Risk factor status and vascular events in patients with symptomatic intracranial stenosis

Design.

We utilized patient data collected during the course of the WASID study for this post hoc analysis. The overall design and primary results of WASID have been published previously.^4,5 In brief, patients with a nondisabling stroke or TIA in the previous 90 days that was attributed to 50 to 99% stenosis of the intracranial carotid, middle cerebral artery, intracranial vertebral, or basilar artery were enrolled in the trial at 59 sites in North America. Patients with persistent blood pressure >180/115 mm Hg were excluded from the study. Patients were randomly assigned to treatment with warfarin or aspirin and followed prospectively.

Vascular and lifestyle risk factors were recorded at baseline and every 4 months thereafter until 1) the patient experienced the primary WASID endpoint (ischemic stroke, brain hemorrhage, or vascular death); 2) the patient died; or 3) the last study follow-up visit. The study was originally designed to enroll 806 patients but was halted after 569 patients had been enrolled due to concerns regarding the increased death and major hemorrhage rates in the warfarin arm.

Risk factors were managed by the local principal investigator in association with the patient’s primary care physician. National guidelines on treatment of risk factors such as hypertension, hyperlipidemia, and diabetes were disseminated to clinical sites.^6,7 Periodic reminders were sent to local sites on the importance of risk factor modification and sites were provided information on the degree of risk factor control within the study as a whole.

Statistical methods.

Since there was no significant difference in the treatment effect of warfarin vs aspirin in the primary efficacy analysis, the two treatment groups were combined. We established cutoff levels for defining good vs poor control of each risk factor according to national guidelines that were available during the trial.^6,7 The criteria for poor control of a risk factor were as follows: systolic blood pressure ≥140 mm Hg, mean arterial pressure ≥106 mm Hg, total cholesterol ≥200 mg/dL, low-density lipoprotein (LDL) ≥100 mg/dL, high-density lipoprotein (HDL) <40 mg/dL, total cholesterol/HDL ratio ≥4.4, hemoglobin A1C >7%, any current smoking, and no current consumption of alcohol. To assess the adequacy of risk factor management in the trial, we compared the percent of patients with good control of each risk factor at the baseline examination vs the year 1 examination, and at the baseline examination vs the year 2 examination using the McNemar test.

To evaluate the relationship between risk factors and subsequent vascular events we utilized two prespecified endpoints: 1) ischemic stroke, myocardial infarction (MI), or vascular death, and 2) ischemic stroke alone. For each patient, available values at the baseline examination and all follow-up examinations were averaged for each continuous risk factor and compared to the cutoff level. As an example, for blood pressure, the baseline value and all values available from the 4 month visits were averaged (baseline, 4, 8, 12, months). The average number of blood pressure readings during the follow-up period was 5.8. For the other risk factors the target for good control was considered met as follows: smoking (never smoked), alcohol consumption (consumed alcohol at some point). For these lifestyle factors, smoking recorded at any visit would remove the patient from the never smoked category and alcohol consumption at any time would qualify as consumed alcohol at some point. Univariate associations between the achievement of the risk factor target level, on average, and time to endpoint were assessed using the log-rank test and Cox proportional hazards regression.

Multivariable analyses were done using stepwise Cox proportional hazards regression. Fifty-seven patients were missing lipid profile information. Compared to the patients for whom this information was available, these 57 patients had a higher rate of stroke, MI, or vascular death (HR = 5.01, p = 0.0001, 95% CI = 3.11 to 8.06) and a higher rate of recurrent ischemic stroke (HR = 5.38, 95% CI = 3.20 to 9.05, p = 0.0001). These 57 patients were also more likely to have an average SBP ≥160 mm Hg (15.8% vs 7.8%, p = 0.0429). Inclusion of a lipid factor (such as cholesterol) in a multivariable analysis from which these patients were deleted would confound the effect of systolic blood pressure. In addition, with these patients deleted, the overall rate of events would be lower, thus decreasing power. Because of this bias in missing lipid data, multiple imputation methods⁸ were used along with Cox proportional hazards regression for estimating multivariable models. Each missing value was replaced with a randomly generated value from the multivariable distribution of the non-missing data. A hazard ratio was determined for each variable using proportional hazards regression. This process of generating possible values and estimating a hazard ratio for each variable was repeated 500 times. A point estimate, 95% CI, and p value for the hazard ratio of each variable was determined the using the mean and SD of the 500 hazard ratios based on a t statistic with 499 degrees of freedom.

Baseline risk factors.

