Etiology, Duration, and Prognosis of Transient Ischemic Attacks: An Analysis From the German Stroke Data Bank

Context  A transient ischemic attack (TIA) has been arbitrarily defined as a focal cerebral ischemic deficit lasting less than 24 hours.

Objective  To determine if TIAs of short duration (<1 hour) and long duration (1 hour to <24 hours) differ from each other and from ischemic stroke (IS).

Design, Setting, and Patients  Inception cohorts of 1429 patients with acute TIAs and 5206 patients with IS were prospectively documented in 15 German medical centers with neurology departments and acute stroke units. Outcome after 3 months was assessed in 72.8% of the patients with TIAs.

Main Outcome Measures  Risk factor distribution, etiology, and prognosis of TIAs and IS.

Results  Patients with TIAs, especially those with symptoms lasting less than 1 hour, were significantly more likely to have a history of TIAs and less likely to have diabetes mellitus, arterial hypertension, or atrial fibrillation at admission compared with those with IS. Cardioembolic etiologies were less frequent and unknown etiologies more frequent among patients with TIAs than those with IS. Functional outcome and mortality did not differ significantly in patients with TIAs of different durations.

Conclusion  This study demonstrates differences in comorbidity and etiology among patients with TIAs of different durations and IS.

THE NATIONAL Institute of Neurological Disorders and Stroke's "Classification of Cerebrovascular Diseases III"¹ defines transient ischemic attacks (TIAs) as brief episodes of focal loss of brain function lasting less than 24 hours that are thought to be due to ischemia, that can usually be localized to a portion of the brain supplied by 1 vascular system (left or right carotid or vertebrobasilar system), and for which no other cause can be found. The incidence of first-time TIAs constitutes roughly one quarter of the incidence of initial cerebrovascular events.^2-8 Although TIAs are generally regarded as significant risk factors for subsequent stroke, the criterion of symptoms lasting less than 24 hours has been criticized as having little relevance for diagnosis and secondary prevention.⁹ To date, only 1 study with small patient numbers has investigated the clinical characteristics of patients with different durations of neurological deficits following cerebral ischemia. It found a higher frequency of cardiac and large-vessel disease among patients with long-duration TIAs.¹⁰

To fulfill one of the predefined aims of the German Stroke Data Bank, we investigated the hypothesis that short- (<1 hour) and long-duration TIAs (1 hour to <24 hours) differ in terms of risk factor distribution, etiology, and prognosis compared with ischemic stroke (IS).

Data were prospectively collected between 1998 and 1999 within the German Stroke Data Bank of the German Stroke Foundation (Stiftung Deutsche Schlaganfall-Hilfe, Gütersloh) in 23 hospitals with neurology departments and acute stroke units. Each one serves a catchment area of more than 100 000 inhabitants and is the main source of care for patients with stroke in its region. Prospectively collected data included age, sex, times of the event and admission, risk factors, vascular comorbidity, prior medication use, National Institutes of Health Stroke Scale (NIHSS) score at admission, functional independence on the Barthel Index and modified Rankin scale before the event and after admission, results of additional clinical investigations such as cerebral imaging, extracranial ultrasonography, and transesophageal echocardiography, Trial of Org 10172 in Acute Stroke Treatment (TOAST) classification,¹¹ complications, length of stay in the documenting hospital, and medications for secondary prevention at discharge. Follow-up interviews after 3 and 12 months assessed functional independence on the Barthel Index and modified Rankin scale or cause of death in patients with TIAs and IS. Data were collected using a standardized questionnaire based on an extensive manual, as well as video training for the assessment of the NIHSS score.

The NIHSS assesses deficits in 15 neurological functions on a scale of 0 (no deficit) to 42. The modified Rankin scale quantifies overall functional status on a score of 0 (no handicap) to 5 (bedridden, incontinent, and requiring constant nursing care and attention).¹² Risk factors were defined as follows: arterial hypertension as a history of elevated blood pressure (>160/90 mm Hg) on 2 independent readings before stroke or the use of antihypertensive medication; diabetes mellitus as a history of elevated blood glucose levels on 2 independent readings before stroke, elevated glycosylated hemoglobin level (>7.5%) at admission, or the use of antidiabetic medication; hypercholesterolemia as a history of elevated total cholesterol levels (>220 mg/dL [>5.69 mmol/L]) on 2 independent readings before stroke or the use of lipid-lowering medication; and smoking currently or during the last 5 years. Information on risk factors and other comorbidity was obtained primarily from the patient; additional information was collected from relatives and primary care physicians.

