Sei sulla pagina 1di 7

Q J Med 2010; 103:679685 doi:10.

1093/qjmed/hcq108 Advance Access Publication 2 July 2010

The ABCD and ABCD2 as predictors of stroke in transient ischemic attack clinic outpatients: a retrospective cohort study over 14 years
J.K. HARRISON1, B. SLOAN1, J. DAWSON2, K.R. LEES2 and D.S. MORRISON1
From the 1Division of Community Based Sciences and 2Division of Cardiovascular and Medical Sciences, Faculty of Medicine, University of Glasgow, Glasgow G12 8RZ, UK
Address correspondence to D.S. Morrison, Director, West of Scotland Cancer Surveillance Unit, Division of Community Based Sciences, University of Glasgow, 1 Lilybank Gardens, Glasgow G12 8RZ, UK. email: d.s.morrison@clinmed.gla.ac.uk
Downloaded from http://qjmed.oxfordjournals.org/ by guest on November 30, 2012

Received 3 March 2010 and in revised form 7 June 2010

Summary
Background: The ABCD and ABCD2 scores have been validated for use as predictors of stroke in community populations up to 90 days after a transient ischemic attack (TIA). TIA outpatient clinics may see a selective group of patients who have not had an early stroke but may be at raised risk in the medium to long term and therefore benefit from preventive treatment. Aim: To describe the prognostic values of the ABCD and ABCD2 scores on long-term stroke risk. Design: Retrospective cohort study of TIA clinic outpatients followed for up to 14 years. Methods: Absolute and relative stroke risks, Kaplan Meier survival curves and cumulative stroke incidence were calculated. Receiver Operating Characteristic curves (ROCs) and areas under the curve were calculated for both scores. Results: Seven hundred and ninety-five patients were included and 138 (17.3%) experienced a stroke within 13.8 years follow-up after first TIA clinic visit, a crude risk of 26.3 per 1000 person-years. Compared with baseline scores of 02, risk ratios for ABCD of 34 were 2.95 (95% CI 1.526.40), and for 56 were 3.42 (95% CI 1.727.54); for the ABCD2, risk ratios for 34 were 2.68 (95% CI 1.375.84), and for 57 were 3.55 (95% CI 1.807.79). Scores of 3 for either ABCD or ABCD2 predicted raised stroke risks at 90 days, 1, 5 and 10 years. Areas under the curve were 0.619 (95% CI 0.5710.668) and 0.630 (95% CI 0.5820.677) for the ABCD and ABCD2 scores, respectively. Conclusion: ABCD and ABCD2 scores of 3 may be clinically useful in identifying TIA outpatients at raised risk of stroke in the medium to long term.

Introduction
Accurate prediction of stroke following a transient ischemic attack (TIA) is important if effective preventive treatment is to be given.1 The ABCD2 uses four clinical variablesage, blood pressure, clinical signs and their durationto assign a score ranging from 0 (lowest risk) to 6 (highest risk). It was derived from the Oxfordshire Community Stroke Project and

predicted 7-day risk of stroke following TIA.2 It has subsequently been validated both for its original 7-day period2 as well as for stroke risks at 2,3 304,5 and 90 days3 in other populations. Its use as a triaging tool in emergency departments has been described.6,7 Some authors have found the ABCD score of limited predictive value, however,8,9 and the additions of computed tomography scans and diabetes mellitus to the score to create the ABCDI4

! The Author 2010. Published by Oxford University Press on behalf of the Association of Physicians. All rights reserved. For Permissions, please email: journals.permissions@oxfordjournals.org

680

J.K. Harrison et al. in both hospital and death records by International Classification of Diseases codes ICD-9 430, 431, 433, 434 and 436 and ICD-10 I60, I61, I63 and I64. The date of last record linkage is 27 December 2006.

and ABCD23 scores, respectively, has been found to deliver better prognostic accuracy compared to the original score.10 Specialist TIA outpatient clinic populations may differ in important ways from patients who have immediately suffered a TIA in the community and from those included in the ABCD2 cohorts. Patients at highest risk of early stroke, including fatal stroke, will be selectively excluded by default. While urgent management must be the aim for all with suspected TIA, subacute prognostication is of greater clinical significance for the majority as only around 5% of patients overall suffer a stroke after TIA at 7 days.11 We therefore evaluated the accuracy of the ABCD and ABCD2 scores in predicting long-term stroke risk among patients attending a TIA outpatient clinic.

