Outcome at 30 Days in the New England Medical Center Posterior Circulation Registry FREE

POSTERIOR CIRCULATION disease has traditionally been considered an illness with high mortality and morbidity since the publication of Kubik and Adams' original article in Brain in 1946¹ concerning the clinical and pathologic findings in fatal cases of basilar artery (BA) occlusions.² Most subsequent outcome studies in patients with vertebrobasilar disease have considered selected groups with specific vascular lesions or clinical states.^{2 - 9} The reported mortality in these studies has been high. In contrast, Bogousslavsky et al¹⁰ studied an unselected, relatively small group of patients with vertebrobasilar disease and found a low mortality rate. We analyzed and report outcomes from the New England Medical Center Posterior Circulation Registry (NEMC-PCR), a prospective collection of patients with posterior circulation disease. The NEMC-PCR included 407 patients who had had vertebrobasilar territory strokes and transient ischemic attacks and who were admitted or referred to the NEMC between January 1, 1988, and December 31, 1996.

The NEMC-PCR was initiated in 1988 and includes complete data for 407 patients. Patients who have had a stroke must have had computed tomographic (CT) or magnetic resonance imaging (MRI) documentation of posterior circulation infarction. Transient ischemic attacks had to be clearly in the vertebrobasilar territory with vascular images that showed vertebrobasilar occlusive lesions. Investigations must have been adequate and include at least neuroimaging (CT and/or MRI) and appropriate cardiac, hematologic, and vascular imaging tests.

The NEMC-PCR data include sociodemographic variables as well as biomedical risk factors including smoking, migraine, obesity, hypertension, hyperlipidemia, illicit drug use, diabetes mellitus, alcohol consumption, oral contraceptive use, previous stroke, coronary artery disease, and peripheral vascular occlusive disease. The results of brain imaging CT, MRI, vascular studies (angiography, ultrasonography of the neck, transcranial Doppler ultrasonography, and/or magnetic resonance angiography), and cardiac studies (electrocardiogram, transthoracic echocardiogram, transesophageal echocardiogram, and/or 24-hour rhythm monitoring) were also recorded. The locations of infarcts were also recorded.

Location was judged clinically and by brain imaging. Brain lesions were categorized as proximal, middle, and distal intracranial posterior circulation territories (Figure 1).² The proximal intracranial posterior circulation territory included regions supplied by the intracranial vertebral arteries (ICVAs)—the medulla oblongata and the posterior inferior cerebellar artery–supplied region of the cerebellum. The middle intracranial posterior circulation territory included the portion of the brain supplied by the BA up to its superior cerebellar artery branches—the pons and the anterior inferior cerebellar artery–supplied portions of the cerebellum. The distal intracranial posterior circulation territory included all of the territory supplied by the rostral BA and its superior cerebellar artery and posterior cerebral artery (PCA), and their penetrating artery branches—midbrain, thalamus, superior cerebellar artery–supplied cerebellum, and PCA territories. The distribution is shown in Figure 1, modeled after a figure published by Caplan² and first used in the Duvernoy atlas.¹¹ A patient with a posterior inferior cerebellar artery territory infarct on MRI who also had a hemianopia (but no occipital infarct on MRI) would be classified as having brain lesions affecting the proximal and distal territories.

Vascular lesions, stroke causes and mechanisms, and outcome data are the remaining factors contained in this database. Most of the large artery vascular lesions (>50% stenosis) were located in the extracranial vertebral artery (ECVA), ICVA, BA, and PCA. Primary stroke mechanisms included large artery disease causing local hypoperfusion, embolism (cardioembolism, intra-arterial embolism, or cardio–intra-arterial embolism), penetrating artery disease, migraine, and other less common causes. Patient outcome was categorized according to a Modified Rankin Disability Scale score as follows: no disability, minor disability, major disability, death from cerebrovascular disease, and death from other cause.