A diagnosis of TIA was made in cases of clinical deficits lasting less than 24 hours regardless of an infarction seen on cerebral imaging scans.¹ The duration of TIAs was stratified into intervals of less than 1 hour, 1 to less than 12 hours, and 12 to less than 24 hours. For the purpose of this analysis, the latter 2 categories respresent long-duration TIAs (1 to <24 hours). The etiological classification was based on the TOAST criteria and was scored according to a standardized protocol by the investigators at each site.^11,13 After a final consistency check with the source data at each site, questionnaires were sent to the data management centers at the University of Essen (Essen, Germany) and the German Stroke Foundation. They were rechecked by 2 physicians for completeness and consistency of the classification with additional diagnostic results. Questions regarding missing or implausible data were directed to the treating neurologist. Data were entered into the database in duplicate by trained personnel. Data quality was further ensured with monthly reports and clinical site visits. All patients gave informed consent if their personal data were to be transferred to the data management center for central follow-up. Otherwise the follow-up was performed in the documenting center, and the results were forwarded to the data management center. To assess outcome after both recurrent TIAs and IS, we considered all information obtained from the patient, relatives, and treating physicians.

A total of 9296 patients with TIAs or IS were recruited for this study. Of these, all 2661 patients from 8 of the participating centers were excluded because we could not rule out a selection bias in these hospitals. Thus, our analyses include 6635 patients from 15 centers. Follow-up after 3 months (median, 102 days; range, 70-150 days) was obtained in 1040 (72.8%) of the patients with TIAs, and the results were considered to indicate the respective patients' status after 3 months. Of the patients with TIAs, 1.4% refused to participate in the follow-up, 7% could not be contacted during the follow-up window, and 18.8% were not contacted at all, partly because of limited funding for central follow-up or lack of staff to ensure local inquiries by the participating centers (Figure 1). Comparison of the patients included in the follow-up and those lost to follow-up showed no significant differences in age, sex, duration of symptoms, prior stroke, prior TIAs, ischemic heart disease, arterial hypertension, or diabetes mellitus.

Statistical analysis was performed using SPSS version 9.0 statistical software (SPSS Inc, Chicago, Ill). Continuous variables are presented as mean and median. Categorical variables are presented as percentages. For statistical comparisons between patients with TIAs of different durations and patients with IS, as well as the comparison of short- and long-duration TIAs, we used the Fisher exact test for categorical variables and the 2-tailed t test for continuous variables. To account for multiple testing, P<.01 was considered significant.

Of the 6635 patients with cerebral ischemia included in this analysis, 1429 patients (21.5%) had a final diagnosis of TIA, and 5206 (78.5%) had a final diagnosis of IS. Among those with TIAs, neurological symptoms were completely resolved in less than 1 hour in 423 patients (29.7%), in 1 to less than 12 hours in 429 patients (30.1%), and in 12 to less than 24 hours in 577 patients (40.5%).

The frequencies of comorbidity and risk factors are listed in Table 1. There were only minor differences in age, sex, and prior functional independence between patients with different durations of TIAs and those with IS. The rates of prior TIAs and hypercholesterolemia were significantly higher in patients with TIAs than in those with IS, whereas the rates of smoking, arterial hypertension, diabetes mellitus, and atrial fibrillation at admission were significantly lower in patients with TIAs. Patients with symptoms lasting less than 1 hour reported the highest rates of prior TIAs and hypercholesterolemia but had the lowest rates of prior stroke and atrial fibrillation at admission.

The median delay between the event and admission was 135 minutes in patients with TIAs compared with 111 minutes in patients with IS. At admission, 57% of patients with TIAs had some neurological deficits according to the NIHSS: usually limb weakness (25%), facial weakness (22%), sensory deficits (21%), or aphasia (11%). The mean NIHSS score at admission was 1.8 (median, 1).

Additional examinations were performed in similar percentages of patients with TIAs and IS. These included cerebral imaging (TIAs and IS: 99.9%), extracranial ultrasonography (TIAs: 97.7%; IS: 95.9%), and transesophageal echocardiography (TIAs: 32.6%; IS: 35.9%). Table 2 indicates the frequency of pathologic findings on these examinations. Patients with TIAs had significantly fewer ipsilateral stenoses with more than a 50% reduction in diameter on extracranial ultrasonography and fewer cardiac sources of embolism on transesophageal echocardiography than patients with IS. In addition, prior infarctions detected using cerebral imaging were significantly less frequent in patients with short-duration compared with long-duration TIAs.