Ethics
Formal ethical committee review was deemed not necessary by the Multicentre Research Ethics Committee B for Scotland. ISDs Privacy Advisory Committee also reviewed and accepted the request for the data to be linked and an anonymised file returned to the investigators.

Risk score calculation


The ABCD and ABCD2 scores each sum the following patient characteristics: age (60 years = 1); blood pressure at clinic assessment (systolic >140 mm Hg or diastolic 90 mm Hg = 1); clinical features (unilateral weakness = 2, speech disturbance without weakness = 1, other = 0) and duration of symptoms in minutes (60 = 2, 1059 = 1, <10 = 0). The ABCD2 adds a further point for the presence of diabetes mellitus. For six patients, we lacked information on their diabetic status and could not allocate an ABCD2 score. Because the study used retrospective clinical audit data, compromises in interpretation of the Duration categories of the original ABCD score were required, such that 60 = 2, 559 = 1 and <5 or not recorded = 0.
Downloaded from http://qjmed.oxfordjournals.org/ by guest on November 30, 2012

Materials and methods


Patient selection and outcome identification
The fast-track TIA clinic at the Western Infirmary, Glasgow assesses all suspected TIA patients referred to the service, predominantly from primary care (approximately 500600 outpatient referrals per year). The Infirmary serves a catchment population of approximately 225 000 people. During the period of this study, the clinic was held twice per week. All new patients are reviewed or discussed with Consultant Stroke Physicians with at least 10 years experience. Patients undergo a standardized assessment including history, examination and investigation. Data from this assessment are prospectively recorded onto a standard form and entered into an electronic database. Patients are informed that their data are collected for audit, research and clinical management purposes, and are given the opportunity to decline entry of their details into the Registry. The Registry holds data on consecutive patients from 1992 onwards. Anonymized records from 1992 to November 2004 were obtained for use in this study. Where patients had more than one clinic attendance only the information from the first clinic visit was used. These were linked to the database of all hospital and death records for Scotland by the Information Services Division of the National Health Service in Scotland (ISD) using a variety of probability matching techniques that are described elsewhere.12 Nine hundred and eighty-eight patients had a clinical diagnosis of TIA of whom 934 (94.5%) linked to hospital or death records. Strokes were identified

Statistical analysis
The relationships between stroke risk and ABCD and ABCD2 scores were described by KaplanMeier survival curves and log-rank tests of independence using the 2 distribution. Person-time at risk of stroke was calculated as the difference between date of TIA (or the date of earliest clinic attendance in 111 cases where this date was not available) and date of death or censoring (7 November 2006). Risk ratios for stroke were calculated using baseline ABCD and ABCD2 scores of 02 and the exact Poisson method used to calculate 95% confidence intervals for them. Proportions of strokes within given time intervals (cumulative incidence), from 2 days to 10 years, were calculated with 95% confidence intervals. All patients were observed for at least 1 year but those who were not followed up for 5 or 10 years (because they entered the study later) were not included in cumulative incidence calculations. Receiver Operating Characteristic curves (ROCs) were drawn and areas under the curve calculated for both scores. SPSS version 15.0 software was used for all analyses except confidence

Validation of the ABCD and ABCD2 in a TIA clinic intervals around risk ratios, which were calculated on Stata version 10 software, and confidence intervals around cumulative risks, which were calculated using Confidence Intervals Analysis software version 2.1.1 using methods described by Altman and others.13

