Elsevier

The Lancet

Volume 369, Issue 9558, 27 January–2 February 2007, Pages 283-292
The Lancet

Fast track — Articles
Validation and refinement of scores to predict very early stroke risk after transient ischaemic attack

https://doi.org/10.1016/S0140-6736(07)60150-0Get rights and content

Summary

Background

We aimed to validate two similar existing prognostic scores for early risk of stroke after transient ischaemic attack (TIA) and to derive and validate a unified score optimised for prediction of 2-day stroke risk to inform emergency management.

Methods

The California and ABCD scores were validated in four independent groups of patients (n=2893) diagnosed with TIA in emergency departments and clinics in defined populations in the USA and UK. Prognostic value was quantified with c statistics. The two groups used to derive the original scores (n=1916) were used to derive a new unified score based on logistic regression.

Findings

The two existing scores predicted the risk of stroke similarly in each of the four validation cohorts, for stroke risks at 2 days, 7 days, and 90 days (c statistics 0·60–0·81). In both derivation groups, c statistics were improved for a unified score based on five factors (age ≥60 years [1 point]; blood pressure ≥140/90 mm Hg [1]; clinical features: unilateral weakness [2], speech impairment without weakness [1]; duration ≥60 min [2] or 10–59 min [1]; and diabetes [1]). This score, ABCD2, validated well (c statistics 0·62–0·83); overall, 1012 (21%) of patients were classified as high risk (score 6–7, 8·1% 2-day risk), 2169 (45%) as moderate risk (score 4–5, 4·1%), and 1628 (34%) as low risk (score 0–3, 1·0%).

Implications

Existing prognostic scores for stroke risk after TIA validate well on multiple independent cohorts, but the unified ABCD2 score is likely to be most predictive. Patients at high risk need immediate evaluation to optimise stroke prevention.

Introduction

About 240 000 transient ischaemic attacks (TIAs) are diagnosed every year in the USA,1 and about 70 000 in the UK,2 with an overall incidence approaching that of ischaemic stroke. Patients with TIA are generally unstable, with recent studies showing that 4–20% will have a stroke within 90 days after a TIA, half within the first 2 days.1, 3, 4, 5, 6, 7, 8 However, most patients with TIA will have a benign short-term course. Identification of those at highest and lowest risk of stroke in the first days and weeks after a TIA would allow appropriate use of costly secondary prevention strategies, including hospital admission.9, 10, 11 Guideline recommendations for admission after TIA are vague11, 12 and practice is highly variable,13 with some institutions admitting most patients and others none.

Two prognostic scores for short-term risk of stroke after TIA have been proposed: the California score3 and the ABCD score.14 Both scores rely on summation of points associated with clinical factors independently predictive of stroke risk, and have several factors in common. However, validation of the California score has only been published in abstract form, and although the ABCD score validated well in two quite small independent Oxfordshire cohorts,14 and in a Greek cohort,15 more validations by independent investigators are needed. Prognostic scores often validate less well than in the original studies when applied to different populations assembled by independent investigators.16 Furthermore, the California score was developed to predict stroke within 90 days and the ABCD score predicts 7-day risk of stroke, whereas the 2-day risk is often most relevant for decisions about necessity of urgent evaluation and observation; complete diagnostic evaluation and treatment would be difficult to implement within 48 h of a TIA in the outpatient setting.

We therefore aimed to validate the two existing prognostic scores in large independently assembled groups from different populations, comparing predictions of stroke risk at 2, 7, and 90 days. To assess the probable generalisability of the validations, we used groups recruited from emergency departments, specialist clinics, and primary care. Furthermore, by combining results from the original groups used to derive the California and ABCD prognostic scores, we sought to generate a new unified score that would improve prediction of risk of stroke in the 2 days after TIA, and thus create a sole standard for use in clinical care. We then validated the new score in four independent cohorts from California and Oxford, UK, and compared its effectiveness with that of the original scores.

Section snippets

Study groups

All study protocols were approved by the appropriate local review boards. To develop and validate generalisable prognostic scores, the study populations included various clinical settings in two dissimilar populations, from the Kaiser-Permanente Medical Care Plan (KPMCP) in Northern California, USA, and from Oxfordshire, UK, studied over a range of periods (table 1). Characteristics of the derivation cohorts from Northern California and Oxford have been detailed elsewhere.3, 4, 14 Each study

Results

Overall, the two derivation groups and four validation groups included 4809 individuals with TIA (table 1) with evaluable data. The validation groups (n=2893) were generally similar to the derivation groups (n=1916). Non-white race, diabetes, and prolonged TIA were more frequent and atrial fibrillation less frequent in California than in Oxford.

Overall, strokes occurred in 442 patients (9·2%) within 90 days of presentation after TIA, 360 (7·5%) at 30 days, 267 (5·5%) at 7 days, and 189 (3·9%)

Discussion

The California score and the ABCD score both reliably predicted short-term risk of stroke after presentation with TIA. They were derived in different populations of patients and for prediction of stroke at different time points, but contained similar components. Both scores predicted stroke risk reliably in independent groups from distinct regions, in patients seen in emergency departments and outpatient clinics, and in population-based and hospital-based groups. In view of the similarities in

References (42)

  • LD Lisabeth et al.

    Stroke risk after transient ischemic attack in a population-based setting

    Stroke

    (2004)
  • M Daffertshofer et al.

    Transient ischemic attacks are more than “ministrokes”

    Stroke

    (2004)
  • GJ Hankey et al.

    Cost-effective investigation of patients with suspected transient ischaemic attacks

    J Neurol Neurosurg Psychiatry

    (1992)
  • MN Nguyen-Huynh et al.

    Is hospitalization after TIA cost-effective on the basis of treatment with tPA?

    Neurology

    (2005)
  • SC Johnston

    Clinical practice. Transient ischemic attack

    N Engl J Med

    (2002)
  • SC Johnston et al.

    National Stroke Association guidelines for the management of transient ischemic attacks

    Ann Neurol

    (2006)
  • SC Johnston et al.

    Practice variability in management of transient ischemic attacks

    Eur Neurol

    (1999)
  • G Tsivgoulis et al.

    Validation of the ABCD Score in identifying individuals at high early risk of stroke after a transient ischemic attack. A hospital-based case series study

    Stroke

    (2006)
  • DG Altman et al.

    What do we mean by validating a prognostic model?

    Stat Med

    (2000)
  • The World Health Organization MONICA Project (monitoring trends and determinants in cardiovascular disease): a major international collaboration

    J Clin Epidemiol

    (1988)
  • LK Newby et al.

    Bedside multimarker testing for risk stratification in chest pain units: The chest pain evaluation by creatine kinase-MB, myoglobin, and troponin I (CHECKMATE) study

    Circulation

    (2001)
  • Cited by (1097)

    View all citing articles on Scopus
    View full text