The frequencies of TOAST classification etiologies are presented in Table 3. The use of this classification system resulted in a high percentage of unknown etiologies. Among patients with TIAs, small vessel disease was significantly more common and cardioembolic causes were significantly less common compared with patients with IS. Compared with long-duration TIAs, short-duration TIAs were classified significantly more often as large artery atherosclerosis and less often as small vessel disease.

The mean length of stay in the documenting hospital was 7.7 days (median, 7 days) for patients with TIAs and 11.9 days (median, 10 days) for patients with IS. At the time of discharge from the hospital, 82% of patients with TIAs were prescribed platelet inhibitors (49% receiving aspirin), 12% were prescribed oral anticoagulants, 4.5% were given intravenous heparin (most of these patients were transferred to another hospital), and 1.5% were discharged without any preventive medication. Of the patients with IS, 69.4% were given platelet inhibitors (43.7% receiving aspirin), 18.4% were prescribed oral anticoagulants, 6.2% received intravenous heparin, and 6.0% were given no preventive medication.

By the 3-month follow-up visit, 18 patients (1.7%) with TIAs had died (Figure 2) compared with 551 patients (14.0%) with IS. Only 3 patients with TIAs had died due to a new stroke. Although more patients with short-duration TIAs experienced a recurrent cerebrovascular event during their hospital stay than those with long-duration TIAs (data not shown), there was no significant difference in functional outcome after 3 months (Table 4).

In this large, hospital-based cohort study, we investigated the hypothesis that TIAs of varying duration differ from each other and from IS in terms of etiology, risk factor distribution, and prognosis. Because this was not an epidemiological study, the population can be considered representative only for specialized stroke care facilities. However, the prevalence of TIAs in all patients with cerebral ischemia (21.5%) was similar to that observed in epidemiological studies.^3,4,8 In contrast, the prevalence of long-duration TIAs (70%) was unexpectedly high. This might be explained by the fact that patients with singular symptoms of shorter duration are less likely to see a physician or be admitted to the hospital. Whereas epidemiological studies to date have not investigated the duration of symptoms in TIAs, other hospital-based studies have reported a mean duration of 207 minutes and symptoms lasting more than 30 minutes in 50% of all patients with TIAs.^14,15

Risk factors such as atrial fibrillation, arterial hypertension, diabetes mellitus, and smoking were significantly less frequent in patients who had TIAs compared with those with IS. An inverse relationship was seen for hypercholesterolemia and prior TIAs. Several population-based studies have shown no significant differences in risk factor distribution between TIAs and IS, which could be because of the exclusion of more severe strokes as well as smaller patient numbers.^16-18

Among patients with short-duration TIAs, the frequency of prior TIAs was significantly elevated, whereas that of prior stroke was significantly reduced. Apart from small vessel disease, which was diagnosed more often in patients with TIAs, known etiologies such as cardioembolic events were reduced; a high percentage of etiologies remained unknown in these patients. However, the classification of large artery atherosclerosis was more frequent among patients who had TIAs with symptoms lasting less than 1 hour and patients with IS than among those who had TIAs with symptoms lasting 1 hour or more. This contrasts with the results of a small study by Kimura et al,¹⁰ which found an increase in pathologic vascular findings with increasing duration of symptoms. Because the diagnostic workup after a TIA was very similar to that after IS, the high percentage of unknown etiologies according to the TOAST criteria might be indicative both of transient pathologic findings and diagnostic uncertainty regarding TIAs.

The mortality rate of patients with TIAs was slightly lower in our study than in a large hospital-based study in San Francisco, Calif, that reported a mortality rate of 2.6% within the first 90 days, of which 44% of deaths were due to a new stroke.¹⁴ The overall 90-day stroke recurrence rate in that study was 10.5%, more than half of which occurred within the first 2 days after the initial TIA. This suggests that TIAs must be considered serious events requiring immediate action. Similarly, a smaller population-based study in Rochester, Minn, reported a 3-month stroke recurrence rate of 10% following TIAs.⁷ A study from Oxford, England, reported stroke in 8.8% of patients with TIAs during the first 6 months after the event and 11.6% during the first year, but it excluded all patients who experienced a stroke prior to inclusion in the study.¹⁹ In our study, recurrent TIAs and stroke could not be reliably differentiated and are therefore the subject of an ongoing study.