681

Results
One hundred and thirty-nine patients were excluded because records contained insufficient data with which to calculate ABCD scores. Seven hundred and ninety-five patients were therefore eligible for inclusion in the study, comprising 453 (57.0%) women and 342 (43.0%) men. Baseline characteristics are described in Table 1. The mean age of participants was 67 years and women were 2.5 years older than men (female 68.1 years and male 65.6 years, independent samples t-test, P = 0.003). Patients were followed from presentation

Table 1 Baseline characteristics of 795 TIA clinic outpatients Variable ABCD score

Age Mean (SD) <60 years, n (%) 60 years, n (%) Blood pressure Systolic, mean (SD) Diastolic, mean (SD) SBP > 140 or DBP  90, n (%) SBP  140 and DBP < 90, n (%) Clinical symptoms, n (%) Motor weakness Speech disturbance Other Duration,a n (%) 0 (<5 min) 1 (559 min) 60 min Diabetes mellitus,b n (%) Yes No Missing data Sex, n (%) Male Female

67.0 (11.9) 197 (24.8) 598 (75.2) 155.1 (26.4) 85.6 (13.5) 232 (29.2) 563 (70.8)

0 1

1 0

240 (30.2) 137 (17.2) 418 (52.6) 236 (29.7) 329 (41.4) 230 (28.9) 65 (8.2) 724 (91.1) 6 (0.8) 342 (43.0) 453 (57.0)

2 1 0 0 1 2 1 0

at clinic for up to 13.8 years between April 1992 and November 2006. The median delay from symptoms to clinic assessment during the study period was 15 days. One hundred and thirty-eight patients (17.3%) experienced a stroke: 26 were fatal events and 112 non-fatal. Mean time to stroke was 2.4 years (SD 2.6). The crude risk of stroke was 26.3 per 1000 person-years. The risk of stroke was significantly greater in patients with medium (34) and high (56) ABCD scores when compared with low (02) scoressee Figure 1. There was no difference in risk between medium and high score groups (2 = 0.871, df = 1, log-rank P = 0.351). A similar pattern was found in survival comparisons for the ABCD2 (Figure 2), with a significant difference in survival between the three groups and a small but statistically non-significant difference between medium and high groups (log-rank 2 = 2.792, df = 1, P = 0.095). Absolute and relative risks with baseline ABCD and ABCD2 scores of 02 in each case are shown in Table 2. Medium and high ABCD and ABCD2 scores were associated with 3-fold increases in stroke risk and whilst high scores predicted greater risk of stroke than medium scores, the differences were clinically small and not statistically significant. Cumulative risks of stroke are presented between 90 days and 10 years in Table 3. The overall absolute risk of stroke after attending an outpatient clinic with a TIA was 7.3% within 1 year, 16.2% within 5 years and 28.0% within 10 years. Patients with high ABCD and ABCD2 scores experienced stroke risks of 89% within 1 year, 2022% within 5 years and 37% within 10 years. The ABCD2 did not predict higher risks in any period compared to the ABCD. The majority of strokes had occurred within 5 years after which there was little difference in risk. In each case, there was no significant difference in stroke risk between patients with medium and high ABCD or ABCD2 scores. The ROC curves for ABCD and ABCD2 scores on predicting stroke risk are described in Figure 3. Areas under the ABCD and ABCD2 curves are 0.619 (95% CI 0.5710.668) and 0.630 (95% CI 0.5820.677), respectively.

Downloaded from http://qjmed.oxfordjournals.org/ by guest on November 30, 2012

Discussion
We found that both ABCD and ABCD2 scores were highly predictive of stroke risk among TIA clinic outpatients. Both scores predicted a 3-fold increase in stroke risk between low (02) and medium or high (3) categories, with statistically significant differences in the risks between medium and high

a Original ABCD categories are <10 min (0), 1059 min (1) and 60 min (2). b Additional ABCD2 score variable.

682

J.K. Harrison et al.

