OUP user menu

Prognosis of vertebrobasilar transient ischaemic attack and minor stroke

Enrico Floßmann, Peter M. Rothwell
DOI: http://dx.doi.org/10.1093/brain/awg197 1940-1954 First published online: 7 July 2003

Summary

Vertebrobasilar (VB) territory transient ischaemic attacks (TIAs) and minor strokes are perceived to have a better prognosis than carotid territory events, and are sometimes managed less aggressively. However, this notion stems mainly from a few small studies in the 1960s and 1970s, and has not been systematically tested. We therefore identified all published studies of prognosis after TIA or minor stroke using MEDLINE and EMBASE, and hand‐searching reference lists and relevant journals. In addition, we attempted to include all available individual patient data (IPD) from studies that had not published outcome data by territory of presenting event. Odds of recurrent events were calculated within studies and combined by fixed‐effects meta‐analysis. Heterogeneity between studies was calculated using the χ2 method. We stratified the analysis by time interval between presenting event and inclusion in study, and by study setting (population‐based, published hospital‐based and unpublished hospital‐based). Public ation bias was tested for by linear regression of the standard normal deviate against precision. Eight hundred and twenty abstracts were reviewed, and 304 papers were considered in detail. Of these, 43 studies representing 36 independent cohorts (12 196 patients) reported outcomes by territory of presenting event. IPD from five studies (4643 patients) were also included. The following results compare relative risks of VB with carotid events. Studies including the acute phase (up to 7 days) after the presenting event found a higher relative risk of subsequent stroke in patients with VB events [odds ratio (OR) 1.47, 95% confidence interval (CI) 1.1–2.0, P = 0.014]. Conversely, studies mainly recruiting after the acute phase found a lower relative risk (OR 0.74, 95% CI 0.7–0.8, P = 0.00001). In published hospital‐based studies, the risk of recurrent stroke was lower for patients presenting with VB events (OR 0.68, 95% CI 0.6–0.8, P < 0.00001). However, there was no difference in hospital‐based IPD (OR 1.02, 95% CI 0.8–1.3, P = 0.91). Moreover, in population‐based studies, patients with VB events had a higher risk of stroke (OR 1.48, 95% CI 1.1–2.0, P = 0.025). There was no within‐stratum heterogeneity. There was no difference in the risk of fatal stroke (OR 1.04, 95% CI 0.8–1.4, P = 0.90). Therefore, we found no evidence that patients presenting with VB events have a lower risk of subsequent stroke or death compared with patients presenting with carotid TIA or minor stroke. Indeed, their risk of stroke is probably higher in the acute phase. Patients with VB events require active preventive treatment.

  • Keywords: cerebral ischaemia; posterior circulation; transient; vertebrobasilar; prognosis
  • Abbreviations: IPD = individual patient data; TIA = transient ischaemic attack; VB = vertebrobasilar

Introduction

Patients presenting with transient ischaemic attack (TIA) or minor stroke are at high risk of major vascular events. Five year risks of stroke, coronary events and death vary from 8 to 30%, 12 to 20% and 8 to 37%, respectively, and 57–79% of deaths are vascular (Whisnant et al., 1973; Simonsen et al., 1981; Muuronen and Kaste, 1982; Heyman et al., 1984a; Howard et al., 1987; Whisnant and Wiebers, 1987; Sørensen et al., 1989; Dennis et al., 1990b; Larsen et al., 1990). However, not all TIAs and minor strokes have the same prognosis. For example, amaurosis fugax is consistently associated with a lower risk of subsequent stroke than carotid territory TIAs (Dennis et al., 1989; Hankey et al., 1991, 1992; Dutch TIA Trial Study Group, 1993; Streifler et al., 1995; Benavente et al., 2001). Such differences in prognosis may influence treatment decisions, and it is important that they are determined reliably.

Vertebrobasilar (VB) territory events account for about 30% of all TIAs and minor strokes. In contrast to carotid events, where research has been stimulated by the development of carotid endarterectomy, there has been relatively little systematic research into the prognosis and risk factors for recurrent vascular events specifically in patients with VB territory TIAs and minor strokes. Despite this, there is a widely held view that VB territory events have a more benign prognosis than carotid territory events (McDowell et al., 1961; Bradshaw and McQuaid, 1963; Marshall, 1964; Baker et al., 1968; Ziegler and Hassanein, 1973; Olsson et al., 1976; Heyman et al., 1984b; Turney et al., 1984; Sivenius et al., 1991; Mohr et al., 1992; Caplan, 1996). This idea stems from the results of a small number of early cohort studies performed in the 1960s and 1970s (Bradshaw and McQuaid, 1963; Marshall, 1964; Acheson, 1971; Baker, 1971; Olsson et al., 1976), many of which do not satisfy modern methodological standards (Kernan et al., 1991). However, partly as a consequence of these early studies, patients with VB events are often investigated less rigorously than patients with carotid events (Caplan, 2000; Culebras et al., 1997) and may not always receive as aggressive preventive treatment against future vascular events (Caplan, 1996; Martin, 1998).