Although several studies have proposed possible predictors for stroke recurrence after TIAs,^14,20-24 only 1 has attempted an external validation.²⁴ Because the prognostic models identified in these studies had low predictive accuracy, we did not analyze possible predictors for recurrent cerebral ischemia.

Some investigators have suggested that the differences between cerebral ischemias of different duration are quantitative rather than qualitative.²⁵ However, the arbitrary definition of TIA as an episode lasting less than 24 hours has little relevance for clinical decision making, nor is it supported by modern imaging techniques.²⁶ Our results support the concept of short-duration TIA as an additional entity, as do those of an earlier study.¹⁰

Our study has several limitations. In most of the participating hospitals, patients were included consecutively and without the application of selection criteria. We cannot exclude the possibility that some patients with short-duration TIAs may have been systematically missed, which could have influenced our results. We were also unable to ensure complete follow-up of all patients. Nevertheless, a comparison between the patients included in the study and those lost to follow-up showed no significant differences in baseline characteristics.

In conclusion, we found differing etiologies and risk factors for short- and long-duration TIAs and IS. Whereas patients with short-duration TIAs showed the highest prevalence of prior TIAs and hypercholesterolemia as strong risk factors for cerebrovascular disease, they had a low prevalence of prior stroke and visible prior infarction. Whether this corresponds to a predominant risk of recurrent TIAs rather than stroke is the subject of an ongoing study. Meanwhile, any TIA should be regarded as a serious event requiring a complete etiologic workup and appropriate medical prevention strategies to prevent recurrent cerebral ischemia with potentially more disastrous consequences.

Accepted for publication April 24, 2002.

Author contributions: Study concept and design (Drs Weimar, Rödl, and Diener); acquisition of data (Drs Weimar, Rödl, Hippe, Harms, and Kloth); analysis and interpretation of data (Drs Weimar, Kraywinkel, and Harms); drafting of the manuscript (Drs Weimar and Kraywinkel); critical revision of the manuscript for important intellectual content (Drs Rödl, Hippe, Harms, Kloth, and Diener); statistical expertise (Drs Diener and Kraywinkel); obtained funding (Drs Weimar and Diener); administrative, technical, and material support (Dr Weimar); study supervision (Drs Weimar and Diener).

This study was supported by the Stiftung Deutsche Schlaganfall-Hilfe, Gütersloh, Germany; Heinz-Nixdorf-Foundation, Essen, Germany; Bayer, Leverkusen, Germany; Bristol-Myers-Squibb, Munich, Germany; Boehringer Ingelheim, Ingelheim, Germany; Glaxo-Wellcome (now GlaxoSmithKline), Munich; Janssen-Cilag, Neuss, Germany; Knoll, Ludwigshafen, Germany; Sanofi-Synthelabo, Berlin, Germany; and Schering, Berlin.

We thank the Stiftung Deutsche Schlaganfall-Hilfe for help with data collection.

Departments of Neurology and Principal Investigators in Germany

Klinikum Minden, Minden: Joerg Glahn, MD. Krankenhaus München-Harlaching, Munich: Martin L. J. Wimmer, MD. University of Essen, Essen: Christian Weimar, MD. University of Homburg, Homburg: Michael P. Roth. University of Leipzig, Leipzig: Dietmar S. Schneider, MD. Benjamin Franklin University, Berlin: Christian Koennecke, MD. Asklepiosklinik Salzhausen, Nidda–Bad Salzhausen: Gerhard-Michael v. Reutern, MD. University of Greifswald, Greifswald: Jochen Machetanz, MD. University of Rostock, Rostock: Antje Kloth, MD. St Katharinen Hospital, Frechen: Rolf Adams, MD. University of Magdeburg, Magdeburg: Michael Goertler, MD. University of Jena, Jena: Joachim Roether, MD. University of Ulm, Ulm: Matthias Riepe, MD. Bürgerhospital Stuttgart, Stuttgart: Elisabeth Schmid, MD. Krankenanstalten Gilead, Bielefeld: Angela Schacker, MD.

Corresponding author and reprints: Christian Weimar, MD, Neurologische Klinik der Universität-GHS Essen, Hufelandstr 55, 45122 Essen, Germany (e-mail: stroke.med@uni-essen.de).