1.0

ABCD low (0-2) medium (3-4) high (5-6) low-censored medium-censored high-censored

0.8

survival probability

0.6

0.4

0.2

Downloaded from http://qjmed.oxfordjournals.org/ by guest on November 30, 2012

0.0 0 2 4 6 8 10 12 14

time (years)

Figure 1. Cumulative survival to stroke by ABCD score, grouped (low, 02; medium, 34; high, 56). One hundred and thirty-eight strokes in 795 patients. 2 = 14.443, df = 2, log-rank P = 0.001.

1.0

0.8

ABCD2
low (0-2) medium (3-4) high (5-7) low-censored medium-censored high-censored

survival probability

0.6

0.4

0.2

0.0 0 2 4 6 8 10 time (years) 12 14

Figure 2. Cumulative survival to stroke by ABCD2 score, grouped (low, 02; medium, 34; high, 57). One hundred and thirty-eight strokes in 789 patients. 2 = 15.500, df = 2, log-rank P < 0.001.

Validation of the ABCD and ABCD2 in a TIA clinic groups found for ABCD2 scores at 1 year or more. Overall risks doubled between 90 days and 1 year after TIA and doubled again in the following 5 years. ABCD and ABCD2 scores each consistently predicted individuals at higher risk of stroke from 90 days to 10 years. ABCD and ABCD2 scores may therefore be most useful for determining stroke risks between 90 days and 10 years in TIA clinic outpatients. We found no significant prognostic advantage from adding diabetes mellitus to the original ABCD score to create the ABCD2. The overall diagnostic accuracy of the ABCD and ABCD2 scores was significantly better than chance, as measured by ROC areas under the curve and their associated 95% CIs being above

683

Table 2 Number, absolute risk per 1000 person-years (p-y) and risk ratio of stroke following TIA by ABCD score ABCD n Strokes per Risk ratio strokes 1000 p-y (95% CI)

ABCD Low (02) 10 Medium (34) 75 High (56) 53 ABCD2 Low (02) 10 Medium (34) 69 High (57) 59 Total 138

9.62 28.36 32.88 9.97 26.73 35.45 26.29

1 2.95 (1.526.40) 3.42 (1.727.54) 1 2.68 (1.375.84) 3.55 (1.807.79)

Baseline for risk ratio is ABCD or ABCD2 = 02.

the null value of 0.5. However, the ABCD and ABCD2 areas under the curve of 0.62 and 0.63, respectively, were poorer than quoted evaluations of the ABCD over shorter periods which are around 0.760.81.2,4,5,10 Our findings may reflect the inevitable exclusion of patients at high risk of early stroke from the TIA outpatient population. The ABCD2 identifies symptoms that may predict non-cerebrovascular as well as cerebrovascular outcomes and its diagnostic accuracy for these conditions (principally syncope, migraine and seizures) has previously been reported to be 0.75 (95% CI 0.730.76).14 Our study has limitations. The diagnostic precision of TIA clinic records has not been validated and misclassification of patients is possible. However, it seems reasonable to assume that the quality of clinical assessment in a specialist tertiary clinic will be satisfactory. A small percentage of clinic records (5.5%) could not be linked to either hospital or death records. This may be because patients truly did not experience either hospital admission or death or because of a technical failure of record linkage. The probability matching techniques used by ISD result in 3% of records in the patient-based hospital and death record database being mismatched.15 However, this results in as many errors of overmatching (where an individual is incorrectly linked to a hospitalization or death that did not occur) as under matching. Thus, it is unlikely to account for a net shortfall in successful links. We relied upon stroke diagnoses appearing correctly in both hospital and death records and it is possible that some strokes were not classified within our

Downloaded from http://qjmed.oxfordjournals.org/ by guest on November 30, 2012

Table 3 Stroke risk (%) at 90 days and at 1, 5 and 10 years after TIA among TIA clinic outpatients, by ABCD (n = 795) and ABCD2 (n = 789) score Risk period n ABCD Low (02) Medium (34) High (56) Total ABCD2 Low (02) Medium (34) High (57) Total 90 days n 1 year n 5 yearsa n 10 yearsb

