Transient ischemic attacks (TIAs) are brief episodes of neurological deficits suspected to result from cerebral ischemia without detectable permanent tissue injury. TIA is a diagnostic challenge. Studies have demonstrated that diffusion-weighted magnetic resonance imaging (DWI-MRI), the current diagnostic standard for a tissue-based definition (TIA versus minor stroke), may reveal ischemic changes in only up to 67% of patients with transient neurological deficits. ^{Reference Brazzelli, Chappell and Miranda1} The 90-day risk of ischemic stroke following TIA or minor cerebral ischemia is estimated to range from 3% to 15%, ^{Reference Johnston, Easton and Farrant2} and it is well established that early initiation of stroke prevention strategies may reduce the total burden of stroke and its complications. ^{Reference Rothwell, Giles and Chandratheva3} However, the high awareness and vigilance may also result in overdiagnosis. It was estimated that TIA mimics comprise 60% of referrals to rule out ischemic etiology. ^{Reference Nadarajan, Perry, Johnson and Werring4} These figures are in line with our local experience. Although new objective and sensitive diagnostic biomarkers are emerging, ^{Reference Serlin, Ofer and Ben-Arie5,Reference Dolmans, Rutten and Koenen6} at present, TIA diagnosis remains highly dependent on the clinical assessment in the emergency department (ED). We tested to what extent a neurological assessment by a trained neurologist can predict the presence of ischemic lesions on MRI. We hypothesize that a comprehensive neurological evaluation may reduce the rate of overdiagnosis and increase the pretest probability for positive DWI findings.

This prospective cohort study was held at a tertiary medical center in southern Israel between 2015 and 2017. We consecutively enrolled adult patients (≥18 years old) who presented to the ED with transient focal neurological deficits. All patients were evaluated by both ED and neurology attendings within less than 24 h, had an initial computed tomography (CT) scan showing no ischemic changes, and underwent brain MRI within <72 h from presentation. Patients with a new or known hemiplegic migraine, focal seizure, atrial fibrillation, structural cardiac anomaly, severe carotid stenosis, carotid dissection, or any acute intracranial lesion during ED workup (routine electrocardiogram and CT angiography of the brain and cervical arteries) were excluded. Patients with neurological deficits (objective or subjective) lasting more than 24 h were excluded. All participants signed an informed consent. The study protocol was approved by the Soroka Medical Center Institutional Review Board.

Two attending neurologists evaluated independently all patients. Assessment included history taking, complete neurological examination, and review of initial CT scans. Each neurologist categorized all patients based on the clinical evaluation as “positive” (the most likely diagnosis is cerebral ischemia, high likelihood of positive DWI findings) or “negative” (diagnosis of ischemic etiology is unlikely). All cases were reviewed immediately following the clinical evaluation and any disagreement was discussed to reach a consensus. All subjects were imaged on a 3T Philips Ingenia MRI. T1- and T2-weighted imaging, FLAIR, diffusion, and susceptibility weighted imaging data were acquired. An experienced neuroradiologist interpreted all MRI scans. Analyses were conducted using IBM SPSS Statistics version 24 (IBM, Armonk, NY, USA). Results are presented as mean ± SD for continuous variables and as percentages for categorical data. Mann–Whitney and chi-square tests were used for comparisons as appropriate. Sensitivity, specificity, predictive values, and positive- and negative-likelihood ratios were calculated to assess the performance of neurological assessment prior to MRI. Statistical significance was set at ≤0.05.

A total of 72 patients were included in the study. The clinical diagnosis prior to MRI was “positive” for cerebral ischemia in 26 patients (36.1%) and “negative” (i.e., likely nonischemic) in 46 (63.9%). No significant differences in demographic characteristics and baseline risk factors were found between groups (Table 1). The clinical assessment showed no association with the ABCD2 score, with specific signs, or symptoms at presentation. Unilateral limb weakness was more common among patients diagnosed as “negative” (Table 1). Acute ischemic lesions were found on MRI scans of 20 patients (27.8%). Restricted diffusion on MRI was more prevalent among patients diagnosed clinically as “positive” (19/26, 73.1%) compared with one patient (1/46, 2.2%) originally suspected to have a nonischemic etiology on clinical evaluation (Table 1). Substantial agreement was demonstrated between the predicted patient status according to the clinical evaluation and the actual DWI findings (Kappa = 0.75). The sensitivity, specificity, positive, and negative predictive values for accurate patient classification based on clinical evaluation alone were 95.0%, 86.5%, 73.0%, and 97.8%, respectively. The positive-likelihood ratio for a patient with restricted DWI lesion to be diagnosed as “positive” for cerebral ischemia prior to MRI was 7.06 (95% CI 3.52–14.16), and the negative-likelihood ratio was 0.06 (95% CI 0.01–0.39).

Table 1 Baseline characteristics, radiological, and clinical findings at presentation

DBP = diastolic blood pressure; DWI = diffusion-weighted imaging; PVD = peripheral vascular disease; SBP = systolic blood pressure.

^a On clinical evaluation at presentation.

In the era of modern neuroimaging, the current diagnosis of TIA still relies on history taking and physical examination. However, a high rate of overdiagnosis and a low interobserver agreement in the ED setting have been reported. ^{Reference Ferro, Falcao and Rodrigues7} In our cohort, clinical evaluation by a neurologist in the ED had a high diagnostic accuracy and patients with positive DWI were seven times more likely to be diagnosed as having an actual ischemic event on clinical evaluation alone. The currently accepted tissue-based definition of TIA/minor stroke on MRI is limited by high variability and availability. ^{Reference Sacco, Kasner and Broderick8} Using the tissue-based definition, 20/72 (27.7%) patients with positive DWI findings were reclassified as cases of minor stroke. Among patients initially diagnosed as likely having cerebral ischemia, 7/26 (26.9%) could be labeled as TIA per se. This discrepancy emphasizes the need for a more objective and sensitive modality to diagnose subtle brain ischemia. ^{Reference Serlin, Ofer and Ben-Arie5} Future stroke prevention is an additional major consideration in the ED setting. A recent meta-analysis showed that the risk of stroke misdiagnosis is much greater when presenting neurologic complaints are transient, nonspecific, or mild (range 24%–60%). ^{Reference Tarnutzer, Lee, Robinson, Wang, Edlow and Newman-Toker9} MRI scans are not yet readily available in all clinical settings. Our findings suggest that clinical assessment by a neurologist has an excellent predicting value for ischemic brain lesion confirmed on MRI. This single-center study is limited by a relatively small sample size. Further replication studies should include larger cohorts and also compare the predictive capacity of clinical evaluation by neurology trainees, primary care practitioners, and ED physicians. Our findings reiterate the value of a neurologist in the ED, ^{Reference Moulin, Sablot and Vidry10} for facilitation of effective patient triage and provision of quality care.