2023 Volume 87 Issue 1 Pages 20-26
Background: There is a dearth of evidence to characterize longitudinal changes in domain-specific cognitive function related to atrial fibrillation (AF).
Methods and Results: This study enrolled 2,844 participants from the Systolic Blood Pressure Intervention Trial (SPRINT). Cognitive function was assessed at baseline and biennially during the follow-up period. Declines in global function and 4 major cognitive domains (i.e., memory, processing speed, language, and executive function) were fitted and compared between participants with and without AF using robust linear mixed-effect models. There were 252 participants with prevalent AF (mean [±SD] age 72.0±8.5 years; 30% women) and 2,592 participants without AF (mean age 67.9±8.4 years; 38% women). The annual decline in global function scores was greater among participants with than without AF (−0.016 vs. −0.012 points); however, the difference was not statistically significant (P=0.33). Processing speed declined faster in participants with prevalent AF, with a distinct difference of –0.013 points/year (95% CI −0.024~−0.001 points/year; P=0.02). For the memory, executive function, and language domains, there were no significant differences in the rate of cognitive decline between participants with and without AF.
Conclusions: In this post hoc analysis of the SPRINT trial, processing speed was the most prominent cognitive domain affected by AF, which may be beneficial for the early screening of cognitive dysfunction.
Atrial fibrillation/flutter (AF) are common arrhythmia whose prevalence are increasing with the growth of the aging population.1,2 There is some evidence suggesting a relationship between AF and cognitive impairment. For example, evidence from prospective studies has shown that AF is associated with an approximate 30–50% increase in the risk of developing dementia.3–5 AF is associated with excess cognitive decline further to age-related cognitive decline, and those with a longer duration of AF exhibit faster cognitive decline.6 The concurrence of the AF and cognitive decline results in a substantial economic burden and poor prognosis. Cognitive impairment in AF patients is associated with higher risks of cardiovascular death and unplanned readmission for any cause.7
Editorial p 27
Global cognition comprises various domain-specific cognitive functions. Global cognitive function is assessed by measuring cognitive domains and related subdomains. Characterization of domain-specific cognitive impairment could provide clues as to the breadth of neuropsychological conditions and potential interventions. Therefore, it is important to determine the pattern of domain-specific cognitive deterioration in patients with AF to prevent further cognitive impairment. Several cross-sectional studies have demonstrated relationships between AF and cognitive declines in immediate and delayed memory, attention, executive function, abstract learning, visuospatial ability, and processing speed.8–12 However, there is limited evidence from prospective studies. Longitudinal analyses of the Atherosclerosis Risk in Communities (ARIC) study and the Framingham cohort suggest AF is related to declines in executive function and processing speed.13,14 However, thus far there is no consistent conclusion regarding the pattern of impairment of cognitive domains related to AF.
The substudy of the Systolic Blood Pressure Intervention Trial (SPRINT) was prospectively designed and used 11 neurocognitive tests.15 This provides a unique opportunity for the comprehensive evaluation of the longitudinal deterioration of major cognitive domains, including memory, executive function, processing speed, and language. In this study we investigated differences in domain-specific cognition declines between patients with and without prevalent AF to explore and characterize AF-related cognitive impairment.
SPRINT data for this study are publicly available at the National Heart, Lung, and Blood Institute BioLINCC data repository and can be accessed online (https://biolincc.nhlbi.nih.gov/studies/sprint/).
Study PopulationThe design, process, and results of SPRINT have been published previously.16,17 Briefly, SPRINT was a multicenter randomized controlled trial that compared the benefits of 2 antihypertension strategies. From November 2010 to March 2013, 9,361 participants aged ≥50 years with coexisting cardiovascular risk were randomized to the intensive treatment group (target systolic blood pressure [SBP] <120 mmHg) or standard treatment group (target SBP <140 mmHg). A subset of 2,921 participants in SPRINT received additional cognitive measurements for global function and domain-specific cognitive function at baseline and at 24 and 48 months (or at the close-out visit if the tests were not performed at 48 months). Follow-up continued until June 2018, after the closeout of SPRINT.15 Participants without cognitive assessments and baseline AF status, as well as those with incident AF during the follow-up period, were excluded from this study, leaving 2,844 participants for analysis (Figure 1). The baseline characteristics of participants included and excluded in this study are presented in Supplementary Table 1.
Flowchart of study design. AF, atrial fibrillation; SPRINT, Systolic Blood Pressure Intervention Trial.
This study was conducted in accordance with the ethical guidelines for medical research on humans laid out in the Declaration of Helsinki. SPRINT was approved by an institutional review board at each participating site, and all participants provided informed consent before randomization.