1/152 0.7 (0.13.6) 5/152 15/403 3.7 (2.36.0) 34/403 8/240 3.3 (1.76.4) 19/240 3.0 (2.04.5) 1/145 13/390 10/254 24/789 0.7 3.3 3.9 3.0 (0.13.8) (2.05.6) (2.27.1) (2.04.5) 5/145 29/390 24/254 58/789

3.3 8.4 7.9 7.3 3.4 7.4 9.4 7.4

(1.47.5) (6.111.6) (5.112.0) (5.79.3) (1.57.8) (5.210.5) (6.413.7) (5.79.4)

7/112 6.3 (3.112.3) 4/32 56/328 17.1 (13.421.5) 24/90 41/203 20.2 (15.326.2) 26/71 104/643 16.2 (13.519.2) 54/193 7/107 6.5 (3.212.9) 4/32 50/318 15.7 (12.120.1) 22/85 47/212 22.2 (17.128.2) 28/75 104/637 16.3 (13.719.4) 54/192

12.5 26.7 36.6 28.0 12.5 25.9 37.3 28.1

(5.028.1) (18.636.6) (26.448.2) (22.134.7) (5.028.1) (17.836.1) (27.348.6) (22.234.9)

95% confidence limits in brackets. a 643 and 637 followed-up for 5 years. b 193 and 192 followed-up for 10 years in ABCD and ABCD2, respectively.

684

J.K. Harrison et al.

Downloaded from http://qjmed.oxfordjournals.org/ by guest on November 30, 2012

Figure 3. ROC curves for ABCD (06) and ABCD2 (07) scores on stroke outcomes in up to 13.8 years follow-up. Reference line indicates null predictive score values.

International Classification of Diseases criteria or were omitted from discharge summaries. While the diagnostic quality of Scottish hospital data is high, with recent audit indicating that 88% of diagnoses were correctly coded in the electronic summary,16 errors still occur. Patients who experienced a non-fatal stroke but did not require hospitalization will be omitted by our methodology. Data from a Scottish community-based incidence study indicated hospitalization rates of >90%.17 We used 5 min, rather than 10, as the threshold to define the shortest TIAs. A 10-min threshold has been reported since the original 7-day follow-up study18 on which the ABCD was based but our clinic records record only categorical information on TIA duration as described above. The proportions of patients in each of the Duration categories was very similar to those reported elsewhere.2 Our data set will also exclude a number of TIA patients who suffered very early recurrent stroke because of the delay to clinic assessment. The role of the ABCD score in this context is well established. While our data may thus underestimate early stroke risk, we believe this apparent limitation actually strengthens our conclusion. The aim of the study was to establish long-term stroke risk in those who do not suffer early stroke and to evaluate long-term predictive ability of commonly used scores for early stroke. Our data are thus relevant to the majority of

TIA patients and remain so, even though medical management of those with TIA has changed over time. Although recommendations in the UK,19,20 Europe21 and the USA1 advise urgency of assessment to minimize early stroke risk, a substantial number of patients will continue to be seen at outpatient clinics days and weeks after the episode due to delays in patient presentation. We also acknowledge that absolute stroke risks after TIA may also change but we have no reason to suspect that risk ratios between ABCD categories will similarly change. In conclusion, we have shown that TIA clinic outpatients remain at risk of stroke and that the simple predictive tools used to identify those at risk of early stroke can also be used to identify patients at significantly raised risk of stroke in the medium term and longer term.

Acknowledgements
K.R.L. and J.D. collected the data. J.K.H., B.S., J.D. and D.S.M. were involved in the design and analysis of the study. J.K.H., J.D. and D.S.M. drafted the paper. There was no external funding source for this study. Conflicts of interest: None declared.

