Hypotensive Episodes Identified on 24-h Ambulatory Blood Pressure and Impaired Cognitive Function: Insights from the SPRINT Study

Abstract

Background:

Hypotensive episodes detected by 24-h ambulatory blood pressure monitoring (ABPM) capture daily cumulative hypotensive stress and could be clinically relevant to cognitive impairment, but this relationship remains unclear.

Methods:

We included participants from the Systolic Blood Pressure Intervention Trial (receiving intensive or standard BP treatment) who had 24-h ABPM measured near the 27-month visit and subsequent biannual cognitive assessments. We evaluated the associations of hypotensive episodes (defined as SBP drops of ≥20 mmHg between two consecutive measurements that reached <100 mmHg) and hypotensive duration (cumulative time of SBP <100 mmHg) with subsequent cognitive function using adjusted linear mixed models. We further assessed 24-h average BP and variability.

Results:

Among 842 participants with treated hypertension (mean age, 71±9 years; 29% women), the presence (versus absence) of recurrent hypotensive episodes (11%) was associated with lower digit symbol coding (DSC) scores (difference in z-scores: −0.249, 95% CI: −0.380, −0.119) and faster declines (difference in z-score changes: −0.128, 95% CI: −0.231, −0.026). A consistent dose-response association was also observed for longer hypotensive duration with worse MoCA and DSC scores. The association with DSC scores remained significant after further adjusting for 24-h average BP and variability, and was not observed for hypotension defined by clinic, orthostatic, or 24-h average BP. Intensive BP treatment increased 24-h hypotensive episodes and modified its association with the decline in DSC score.

Conclusion:

24-h hypotensive episodes were associated with worse cognitive function, especially in processing speed, and could be a novel marker for optimal BP control and dementia prevention.

Graphical Abstract

Introduction

Hypertension is a leading cause of cardiovascular events and a major modifiable risk factor for Alzheimer’s disease and related dementias, contributing significantly to premature death, disability, and dependency in older adults.^1,2 Hypotensive episodes, such as orthostatic, postprandial, and transient hypotension, frequently occur in older hypertensive patients due to impaired BP regulation and are also linked to increased risk of stroke, syncope, falls, dementia, and death.^3–6 Previous studies have also observed a U-shaped association between late-life clinic BP and adverse cognitive outcomes as well as faster cognitive decline associated with low BP.⁷ The Systolic Blood Pressure Intervention Trial (SPRINT) demonstrated that intensive BP control prevented mild cognitive impairment.⁸ However, the increased incidence of hypotension and syncope associated with aggressive BP lowering may limit the overall benefit and could be targeted to improve BP management.⁹ Increased BP variability over various periods, capturing large rises and falls in BP, is linked to cardiovascular disease, stroke, dementia, and mortality independent of BP level in older adults.^10–18 These observations indicate that, in addition to hypertension, older adults may be particularly vulnerable to hypotensive episodes due to impaired mechanisms to maintain stable BP (e.g., baroreflex sensitivity) and stable blood flow (e.g., cerebral autoregulation) associated with hypertension and aging.^19–21 These age-related physiological changes could predispose older adults to hypotension-induced cerebral hypoperfusion and cerebral ischemia, and subsequently, stroke and dementia.

Ambulatory BP monitoring (ABPM) over a continuous 24 hours may better capture hypotensive episodes than clinic BP and their association with cognitive impairment and dementia.^22–24 However, the relationship between 24-h hypotensive episodes and cognitive function in hypertensive patients is undetermined, and it is unknown whether hypotensive episodes explain the association between BP variability and cognitive impairment. The implication of intensive BP treatment in this putative relationship also remains unclear.

In this post-hoc analysis of the SPRINT trial, we investigated the associations of hypotensive episodes obtained from 24-h ABPM with cognitive function. We hypothesized that 24-h hypotensive episodes are associated with cognitive impairment and that this association partly explains the relationship between large BP variability and adverse cognitive outcomes.

Methods

Data Source

The data that support the findings of this study are available from the National Heart, Lung, and Blood Institute Biologic Specimen and Data Repositories (https://biolincc.nhlbi.nih.gov/home/).

Study Population

SPRINT was an NIH-funded, multicenter randomized clinical trial conducted among 9361 participants in the United States and Puerto Rico. The design and protocol of SPRINT have been published previously.^9,25 Briefly, participants were invited to participate if aged 50 years or older with a screening systolic BP between 130 and 180 mmHg and had increased cardiovascular risk. Participants without diabetes, dementia, heart failure, or a history of stroke were randomized to either intensive BP lowering (systolic BP target < 120 mmHg) or standard BP lowering (systolic BP target < 140 mmHg). Intensive BP control reduced mean SBP and the risk of CVD over a median follow-up of 3.26 years,⁹ and reduced the risk of mild cognitive impairment over a median follow-up of 5.11 years.⁸ The ambulatory BP ancillary study was conducted at 27 months of follow-up among 897 SPRINT participants (453 in the intensive group and 444 in the standard group) at 15 clinical sites.²⁶ After excluding participants who did not complete cognitive assessments after ambulatory blood pressure monitoring (ABPM), 842 (94%) participants were included in the current study (see flowchart and study timeline illustrated in Figure S1).

