Methods
We conducted a systematic review with enhanced processes and automation tools.14 The systematic review is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement.15 Our protocol was shared on the Open Science Framework (https://osf.io/e8jdy) on 2 March 2022.
We searched the PROSPERO and Open Science Framework databases to exclude similar reviews. We then searched PubMed, Embase, and Cochrane covid-19 trials for published studies, and Europe PubMed Central (Europe PMC) for preprints, from 1 January 2020 to 3 August 2022. A search string of medical subject headings terms and words was developed in PubMed and translated to run in other databases with the Polyglot search translator.16 Online supplemental file 1 shows the search strategies for all databases.
We also conducted forward and backward citation searches of the included studies. For registered studies, we searched ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform. Searches were run from inception to 3 August 2022 (appendix 1). We also checked the VIEW-hub database (www.view-hub.org), a collaboration between the International Vaccine Access Centre and Johns Hopkins Bloomberg School of Public Health. No publication type or language restrictions were applied. We also contacted authors of large vaccine trials for any unpublished data on long covid.
We included randomised controlled trials, cohort studies (retrospectively or prospectively assembled), interrupted time series, and case-control studies. We excluded case reports, case series, cross sectional studies, and modelling studies. We searched for studies that assessed vaccination status and the emergence of long covid (history of confirmed or probable covid-19 within the past three months and symptoms that lasted at least two months that could not be explained by an alternative diagnosis). Studies conducted in the community, primary care, and hospital settings were included.
Our inclusion criteria were people of all ages who were eligible to receive a covid-19 vaccine. The interventions were any dose of a covid-19 vaccine recognised by WHO (ie, BNT 162b2 (tozinameran, Pfizer-BioNTech), mRNA-1273 (elasomeran, Moderna), ChAdOx1 nCoV-19 (Oxford-AstraZeneca), and Ad26.COV2.S (Janssen or Johnson & Johnson)), before or after the first SARS-CoV-2 infection, or after a diagnois of long covid. Comparators were no vaccination, an active non-covid-19 vaccine control (eg, influenza vaccine), or placebo.
The primary outcomes were patients with a diagnosis of long covid, according to the WHO definition (ie, history of confirmed or probable covid-19 within the past three months and symptoms that lasted at least two months that could not be explained by an alternative diagnosis), and remission or resolution of long covid in patients who were vaccinated after a diagnosis of long covid. The secondary outcome was prevalence of individual symptoms of long covid, such as prolonged fatigue, shortness of breath, cognitive difficulties, and loss of sense of smell. We excluded protocols, studies that did not report long covid outcomes, and studies with uncertain vaccination status at the time of infection (figure 1).
Figure 1Screening and selection of studies
Study selection and screening
Two of the authors (OB and PS) independently screened the titles and abstracts, and full text articles were retrieved for potentially eligible articles. The full texts were then reviewed against the inclusion criteria. Discrepancies were resolved by referring to a third author (PG). Figure 1 summarises the screening process. Online supplemental file 2 lists the excluded articles and reasons for exclusion.
Data extraction
Two of the authors (OB and PS) extracted the data with Microsoft Excel. Study characteristics and outcomes extracted from each study were: methods (study authors, year, country, study design, length of follow-up, and setting); participants (number of participants, age, sex, and any co-comorbidities); interventions (type of intervention, dose, and frequency) and type of comparators (no treatment, other non-covid-19 vaccine, or placebo); and outcomes (patients with long covid (primary outcome) and prevalence of individual symptoms (secondary outcome)).
Assessment of risk of bias
Risk of bias was assessed with the ROBINS-I tool, which can assess both randomised and non-randomised studies on a common template.17 Two of the authors (OB and PS) independently assessed the risk of bias for each study.
Data analysis
We did not conduct meta-analyses because of the high heterogeneity of the data. For dichotomous outcomes, the effect of the intervention was calculated with odds ratios. For one study, we calculated the odds ratio from the reported mean differences.18 We used individual participants as the unit of analysis. When data were missing or unclear, the study investigators were contacted. We found no registered trials for vaccines and long covid. We could only present subgroups by dose of vaccine and timing of vaccine dose.
Patient and public involvement
Patients and the public were not involved in this review. Systematic reviews identify and analyse relevant primary studies to answer a specific research question, but they are not conducted on patients or public directly. We plan to disseminate our results through open access publication, our institute’s monthly newsletter, and preprint database update.