Covid-19 vaccine effectiveness against post-covid-19 condition among 589 722 individuals in Sweden: population based cohort study

Findings in context

The few earlier studies on vaccine effectiveness against the long term effects of covid-19 have mostly shown protective effects, with a wide range of effect estimates,2627 but some failed to show an overall protective effect.2829 The methodology and data included in the earlier studies were heterogeneous and had limitations. Study populations have rarely been population based and often have included a small number of participants.3031 Analyses of different effects for different numbers of vaccine doses before covid-19 have not always been performed.2732 Because vaccination during follow-up has often not been a criterion for censoring, both vaccinated and unvaccinated individuals have been included in the unvaccinated group. In the present study, we used population based survival data of 589 722 individuals, censoring at both vaccination and reinfection, and report vaccine effectiveness separately for any dose, one dose, two doses, and three or more doses. Earlier studies have generally lacked a clear definition of PCC, and symptoms have often been self-reported,29303133 whereas we used register based clinical diagnoses of PCC as the outcome. Furthermore, in earlier studies follow-up duration has often been short,29 whereas in our study the median follow-up was 129 days from 28 days after a first registered infection. A recent systematic review concluded that being vaccinated against covid-19 before infection had a protective effect on PCC in 10 of the 12 included studies, with effect estimates ranging from 0.48 to 0.87 for any vaccine dose given before infection.11 Owing to high heterogeneity between studies and the low certainty of evidence, no meta-analysis was performed. In other systematic reviews, meta-analyses have, however, included several of these studies, but the results should be interpreted with caution.3435 One of these meta-analyses concluded that receiving two doses of vaccine before covid-19 was associated with a lower risk of PCC compared with no vaccination, with an odds ratio of 0.64,34 and that the odds ratio was 0.71 with at least one dose before infection.35

Using register data from the whole adult population in the two largest regions of Sweden, we showed that vaccination against covid-19 before infection was associated with a decreased risk of receiving a diagnosis of PCC. When stratifying by the median time between last vaccination and infection (126 days) to assess the potential different effects of recent versus earlier vaccination, we found that receiving the last vaccine dose more than 126 days before covid-19 was still associated with a relatively high vaccine effectiveness against PCC, and only slightly lower than in the main analysis. In addition, to ensure sufficient time between vaccination and the acute infection, in a sensitivity analysis we restricted the vaccinated population to those who received their last vaccine dose more than 14 days before covid-19, and the estimated vaccine effect did not markedly change from the main analysis. Furthermore, in the main analyses we only considered the first PCC diagnosis at least 28 days from infection, but in sensitivity analyses we required at least 90 days from infection, with similar results.

Studies have shown that women may develop greater immune responses to vaccination than men,36 although this does not necessarily translate to better protection against the disease. In our study, men showed a higher vaccine effectiveness against PCC than women. It has not yet been fully established whether PCC is more likely to occur with particular variants. Available data suggest, however, that individuals infected with the omicron variant are at lower risk of developing long term effects of covid-19 than individuals infected with the other variants.373839 Nonetheless, it is difficult to determine if this lower risk is associated with the specific variant or is the result of immunity from previous infections or vaccinations, or as a result of shorter follow-up durations. A small study evaluating the protective effect of covid-19 vaccines against PCC, which included individuals with infections during the omicron period as well as the earlier periods, did not show significantly different results between the variants.30 In the present study, the study population included individuals with infections at the time when the alpha and delta variants predominated, and also during part of the pre-alpha and omicron periods. Although vaccination coverage was not evenly distributed during these periods, stratifying on the period of dominant variant at the time of infection showed only a slightly lower vaccine effectiveness against PCC in the omicron period than in the pre-alpha and alpha periods. As we did not have access to virus sequencing data in our analysis, we used the time of infection as a proxy for variant. Consequently, the variant causing acute infection in some of the study individuals might have been misclassified.