24-h Ambulatory and Clinic Blood Pressure Measurements

The protocol for 24-h ABPM has been described in previous publications.²⁶ Briefly, ambulatory BP was measured within 3 weeks of the 27-month study visit using SpaceLabs Medical Model 90207 monitors (Snoqualmie, WA) with a standard 24-h protocol that recorded BP every 30 minutes on the participant’s nondominant arm. A recording with at least 14 readings between 6:00 am and 12:00 midnight and at least 6 readings between 12:00 midnight and 6:00 am was deemed to be acceptable for analysis.³⁰ Clinical BP was measured at each research visit by trained clinical staff using an automated measurement system (HEM-907 XL, Omron Healthcare, Lake Forest, IL). A mean of three consecutive BP measurements obtained at one-minute intervals after 5 minutes of quiet rest was used as the office BP for analyses.

Assessment of 24-h Hypotensive Episodes and Other BP Phenotypes

Our primary exposures are the presence of recurrent (≥ 2) hypotensive episodes and the hypotensive duration. We defined a hypotensive episode as a systolic BP drop ≥ 20 mmHg between two consecutive ABPM measurements that reached < 100 mmHg. We defined the hypotensive duration as the cumulative time with systolic BP < 100 mmHg over a 24-h period, based on kernel interpolation. The cutoff is consistent with previous studies of ambulatory hypotension.^22–24 Figure S2 shows examples of these hypotensive episodes based on 24-h individual BP readings. In secondary analyses, we also defined hypotensive episodes based on diastolic BP (presence of a diastolic BP drop ≥ 10 mmHg that reached < 50 mmHg and cumulative time of diastolic BP < 50 mmHg over a 24-h period) and assessed recurrent diastolic hypotensive episodes as the exposure. The criteria for defining hypotensive episodes and hypotensive duration based on 24-h ambulatory systolic BP and diastolic BP were detailed in Table S1. We also assessed these measures for nighttime (from 1 am to 6 am) and daytime (from 9 am to 9 pm), respectively.²⁹ We assessed nocturnal dipping patterns using the night/day ambulatory systolic BP ratio: extreme dipping (< 0.8), normal dipping (≥ 0.8 and < 0.9), non-dipping (≥ 0.9 and < 1), and reverse dipping (≥ 1).^29,30 24-h BP variability was defined by the coefficient of variation of individual BP readings over 24 hours.³¹

Cognitive Assessments

Cognitive assessments in SPRINT have been described previously.⁸ Briefly, at SPRINT baseline and during follow-up (year 2, year 4, and year 6), participants underwent in-person cognitive screening assessments by trained examiners that included the Montreal Cognitive Assessment (MoCA) for global cognitive function; the Logical Memory form I (LM I) and II (LM II) subtests of the Wechsler Memory Scale for learning and memory cognitive domain; and the Digit Symbol Coding (DSC) Test of the Wechsler Adult Intelligence Scale for processing speed domain. Each test score was standardized by subtracting the baseline median and dividing by the baseline interquartile range to get comparable scales and used in all the subsequent analyses. In the current study, we used cognitive function assessed after 24-h ABPM (approximately 27 months of follow-up). 99% of participants had cognitive assessments at 4 years of follow-up, and 60% repeated cognitive measurements at 6 years of follow-up (Figure S1). Our primary outcome was the average cognitive function after a 24-hour ABPM assessment. Cognitive decline after a 24-hour ABPM assessment was assessed as a secondary outcome, as it was only available in a subset (60%) of participants.

Other Covariates Measurements

We used covariate information collected before and closest to 24-h ABPM. Specifically, clinical BP and antihypertensive medication use were assessed at 27 months of follow-up. Body mass index, smoking status, estimated glomerular filtration rate (eGFR, calculated using the 4-variable Modification of Diet in Renal Disease equation),⁹ and fasting plasma glucose were assessed at 24 months of follow-up. Information on age, sex, race, education, history of cardiovascular disease (CVD), and statin use was obtained from the SPRINT baseline survey. Missing values in covariates (ranging from 0.1% to 4.5%) were imputed using random forest methods and used for subsequent analyses.