To determine whether or not the prognosis of VB territory TIAs and minor strokes does differ from that of carotid territory events, we reviewed all published studies which reported outcome events according to the vascular territory of the presenting event and included all available individual patient data (IPD) from studies that had not published outcome data by territory of presenting event, and determined the odds of recurrent TIAs, strokes and death.

Material and methods

Search strategy

We aimed to identify articles which reported follow up data on patients presenting with TIA or minor stroke and which reported these separately for patients presenting with VB and carotid events. Studies were identified by a single observer from MEDLINE®+ and EMBASE® (Silverplatter Winspirs 4.0 on‐line and Entrez‐PubMed NIH August 5, 2001 for the period 1966 to week 1 August 2001) with the search terms: [posterior circulation OR vertebrobasilar] AND [stroke OR CVA OR TIA OR cerebrovascular OR transient isch(a)emic], and TIA AND [natural history OR prognosis]. The search was limited to human studies only. No restriction was made on the language of publication. Journals that yielded more than 10% of all studies identified electronically were systematically hand‐searched for further relevant studies. The reference lists of all papers that met the inclusion criteria were also searched. By contact with principal investigators we attempted to obtain all available IPD from cohort studies containing outcome events for patients with both carotid and VB events that had previously not published outcome data by vascular territory of the presenting event.

Inclusion criteria

Articles were included in the review if they fulfilled the following criteria: (i) prospective or retrospective cohort study; (ii) included patients with TIA, reversible ischaemic attack, reversible ischaemic neurological deficit or minor stroke (defined as reversible, non‐disabling, Rankin scale grade not worse than 3, or similar) in VB and carotid territories; (iii) presenting events and outcome data reported separately according to the vascular territory of presenting event; (iv) outcome data reported for one or more of the following: recurrent TIA, stroke (all strokes, major stroke or fatal stroke), myocardial infarction, vascular death or overall mortality.

Data extraction

The following information was extracted from eligible reports:

(i) Aims of study. (ii) Type of study (cohort study, non‐randomized treatment trial, randomized controlled trial). (iii) Setting of the study (population‐based, hospital‐based). (iv) Inclusion and exclusion criteria, and number of excluded patients. (v) Basic demographic data for patients with events in different vascular territories. (vi) Criteria used to determine the vascular territory of the qualifying event. (vii) Time interval between qualifying event and inclusion in study. Studies were grouped into studies including the acute phase, studies recruiting mainly during the subacute phase and studies recruiting mainly during the chronic phase if the majority of patients were enrolled within 7 days, >7 days but <1 month, and >1 month, respectively, after the qualifying event. If this was not stated or not implicit in the study methods it was considered as unknown. (viii) Baseline investigations performed. (ix) Method and frequency of follow up. (x) Length of follow‐up and numbers of patients lost to follow up. (xi) Number of patients presenting with TIA or minor stroke (numbers were recorded separately if possible), and number of patients with events in the carotid, VB, both or uncertain territories. (xii) Number of patients with one or more outcome events according to the territory of the qualifying event. Where analyses were reported for more than one period of follow‐up, the longest follow up period was used for analysis. Only ischaemic strokes were counted as outcome events if the study discriminated between ischaemic and haemorrhagic strokes. Where possible, numbers of events were extracted from survival curves if no actual numbers were given in the text.

Analysis

Odds of recurrent events in relation to the territory of the presenting event were calculated within individual studies. Where appropriate, odds ratios (ORs) from separate studies were combined by fixed‐effects meta‐analysis according to the Mantel–Haenszel method. Heterogeneity between estimates from individual studies was calculated using the χ2 method. Proportional hazards were assumed. Publication bias was tested for by linear regression of the standard normal deviate against precision (Egger et al., 1997). The characteristics of studies that mainly accounted for heterogeneity were identified by stepwise weighted (by total number of events) linear regression of the natural logarithm of the OR versus year of publication, study setting, length of follow‐up, prospective/retrospective design, type of study, time from event until enrolment, and use of strict criteria for diagnosis of VB events. Our analysis was then stratified by the main determinants of heterogeneity. For graphical representation and confidence interval (CI) estimation, 0.5 was added to all four cells in the 2 × 2 table according to established practice where there was no outcome event in one of the cells (Sterne et al., 2001). Duplicate publications on the same cohort and overlapping cohorts were identified to avoid multiple inclusion of patients.

Results

The electronic literature search identified 2687 publications with the search terms [posterior circulation OR vertebrobasilar] AND [stroke OR CVA OR TIA OR cerebrovascular OR transient isch(a)emic], and 1051 publications with the search terms TIA AND [natural history OR prognosis]. Duplicate records were removed. After exclusion of publications that were clearly not relevant, 820 publications were reviewed and 66 potentially eligible publications identified. Review of reference lists of these papers identified a further 236 potentially eligible reports. Two further studies were identified by hand‐searching the journal Stroke, which was the only journal meeting the criterion for hand‐searching. A total of 304 papers were considered in detail. Of these, 278 were published in English, nine in German, seven in French, five in Spanish, and one each in Italian, Romanian, Dutch, Japanese and Polish.