A stroke service senior neurologist (L.R.C., M.S.P., or L.D.D.) made each diagnosis following clinical evaluation and after reviewing all available data. Each case was reviewed on multiple occasions with the entire stroke service staff. The criteria for registry diagnoses have been published² and are listed in Table 1.

Data analyses were conducted in 2 stages. First, we examined the associations between predictor variables and the risk of poor outcome at 30 days (defined as death or severe disability) using relative risks (RRs). To investigate whether significant associations seen in the bivariate case were independent of the effect of age, baseline risk factors for poor outcome, and severity, we fitted a series of multivariate logistic regression models. These models estimate the log of the odds of death or major disability at 30 days as a linear function of covariates. Model 1 includes age and 2 additional risk factors that were found (in previous analysis not shown) to be of at least borderline significance in preliminary analyses—history of previous stroke and history of alcohol abuse. Exploratory data analyses showed that the association between age and poor outcome was nonlinear. A series of binary variables is included in all models to capture this nonlinear association. Model 2 adds to this baseline model the number of intracranial territories involved (2 or 3 with 1 territory serving as the reference category) and the location of the lesion (proximal, middle, and distal, with "other" serving as the reference category). Model 3 tests the hypothesis that the relationship between large artery hemodynamic mechanism and 30-day outcome is independent of age, risk factors, and severity (as measured by the number of involved territories). In models 3 through 5, the variables related to vascular territory are dropped because they are not independent of the vessels involved or the underlying mechanism. For example, 77% of those with a cardioembolic mechanism had a distal infarct. Model 4 tests the hypothesis that the relationship between cardioembolic mechanism and 30-day outcome is independent of age, risk factors, and severity. Finally, model 5 tests the hypothesis that the association between BA involvement and 30-day outcome is independent of age, risk factors, and severity. To capture the joint influence of several variables numerous interaction terms were also fitted. With the exception of 2 (middle×distal territory and middle territory×BA involvement), the presence of empty observed cells made these interaction terms impossible to estimate. Neither of the 2 interaction terms that were tested were significant.

The NEMC-PCR contains data for 407 prospectively entered patients with ischemia affecting posterior circulation. Forty-six patients (11%) were excluded from analyses involving outcome owing to incomplete 30-day outcome assessments. Excluded patients were more likely to have smoked, used alcohol, had hyperlipidemia, and had more than 1 focal brain lesion. The 46 excluded patients were no different from the included cases with complete outcome data on other variables of interest including age, race, sex, obesity, previous stroke, history of hypertension, diabetes mellitus, coronary artery disease, and peripheral vascular disease.

Table 2 summarizes the NEMC-PCR for sociodemographic characteristics, outcome, frequency of vascular occlusive lesions, and primary mechanism. Consistent with the epidemiology of stroke and the demographic characteristics of the population in Boston, Mass, cases included in the registry are disproportionately white and male, with an average (SD) age of 60.5 (16) years.

The Modified Rankin Disability Scale categorizes outcome at 30 days among the 361 patients for whom outcome was known. Overall mortality was very low: 13 deaths (3.6%) with 7 (1.9%) due to cerebrovascular disease and 6 (1.7%) from other causes. In total, 77 patients (21%) died or were left with a major disability. However, most patients (79%) were left with either no disability or minor disability. A substantial percentage (77%) of these patients were studied using CT and/or MRI. The use of other vascular and radiological studies was also common. Of the 260 patients with an abnormal vascular study result, lesions in the ECVA were the most common (32%), followed by the BA (30%), and the ICVA (22%). The most frequently diagnosed mechanism (listed both as the primary mechanism and all of the 3 possible diagnoses) was embolism (40%), with embolus of cardiac origin having been the most frequently diagnosed pattern (24%). Large artery hemodynamic disease was the primary diagnosis in one third of the patients.