Statistical analyses

We used linear mixed models to estimate the associations between hypotensive episodes and cognition function. Models included random effect for participant and clinic site and were adjusted for years of cognitive assessment since ABPM, age, sex, race, treatment assignment, education level, smoking status, body mass index (BMI), eGFR and fasting glucose levels, and number of antihypertensive medications, morning dose and evening dose of antihypertensive medication at ABPM, and history of CVD and statin use at randomization. In the primary analysis, we modeled the recurrence (yes vs no) and hypotensive duration (per hour) as the main exposures. The covariate selection was based on prior literature on these factors as potential confounding factors in the association between BP and cognitive outcomes. To address potential over-adjustment, we constructed a model only adjusting for age, sex, race, education, treatment assignment, and years of cognitive assessment. We also estimated the dose-response relationships between 24-h hypotensive duration and cognitive function by modeling hypotensive duration as ordered categorical variables (0, 0–2, 2–4, 4–6, ≥6 hours) based on the skewed distributions and we additionally used penalized spline to characterize the dose-response relationship. We further estimated the associations between hypotensive episodes and rate of cognitive decline, which was measured by incorporating an interaction term between hypotensive episodes and the follow-up visits of cognitive assessment after BP measurements in the linear mixed models.

We next compared 24-h recurrent hypotensive episodes with 24-h BP levels and variability in their associations with cognitive function, first in separate models and then with mutual adjustment, in addition to covariates specified in the primary models. To evaluate whether 24-h hypotensive episodes better capture cumulative hypotensive stress and its distinct association with cognitive function compared to other common BP phenotypes, we compared 24-h recurrent hypotensive episodes with hypotension defined based on orthostatic BP (defined as a drop in systolic BP of at least 20 mmHg or in diastolic BP of at least 10 mmHg at 1 minute after standing up compared with the value obtained when sitting; diagnosed before ABPM), clinic BP, 24-h average and nighttime average BP (defined as a systolic BP value lower than 100 mmHg), as well as nocturnal systolic BP dipping patterns.

In the secondary analysis, we estimated the associations of daytime and nighttime hypotensive episodes with cognitive function and evaluated hypotensive episodes defined based on 24-h diastolic BP readings. Given that the intensive BP treatment group had an increased risk of hypotension in SPRINT,⁹ we further assessed if intensive BP treatment modified the association between hypotensive episodes and cognitive function. To examine whether participants who underwent intensive BP treatment and experienced hypotensive episodes had worse cognitive function, we combined groups of randomized treatment assignment and hypotensive episodes in the analyses. Given that subclinical neurodegeneration and impaired cognitive function might impact central BP regulation (i.e., potential reverse causation), we conducted several sensitivity analyses: 1) additionally adjusting for cognitive function at ABPM assessment, 2) additionally adjusting for history of depression, and 3) excluding participants with probable dementia or mild cognitive impairment diagnosed before ABPM (n = 55).

We used false discovery rate (FDR) method to adjust the P values for multiple comparisons due to the multiple cognitive tests. An FDR-adjusted two-sided P value (FDR-P) lower than 0.05 was considered statistically significant. Data analyses were performed using R software (Version 4.3.1; R Core Team).

Results

Characteristics of study participants

Table 1 shows the ABPM baseline characteristics of 842 participants stratified by the presence of recurrent hypotensive episodes on 24-h ABPM. Age and race did not differ significantly between the two groups. In contrast, recurrent hypotensive episodes were more prevalent in men, those receiving intensive BP treatment, and individuals with higher BMI and lower DSC test scores at ABPM. Individuals with recurrent hypotensive episodes had both lower clinic and 24-h average BP levels, as well as higher 24-h BP variability, compared to those without. Characteristics were similar when further stratified by treatment assignment (Table S2).

Table 1.

Characteristics of study participants by recurrent hypotensive episodes identified on 24-hour ambulatory systolic BP.