A total of 56 articles satisfied our inclusion criteria. Of these, 13 were excluded from the analysis either because they reported only qualitative or quantitative relative risks of outcomes for patients with carotid versus VB events (i.e. no absolute numbers of events were given) (Canadian Cooperative Study Group, 1978; Gent et al., 1980; Candelise et al., 1982; Heyman et al., 1982; Whisnant and Wiebers, 1987; Lee et al., 1990; Carolei et al., 1992; Puranen et al., 1998; Marini et al., 1999), or because they reported only combined outcomes (e.g. stroke and TIA; or stroke and death) (Sørensen et al., 1983, 1989; Matias‐Guiu et al., 1987; ESPS Group, 1990). Thus, 43 papers reporting data from 36 independent cohorts were included in the analysis.

We were able to obtain IPD from seven studies including 8447 patients containing outcome data for patients presenting with carotid or VB TIA or minor stroke, five of which, including 4643 patients, had not previously published data on the prognosis according to the vascular territory of the qualifying event (Dennis et al., 1989; Farrell et al., 1991; Hankey et al., 1991; Davenport et al., 1996; Mead et al., 2002; Coull et al., 2003). Two studies had previously published data on the risk of stroke for carotid and VB territory events separately, but they had not published data on the risk of fatal stroke and overall death (Candelise et al., 1986; Dutch TIA Trial Study Group, 1993).

Table 1 summarizes the important characteristics of the eligible reports. Seven papers reported data from five independent population‐based studies. The remainder were hospital‐ and/or office‐based. Thirty‐one papers reported data from 26 independent prospective cohort studies or treatment trials, and 13 papers reported data from 10 independent retrospective cohort studies. Only six studies specifically intended to determine the prognosis according to the vascular territory of the qualifying event as one of their primary aims (David and Heyman, 1960; Drake and Drake, 1968; Acheson, 1971; Whisnant et al., 1978; Heyman et al., 1984b; Howard et al., 1987), two of which only reported the prognosis for overall mortality. Nine papers reporting data from six independent cohorts reported outcomes for individual patients who had events in both carotid and VB territories (Baker et al., 1968; Drake and Drake, 1968; Friedman et al., 1969; Ziegler and Hassanein, 1973; Cartlidge et al., 1977; Whisnant et al., 1978; Fieschi et al., 1981; Candelise et al., 1986; Keith et al., 1987). Table 2 summarizes which papers reported each specific outcome.

View this table:
Table 1.