The pathogenesis of PCC has not yet been clarified, but several mechanisms have been proposed relating to the different symptom manifestations and it has become increasingly evident that patients with PCC are a heterogenous group. Potential mechanisms include organ damage, abnormal immune activation during acute infection, reactivation of other viruses, altered systemic immunity, autoimmunity, and sustained immune activation due to viral persistence.40 Determining the pathogenesis might suggest potential pathways for the protective effect of the vaccines—for example, a reduced viral load during the acute infection after vaccination could reduce the virus’s persistence with lasting immune activation. Different symptom clusters of PCC may have different pathogeneses and therefore different mechanisms for the vaccine effect. We have shown that almost 37% of patients with covid-19 treated in the ICU subsequently have a PCC diagnosis.16 Covid-19 vaccines have been shown to protect against hospital admission with covid-19,41 which could be one pathway for the vaccines to exert a protective effect against PCC. In our analysis, vaccine effectiveness against PCC seemed to be only partly explained by a decreased risk of hospital admission. In addition, analyses stratified on severity of acute disease as indicated by the need for hospital admission showed that vaccine effectiveness was similar in both the group admitted to hospital without ICU admission and the group with no hospital admission. Furthermore, a study showed that those who were vaccinated after covid-19 had a lower risk of developing PCC compared with those who were unvaccinated 12 weeks after covid-19.26 This, together with the findings in the present study, support the hypothesis of pathways beyond the protective effect against hospital admission that may contribute to the protective effect of covid-19 vaccines against PCC. It is also important to note that symptoms of PCC are frequently observed not only in patients with confirmed covid-19 but also in those without a positive SARS-CoV-2 PCR test result.42

Strengths and limitations of this study

The present study has several strengths. Firstly, we used register based data collected from high quality registers, resulting in essentially no loss to follow-up and a low risk of self-reporting bias. In Sweden, it is mandatory and regulated by law to register every administered covid-19 vaccine dose in the national vaccination register. Therefore the exposure data (vaccination) are particulary comprehensive and accurately measured. Secondly, we had access to individual level data from primary healthcare as well as inpatient and outpatient specialist healthcare. This is of importance when studying the diagnosis of PCC, since we have previously shown that most (>85%) patients with PCC in Sweden received their diagnosis in primary healthcare.16 In addition, to fully account for health seeking behaviour and the potential that the PCC group is a biased group of healthcare seekers, the number of healthcare contacts in 2019 was included as a confounder in the full model. Futhermore, the study was population based, covering the two largest regions of Sweden (Region Stockholm and Västra Götaland, 40% of the total Swedish population). Lastly, most previously published studies investigating the protective effect of vaccination before covid-19 against PCC have not been able to account for vaccinations given after infection. By not considering these vaccinations, the total protective effect against PCC will potentially be diminished as a result of the groups becoming more similar to each other. By using survival data in combination with data on vaccinations from the national vaccination register, we were able to censor individuals at vaccinations given after the acute infection.

The limitations of the present study include that both PCC and the ICD-10 diagnosis code, U09.9, are relatively new and the code has not yet been validated in a Swedish setting. It is possible that PCC might be overdiagnosed as well as underdiagnosed, which could affect both the sensitivity and the specificity of PCC as an outcome measure. If this affects both unvaccinated and vaccinated individuals fairly equally, this would lead to a non-differential misclassification of the outcome, which on average would result in some bias towards the null. However, we cannot fully rule out the possibility that vaccinated individuals are less likely than unvaccinated individuals to receive a PCC diagnosis owing to expectations from both patients and healthcare providers about the protective effect of vaccination—although it may be less likely that this bias would increase with increasing number of vaccine doses and show the strong dose-response association in our results. A recent paper from Sweden investigating healthcare use after covid-19 among patients with the PCC diagnosis code, in comparison with controls matched on age, sex, and number of healthcare contacts before infection, showed that the PCC group had significantly more healthcare contacts after covid-19.24 Therefore, we believe that the specificity of the PCC diagnosis code might be good, while its sensitivity remains less clear. In addition, it is possible that vaccine effectiveness differs in patients who experience a specific symptom compared with those who experience another symptom within the PCC spectrum. However, if the protective effect of the vaccines would be valid for a few specific symptoms within the PCC spectrum only, the relatively strong effect on the PCC diagnosis we see in the present study would be less likely to occur. A few studies have also investigated the impact of vaccination on existing PCC, showing both no effect as well as alleviation and aggravation of PCC symptoms.434445 The register based data used in the present study had limited data on symptoms and therefore it would be difficult to assess changes in symptoms of an already existing PCC. Furthermore, although PCC is diagnosed on a specific date, the condition and symptoms usually have been present before the date of diagnosis. Lastly, our results are based on first SARS-CoV-2 infections, whereas reinfections might represent most of the infections today. The potential impact of reinfections on the covid-19 vaccine effectiveness of PCC remains to be elucidated.