	Recurrent hypotensive episodes (n = 842)
Characteristics at ABPM baseline†	Yes (n = 95)	No (n = 747)	P value*
Age, mean (SD), yrs	71.5 (9.5)	71.3 (9.4)	0.81
Age ≥ 75 yrs	36 (37.9%)	293 (39.2%)	0.89
Female	16 (16.8%)	227 (30.4%)	0.009
Black race	25 (26.3%)	217 (29.0%)	0.66
Education			0.016
Lower than high school	1 (1.1%)	49 (6.6%)
High school graduate	14 (14.7%)	98 (13.1%)
Some college	44 (46.3%)	273 (36.5%)
College graduate	7 (7.4%)	129 (17.3%)
Above college	29 (30.5%)	198 (26.5%)
Intensive blood pressure control	66 (69.5%)	360 (48.2%)	<0.001
Clinical SBP, mean (SD), mmHg	119.9 (14.4)	128.5 (15.4)	<0.001
Clinical DBP, mean (SD), mmHg	66.0 (11.2)	70.2 (11.9)	0.002
24-h average systolic BP, mean (SD), mmHg	115.6 (10.1)	129.7 (12.7)	<0.001
24-h average diastolic BP, mean (SD), mmHg	65.3 (6.9)	72.5 (9.5)	<0.001
24-h systolic BP coefficient of variation, mean (SD)	0.14 (0.03)	0.11 (0.03)	<0.001
24-h diastolic BP coefficient of variation, mean (SD)	0.17 (0.04)	0.14 (0.04)	<0.001
24-h hypotensive duration, median (IQR), hour	3.7 (1.7, 5.9)	0.0 (0.0,1.0)	<0.001
BMI, mean (SD), kg/m2	30.8 (5.4)	29.3 (5.5)	0.013
Never smoked	43 (45.3%)	344 (46.1%)	0.93
Estimated GFR, mean (SD), mL/min/1.73 m2	68.6 (20.0)	70.9 (20.9)	0.20
Estimated GFR< 60 mL/min/1.73 m2	32 (33.7%)	234 (31.3%)	0.77
Fasting plasma glucose, mean (SD), mg/dL	101.5 (22.0)	98.5 (14.7)	0.13
History of cardiovascular disease at trial randomization	22 (23.2%)	156 (20.9%)	0.71
Statin use at trial randomization	45 (47.4%)	300 (40.2%)	0.26
Number of antihypertensive medications, mean (SD)	2.6 (1.2)	2.3 (1.3)	0.047
Morning dose of antihypertensive medicationa	70 (73.7%)	516 (69.1%)	0.42
Evening dose of antihypertensive medicationb	32 (33.7%)	269 (36.0%)	0.74
Montreal cognitive assessment, median (IQR)c	23 (21, 25)	24 (21, 26)	0.26
Logical memory form I, median (IQR)d	21 (18, 23)	21 (17, 24)	0.85
Logical memory form II, median (IQR)e	9 (7, 12)	9 (7, 12)	0.93
Digit Symbol Coding Test, median (IQR)f	45 (40, 56)	50 (42, 60)	0.019
Probable dementia or mild cognitive impairmentg	9 (9.5%)	46 (6.2%)	0.32

Data are presented as mean (SD) or median (IQR) for continuous variables, and as N (%) for categorical variables.

Abbreviations: ABPM, ambulatory blood pressure monitoring; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); DBP, diastolic blood pressure; GFR, glomerular filtration rate; SBP, systolic blood pressure; SD, standard deviation; SPRINT, Systolic Blood Pressure Intervention Trial.

^†ABPM baseline was defined as the time of ABPM measurement (27-month visit since randomization). The characteristics data were collected at or closest before ABPM baseline. If not specified, data were measured at either 24-month or 27-month visit since randomization.

^*The difference was compared using chi-square test for the categorical variables and Kruskal-Wallis test for the continuous variables. P value < 0.05 indicates significant difference between groups of recurrent and non-recurrent hypotensive episodes.

^aMorning dose of antihypertensive medication was from 4 am to 10 am.

^bEvening dose of antihypertensive medication was from 6 pm to 2 am.

^cScores range from 0 to 30, with higher scores denoting better cognitive function.

^dScores range from 0 to 28, with higher scores denoting better cognitive function.

^eScores range from 0 to 14, with higher scores denoting better cognitive function.

^fScores range from 0 to 135, with higher scores denoting better cognitive function.

^gDiagnosed since randomization and before ABPM.

Distribution of hypotensive episodes

Both frequency and duration of 24-h hypotensive episodes were highly skewed in the study population, with approximately 60% of participants having no hypotensive episodes (Figure S3). As shown in Figure 1, the proportion of 24-h recurrent hypotensive episodes in the intensive group was significantly higher than that in the standard group (15.5% vs 7.0%, P < 0.001). Similarly, the proportion of 24-h hypotensive duration ≥ 1 hour in the intensive group was also significantly higher than the standard group (45.8% vs 18.5%, P < 0.001). Similar patterns were observed for daytime and nighttime measurements (Figure 1 and Table S3).

Figure 1. Percentage of participants with recurrent hypotensive episodes (A) and hypotensive duration ≥ 1 hour (B) in intensive and standard treatment groups during 24h, daytime and nighttime.

Daytime defined based on clock time from 9:00 AM to 9:00 PM, and nighttime defined based on clock from 1:00 AM to 6:00 AM. Difference between intensive and standard group were examined using chi-square test.