We analyzed outcome according to vascular occlusive lesion, brain infarct location, and stroke mechanism. Frequency of poor outcome by vascular lesions, listed in Table 3, indicated that occlusion of the BA led to the worst outcome. Basilar artery occlusive disease was responsible for the greatest risk of mortality and major disability, followed by disease in the ICVA. Involvement of the ECVA and the PCA accounted for the remaining cases of death or major disability.

Table 3 lists the frequency of poor outcome by brain location. Involvement of both the distal and middle regions (either alone or in combination) results in a greater than 30% risk of death or major disability. Most patients who died or were left with major disability (77%) had lesions involving some part of the distal region (59 of 77). Of the 185 patients with a stroke involving the distal region, 32% had poor outcome. The risk of poor outcome was greatest among those with middle and distal involvements (52%). The risk was nearly as great for those with involvement of proximal, middle, and distal regions (50%) as well as cases involving the border zone region (50%), although both groups contributed only modestly to the total burden of morbidity and mortality. Stroke including any of the middle region accounted for the next largest percentage of total mortality or major disability and was associated with a similar overall risk. Twenty percent of the patients with middle territory lesions only had poor outcome. The lowest risk of poor outcome was associated with proximal only infarcts (5%) and no infarcts within the posterior circulatory system.

Table 4 gives the frequency of poor outcome at 30 days by primary stroke mechanism. The single most likely stroke mechanism is shown as well as the frequency of all potential mechanisms (N = 361). Embolism was the most common mechanism associated with mortality or major disability as well as the mechanism associated with the greatest risk of poor outcome. Embolism caused death or major disability in 42 (55%) of the 77 patients who had poor outcome. Embolism caused 41% (147 cases) of stroke but was responsible for 55% of mortality and major disability (42 of 77 cases). Of this number, poor outcome was observed in 33% of the cases judged to involve cardioemboli, 18% of the cases with an intra-arterial embolus, and 50% of the patients who had both potential cardiac and arterial sources of emboli. Other mechanisms, including large artery disease (16 cases) and penetrating artery disease (7 cases), contributed 20% and 9%, respectively, to the overall burden of poor outcome.

Table 5 lists unadjusted RRs for poor outcome (death or severe disability) at 30 days in the NEMC-PCR. In analyses not shown, predictors of poor outcome did not differ by outcome (death vs major disability). Therefore, for ease of presentation, the 2 outcomes were combined. In unadjusted analyses, none of the sociodemographic or medical risk factors was associated with increased risk of poor outcome. Age, alcohol abuse, and previous stroke were identified as factors associated with increased risk of poor outcome (although not significant) that might modify or confound the effect of other variables of interest. Features of the stroke that were associated with increased risk of poor outcome in unadjusted analyses included lesions in the middle (RR,1.88; 95% confidence interval [CI], 1.28-2.79) and distal regions (RR,3.12; 95% CI, 1.92-5.07) and a cardioembolic mechanism (RR,1.89; 95% CI, 1.28-2.80). Those with more than 1 infarct were also at increased risk for poor outcome. Large artery hemodynamic mechanism was associated with a significantly lower risk of poor outcome (RR, 0.59; 95% CI, 0.36-0.98).

Table 6 summarizes the results of multiple logistic regression analyses designed to examine those factors associated with the risk of poor outcome after adjusting for potential confounders. Model 1 adjusts for factors thought to be potential confounders from preliminary analyses. Aged older than 75 years as well as history of alcohol abuse were seen to be associated with a significant increase in the odds of poor outcome. Model 2 shows that patients with 3 infarct territories had significantly worse outcome compared with those with 1 territory of involvement independent of age and other risk factors. In addition, middle and distal lesions were associated with worse outcomes after adjusting for age, risk factors, and severity (defined as the number of infarct territories involved). In model 3, the protective effect of intra-arterial embolism seen in unadjusted analyses is no longer significant, suggesting that this association is explained by other factors in the model. However, the poor prognosis associated with a cardioembolic source (model 4) and involvement of the BA (model 5) survived the addition of potential confounders.