The important characteristics of the eligible reports

StudyNo. of patientsLength of follow upProspective or retrospective studyStudy typeWere demo graphic data provided for each vascular territory separately?Time between event and enrolment in study in majority of patients**Were specific diagnostic criteria applied for the diagnosis of vertebro basilar events?Were patients with Amaurosis fugax included?Were patients with a likely cardio embolic source included?Were patients who had carotid endarter ectomy included?Were only first ever TIAs/MSs recruited?Were patients with prior strokes excluded?Were patients selectively recruited?
Hospital‐based studies
 Prencipe et al., 1998322 MS10 yearsProspectiveCohortNoAcuteCTNoYesNon/aYesNo
 López‐Gastón et al.,  199455 TIA3 yearsRetrospectiveCohortNoChronicYes?YesYes??No
 Cillessen et al., 19933150 TIA or MS2.6 yearsProspectiveRCTNoSub‐acuteYesYesNo???Yes
 Grotta et al., 19923034 TIA, RIND, MS2‐6 yearsProspectiveRCTNoChronicNoYesNoNoNoNoYes
 Sivenius et al., 1991;  ESPS, 19902500 TIA, RIND, CVA2 yearsProspectiveRCTSexChronicYesYesYes???Yes
 Larsen et al. 199046 TIA 10 yearsRetrospectiveCohortNoSub‐acuteNoYesYes??YesYes
 Baumgartner et al. 1989159 TIA 71 monthsRetrospectiveCohortNoAcuteNo?Yes?NoYesNo
 Biller et al. 198962 TIA3–15 daysProspectiveRCTNoAcuteYesYesYes?No?Yes
 Sørensen et al.,  1989, 1983203 RIA25 monthsProspectiveRCTNoSub‐acuteYesYesYesNoNoNoYes
 Howard et al., 1987451 TIA≤8 yearsProspectiveCohortNoSub‐acuteNo?Yes?NoNoNo
 Candelise et al., 1986;  Fieschi et al., 1981462 RIA4.6 yearsProspectiveCohortNoChronicYes?YesYesNo?No
 Putman and Adams,  198574 TIA6 daysProspectiveNRTNoAcuteYesYesNoNoNoNoNo
 Eriksson, 1985188 TIA40 monthsProspectiveNRTSexSub‐acute?YesNoYesNoNoYes
 Calandre and Molina,  198562 TIA, 60 RIND, 57 MS3 monthsRetrospectiveNRTNoAcuteNo?NoNoNo?No
 Heyman et al., 1984a424 TIA5 yearsProspectiveCohortYesSub‐acuteNoYes#NoYesNoNoNo
 Bousser et al., 1983604 (16% transient)3 yearsProspectiveRCTNoChronicYesYesNoNo??Yes
 Tsuda et al., 198345 TIA12.9 monthsRetrospectiveCohortNoUnknownYes?YesYesNo?Yes
 Muuronen and Kaste,  1982314 TIA7.8 yearsRetrospectiveCohortNoSub‐acuteYes?Yes?YesYesNo
 Simonsen et al., 1981243 TIA6 yearsRetrospectiveCohortYesSub‐acuteNo?YesYesYesYesNo
 Buren and Ygge, 1981125 TIA24 monthsProspectiveNRTNoUnknownYesYesYesYes??Yes
 Herskovits et al., 1981a66 TIA12 monthsProspectiveRTNoSub‐acuteYes?YesYes??Yes
 Olsson et al., 1980156 TIA or RIND∼12.5 monthsProspectiveRTYesSub‐acuteYes?YesNoYesNoYes
 Reuther et al., 198058 TIA or RIND2 yearsProspectiveRCTNoChronicNoYes#NoNoNo?Yes
 Haerer et al., 1977∼1300 TIA14.3 monthsProspectiveCohortYesSub‐acute?YesYesYes??No
 Omae, 197628 TIA48 monthsProspectiveCohortNoAcute?NoNo????
 Olsson et al., 1976112 TIA, 66 TIA‐IR21 monthsProspectiveNRTYesSub‐acuteYesYesNoYesNosame territoryYes
 Ziegler and Hassanein,  1973144 TIA3 yearsProspectiveNRTNoUnknownNoYesYesNoNoYesYes
 Baker 1971/Baker et al.  1968, 196679 TIA41 monthsProspectiveCohortSexChronicYesNoYesNoNoNoYes
 Acheson 1971; Acheson  and Hutchingson, 1964151 TIA4.6 yearsProspectiveCohortSexChronicYes?Non/aNoYesYes
 Drake and Drake, 196859 TIA/RIND, 30 CVA?ProspectiveCohortYesUnknownNoYesNoYes??Yes
 Marshall and Reynolds,  196526 TIA3–38 monthsProspectiveNRTSexUnknownNo?Non/aNo?Yes
 Siekert et al., 1961*230 TIA1‐5 yearsRetrospectiveNRTNoUnknown??Yes?No?Yes
 David and Heyman,  1960100 TIA or CVA2 yearsProspectiveCohortNoChronic??Yesn/aNoNoNo
Population‐based studies
 Ueda et al., 198718 TIA20 yearsProspectiveCohortYesAcuteYesYesYesNoYesYesNo
 Keith et al., 1987*289 TIA30 daysRetrospectiveCohortNoSub‐acuteYesYesYesYesYesYesNo
 Wiebers et al., 1982120 RIND?RetrospectiveCohortNoAcuteYesYesYes?No?No
 Cartlidge et al., 1977;  Whisnant et al., 1978199 TIA?RetrospectiveCohortNoAcuteYesYesYesYes?YesNo
 Ostfeld et al., 1973176 TIA3 yearsProspectiveCohortNoChronicYesYesNon/aNoYesNo
 Friedman et al., 196960 TIA27.4 monthsRetrospectiveCohortNoSub‐acuteYes?YesYesNoNoNo
Hospital‐based IPD
 Dutch TIA trial3150 TIA or MS31 monthsProspectiveRCT YesSub‐acuteYesYes?Yes??Yes
 Lothian stroke register1762 TIA or MS18 monthsProspectiveCohortYesSub‐acuteYesYes?Yes??No
 Italian Multicenter  Study654 TIA or MS46 monthsProspectiveCohortYesUnknownYes??YesNoNo?
 Oxford hospital  referred series451 TIA or MS50.5 monthsProspectiveCohortYesSub‐acuteYesYesYesYesNoNoNo
 UK‐TIA Aspirin  Trial2066 TIA or MS53 monthsProspectiveRCTYesChronicYesYesYesYesNoNoYes
Population‐based IPD
 Oxford Vascular  Study  (OXVASC)100 TIA or MS4 monthsProspectiveCohortYesAcuteYesYesYesYesYesYesNo
 Oxford Community  Stroke Project383 TIA or MS40 monthsProspectiveCohortYesAcuteYesYesYesYesYesYesNo

*Compare Cartlidge et al. (1977) and Whisnant et al. (1978); #Amaurosis fugax included separately; †mean values; **acute‐within 7 days, sub‐acute‐more than 7 days but within 1 month, chronic‐longer than 1 month; CVA, stroke; MS, minor stroke; n/a, not applicable; NRT, non‐randomised trial; RCT, randomised controlled trial; RIA, reversible ischaemic attack; RIND, reversible ischaemic neurological deficit; RT, randomised trial; TIA, transient ischaemic attack; TIA‐IR, TIA (incomplete recovery).

View this table:
Table 2.

