and regional registers through the country’s unique personal identification number. The team, led by Lisa Morris and colleagues at the University of Gothenburg and collaborating institutions, focused on adults living in Stockholm and Västra Götaland, the two largest regions in Sweden, which together account for nearly 40 percent of the national population. Because post COVID-19 condition is largely managed in primary care, access to primary healthcare databases was a decisive advantage, allowing the researchers to capture diagnoses that would be missed in studies restricted to hospital records.

The methodological design was a sequential propensity score-matched cohort study, an approach well suited to observational vaccine research. Rather than comparing vaccinated with unvaccinated people, a comparison now nearly impossible in highly immunised populations, the investigators compared individuals who received a fourth dose with closely matched counterparts who had received three doses but had not yet received a fourth at the same calendar time. Propensity scores were estimated using a Cox proportional hazards model that incorporated a remarkably rich set of covariates: age, gender, education, occupation, income, marital status, urbanicity, country of birth, parents’ country of birth, nursing home residence, home care services, prior healthcare contacts, pregnancy, a wide range of comorbidities, medication use, and previous COVID-19 infection classified as non-hospitalised or hospitalised. Matching was performed within strata defined by gender, age group, the week of the third dose, and the presence or absence of risk factors for severe acute COVID-19.

This procedure produced two groups of 801,869 individuals each, drawn from a total of roughly 1.4 million unique people. Balance between the groups was excellent, with all standardised mean differences below the conventional threshold of 0.1, meaning the three-dose and four-dose groups were statistically indistinguishable across the measured characteristics. Each participant was followed from the index date, the day the matched exposed individual received the fourth dose, until a post COVID-19 condition diagnosis, a new vaccination, migration, death, six months of follow-up, or the end of the study in February 2024. When one member of a matched pair was censored because of new vaccination, the entire pair was censored to prevent informative censoring from skewing the comparison.

The outcome was defined as a registered diagnosis of post COVID-19 condition using the ICD-10 code U09.9, recorded in either primary or specialist care. Sweden implemented this diagnosis code clinically in October 2020, and the National Board of Health and Welfare recommends that physicians use it when a condition is judged to be associated with a previous COVID-19 infection. During follow-up, 388 individuals in the three-dose group and 268 in the four-dose group received such a diagnosis. The incidence rate was 1.67 per 1000 person-years among those with three doses and 1.14 per 1000 person-years among those with four doses, an absolute difference of just over half a case per 1000 person-years.

In the Cox regression analysis, the hazard ratio for post COVID-19 condition after a fourth dose was 0.69, corresponding to a vaccine effectiveness of 31 percent. The researchers then asked whether this protection differed between people with and without risk factors for severe acute COVID-19, a group defined by the Swedish Public Health Agency as including people aged 65 or older, nursing home residents, those receiving home care, pregnant women, and people with conditions such as diabetes, cardiovascular disease, chronic respiratory disease, obesity, renal failure, liver disease, Down’s syndrome, neuromuscular disease, or immunodeficiency. A statistically significant interaction confirmed that the effect did indeed differ. Among individuals with risk factors, vaccine effectiveness reached 38 percent, with an incidence rate difference of minus 0.82 per 1000 person-years. Among those without risk factors, effectiveness was a more modest 18 percent, and the absolute difference of minus 0.21 per 1000 person-years did not reach statistical significance.

The authors subjected their findings to an unusually thorough battery of sensitivity analyses. Repeating the matching with replacement, which allowed 99 percent of the exposed population to be included, produced essentially identical results. Alternative approaches to handling informative censoring, including inverse probability censoring weights and removing censoring for new vaccination altogether, yielded slightly lower estimates but preserved the overall pattern. Broadening the outcome to include the post-viral fatigue code G93.3 increased sensitivity and produced larger incidence rate differences, while excluding individuals diagnosed with other severe acute respiratory infections, a step designed to sharpen the specificity of the outcome, left the results unchanged. Restricting the analysis to index dates from April 2022 onwards, when the Omicron variant dominated, also gave findings consistent with the main analysis, suggesting the protection observed applies to the variant landscape that prevails today.

The study arrives at a moment when vaccination policy is being recalibrated across the world. A recent systematic review and meta-analysis had suggested that third or fourth doses reduce the odds of long COVID compared with a primary series alone, but the certainty of that pooled evidence was rated very low, and the underlying studies relied on self-reported vaccination status and symptoms. The Swedish register data, by contrast, capture every administered vaccine dose through a legally mandated national vaccination register, making exposure misclassification negligible. Earlier work by the same team had already shown that vaccination before a first SARS-CoV-2 infection reduces the risk of post COVID-19 condition in a dose-response fashion, but whether boosters still helped once most of the population carried pre-existing immunity remained genuinely uncertain. The new findings close that gap, showing that a fourth dose retains measurable, if modest, protective value even in this setting.

Interpretation of the mechanism requires care. Because infection is a prerequisite for post COVID-19 condition, and severe acute disease is a well-established risk factor for lingering symptoms, part of the observed protection likely reflects the vaccines’ continued ability to reduce infection and severe illness rather than a unique effect on the post-acute syndrome itself. Waning protection from the third dose may also have contributed to the excess of cases in the comparison group. The authors acknowledge other limitations: the U09.9 diagnosis code has not been formally validated, the registers cannot distinguish mild from severe long COVID, and only people who sought care and received a diagnosis are captured, probably missing milder cases. As an observational study, residual confounding and healthy vaccinee bias cannot be entirely excluded, although the design, in which most three-dose individuals eventually received a fourth dose, mitigates some of these concerns.

For policymakers, the practical message is nuanced. The absolute risk reduction of roughly half a case per 1000 person-years is small in the general population, and the benefit concentrates among those at higher risk of severe acute disease, precisely the group for whom the Swedish Public Health Agency’s 2025 health economic analysis judged continued vaccination most reasonable. With SARS-CoV-2 still circulating and reinfections carrying documented risks of hospitalisation and organ-specific sequelae, the study suggests that booster decisions should weigh not only protection against acute illness but also the emerging, quantifiable protection against long COVID, particularly for older and medically vulnerable individuals. The researchers call for further studies in other settings and with different measures of post COVID-19 condition to confirm and extend their findings.

Subject of Research: Effectiveness of a fourth COVID-19 vaccine dose against post COVID-19 condition in individuals with pre-existing immunity

Article Title: Booster dose of COVID-19 vaccine and its effectiveness against post COVID-19 condition in individuals with pre-existing immunity: a sequential propensity score-matched cohort study

Article References: Booster dose of COVID-19 vaccine and its effectiveness against post COVID-19 condition in individuals with pre-existing immunity: a sequential propensity score-matched cohort study. (n.d.). https://doi.org/10.1016/j.eclinm.2026.104215

Image Credits: AI Generated

DOI: 10.1016/j.eclinm.2026.104215

Keywords: COVID-19, long COVID, post COVID-19 condition, booster dose, vaccine effectiveness, SARS-CoV-2, Sweden, cohort study, propensity score matching, epidemiology, vaccination policy, public health