24-h hypotensive episodes and cognitive function

Table 2 and Figure 2 show that recurrent hypotensive episodes and longer hypotensive duration were significantly associated with a lower DSC score (indicating worse processing speed) after adjusting for potential confounding factors and multiple testing correction. Compared to the participants without recurrent hypotensive episodes, those with recurrent hypotensive episodes had a significantly lower standardized DSC score (difference: −0.249, 95% CI: −0.380 to −0.119, Table 2). Each hour increase in hypotensive duration was associated with worse standardized DSC scores in a dose-response manner (difference: −0.029; 95% CI: −0.046 to −0.012, P for trend = 0.002, Table 2, Figure 2). The penalized spline analysis indicated that the DSC score appears to be even lower with longer hypotensive duration (Figure S4). We also observed that longer hypotensive duration was associated with lower global cognitive function assessed by MoCA (P for trend = 0.036, Figure 2 and Table S4). Secondary analysis also showed that 24-h recurrent hypotensive episodes and longer hypotensive duration were both significantly associated with faster decline in processing speed assessed by DSC test (P < 0.05, Figure 3). No significant associations were observed between hypotensive episodes with learning and memory domain assessed by logical memory forms I and II.

Table 2.

Associations between ABPM-identified hypotensive episodes and cognitive function.

		Recurrent hypotensive episodes (Yes vs No)				24-h hypotensive duration (per hour increase)
Cognitive function		β (95% CI)	P value	FDR-P *		β (95% CI)	P value	FDR-P *
Basic model†
MoCA		−0.088 (−0.202, 0.025)	0.126	0.252		−0.017 (−0.031, −0.002)	0.026	0.052
LM I		0.018 (−0.109, 0.145)	0.785	0.944		−0.006 (−0.022, 0.010)	0.463	0.590
LM II		−0.004 (−0.127, 0.118)	0.944	0.944		−0.004 (−0.020, 0.011)	0.590	0.590
DSC		−0.233 (−0.364, −0.103)	< 0.001	0.002		−0.029 (−0.046, −0.012)	<0.001	0.003
Fully adjusted model‡
MoCA		−0.080 (−0.194, 0.033)	0.166	0.332		−0.016 (−0.031, −0.002)	0.030	0.060
LM I		0.018 (−0.110, 0.146)	0.783	0.872		−0.006 (−0.022, 0.011)	0.491	0.610
LM II		−0.010 (−0.133, 0.113)	0.872	0.872		−0.004 (−0.020, 0.012)	0.610	0.610
DSC		−0.249 (−0.380, −0.119)	<0.001	0.001		−0.029 (−0.046, −0.012)	<0.001	0.003

Abbreviations: ABPM, ambulatory blood pressure monitoring; MoCA, Montreal cognitive assessment; LM I, logical memory form I; LM II, logical memory II; DSC, digit symbol coding test; FDR, false discovery rate.

^†Estimates were based on linear mixed models adjusted for years of cognitive assessment since ABPM, treatment assignment, age, sex, race, and education, with random effects for participant and site.

^‡Estimates were based on linear mixed models adjusted for years of cognitive assessment since ABPM, treatment assignment, age, sex, race, education, smoking status, BMI, eGFR and fasting glucose levels, number of antihypertensive medications, morning dose and evening dose of antihypertensive medication, history of CVD and statin use at randomization, with random effects for participant and site.

^*P values were adjusted for multiple comparisons using false discovery rate (FDR) method.

Figure 2. Dose-response associations between 24-h hypotensive duration and cognitive function.

Number (%) participants in each group of 24-h hypotensive duration were listed in the table at right. Estimated from linear mixed model adjusted for days of cognitive function measurement since ABPM, treatment assignment, age, sex, race, education, smoking status, BMI, eGFR and fasting glucose level, number of antihypertensive medications, morning dose and evening dose of antihypertensive medication, history of CVD and statin use, with random effects for participant and site. The points indicate the effect estimates, with error bars showing the 95% confidence intervals. P values for trend were adjusted for multiple comparisons using false discovery rate (FDR) method and P < 0.05 indicates statistically significant.

Figure 3. Associations of recurrent hypotensive episodes (A) and hypotensive duration (B) with cognitive decline.

833 participants had available cognitive data in year 2 follow-up and 502 in year 4 follow-up were include in analyses. Difference in cognitive decline were estimated from linear mixed model adjusted for treatment assignment, age, sex, race, education, smoking status, BMI, eGFR and fasting glucose level, number of antihypertensive medications, morning dose and evening dose of antihypertensive medication, history of CVD and statin use, with random effects for participant and site. The points indicate the mean values, with error bars showing the standard errors.