Papers that reported each specific outcome

OutcomeNo. of papersReferencesIPD
Recurrent TIA13Baker, 1971; Baumgartner et al., 1989; Biller et al., 1989; Haerer et al., 1977; Herskovits et al., 1981a; López‐Gastón et al., 1994; Marshall and Reynolds, 1965; Olsson et al., 1980; Olsson et al., 1976; Putman and Adams, 1985; Reuther et al., 1980; Siekert et al., 1961; Tsuda et al., 1983None
Ischaemic stroke17Baker et al., 1968; Baker, 1971; Biller et al., 1989; Bousser et al., 1983; Burén and Ygge, 1981; Cartlidge et al., 1977; Cillessen et al., 1993; Friedman et al., 1969; Haerer et al., 1977; Larsen et al., 1990; López‐Gastón et al., 1994; Muuronen and Kaste, 1982; Olsson et al., 1980; Omae, 1976; Ostfeld et al., 1973; Siekert et al., 1961; Whisnant et al., 1978All studies, except Italian Multicenter Study
Stroke24Acheson 1971; Acheson and Hutchinson, 1964; Baker et al., 1966; Baumgartner et al., 1989; Calandre and Molina, 1985; Candelise et al., 1986; Cartlidge et al., 1977; Drake and Drake, 1968; Eriksson, 1985; Fieschi et al., 1981; Grotta et al., 1992; Heyman et al., 1984b; Keith et al., 1987; Olsson et al., 1976; Ostfeld et al., 1973; Prencipe et al., 1998; Putman and Adams, 1985; Reuther et al., 1980; Sivenius et al., 1991; Tsuda et al., 1983; Ueda et al., 1987; Wiebers et al., 1982; Whisnant et al., 1978; Ziegler and Hassanein, 1973All studies
Fatal stroke7Burén and Ygge, 1981; Cartlidge et al., 1977; Eriksson, 1985; Muuronen and Kaste, 1982; Olsson et al., 1976; Siekert et al., 1961; Simonsen et al., 1981All studies
MI3Heyman et al., 1984b; Larsen et al., 1990; Muuronen and Kaste, 1982All studies
Cardiac mortality6Cartlidge et al., 1977; Eriksson, 1985; Larsen et al., 1990; Muuronen and Kaste, 1982; Olsson et al., 1980; Simonsen et al., 1981All studies
Cardiovascular mortality2López‐Gastón et al., 1994; Simonsen et al., 1981All studies
Deaths12Baker et al., 1966; Baumgartner et al., 1989; Cartlidge et al., 1977; David and Heyman, 1960; Eriksson, 1985; Haerer et al., 1977; Larsen et al., 1990; Muuronen and Kaste, 1982; Prencipe et al., 1998; Putman and Adams, 1985; Siekert et al., 1961; Simonsen et al., 1981All studies
Stroke or death6Cartlidge et al., 1977; Eriksson, 1985; ESPS Group, 1990; Sivenius et al., 1991; Sørensen et al., 1983; Sørensen et al., 1989All studies
Survival4Cartlidge et al., 1977; Howard et al., 1987; Muuronen and Kaste, 1982; Simonsen et al., 1981All studies

Risk of stroke

A total of 32 independent studies reported the number of strokes on follow‐up. Of these, 16 studies reported ischaemic strokes separately. Fatal strokes were reported for six cohorts. All IPD included data on follow‐up strokes and fatal strokes, and all except the Italian Multicenter Study discriminated between ischaemic and haemorrhagic strokes.

There was significant heterogeneity between studies in the odds of subsequent stroke (P = 0.02). Time from event until enrolment and study setting were both significantly associated with the relative risk of stroke and therefore accounted for some of this heterogeneity (P = 0.002 and 0.005, respectively, in the model including one predicting variable). When both variables were tested together, only time from event remained significant (P = 0.027), whereas study setting did not (P = 0.090).

We therefore stratified the analysis first into studies that included the acute phase, studies that mainly recruited during the subacute phase, and studies that mainly recruited during the chronic phase, and secondly, into hospital‐ versus population‐based studies, and substratified hospital‐based studies into unpublished IPD versus published studies without calculating an overall OR estimate. There was no evidence of heterogeneity within the resulting strata (P = 0.38 for studies including the acute phase; P = 0.55 for studies mainly recruiting during the sub‐acute phase; P = 0.29 for studies mainly recruiting during the chronic phase; P = 0.47 for published hospital‐based studies; P = 0.84 for unpublished hospital‐based studies; P = 0.54 for population‐based studies) The distribution of a plot of ORs of stroke risk for VB versus carotid events against the variance of the OR in each study was symmetrical, and there was no evidence of overall publication bias on regression analysis (P = 0.79). The proportional hazard assumption appeared valid with no correlation between the OR and the mean length of follow up in each study (r = 0.099, P = 0.578).

Patients presenting with VB events had a higher risk for subsequent strokes in studies including the acute phase compared with patients presenting with carotid events (OR 1.47, 95% CI 1.1–2.0, P = 0.014; 1543 patients). In studies mainly recruiting after the acute phase however, patients with VB events had a lower risk of subsequent stroke (OR 0.81, 95% CI 0.6–1.0, P = 0.077; 6752 patients) for studies recruiting mainly during the subacute phase and (OR 0.75, 95% CI 0.6–0.9, P = 0.00026; 8252 patients) for studies recruiting mainly during the chronic phase (Fig. 1).

