Effect of Intensive Blood Pressure Control on Incident Stroke Risk in Patients with Mild Cognitive Impairment

Abstract

Background

Patients with mild cognitive impairment may be at higher risk of incident stroke, but the effect of intensive blood pressure (BP) control on that risk has not been explored.

Methods

We performed a post-hoc analysis of the SPRINT trial and included patients with a baseline Montreal Cognitive Assessment (MoCA) score of 19-25 and without a prior history of stroke. The primary outcome was incident stroke (ischemic and hemorrhagic) during follow-up. We report the unadjusted cumulative risk our primary outcome by SPRINT randomization arm (intensive vs. standard BP control) and also fit Cox models to the primary outcome and adjusted for patient age at randomization, race/ethnicity, sex, baseline blood pressure, atrial fibrillation, diabetes, and smoking.

Results

We included 5091 patients (mean age 68.2, 44% female, 56.7% non-Hispanic white, and 50.2% randomized to intensive BP control), of which 95/5091 (1.9%) had an incident stroke during a mean of 3.8±0.9 years of follow-up. The risk of incident stroke in patients randomized to standard BP control was 57/2536 (2.3%) and to intensive BP control was 38/2555 (1.5%) (p=0.045). In the adjusted Cox model, the hazard ratio for incident stroke events with intensive BP control was 0.65 (95% CI 0.43-0.98, p=0.040).

Conclusion

Although the SPRINT trial failed to show a reduction in stroke with intensive BP control for all subjects, those with a MoCA score consistent with mild cognitive impairment at baseline had an association between intensive BP control and lower risk of incident stroke. Future trials of primary prevention of stroke may benefit from enrichment using baseline vascular biomarkers of elevated risk, such as mild cognitive impairment.

Graphical Abstract

Introduction

Meta-analyses confirm that mild cognitive impairment (MCI) and dementia increase the risk of future ischemic and hemorrhagic stroke.^1,2 Shared vascular risk factors, primarily hypertension, have emerged as central pathophysiologic mechanisms for both cognitive impairment and stroke.^3–5 However, prior research has not identified if intensive blood pressure control in hypertensive patients with MCI results in lower risk of stroke. Given potential vascular contributions to MCI, those with MCI may represent a unique subset of hypertensive individuals with evidence of subclinical cerebrovascular disease. We therefore hypothesized that intensive BP control would reduce the risk of clinical stroke in subjects with MCI enrolled in the Systolic Blood Pressure Intervention Trial (SPRINT).⁶

Methods

Study design

We performed a post-hoc analysis of the SPRINT trial, using a publicly available deidentified dataset supplied by the NHLBI that did not require IRB approval and is available at https://biolincc.nhlbi.nih.gov/home.⁶ We included stroke-free participants with MCI at the study baseline visit, defined as a Montreal Cognitive Assessment (MoCA) of 19-25. Prior research has shown that of eleven commonly used brief cognitive screening tools, MoCA is the best for identification of MCI.⁷ We used the MoCA range of 19-25 for MCI based on recommendations from a Cochrane Review and a study of the Alzheimer’s Disease Neuroimaging Initiative.^8,9 While other ranges have been suggested, specific to race/ethnicity, education, and age, such approaches typically improve specificity at the expense of sensitivity and utility.¹⁰

Outcome ascertainment

The primary outcome is stroke events, both ischemic and hemorrhagic, during follow-up in the SPRINT trial. The adjudication of stroke in SPRINT has been previously described but was a robust process that included routine follow-up query of change in health status, record review, and central adjudication.¹¹

Exposure ascertainment

The study exposure is the SPRINT randomization arm, which was either an intensive control arm (goal systolic blood pressure <120 mm Hg) versus standard control arm (goal systolic blood pressure <140 mm Hg). We report demographics and the cumulative risk of stroke in both study arms and tested for intergroup differences with the chi-squared test for binary variables, the Wilcoxon rank sum test for ordinal variables, and Student’s t-test for continuous variables. We fit time-to-event models with randomization as the predictor and report the p-value for the log-rank test. Finally, we fit Cox proportional hazards models that were a priori adjusted for potential confounders known to affect stroke risk including baseline patient age, race/ethnicity, sex, baseline systolic blood pressure, atrial fibrillation, diabetes, and smoking.^12,13 We verified the proportional hazards assumption of the multivariable Cox model and also tested interaction terms between the covariates in the Cox model and the exposure, to determine if there was heterogeneity of treatment effect.

Results

We included 5091 patients (Figure S1) (median MoCA score 23, mean age 68.2, 66.0% male, 56.7% non-Hispanic white, and 50.2% randomized to intensive BP control), of which 95/5091 (1.9%) had an incident stroke during a mean of 3.8±0.9 years of follow-up (Table 1). Despite being a subgroup of the full SPRINT cohort, there were not significant differences in demographics between those in the intensive versus standard blood pressure (BP) control arm (Table 1). The proportion with incident stroke in patients randomized to standard BP control was 2.3% (57/2536) and to intensive BP control was 1.5% (38/2555) (p=0.045). The Kaplan-Meier curve by randomization arm is seen in Figure 1, which revealed a log-rank p value of 0.042. The separation between the treatment arms increased over time. In the adjusted Cox model, the hazard ratio for incident stroke events with intensive BP control was 0.65 (95% CI 0.43-0.98, p=0.039). The p value for the test of the proportional-hazards assumption was 0.924. The interaction terms between the covariates in the main model and randomization arm were not significant (p>0.1). The association remained significant when adjusting for other potential confounders, although data missingness led to the null in one model (Table 2).