Similar association patterns were observed for hypotensive episodes defined by diastolic BP (Table S5). When stratified by treatment assignment, the association between recurrent hypotensive episodes and DSC scores was more pronounced and only remained significant in the intensive BP treatment group (β = −0.279, 95% CI, −0.449 to −0.110, Table S6). Moreover, BP treatment significantly modified the association between recurrent hypotensive episodes and decline in DSC scores (P for interaction = 0.02) and recurrent hypotensive episodes was associated with faster decline in DSC scores only among participants receiving intensive BP treatment (β = −0.197, 95% CI, −0.343 to −0.051, Table S6). Additionally, participants who underwent intensive BP treatment and experienced recurrent hypotensive episodes had the lowest DSC levels and fastest DSC decline compared to those who underwent standard treatment and without recurrent hypotensive episodes (Figure S5 and Figure S6).

In the sensitivity analysis additionally adjusting for baseline DSC scores, the association between recurrent hypotensive episodes and DSC z-scores was moderately weakened but remained significant (difference: −0.135, 95% CI: −0.222 to −0.048, Table S7). The sensitivity analyses additionally adjusting for history of depression and excluding participants with dementia or mild cognitive impairment diagnosed before ABPM did not significantly alter the main results (Table S8 and Table S9). When daytime and nighttime BP were assessed separately, recurrent hypotensive episodes during daytime appeared to have more pronounced associations with worse cognitive function than those during the nighttime (Figure S7).

24-h recurrent hypotensive episodes, average BP levels, BP variability, and cognitive function

As shown in Table 3, when assessed in separate models, 24-h recurrent hypotensive episodes and BP variability (coefficient of variation), but not 24-h average BP levels, were associated with worse processing speed after adjusting for potential confounding factors. The associations with cognition for both recurrent hypotensive episodes and BP variability remained significant and essentially unchanged after further adjusting for 24-h average BP. In contrast, when 24-h recurrent hypotensive episodes and BP variability were included in the same model, the association with cognition for recurrent hypotensive episodes remained significant (P = 0.003), whereas the association for 24-h BP variability was attenuated and no longer statistically significant (P = 0.120), with hypotensive episodes explaining 40% of the association between 24-h BP variability and cognitive function. When all three BP phenotypes were included in the same model, only the association of 24-h recurrent hypotensive episodes with worse processing speed remained statistically significant (P = 0.004).

Table 3.

Associations of 24-hour recurrent hypotensive episodes, average SBP levels, and SBP variability with standardized DSC scores.

24-hour ABPM phenotype	β (95% CI)	P value	Change in β estimate, %*
24-hour recurrent hypotensive episodes, yes vs no †
Primary model	−0.249 (−0.380, −0.119)	<0.001	0 (Ref)
Primary model + 24-hour average SBP level	−0.242 (−0.380, −0.105)	<0.001	−2.81
Primary model + 24-hour SBP coefficient of variation	−0.212 (−0.350, −0.073)	0.003	−14.86
Primary model + 24-hour average SBP level + 24-hour SBP coefficient of variation	−0.211 (−0.354, −0.068)	0.004	−15.26
24-h average SBP level, per 10 mmHg increase †
Primary model	0.026 (−0.009, 0.061)	0.145	0 (Ref)
Primary model + 24-hour recurrent hypotensive episodes	0.006 (−0.031, 0.043)	0.760	−79.92
Primary model + 24-hour SBP coefficient of variation	0.015 (−0.021, 0.051)	0.422	−42.31
Primary model + 24-hour recurrent hypotensive episodes + 24-hour SBP coefficient of variation	0.001 (−0.037, 0.038)	0.964	−96.15
24-hour SBP coefficient of variation, per SD increase †
Primary model	−0.058 (−0.100, −0.016)	0.006	0 (Ref)
Primary model + 24-hour average SBP level	−0.054 (−0.097, −0.010)	0.015	−6.90
Primary model + 24-hour recurrent hypotensive episodes	−0.035 (−0.079, 0.009)	0.120	−39.66
Primary model + 24-hour average SBP level + 24-hour recurrent hypotensive episodes	−0.035 (−0.080, 0.010)	0.127	−39.66

Abbreviations: ABPM, ambulatory blood pressure monitoring; DSC, digit symbol coding test; SBP, systolic blood pressure; SD, standard deviation.

^†Primary linear mixed model was adjusted for years of cognitive function measurement since ABPM, treatment assignment, age, sex, race, education, smoking status, BMI, eGFR and fasting glucose levels, number of antihypertensive medications, morning dose and evening dose of antihypertensive medication, history of CVD and statin use, with random effects for participant and site. Other models were additionally adjusted models based on primary model.

^*The percent change in β estimate for each feature in the additional adjusted models was calculated by subtracting the primary adjusted β estimate from the additional adjusted β estimate, then divided this difference by the primary adjusted β estimate and multiplied by 100.