Validation of the ABCD and ABCD2 in a TIA clinic

685

References
1. Wolf PA, Clagett GP, Easton JD, Goldstein LB, Gorelick PB, Kelly-Hayes M, et al. Preventing ischemic stroke in patients with prior stroke and transient ischemic attack. Stroke 1999; 30:19914. 2. Rothwell PM, Giles MF, Flossman E, Lovelock CE, Redgrave JNE, Warlow CP, et al. A simple score (ABCD) to identify individuals at high early risk of stroke after transient ischemic attack. Lancet 2005; 366:2936. 3. Johnston SC, Rothwell PM, Nguyen-Huynh MN, Giles MF, Elkins JS, Bernstein AL, et al. Validation and refinement of scores to predict very early stroke after transient ischemic attack. Lancet 2007; 369:28392. 4. Sciolla R, Melis F. for the SINPAC Group. Rapid identification of high-risk transient ischemic attacks. Prospective validation of the ABCD Score. Stroke 2008; 39:297302. 5. Tsivgoulis G, Spengos K, Manta P, Karandreas N, Zambelis T, Zakopoulos N, et al. Validation of the ABCD Score in identifying individuals at high early risk of stroke after a transient ischemic attack. Stroke 2006; 37:28927. 6. Byrne A, Daly C, Rocke L, Gray J. Can risk stratification of transient ischemic attacks improve patient care in the emergency department? Emerg Med J 2007; 24:63740. 7. Bray JE, Coughlan K, Bladin C. Can the ABCD Score be dichotomised to identify high-risk patients with transient ischemic attack in the emergency department? Emerg Med J 2007; 24:925. 8. Cucchiara BL, Messe SR, Taylor RA, Pacelli J, Maus D, Shah Q, et al. Is the ABCD Score useful for risk stratification of patients with acute transient ischemic attack? Stroke 2006; 37:17104. 9. Molina CA, Montaner J, Alvarez-Sabn J. Absence of usefulness of ABCD Score in the early risk of stroke of transient ischemic attack patients. Stroke 2007; 38:8556. 10. Tsivgoulis G, Vassilopoulou S, Spengos K. Potential applicability of the ABCD2 score in triaging TIA patients. Lancet 2007; 369:1082.

11. Giles MF, Rothwell PM. Risk of early stroke after transient ischemic attack: a systematic review and meta-analysis. Lancet Neurol 2007; 6:106372. 12. Kendrick SW, Clarke JA. The Scottish Medical Record Linkage System. Health Bull 1993; 51:729. 13. Newcombe RG, Altman DG. Proportions and their differences. In: Altman D, Machin D, Bryant T, Gardner S, eds, Statistics with Confidence. Chapter 6, 2nd edn. BMJ Books, London, 2000. 14. Quinn TJ, Cameron AC, Dawson J, Lees KR, Walters MR. ABCD2 scores and prediction of noncerebrovascular diagnoses in an outpatient population: a case-control study. Stroke 2009; 40:74953. 15. Walsh D, Smalls M, Boyd J. Electronic health summaries: building on the foundation of Scottish Record Linkage system. Medinfo 2001; 10(Pt 2):12126. 16. NHS Hospital Data Quality. Towards better data from Scottish Hospitals. An assessment of SMR01 and associated data 20042006. Edinburgh: ISD Scotland Publications, 2007. 17. Syme PD, Byrne AW, Chen R, Devenny R, Forbes JF. Community-based stroke incidence in a Scottish population. Stroke 2005; 36:183743. 18. Johnston CC, Gress DR, Browner WS, Sidney S. Short-term prognosis after emergency department diagnosis of TIA. JAMA 2000; 284:29016. 19. Scottish Intercollegiate Guidelines Network. Management of Patients with Stroke or TIA, Guideline number 108. Edinburgh, 2008. 20. National Institute for Health and Clinical Excellence. Guideline 68. Stroke. Diagnosis and initial management of acute stroke and transient ischaemic attack(TIA). NICE, London, July 2008. 21. The European Stroke Organisation (ESO) Executive Committee and the ESO Writing Committee. Guidelines for Management of Ischaemic Stroke and Transient Ischaemic Attack 2008. Cerebrovasc Dis 2008; 25:457507.

Downloaded from http://qjmed.oxfordjournals.org/ by guest on November 30, 2012

Potrebbero piacerti anche