Ascertainment of AFParticipants with a history of AF or evidence of AF on an electrocardiogram (ECG) within the first 7 days after randomization were defined as “prevalent AF”. All ECGs were visually inspected for image quality before automatic processing (12-SL Marquette version 2001; GE, Milwaukee, WI, USA). All ECG readings were adjudicated centrally at the Epidemiological Cardiology Research Center, Wake Forest School of Medicine (Winston-Salem, NC, USA).18
Neuropsychological MeasuresGlobal cognitive function and 4 cognitive domains (i.e., memory, processing speed, language, and executive function) were evaluated by 11 neuropsychological tests: Montreal Cognitive Assessment (MoCA), Digit Symbol Coding test, Logical Memory tests (I and II), Hopkins Verbal Learning Test, Trail Making Test (Parts A and B), digit span test, Boston Naming Test, Modified Rey-Osterrieth Complex Figure, and Category Fluency-Animals. Global function was assessed by all these tests except MoCA. Each domain was quantified by composite scores of the relevant cluster of neurocognitive tests. Details of the neuropsychological tests used to evaluate global cognitive function are provided in Supplementary Table 2.
Statistical AnalysisCognition scores had been processed in the original dataset of SPRINT.15 Trail Making Test scores had been transformed to speed (score=1/time [s]). For separate neurocognitive tests, raw scores from which the baseline median was subtracted were later divided by the baseline interquartile range, forming a standardized score. Next, the mean of standardized scores for each component test were summed, followed by further standardization to achieve similar scales as described previously.15
To analyze repeated-measures data, the robust linear mixed model was fitted to examine the longitudinal change in domain-specific cognitive function between the AF and no-AF groups.15 The interaction between time and AF was added as a fixed effect, and the participant and clinic sites were set as random effects. The bounded function was additionally tuned to attenuate the effect caused by ceiling or floor effects. In addition, adjustments were made for the following confounding factors: age, female sex, body mass index, education, smoking status, self-report history of alcohol abuse, the presence of chronic kidney diseases, the presence of cardiovascular diseases, SBP, diastolic blood pressure, cholesterol, whether aspirin was used, and whether intensive treatment was applied. We further estimated annual changes in the total brain volume (TBV) and white matter lesion (WML) volume of participants with and without AF by fitting the robust linear mixed model. Multivariable adjustments were also performed using the variables mentioned above. Because of the skewed distribution of WML volume, we applied an inverse hyperbolic sine transformation (asinh).19 For sensitivity analysis, we repeated the robust linear mixed models among participants without incident stroke during the follow-up period.
The threshold for statistical significance was set at 2-sided P<0.05. All analyses were performed using R version 4.0.5 (https://www.R-project.org/). The robust linear mixed model was implemented using the R Package robustlmm.20
There were 252 participants with and 2592 without prevalent AF (Figure 1). At baseline, participants with AF were approximately 4 years older (72.0 vs. 67.9 years), less likely to be female (30.2% vs. 37.6%), and with higher morbidity of chronic kidney disease (38.5% vs. 29.2%) and cardiovascular disease (40.1% vs. 17.7%) than those without AF. The use of aspirin (65.5% vs. 50.2%) and statins (58.4% vs. 42.2%) was higher among participants with prevalent AF than those without prevalent AF. In the AF and no-AF groups, 95.3% and 91.1% of participants, respectively, had a high school graduate or higher diploma (Table 1). There were no significant differences in baseline scores for the 4 domains (memory, processing speed, language, and executive function) between the AF and no-AF groups (Supplementary Table 3).
