Stage 1 statement
This submission is a Stage 1 Registered Report. No confirmatory analyses have been conducted beyond feasibility checks.
Rationale for Registered Report: this study uses observational registry data in which cohort definitions, outcome operationalization, covariate adjustment choices, and missing data handling can substantially influence conclusions. Registering hypotheses and the analysis plan before confirmatory analyses begin and prior to access to national registry data reduces analytic flexibility and selective reporting. This is particularly relevant for sex-difference research, where multiple plausible modelling choices can produce divergent inferences.
Stage 1 in-principle acceptance therefore strengthens the credibility of the confirmatory findings irrespective of statistical significance.
Introduction
Atrial fibrillation (AF) is one of the most important modifiable risk factors for ischaemic stroke, particularly among older adults.1 Oral anticoagulants (OACs), including vitamin K antagonists (VKAs) and direct oral anticoagulants (DOACs), significantly reduce stroke risk and mortality in patients with AF.2,3 However, numerous studies indicate that women are less likely than men to receive adequate anticoagulation despite a higher baseline stroke risk and poorer outcomes after ischaemic events.4-6
Large vessel occlusion (LVO) strokes, caused by the blockage of a major cerebral artery, are the most severe stroke subtypes and often require endovascular treatment (EVT). Sex-based disparities in anticoagulation status among patients presenting with LVO strokes may contribute to differences in stroke severity, access to treatment and clinical outcomes. In preliminary observations from Stavanger University Hospital, women with AF were more often under-anticoagulated prior to stroke compared with men, even though EVT treatment rates and outcomes appeared similar.
This study aims to test specific, preregistered hypotheses regarding sex differences in anticoagulation, stroke management and outcomes in patients with AF. We will use a preregistered confirmatory framework, with national registry data used to explore the generalizability of the confirmatory findings.
Study aims and hypotheses
Primary aim (confirmatory)
To assess whether sex differences exist in stroke risk factors, clinical presentation, treatment and outcomes among patients with AF, using a preregistered confirmatory design.
Primary analyses will be conducted in patients with AF presenting with LVO stroke treated with EVT, as this represents a clinically homogeneous group in which treatment disparities are highly relevant. To assess generalizability to the broader AF stroke population, exploratory analyses will also include all patients with AF and ischaemic stroke recorded in the Norwegian Stroke Registry.
Primary hypothesis (H1)
Among adults with AF presenting with large vessel occlusion ischaemic stroke treated with endovascular thrombectomy, women differ from men in prespecified stroke risk factors, clinical presentation and clinical outcomes, including stroke severity at admission, 3-month functional outcome, and mortality.
Null hypothesis (H0)
There are no sex-based differences in prespecified stroke risk factors, clinical presentation, treatment or outcomes among adults with AF presenting with LVO stroke treated with EVT.
Secondary aims (exploratory)
Materials and methods
Ethics statement
Ethical approval has been granted by the Regional Committee for Medical and Health Research Ethics (REK Vest, reference 217700), the Norwegian Centre for Research Data (NSD) and the data protection officer at Stavanger University Hospital for the retrospective analysis of stroke patients with AF. This approval covers the use of local data from Stavanger University Hospital and linkage with the Norwegian Stroke Registry and the Norwegian Prescription Registry, with a focus on age- and sex-based differences in anticoagulation, treatment and outcomes.
Following Stage 1 in-principle acceptance, a formal amendment will be submitted to REK to expand the ethical approval to include analyses of long-term mortality (1, 2 and 3 years) through linkage with the Norwegian Cause of Death Registry and to clarify the use of a Registered Report methodology. No confirmatory analyses involving these extended data will be conducted before the updated REK approval has been granted.
Study design
This is a retrospective cohort study with a confirmatory primary analysis and several exploratory secondary analyses. Data will be obtained from two sources:
The local dataset includes detailed clinical timelines, imaging results and EVT procedure data. The national dataset provides population-level coverage for exploratory assessment of generalizability and broader patterns.
Open Science Framework Preregistration: The full protocol, hypotheses and statistical analysis plan will be preregistered in the ‘Stiftelsen Dam’ registry on the Open Science Framework before confirmatory analyses begin.
Statistical analyses summary
Confirmatory analyses will be performed in adults with AF presenting with LVO stroke treated with EVT. The primary confirmatory endpoint is functional outcome at 3 months (modified Rankin Scale (mRS) 0–2 versus 3–6) modelled using logistic regression with prespecified adjustment for age, baseline stroke severity, comorbidities and acute treatments. Stroke severity at admission (NIHSS) and mortality endpoints will be analysed as secondary confirmatory outcomes using appropriate regression models, including logistic regression at prespecified time points and, where dates are available, Cox proportional hazards models.
Missing data will be handled using multiple imputation if key variables exceed 5% missingness, with complete-case sensitivity analyses. Multiplicity across confirmatory outcomes will be controlled using Holm–Bonferroni adjustment (two-sided family-wise alpha 0.05), and effect sizes with 95% confidence intervals will be reported for all endpoints.
An overview of the study timeline, data sources, eligibility criteria, exclusions and final analytic cohorts is shown in Fig. 1.
Participants
Inclusion criteria (local and national data)
Exclusion criteria
Primary analyses will be restricted to patients with AF and LVO stroke treated with EVT. As a planned exploratory extension, we will also examine all patients with AF and ischaemic stroke in the Norwegian Stroke Registry to assess whether observed sex differences generalize beyond the LVO population.