There are 2 major outcome findings of this review of the NEMC-PCR, the largest case series of posterior circulation strokes to date: (1) the low (21%) risk of poor outcome (defined as either mortality or major disability) occurring 30 days after stroke among the 361 patients with complete data and (2) the high frequency of patients with infarction due to embolism (147 [41%] of 361 patients). The low risk of poor outcome occurring 30 days postonset suggests that long-standing pessimism about posterior circulation occlusive disease is unwarranted. These results corroborate the analysis of Bogousslavsky et al¹⁰ of 1000 patients from the Lausanne Stroke Registry showing a similarly low mortality rate (5.9%) among patients with vertebrobasilar territory infarcts (4% died within 3 weeks).¹⁰ The 30-day mortality rate among the patients in the NEMC-PCR was only 3.6%.^{10 ,12}

These low mortality rates diverge from prior reports that considered only selected groups of patients.^{2 - 9} Millikan et al³ and McDowell et al⁴ examined the effects of anticoagulant treatment vs no particular treatment; 44 patients with vertebrobasilar ischemia and possible BA occlusion and 50 patients with vertebrobasilar occlusion were studied, respectively. Millikan et al found an overall mortality rate of 30% and McDowell et al found an overall mortality rate of 38% (8 patients who received no specific treatment died within the first month). The retrospective study by Fogelholm and Aho⁵ selected 141 patients with ischemic brainstem infarcts; the overall mortality rate was 28%, with 5% mortality within the first month. Jones et al⁶ found a 27% mortality rate within the first week among 37 consecutive patients with acute vertebrobasilar territory infarcts. Patients were included if they were admitted to the hospital within 36 hours of the onset of symptoms. Hacke et al⁷ described 43 patients, most of whom were comatose or tetraplegic, who had angiographically proven occlusions of the ICVAs or BA, and were considered for intra-arterial thrombolysis. For these specific patients the mortality rate was very high (70%).⁷ Archer and Horenstein¹³ also selected very severe cases, patients with vertebrobasilar occlusions confirmed by angiography. Among 20 patients, 11 of whom were comatose at the outset or by the time of undergoing angiography, they found an 80% mortality rate. Zeumer et al⁸ and Von Kummer et al⁹ also analyzed outcome among patients with angiographically confirmed ICVA and BA occlusions, who were considered for intra-arterial thrombolysis, selecting only the most seriously ill patients for treatment. They recorded mortality rates of 36% (10 patients) and 67% (27 patients), respectively. The fewer patients evaluated in these prior reports showed that the high mortality rates were explained by the inclusion of patients with pessimistic prognoses and unfavorable outcomes. Fields et al¹⁴ and Caplan¹⁵ described the results of patients selected because of their excellent outcomes after posterior circulation strokes.

In the NEMC-PCR, all patients determined to have ischemia involving posterior circulation were included and analyzed; they were not selected on the basis of infarct, vascular lesions, presenting features, or outcome. The findings in this registry may not be representative of the general population since NEMC physicians were known to specialize in posterior circulation disease; referral bias cannot be excluded, although typically this type of bias results in a more (rather than less) severely affected group of patients.

Data from the NECM-PCR show that stroke outcome (mortality and major disability) is dependent on lesion location, mechanism, and arteries involved. Involvement of the distal region was responsible for 77% of the 77 cases that resulted in mortality or major disability after the first 30 days, while middle and proximal locations accounted for 40% and 23%, respectively. Overall, the risk of poor outcome was highest (about half) in cases involving both the middle and distal territories. Distal territory involvement is most likely a marker for embolism since most emboli from both cardiac and arterial sources involved the distal territory. The distal BA supplies the major blood flow to the tegmentum of the pons and midbrain. The higher mortality rate and major disability of distal territory infarcts could reflect the effects of tegmental brainstem infarction.