Comparisons of recurrent hypotensive episodes with other hypotensive phenotypes

Table S10 shows the prevalence of different hypotension phenotypes and their associations with cognition. Recurrent hypotensive episodes based on 24-h individual BP readings had the highest proportion (11.3%) compared with orthostatic hypotension (8.6%) and hypotension defined based on average BP during clinic visits (1.9%), over 24 hours (0.8%), during daytime (0.7%) and nighttime (7.4%), respectively, and was the only phenotype that was associated with worse cognition (P < 0.001). Nocturnal dipping patterns and nocturnal hypertension were not associated with cognition (Table S11).

Discussions

In this post-hoc analysis of the SPRINT trial among older adults with hypertension, we found that recurrent hypotensive episodes and longer hypotensive duration identified through 24-h ambulatory BP monitoring (ABPM) were associated with worse processing speed and a faster decline. The association was independent of clinic BP, 24-h average BP levels, and 24-h BP variability. This suggests that hypotensive episodes identified based on 24-h individual BP readings may be a more sensitive BP phenotype clinically relevant to cognitive impairment and could help identify hypertensive patients at higher risk of adverse cognitive outcomes. Intensive BP treatment increased the occurrence of 24-h hypotensive episodes and modified its association with cognitive decline, indicating the potential of targeting hypotensive episodes for better BP treatment and dementia prevention.

Our study provides new insights into the complex relationship between BP and worse cognitive function in older adults with hypertension. A previous cross-sectional study involving 91 nursing home residents reported a higher prevalence of dementia among those with 24-h hypotensive episodes compared to those without.²² Our study extends this evidence by demonstrating the associations between 24-h hypotensive episodes and subsequent lower digit symbol coding (DSC) scores indicative of worse processing speed and its decline in a larger sample size. This adds to the growing body of evidence linking hypotension to cognitive impairment and dementia in older adults.^32–35 Moreover, our findings also extend previous observations by showing that 24-h hypotensive episodes could partly explain the association of higher 24-h BP variability with impaired cognitive function and dementia.^36,37 Our findings are consistent with a previous study showing more rapid cognitive decline in patients with mild cognitive impairment and lower treated daytime BP.³⁸ These data suggest the potential importance of managing hypotension in populations at high risk of cognitive impairment.

Nearly 11% of participants experienced recurrent hypotensive episodes based on 24-h individual BP readings, including 16% receiving intensive BP treatment, and the median hypotensive duration was 3.7 hours (out of 24 hours) in these participants. However, fewer than 2% of the participants were classified as having hypotension according to clinic BP and 24-h average BP. Moreover, none of these other measures of hypotension were associated with cognitive function. These observations suggest that most hypotensive episodes were missed in routine clinic snapshot BP assessments, and individual readings of 24-h ABPM can serve as a more sensitive tool for detecting hypotensive episodes that are potentially clinically relevant, especially for cognitive outcomes, in older adults.

We observed that participants receiving intensive BP treatment had a significantly higher proportion of recurrent hypotensive episodes compared to the standard treatment. This is in line with a previous report in SPRINT showing that participants receiving intensive (versus standard) BP treatment tended to have a faster decline in DSC cognitive scores.³⁹ These observations suggest that the cognitive benefits of intensive BP treatment might be offset by the higher occurrence of hypotensive episodes with intensive BP treatment. We observed that intensive BP treatment assignment modified the association between hypotensive episodes and cognitive decline. We also observed that the association between hypotensive episodes and cognition was more pronounced and remained significant only in the intensive treatment group and that participants who underwent intensive BP treatment and experienced hypotensive episodes had the lowest cognitive scores and fastest cognitive decline compared to those who underwent standard treatment and no hypotensive episodes. These data together indicate the potential importance of monitoring and reducing hypotensive episodes in hypertensive patients for better cognitive outcomes, especially in individuals at higher risk of hypotension (e.g., in men and those with higher BMI). Intensive BP treatment might be one of the triggers of hypotensive episodes. Further evidence is required to establish whether the observed relationship between hypotensive episodes and cognitive outcomes is causal, due to reverse causation, or involves both in a vicious feedback loop.

We also observed that the association of hypotensive episodes with worse cognition appears to be more pronounced during daytime compared to nighttime, which could be due to the higher cerebral metabolic demand during daytime (e.g., physical exercise). The upright position during the day also requires higher BP to overcome the height difference between the brain and the heart and maintain a sufficient brain blood supply.⁴⁹ Additionally, cerebrovascular function during daytime and nighttime exhibits different patterns.⁴² These factors could make the brain more susceptible to daytime hypotensive episodes. Future studies assessing real-time triggers of 24-h hypotensive episodes, such as through simultaneous monitoring of physical activity and sleep, would further elucidate the underlying mechanisms and preventive strategies.