Fig. 1 Odds of stroke for patients with vertebrobasilar (VB) as opposed to carotid territory TIAs/minor strokes stratified into studies recruiting mainly during the acute, subacute and chronic phase. Studies with unknown timing of enrolment after the presenting event are not included. het = heterogeneity; sig = significance.

In published hospital‐based studies, patients presenting with VB events had a lower risk of subsequent stroke (OR 0.69, 95% CI 0.6–0.8, P < 0.00001, 27 studies; 11 664 patients) than patients presenting with carotid events. In previously unpublished hospital‐based IPD, however, there was no difference in the subsequent risk of stroke between carotid and VB events (OR 1.02, 95% CI 0.8–1.3, P = 0.91; three studies, 4149 patients). Moreover, in population‐based studies patients with VB events had a higher risk of subsequent stroke compared with patients with carotid events (OR 1.48, 95% CI 1.1–2.0, P = 0.025; seven studies, 1026 patients) (Fig. 2).

Fig. 2 Odds of stroke for patients with vertebrobasilar (VB) as opposed to carotid territory TIAs/minor strokes stratified by study setting: hospital‐ versus population‐based. het = heterogeneity; sig = significance.

The lower risk of stroke in patients presenting with VB events in the published hospital‐based studies was independent of the diagnostic criteria used for defining the territory of the presenting event: 15 studies used strict ‘Ad Hoc Committee National Institute of Neurological and Communicative Disorders and Stroke (1975)’ or similar criteria (OR 0.65, 95% CI 0.5–0.8, P = 0.00003, heterogeneity P = 0.21) versus 12 studies with less strict criteria (OR 0.72, 95% CI 0.6–0.9, P = 0.003, heterogeneity P = 0.56). There was no significant difference between studies published before 1980 (OR 0.65, 95% CI 0.5–0.9, P = 0.04, heterogeneity P = 0.09; seven studies) and studies published after 1980 (OR 0.69, 95% CI 0.6–0.8, P < 0.00001, heterogeneity P = 0.59; 20 studies). The findings in the three largest studies (OR 0.66, 95% CI 0.5–0.8, P = 0.00003, heterogeneity P = 0.23) (ESPS Group, 1990; Sivenius et al., 1991; Grotta et al., 1992; Cillessen et al., 1993) did not differ from the remainder (OR 0.71, 95% CI 0.6–0.9, P = 0.003, heterogeneity P = 0.45), and the results were also consistent when the analysis was confined to patients presenting with TIAs only (OR 0.68, 95% CI 0.5–0.8, P = 0.004, heterogeneity P = 0.13; 18 studies).

Studies specifically intending to investigate differences between carotid and VB events found no difference in the risk of subsequent stroke between the vascular territories (OR 1.02, 95% CI 0.7–1.4, P = 0.99, heterogeneity P = 0.04; four studies). There was no difference in the ratio of subsequent disabling strokes versus all strokes between patients initially presenting with VB versus carotid events (OR 0.8, 95% CI 0.6–1.2, P = 0.41, heterogeneity P = 0.37; six cohorts) (Baker et al., 1968; Dennis et al., 1989; Dutch TIA Trial Study Group, 1991; Farrell et al., 1991; Hankey et al., 1991; Mead et al., 2002).

Patients presenting with events in both carotid and VB territories had the highest risk of subsequent strokes (OR 2.01, 95% CI 1.3–3.1, P = 0.02, heterogeneity P = 0.41; six studies) versus carotid events alone (OR 2.93, 95% CI 1.8–4.8, P = 0.0008, heterogeneity P = 0.11) and versus VB events alone.

Risk of fatal events

In contrast to the risk of any stroke on follow‐up, there was overall agreement between the different strata for the risk of fatal stroke. Patients presenting with carotid and VB territory events had the same risk of fatal stroke (OR 0.89, 95% CI 0.7–1.3, P = 0.55, heterogeneity P = 0.64; 12 studies, 9533 patients) (Table 3). We were able to calculate case fatality of stroke in eight independent studies (Cartlidge et al., 1977; Muuronen and Kaste, 1982; Eriksson, 1985; Dennis et al., 1989; Dutch TIA Trial Study Group, 1991; Farrell et al., 1991; Hankey et al., 1991; Mead et al., 2002). There was no difference between patients with VB and carotid events (OR 0.99, 95% CI 0.7–1.4, P = 0.95, heterogeneity P = 0.02). The total number of deaths during follow‐up was reported for 20 independent cohorts. There was no difference in the risk of death between patients presenting with VB and carotid events (OR 0.96, 95% CI 0.8–1.1, P = 0.57, heterogeneity P = 0.38; 10 749 patients; Table 3).

View this table:
Table 3.