Table 1.

Demographics in the SPRINT randomization arms in our cohort of patients with mild cognitive impairment at baseline.

Variable	Standard blood pressure control (n=2555)	Intensive blood pressure control (n=2536)	P value
Age (years)	68.1±9.6	68.3±9.6	0.510
Male sex	1699 (67.0%)	1662 (65.1%)	0.143
Race/ethnicity			0.345
Non-Hispanic White	1429 (56.4%)	1456 (57.0%)
Non-Hispanic Black	823 (32.5%)	785 (30.7%)
Hispanic	245 (9.7%)	263 (10.3%)
Other	39 (1.5%)	51 (2.0%)
<College education	906 (35.7%)	923 (36.1%)	0.766
Full-time employment (n=5090)	512 (20.2%)	510 (20.0%)	0.833
Vigorous activity (n=5072)			0.946
≤ 1/week	1354 (53.6%)	1353 (53.1%)
2-4/week	820 (32.5%)	836 (32.8%)
≥5/week	352 (13.9%)	357 (14.0%)
Diabetes	45 (1.8%)	36 (1.4%)	0.297
Atrial fibrillation	201 (7.9%)	224 (8.8%)	0.278
Current smoking	372 (14.7%)	350 (13.7%)	0.321
Aspirin use (n=5077)	1285 (50.8%)	1315 (51.6%)	0.589
Baseline systolic blood pressure (mm Hg)	145.0±11.0	145.2±11.2	0.389
Baseline LDL cholesterol (n=5045)	111.5±34.5	111.7±35.2	0.825
Baseline MoCA	22, 21-24	23, 21-24	0.008
Average follow-up (days)	1364±342	1378±332	0.199
Stroke during follow-up	57 (2.3%)	38 (1.5%)	0.045

^*Binary variables shown as n (%), ordinal variables as median (IQR), and interval variables as mean±SD.

Figure 1.

Kaplan-Meier curve, showing stroke events by randomization arm.

Table 2.

Association between SPRINT randomization and incident stroke events.

Model	Hazard ratio with intensive blood pressure control	95% CI	P value
Unadjusted	0.66	0.44-0.99	0.044
Adjusted for age, sex, race/ethnicity	0.65	0.43-0.98	0.040
Fully adjusted*	0.65	0.43-0.98	0.040
Further adjusted for baseline MoCA	0.65	0.43-0.98	0.041
Further adjusted for baseline aspirin use**	0.65	0.43-0.98	0.041
Further adjusted for aspirin use, LDL cholesterol, employment status, education, and frequency of vigorous activity***	0.67	0.44-1.01	0.055

^*Adjusted for baseline patient age, race/ethnicity, sex, baseline systolic blood pressure, atrial fibrillation, diabetes, and smoking

^**n=5077

^***n=5025

Discussion

Although the main results of the SPRINT trial did not show that intensive BP control significantly lowered incident stroke risk (HR 0.89, 95% CI 0.63-1.25),⁶ in a post-hoc analysis restricted to patients with baseline MoCA score consistent with MCI, we report an association between intensive BP control and lower risk of incident stroke (HR 0.65, 95% CI 0.43-0.98). This finding is consistent with the American Heart Association 2014 guideline on primary stroke prevention, which reported a relative risk of incident stroke of 0.80 (95% CI 0.70-92) with a blood pressure goal of <130 mm Hg versus 130-139 mm Hg.¹⁴ However, the meta-analysis relative risk of 0.80 in the AHA guideline is a smaller effect size than we found in our analysis, perhaps reflecting higher baseline stroke risk in patients with MCI or the lower BP goal in SPRINT (<120 mm Hg).

The novel association reported in this analysis is hypothesis generating, but addresses an important element of shared disease processes. While MCI can be the result of neurodegeneration, it is more frequently a sequelae of vascular disease.^4,15 Accordingly, patients with MCI have a known higher risk of incident stroke.^1,2 Because hypertension accounts for a high proportion of the attributable risk of both MCI and incident stroke, there is biologic plausibility that intensive BP control could mitigate the risk of developing incident stroke in those with MCI.

However, our study has important limitations that warrant discussion. The first is that SPRINT was not designed to answer this specific research question and there is unmeasured confounding in subgroup analyses, despite there being no difference in known major stroke risk factors between the BP randomization arms. The second is that while MoCA is the preferred screening tool for MCI diagnosis,⁷ it is not in itself a diagnosis. Thus, the results of our analysis are not definitive and will have to be replicated in prospective randomized clinical trials of intensive BP control. Finally, because patients with prior stroke were excluded from SPRINT, we are not able to report the effect of intensive BP control on the risk of recurrent stroke in stroke patients with MCI, which is an important knowledge gap.

Conclusion

Although the SPRINT trial failed to show a reduction in stroke with intensive BP control for all subjects, those with a MoCA score consistent with MCI at baseline had an association between intensive BP control and lower risk of incident stroke. Future trials of stroke primary prevention may benefit from enrichment using baseline vascular biomarkers of elevated risk, such as MCI.