Several possible mechanisms may link hypotensive episodes to cognitive impairment. First, a causal explanation is plausible. Baroreflex sensitivity and cerebral autoregulation may be impaired with aging and hypertension, which could place older adults at higher risk of unstable BP, especially hypotensive episodes, and increased vulnerability to cerebral hypoperfusion and dementia risk.⁴³ This, in turn, could lead to hypoxia, brain ischemia, cerebral metabolism disorders, and cognitive impairment associated with hypotensive episodes.⁴⁴ This explanation is congruent with our observation of the association between hypotensive episodes and worse cognitive performance in the processing speed domain, which was also observed in another study among patients with neurogenic orthostatic hypotension.⁴⁵ Processing speed is a domain at an upper level of cognitive functioning conceptualization potentially involving the frontal lobe,^46,47 which was suggested to have the highest ischemic vulnerability to hypoperfusion.⁴⁸ The assessment of neuroimaging biomarkers including whiter matter lesions in future studies will help inform the mechanistic link between hypotension and cognitive impairment. Alternatively, the observed association could be due to reverse causation, although we observed similar findings after excluding participants with prevalent mild cognitive impairment or probable dementia and after further adjusting for baseline cognitive levels. Subclinical neurodegeneration underlying early cognitive alterations may also impact the central nervous regulation of BP, leading to hypotensive episodes. Future studies with longer follow-ups in healthier individuals could help further clarify the temporal relationship between hypotensive episodes and cognitive impairment.

The strengths of our study include comprehensive assessments of cognitive function across multiple domains, the use of 24-h individual BP readings to define hypotensive episodes and hypotensive duration, and the integration of various BP phenotypes indicative of age-related impaired BP dynamics in the analyses. The randomized trial design of the SPRINT study also enabled us to assess the clinical implication of hypotensive episodes for intensive BP treatment. Several limitations should also be noted. In this study, 24-h ABPM was only assessed 2 years after BP control intervention, which limits our ability to evaluate the actual effect of BP treatment on the occurrence of hypotensive episodes and to determine the relationship between new-onset hypotensive episodes and cognitive impairment. Given the 30-minute intervals between BP readings during the 24-hour BP monitoring, it is possible that some brief hypotensive episodes may not have been captured, potentially underestimating the frequency and duration of these events. Future studies assessing BP dynamics in shorter intervals would identify hypotensive episodes more accurately. Furthermore, our analysis in treated hypertensive patients also limited the generalizability of our results to the older population without hypertension (approximately 40% of general older adults).⁴⁹ In SPRINT, patients with diabetes, heart failure, institutionalized older patients, and individuals with low standing BP (who might have increased susceptibility to hypotension) were excluded, which further limits the generalizability of our findings. The issue of residual confounding, such as sleep hygiene, is another limitation. Additionally, although our analyses were based on a priori hypotheses, the results should be viewed as exploratory due to their post-hoc nature.

Perspectives

Hypotensive episodes identified on 24-h ABPM were associated with worse cognitive function and faster cognitive decline in the processing speed domain among older adults with hypertension. Intensive BP control increased hypotensive episodes compared to standard control and modified its association with cognitive decline in processing speed. Monitoring and intervening on hypotensive episodes detected from 24-h individual BP readings may help identify individuals at higher risk of cognitive impairment and potentially inform personalized BP control and the early detection and prevention of dementia in clinical practice.

Novelty and Relevance.

What is New?

• In hypertensive patients, both recurrent hypotension and longer hypotensive duration over 24 hours were associated with worse cognitive function and accelerated cognitive decline, especially in processing speed domain.

• The association of hypotensive phenotypes derived from 24-hour individual blood pressure (BP) readings with worse cognition remained even after further adjustment for average 24-hour BP level and BP variability, and it was not observed for hypotension defined by clinic BP.

• Intensive BP treatment increased the 24-h hypotensive episodes and modified its associations with cognitive decline in processing speed domain.

What is Relevant?

• Hypertensive patients who underwent intensive BP treatment and experienced hypotensive episodes had lowest cognitive scores and fastest cognitive decline compared to those underwent standard treatment and no hypotensive episodes.

Clinical/Pathophysiological Implications?

• Monitoring and intervening on hypotensive episodes detected from 24-h individual BP readings might help identify individuals at higher risk of cognitive impairment and potentially inform personalized BP control and dementia prevention in clinical practice.

• Impaired BP dynamics, in addition to conventional BP levels, are potentially important factors associated with cognitive health.

Abbreviations:

ABPM

ambulatory blood pressure monitoring

BP

blood pressure

CVD

cardiovascular disease

DBP

diastolic blood pressure

DSC

digit symbol coding

FDR

false discovery rate

GFR

glomerular filtration rate

LM I

logical memory form I

LM II

logical memory II

MoCA

Montreal cognitive assessment

SBP

systolic blood pressure

SPRINT

Systolic Blood Pressure Intervention Trial