This systematic review and meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement. Data not published within the article are available from the corresponding author on reasonable request. We systematically searched PubMed, Embase, Emcare, Web of Science, and the Cochrane Library for published articles from their inception until May 2020, to identify studies that reported comparisons between men and women in acute stroke symptomatology. The complete search strategy is provided in Supplemental Material.
Classification of Stroke or TIA Symptoms
The definition of nonfocal symptoms and focal symptoms was based on a previous classification. We added lightheadedness to the classification nonfocal symptoms and unilateral numbness, discoordination/ataxia, and paresis/hemiparesis to the classification focal symptoms.
Nonfocal symptoms thus included lightheadedness, mental status change/change in level of consciousness (confusion, altered mentality, mental status change, disorientation, and drowsiness defined as Glasgow Coma Scale [GCS] score ≤14; coma/stupor/loss of consciousness/unconsciousness defined as GCS score ≤8), headache including migraine, aspecific neurological or other neurological symptoms (nonrotatory dizziness and non-neurological symptoms), and atypical symptoms including chest pain, palpitations, shortness of breath, nausea, hiccups, and generalized weakness (Table 1).
Focal symptoms included unilateral numbness, discoordination/ataxia, paresis/hemiparesis, aphasia, dysarthria, gait disturbance, imbalance, facial weakness, vertigo with or without nausea/vomiting, diplopia, other focal visual disturbances, and pain of neurological origin (face or hemibody pain; Table 2).
Rare stroke symptoms such as binocular blindness, speaking with a foreign accent, hemiballismus, and alien hand syndrome were excluded.
We applied the following selection criteria for inclusion of studies: (1) patients admitted consecutively because of a diagnosis or suspicion of acute stroke including TIA; (2) use of a cohort, cross-sectional, case-control, or randomized controlled trial design; (3) diagnosis of stroke based on neurological examination and neuroimaging (either computed tomography or magnetic resonance imaging). The diagnosis of TIA had to be based on the anamnesis, neurological examination performed by an emergency medicine doctor or a neurologist (in training) in combination with normal neuroimaging. Patients with TIA had to be examined at specialized TIA centers (TIA clinic or emergency department of a stroke center). A TIA was defined as acute focal neurological symptoms lasting <24 hours. Preferably, analyses for intracerebral hemorrhage (ICH) and subarachnoid hemorrhage (SAH) were performed separately; (4) included patients were ≥18 years of age; (5) presence of ≥1 nonfocal or focal stroke symptoms is quantified in the article; (6) possible differences in stroke symptoms between women and men could be retracted; and (7) articles written in English, Dutch, German, French, Spanish, or Italian. If multiple publications originated from the same cohort, we included the study reporting on the largest number of symptoms by sex. In case similar data were presented, we included the study with the largest population.
Two reviewers (M.A. and N.v.d.W.) independently performed 2 rounds of screening: (1) title and abstract and (2) full-text versions of the remaining studies after the first selection. Data extraction forms included details on the study characteristics (publication year, country of study population, setting, study period, stroke subtype, study design, study size, proportion women, and mean or median age). In cases of doubt, a consensus meeting was held with a third reviewer to determine whether articles met the inclusion criteria.
Assessment of Risk of Bias
The quality assessment of included studies was conducted with the Newcastle-Ottawa Scale.[14,15] For this study, we developed a customized version of the Newcastle-Ottawa Scale with adjustments for the assessment of stroke and TIA symptoms. Studies were scored low risk, high risk, or possible/unclear on the domains (1) validation of diagnosis, (2) assessment of symptoms, (3) adjustment for confounding, and (4) generalizability (Supplemental Methods).
Statistical analyses were performed using R. We weighted the log of the odds ratios (ORs) by the inverse of their variance to obtain pooled estimates. We used a random-effects model because we expected heterogeneity to be present due to known differences in stroke patient characteristics and symptom definitions.
We used the sex-specific number of symptom occurrences provided in the study, along with the total number of women or men studied to calculate ORs. We also used this information to calculate the corresponding 95% CIs and the degree of overlap in symptom presentation between men and women in percentages.
In case a patient presented with multiple symptoms, we included this patient in multiple analyses relating to these symptoms. We intended to use the adjusted effect estimates. If not available, the crude effect estimates were pooled.
Meta-analyses per symptom were conducted if at least 2 independent studies quantified the same symptom.
The overall effect estimate and 95% CI of the forest plots per symptom were used to create summary forest plots for both nonfocal and focal symptoms.
Statistical heterogeneity was assessed by the Higgin I2 statistics. We considered study-level estimates to be heterogeneous if the I2 value was >50%. I2 from 50% to 75% was considered as substantial heterogeneity, and I2 >75% was indicated as considerable heterogeneity.
We used funnel plots to examine potential publication bias of symptoms that were reported by at least 10 studies. Furthermore, we intended to perform subgroup analyses for stroke subtypes.
Stroke. 2022;53(2):345-354. © 2022 American Heart Association, Inc.