Based on prior Norwegian and international data, LVO accounts for ∼20%–30% of ischaemic strokes, corresponding to an estimated 1500–3000 cases annually in Norway. We expect ∼40–45% female and 55–60% male participants in the study cohort.
Variables
Exposure variable
Patient sex (female or male) is recorded in the hospital and registry datasets. Information on gender identity is not available; analyses will therefore examine sex differences.
Confirmatory outcomes:
Secondary exploratory outcomes
Covariates
Dissemination plan
The Stage 2 Registered Report will present all preregistered confirmatory analyses in a single integrated manuscript, irrespective of the direction or statistical significance of the results. No preregistered confirmatory analyses will be split across multiple publications. Following the completion of Stage 2 reporting, additional analyses using the same datasets may be undertaken to address secondary or exploratory research questions.
These manuscripts will be clearly identified as exploratory, will explicitly reference the Registered Report preregistration and will not re-label post hoc analyses as confirmatory.
Statistical analysis plan
Primary (confirmatory) analysis
Confirmatory endpoint hierarchy and multiplicity: The primary confirmatory endpoint is a functional outcome at 3 months, defined as binary mRS score 0–2 versus 3–6. NIHSS at admission and mortality endpoints (in-hospital, 3 months, 1 year, 2 years and 3 years) are secondary confirmatory outcomes. To control the family-wise error rate across the confirmatory outcome family, we will apply Holm–Bonferroni adjustment at a two-sided alpha of 0.05 and report both adjusted and unadjusted P-values alongside effect sizes and 95% confidence intervals.
Baseline sex-based differences in stroke severity, clinical presentation, treatments and outcomes were assessed using chi-square tests for categorical variables and t-tests or Mann–Whitney U-tests for continuous variables, as appropriate. Logistic regression will be used to assess the association between sex and favourable functional outcome (mRS 0–2 at 3 months), adjusting for prespecified age, comorbidities, stroke severity and treatment variables.
Mortality analyses: logistic regression models will assess the association between sex and mortality at 3 months, 1 year, 2 years and 3 years. Where exact dates of death are available, time-to-event (survival) analyses will be conducted using Cox proportional hazards models, adjusted for age, stroke severity, comorbidities and treatment variables.
Secondary (exploratory) analyses
Power calculation
Preliminary data suggest an approximate 10% difference in adequate anticoagulation rates between sexes (70% in men versus 60% in women). With α = 0.05 and 80% power, ∼800 patients with AF would be required to detect this difference. The Norwegian Stroke Registry includes a sufficiently large population to support this level of precision.
This calculation is intended to demonstrate feasibility and expected precision rather than to define power for all confirmatory outcomes.
Deviations from the preregistered analysis plan
Any deviations from this preregistered analysis plan will be transparently documented and justified in the Stage 2 Registered Report. All changes will be clearly distinguished from preregistered confirmatory analyses and reported as exploratory analyses.
Missing data
Multiple imputations will be conducted if more than 5% of values are missing for prespecified key variables, under a missing-at-random (MAR) assumption. Sensitivity analyses will compare complete-case and imputed results.
Quality checks
To ensure data validity and the feasibility of the planned analyses, the following checks will be conducted prior to confirmatory analyses:
Language editing and use of AI tools
Artificial intelligence tools (ChatGPT, OpenAI) will be used for language editing and refinement of the manuscript text. The tool will be applied only for proofreading and improving clarity of language; all scientific content, analyses and interpretations will be produced by the authors.
Summary of design rationale
This study is designed to test preregistered hypotheses that there are sex differences in stroke characteristics, management and outcomes among AF patients using a confirmatory Registered Report framework. The use of a large, routinely collected registry data allows for the systematic evaluation of these questions while minimizing selection bias and enhancing generalizability. By focusing the primary confirmatory analyses on patients with AF presenting with LVO stroke treated with EVT, the study targets a clinically homogeneous population with well-defined treatment pathways.
This design reduces clinical heterogeneity and strengthens interpretability of observed differences. To assess broader relevance beyond the EVT population, exploratory analyses will additionally include all patients with AF and ischaemic stroke recorded in the Norwegian Stroke Registry. Secondary exploratory aims will examine age, comorbidity burden, treatment delays and health system factors as potential contributors to observed sex differences.
Together, this design enables transparent and rigorous assessment of whether sex differences are present in stroke care and outcomes, while ensuring that both confirmatory and exploratory findings are interpreted within a clearly prespecified analytic framework.
Design considerations
A structured overview of hypotheses, outcomes, covariates and planned statistical models is provided in Table 1 (Analysis Overview). Additional details on theoretical rationale, sensitivity considerations and interpretation of potential outcomes are provided in the Supplementary material.
Limitations
This is a retrospective, observational study, which limits causal inference. While the Norwegian Stroke Registry offers broad coverage, variation in data completeness and accuracy across hospitals may influence results. Classification of adequate anticoagulation depends on available documentation and may be affected by misclassification, and some patients may have clinical contraindications to anticoagulation that are not captured in registry data.
Residual confounding from unmeasured variables may persist despite multivariable and mediation analyses. Follow-up for 1-, 2- and 3-year mortality will depend on successful linkage to the Norwegian Cause of Death Registry and/or availability of data in hospital records.
Supplementary Material