Odds of death and fatal stroke for patients with vertebrobasilar (VB) as opposed to carotid TIAs/MSs

Odds Ratio for death (95% CI), significance, heterogeneityOdds Ratio for fatal stroke (95% CI), significance, heterogeneity
Acute phase studies1.00 (0.7–1.4), P = 0.94, P = 0.140.95 (0.5–1.9), P = 0.95, P = 0.56
5 studies, 351 events, 1042 patients2 studies, 39 events, 569 patients
Sub‐acute phase studies1.00 (0.8–1.2), P = 0.99, P = 0.460.96 (0.6–1.6), P = 0.97, P = 0.35
9 studies, 790 events, 6847 patients7 studies, 111 events, 6119 patients
Chronic phase studies0.91 (0.7–1.1) P = 0.45, P = 0.271.00 (0.5–2.0), P = 0.87
3 studies, 542 events, 2206 patients1 study, 53 events, 2066 patients
Hospital‐based studies: published0.94 (0.7–1.2), P = 0.70, P = 0.290.67 (0.3–1.6), P = 0.45, P = 0.45
11 studies, 429 events, 2227 patients5 studies, 32 events, 1014 patients
Hospital‐based studies: unpublished0.96 (0.8–1.1), P = 0.70, P = 0.520.94 (0.6–1.4), P = 0.86, P = 0.38
5 studies, 1096 events, 7953 patients5 studies, 160 events, 7950 patients
Population‐based studies0.98 (0.7–1.4), P = 0.57, P = 0.380.95 (0.5–1.9), P = 0.95, P = 0.56
2 studies, 241 events, 569 patients2 studies, 39 events, 569 patients
Total0.96 (0.8–1.1), P = 0.57, P = 0.380.89 (0.6–1.2), P = 0.55, P = 0.64
18 studies, 1739 events, 10,749 patients12 studies, 231 events, 9533 patients

Risk of cardiac events

Thirteen cohorts recorded myocardial infarctions or cardiac deaths during follow‐up (Table 2). There was no significant difference in risk for cardiac events in patients presenting with VB and carotid events (OR 0.85, 95% CI 0.7–1.0, P = 0.07, heterogeneity P = 0.05; 9781 patients).

Risk of recurrent TIA

The number of patients with recurrent TIA(s) was reported in 13 cohorts (Table 2). Patients presenting with VB events were more likely to have a recurrent TIA than patients with carotid events (OR 1.70, 95% CI 1.3–2.2, P = 0.00007, heterogeneity P = 0.71; 1445 patients).

Discussion

The widely held view that VB territory TIAs and minor strokes have a better prognosis than carotid territory events was supported by some studies of stroke risk (Siekert et al., 1961; Marshall, 1964; Baker et al., 1968; Drake and Drake, 1968; Friedman et al., 1969; Acheson, 1971; Ziegler and Hassanein, 1973; Herskovits et al., 1981a, b; Simonsen et al., 1981; Heyman et al., 1984b; Sivenius et al., 1991; Grotta et al., 1992), but not others (Acheson and Hutchinson, 1964; Baker et al., 1966; Friedman et al., 1969; Ostfeld et al., 1973; Cartlidge et al., 1977; Toole et al., 1978; Heyden, 1979; Heyman et al., 1982; Muuronen and Kaste, 1982; Wiebers et al., 1982; Candelise et al., 1986; Keith et al., 1987; Whisnant and Wiebers, 1987; Baumgartner et al., 1989; Sacco et al., 1989; Cillessen et al., 1993). However, most of these studies were small and did not have the statistical power to determine differences in prognosis reliably. We found that the relative ORs of stroke for patients with VB versus carotid events depended on whether studies had included the acute phase (i.e. whether studies enrolled the majority of patients within days after the event) and also on whether the studies were hospital‐ or population‐based.

Our multivariate analysis suggested that the higher odds of stroke in population‐based studies was accounted for by the fact that these studies were more likely to have included the acute phase than hospital‐based studies. Hospital‐based studies tended to be outpatient‐based and did not usually include recurrent strokes that occurred after the presenting event prior to being seen in clinic. Interestingly, the only hospital‐based study that found a significantly increased stroke risk in patients with VB events versus carotid events had enrolled patients very early after their event (Biller et al., 1989).

Since we compared the risk of recurrent stroke in patients with VB events with that in patients with carotid events, our findings indicate that patients with VB TIAs and minor strokes have a higher risk of stroke than patients with carotid events in the acute phase, but that this is reversed in the subacute and chronic phase. Unfortunately, we were unable to compare directly the very early risk of stroke between patients with carotid and VB events, because few studies specifically reported the very early occurrence of strokes. A high early risk of stroke recurrence in patients with posterior circulation infarction compared with carotid territory strokes was reported in a previous population‐based stroke incidence study (Hankey et al., 1998), but there are very few other published data. One previous imaging study found a higher prevalence of silent infarcts in patients presenting with VB TIA or minor stroke compared with patients with carotid events (Herderschee et al., 1992), but this was only a cross‐sectional observation.