Nadeau et al¹⁶ used a sectorial approach, based on the brainstem vascular–anatomical pattern, to classify patients admitted with acute ischemic strokes. The brainstem and cerebellum were divided into sectors: median, paramedian, lateral, and dorsal that were supplied by defined groups of vessels. Generally, single-sector involvement reflected small-vessel disease, while multisector involvement reflected large-vessel occlusive disease. The top of the basilar was regarded as multisector involvement, while lateral medullary infarct was considered single-sector involvement. The constellation of clinical signs and radiological (CT) findings defined the number of sectors involved. Nadeau et al¹⁶ showed that involvement of a single sector was more likely to result in better outcomes in contrast with poorer outcomes associated with multisector lesions. The present data from the NEMC-PCR supports this finding, showing that cardioemboli, multisector lesions, and distal territory lesions all led to a higher risk of poor outcome. Our findings, along with those of Nadeau et al, indicate that a stroke involving more than 1 region results in the highest risk of mortality or major disability. Findings from the multiple logistic regression models confirm that the increased risk of poor outcome associated with involvement of more than 1 intracranial territory, especially if middle or distal, is not confounded by age or other risk factors marginally associated with the outcome (Table 6).

Embolism is the single most common stroke mechanism in the NEMC-PCR (between 40% and 54%). Owing to increased risk of poor outcome in patients with cardioembolic source, this mechanism accounts for a substantial proportion of mortality and major disability (55%). This finding concurs with those from the Stroke Data Bank in which the second and third most frequent diagnosis of stroke was embolism.¹⁷ Timsit et al¹⁸ found that embolism (cardioembolic and intra-arterial) accounted for 24.4% of cerebral infarction in the Stroke Data Bank. Major international stroke registries, like ours, have found embolism to be one of the main causes of mortality, major disability, and stroke, despite differing definitions of embolism.^{13 ,17 ,19 - 21} The analysis of brain location by stroke mechanism was also consistent with other findings from the NEMC-PCR; the distal region was closely linked to the most common mechanism—embolism.

In the NEMC-PCR, vascular lesions were most commonly seen in the ECVA (82) followed by the BA (77), ICVA (57), and PCA (21). Consistent with previous reports, we found that occlusion of the BA or ICVA resulted in the most grave outcomes, and involvement of the ECVA or PCA was generally more benign.^{1 - 10 ,13 - 14} Thirteen patients (3.6%) died during the first 30 days. Seven died of cerebrovascular disease; 3 of these had lesions in the ICVA indicating that lesions in this vessel potentially carry more risk for death during the short-term period. The 7 cerebrovascular deaths included the following: a BA embolic occlusion from the heart, bilateral ICVA lesions, intra-arterial embolism from an ECVA to the top of BA, and cardioembolism to the top of BA. Three of the patients who died of cerebrovascular disease did not have sufficient data to determine the precise cause of death. The 6 additional deaths were due to myocardial infarction (1 patient), endocarditis (2 patients), congestive heart failure (2 patients), and human immunodeficiency virus HIV and coagulopathy (1 patient).

Accepted for publication November 28, 2001.

Author contributions:Study concept and design (Drs Glass, Wityk, and Caplan); acquisition of data (Drs Pazdera, Chang, Wityk, Dewitt, Pessin, and Caplan); analysis and interpretation of data (Drs Glass, Pazdera, Chang, Wityk, Dewitt, Pessin, and Caplan, and Ms Hennessey); drafting of the manuscript (Drs Glass, Pazdera, Chang, Dewitt, Pessin, and Caplan, and Ms Hennessey); critical revision of the manuscript for important intellectual content (Drs Glass, Wityk, and Caplan, and Ms Hennessey); statistical expertise (Dr Glass); study supervision (Dr Caplan).

Corresponding author and reprints: Louis R. Caplan, MD, Division of Cerebrovascular Disease, Beth Israel Deaconess Medical Center, 330 Brookline Ave, Boston, MA 02215 (e-mail: lcaplan@caregroup.harvard.edu).