When the study was halted, 569 patients had been enrolled and followed for a mean of 1.8 years. Key baseline features of the study patients are presented in table 1. The mean systolic and diastolic blood pressure at randomization was 139.8 mm Hg and 76.8 mm Hg. The mean systolic pressure at year 1 was 139.9 mm Hg (paired t test p = 0.45) and it was 137.3 mm Hg at year 2 (p = 0.71). The mean glycosylated hemoglobin in patients with diabetes was 8.4%. For lipid values, the mean LDL was 123.4 mg/dL and the mean HDL was 44.4 mg/dL.

Pharmacologic therapy for risk factor management.

Pharmacologic treatment of risk factors was widely used in the trial. Of patients with a history of a lipid disorder at study entry, 91% were prescribed lipid lowering therapy during follow-up, with 99% of these patients on a statin, 8% of whom were also prescribed a second agent (e.g., bile acid sequestrant, fibrate). In the study patients as a whole, 61% of patients were on a statin at baseline and 82.4% took a statin at any time during the trial. For patients with a history of diabetes at study entry, 91% were prescribed insulin or an oral hypoglycemic agent during follow-up. Of patients with a history of hypertension at baseline, 96% were prescribed an antihypertensive agent during follow-up: 53% of patients were on three or more agents, 27% were on two agents, and 16% were on a single agent. The types of antihypertensive agents used included the following: ACE inhibitor 67%, β-blocker 54%, calcium channel blocker 51%, diuretic 49%, vasodilator 19%, angiotensin receptor blocker 18%, and α-agonist 12%.

Effectiveness of risk factor management.

Table 2 shows the percent of patients with good control of each risk factor at baseline vs year 1 and at baseline vs year 2. These data indicate that there were significant improvements in several lipid values (total cholesterol, LDL, HDL, cholesterol/HDL ratio), diabetes control, and smoking cessation over the first 2 years of follow-up. Only 10% of patients achieved the high-risk LDL target of <70 mg/dL at the 2-year mark, although this was not an explicit target in WASID.

Relationship between risk factor control and ischemic stroke, MI, or vascular death.

There were 137 outcome events. These included 106 ischemic strokes, 18 vascular deaths, and 13 MI events. Univariate analysis for the association between risk factors and stroke, MI, or vascular death is shown in table 3. Systolic blood pressure ≥140 mm Hg (p = 0.0005), mean arterial pressure ≥106.7 mm Hg (p < 0.0001), cholesterol ≥200 mg/dL (p = 0.0223), and no alcohol consumption (p = 0.0020) were associated with an increased risk of major vascular events.

During the mean 1.8 years of follow-up, 30.7% of patients with a mean SBP ≥140 mm Hg had a stroke, MI, or vascular death, compared to 18.3% of patients with a mean SBP <140 mm Hg (p < 0.0005). Over the same period, 25.0% of patients with a LDL ≥115 mg/dL (the median LDL) had a stroke, MI, or vascular death, compared to 18.5% of patients with a mean LDL <115 mg/dL (p = 0.029).

Multivariable analysis showed that systolic blood pressure ≥140 mm Hg (HR = 1.79, 95% confidence limits 1.27 to 2.52, p = 0.0009), no alcohol consumption (HR 1.69, 1.21 to 2.39, p = 0.002), and cholesterol ≥200 mg/dL (HR 1.44, 1.00 to 2.07, p = 0.048) were associated with an increased risk of stroke, MI, or vascular death (figures 1 and 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Risk of stroke, myocardial infarction (MI), or vascular death according to mean systolic blood pressure (SBP)

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Risk of stroke, myocardial infarction (MI), or vascular death according to mean cholesterol (CHOL) level

Relationship between risk factor control and ischemic stroke.

Patients were more likely to have a recurrent ischemic stroke if during follow-up they had SBP ≥140 mm Hg (p = 0.0120, HR 1.63), MAP ≥106 mm Hg (p < 0.0001, HR 2.83), total cholesterol ≥200 mg/dL (p = 0.0006, HR 2.06), LDL ≥100 mg/dL (p = 0.0326, HR 1.72), non-HDL ≥130 mg/dL (p = 0.0109, HR 1.94), total cholesterol/HDL ≥4.4 (p = 0.0043, HR 1.89), and no alcohol consumption (p = 0.0033, HR 1.78). Multivariable analysis showed that systolic blood pressure ≥140 mm Hg (HR = 1.58, 1.07 to 2.32, p = 0.02), no alcohol consumption (HR 1.76, 1.19 to 2.59, p = 0.005), and total cholesterol ≥200 mg/dL (HR 1.95, 1.29 to 2.97, p = 0.002) were the most important predictors of recurrent ischemic stroke.