Other possible explanations for the differences in stroke risk between hospital‐ and population‐based studies are first publication and reporting bias particularly as results differed between published hospital‐based studies and unpublished IPD. However, we found no evidence for this. Secondly, differences in the prevalence of important risk factors could have influenced the results, but few studies reported these data by the vascular territory of the presenting event and there were no consistent differences among the seven studies that reported sex and age by vascular territory of the presenting event (Drake and Drake, 1968; Olsson et al., 1976, 1980; Haerer et al., 1977; Simonsen et al., 1981; Heyman et al., 1984b; Ueda et al., 1987), and no differences in those studies that reported data on co‐morbidity and risk factors (Olsson et al., 1976; Dyken et al., 1977; Fieschi et al., 1981; Simonsen et al., 1981; Heyman et al., 1984b). Moreover, the lack of a difference between patients with VB events and carotid events in overall mortality or in risk of coronary vascular events suggests that they had similar risk factor profiles. Thirdly, a varying degree of thoroughness on follow‐up between hospital‐ and population‐based studies might have influenced the numbers of follow‐up strokes if stroke severity differed between the vascular territories. Some studies have reported a better functional outcome in patients with completed posterior circulation strokes compared with carotid strokes (Check, 1982). We were unable to reliably compare the level of disability that resulted from strokes during follow‐up in patients presenting with VB events versus carotid events, because insufficient data were reported. However, we found no evidence that the risk of fatal stroke was lower in patients presenting with VB events. Moreover, case fatality due to stroke during follow‐up did not differ. Similarly, although some studies have reported an overall lower mortality during follow‐up for patients with VB events than for patients with carotid events (Acheson and Hutchinson, 1964; Marshall, 1964; Ziegler and Hassanein, 1973; Simonsen et al., 1981; Howard et al., 1987; Turney et al., 1984), our analysis shows that there is no difference in mortality.

Methodological issues

There are a number of methodological issues that require discussion. First, there was a wide variation in study methodology, and most studies of the studies that we included fell short of the gold standard of an inception cohort with all patients enrolled at a similar stage of their illness in a similar setting (Kernan et al., 1991). The time allowed between qualifying event and enrolment differed widely between studies (range 12 h to 1 year), and the length of follow‐up varied from a few days to >10 years. The studies included ranged from the mid‐1950s to the late 1990s. Absolute risks of stroke and death are likely to have diminished since the introduction of effective preventive therapies, such as anti‐platelet agents, treatment of hypertension and hyper‐cholesterolaemia. We therefore limited our analyses to the relative odds of recurrent events between patients with VB and carotid events within studies, and did not attempt to estimate overall absolute risks. There was no evidence in any of the studies of systematic differences between patients presenting with VB and carotid events in the timing of presenting events, the time‐course of outcome events, the length of follow‐up or the use of pharmacological treatments.

Secondly, we combined data on patients presenting with TIAs and minor ischaemic strokes because they have a similar underlying pathology and prognosis (Wiebers et al., 1982; Calandre and Molina, 1985; Sørensen et al., 1989; Dennis et al., 1990a; Carolei et al., 1992; Koudstaal et al., 1992), and patients with TIA frequently have appropriately localized infarcts on brain imaging (Murros et al., 1989; Kidwell et al., 1999).

Thirdly, there was considerable variation between studies in what proportion of all events were in the VB territory, ranging from 17 to 56% in population‐based studies (Ostfeld et al., 1973; Wiebers et al., 1982; Keith et al., 1987), and 15 to 72% in hospital‐based studies. This variation will have been due partly to differences in patient selection, and partly to variation in diagnostic criteria. Studies published after the introduction of the Ad Hoc Committee National Institute of Neurological and Communicative Disorders and Stroke (1975) diagnostic classification were more consistent with 15–35% of patients reported to have presented with VB events. However, differences in diagnostic criteria or patient selection did not appear to have influenced our results. There was no difference in the relative odds of stroke in patients presenting with VB versus carotid events between studies that used the strict diagnostic criteria and those that did not, and no correlation between the relative odds and the proportion of the study sample that presented with VB events.

Finally, some studies included patients who underwent carotid endarterectomy during follow‐up. However, endarterectomy was only usually performed on a small proportion of patients with carotid events (mostly <10%) and this is unlikely to have had a major effect on the overall risk of recurrent stroke. Few studies reported patients presenting with amaurosis fugax separately from carotid territory cerebral events. It was therefore not possible to assess what impact their better prognosis had.

Conclusions

Our analyses are not necessarily helpful in deciding how best to manage individual patients with VB TIA or minor stroke. This can be very difficult, not least because of the difficulty in reliably distinguishing clinically between ischaemic events affecting the different territories within the posterior circulation (Caplan, 2000). However, our results do provide useful data to guide treatment policy for patients presenting with VB events. Despite the variations in study methodology and patient selection, our findings were remarkably consistent. Compared with patients presenting with carotid TIA or minor stroke, there is no evidence that patients with VB events have a lower risk of subsequent stroke or death. Indeed, their risk of stroke is probably higher in the acute phase. Patients with VB events require active preventive treatment.

There were no differences however, in the risk of fatal strokes, cardiac events or all cause mortality (mainly vascular), and there was a higher risk of recurrent TIAs for patients with VB events. Overall, this suggests that patients presenting with VB events are no less likely to have serious underlying vascular pathology.

Acknowledgements

We wish to thank Dr S. Howard for statistical advice and Dr S. Guthikov and Mr R. Bond for assistance with analysis of data. E.F and P.M.R. are funded by the UK Medical Research Council.

References

  1. 3.
View Abstract