| No AF (n=2,592) |
AF (n=252) |
P value | |
|---|---|---|---|
| Age (years) | 67.9±8.4 | 72.0±8.5 | <0.001 |
| Female sex | 974 (37.6) | 76 (30.2) | 0.02 |
| Ethnicity | <0.001 | ||
| Black | 815 (31.4) | 47 (18.7) | |
| Hispanic | 236 (9.1) | 8 (3.2) | |
| White | 1,484 (57.3) | 194 (77.0) | |
| Other | 57 (2.2) | 3 (1.2) | |
| EducationA | 0.09 | ||
| Lower than high school | 230 (8.9) | 12 (4.8) | |
| High school graduate | 433 (16.7) | 39 (15.5) | |
| Post-high school training | 912 (35.2) | 88 (35.0) | |
| College graduate or higher | 1,017 (39.2) | 113 (44.8) | |
| BMI (kg/m2) | 29.8±5.7 | 29.8±5.5 | 0.84 |
| Smoking status | 0.001 | ||
| Never smoked | 1,155/2,589B (44.6) | 96 (38.1) | |
| Previous smoker | 1,102/2,589B (42.5) | 138 (54.8) | |
| Current smoker | 332/2,589B (12.8) | 18 (7.1) | |
| Alcohol abuseC | 102 (3.9) | 5 (2.0) | 0.17 |
| CKD | 757 (29.2) | 97 (38.5) | 0.003 |
| CVD | 458 (17.7) | 101 (40.1) | <0.001 |
| Cancer | 298 (11.5) | 38 (15.1) | 0.12 |
| Orthostatic hypotension | 549 (21.2) | 66 (26.2) | 0.08 |
| SBP (mmHg) | 138.6±15.9 | 139.1±16.5 | 0.69 |
| DBP (mmHg) | 77.7±11.6 | 75.1±12.4 | <0.001 |
| No. antihypertensive agents used | 0.007 | ||
| 0 | 247 (9.5) | 12 (4.8) | |
| 1 | 749 (28.9) | 63 (25.0) | |
| 2 | 909 (35.1) | 89 (35.3) | |
| 3 | 541 (20.9) | 62 (24.6) | |
| ≥4 | 146 (5.6) | 26 (10.3) | |
| MoCA score | 23.00 [21–26] | 24.00 [21–26] | 0.24 |
| Use of statin | 1,089 (42.2) | 146 (58.4) | <0.001 |
| Use of aspirin | 1,299 (50.2) | 165 (65.5) | <0.001 |
| eGFR (mL/min/1.73 m2) | 71.3±20.7 | 66.6±21.9 | <0.001 |
| Urine albumin : creatinine ratio (mg/g) | 9.4 [5.7–21.4] | 13.0 [6.4–35.7] | 0.002 |
| Cholesterol (mg/dL) | 190.7±40.7 | 179.5±42.3 | <0.001 |
Unless indicated otherwise, data are given as the mean±SD, median [interquartile range], or n (%). ALower than high school included those who did not go to school/grade school (1–4 years), those who went to grade school (5–8 years), or those with some high school education (9–11 years). High school graduates were those with a high school diploma or General Educational Development certificate. Post-high school training included business/vocational training, some College education (no degree obtained), and associate degrees (AD or AA). College graduates and higher included those with some College or professional education after college graduation and those with Master’s and/or doctoral degrees (PhD, MD, JD, etc.). BAmong participants without prevalent atrial fibrillation (AF), smoking status was unknown for 3. CSelf-report history of alcohol abuse. BMI, body mass index; CKD, chronic kidney disease; CVD, cardiovascular disease; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; MoCA, Montreal Cognitive Assessment; SBP, systolic blood pressure.
As shown in Figure 2, the global function of participants with AF declined at a speed of −0.016 points/year, whereas the annual cognitive decline in global function for those without AF was −0.012 points; the difference was not statistically significant (P=0.33). For all domains except memory, the domain-specific cognitive status consistently exhibited declining trends for both participants with and without AF. For AF participants, there was an upward trend in the memory domain (0.008 points/year; 95% CI −0.003~0.021 points/year). In contrast, there was a decline in the memory domain for participants without AF (−0.004 points/year; 95% CI −0.008~−0.0001 points/year). After adjustment for confounders, the difference in the rate of cognitive decline between the 2 groups was not significant (P=0.06). The decline in processing speed was faster for those with than without AF, with an annual rate difference of −0.013 points/year (95% CI −0.024~−0.001 points/year), and this difference remained significant after adjustment (P=0.02). There was a trend for a quicker decline in executive function among participants with than without AF; however, the 95% CIs overlapped, and the difference was −0.012 points/year (95% CI −0.026~0.002 points/year; P=0.09). In the language domain, declines were seen in both AF (−0.017 points/year; 95% CI −0.029~−0.005 points/year) and non-AF (−0.013 points/year; 95% CI −0.018~−0.009 points/year) participants, and the difference was not statistically significant (P=0.58; Table 2). Similar results were seen for participants without incident stroke during the follow-up period (Supplementary Table 4).
Unadjusted changes in (A) global and (B–E) domain-specific cognitive function during follow-up of participants with and without atrial fibrillation (AF). Solid lines denote the estimated mean for each group calculated by robust linear mixed effect model. Shading indicates 95% confidence intervals.
| Annual changeA (95% CI) | Unadjusted mean difference (95% CI) |
P valueB | Adjusted mean differenceC (95% CI) |
P valueD | ||
|---|---|---|---|---|---|---|
| No AF (reference group) |
AF | |||||
| Global function | −0.012 (−0.015~−0.011) |
−0.016 (−0.024~−0.008) |
−0.005 (−0.012~0.004) |
0.33 | −0.005 (−0.014~0.006) |
0.33 |
| Memory | −0.004 (−0.008~−0.0001) |
0.008 (−0.003~0.021) |
0.012 (−0.001~0.026) |
0.05 | 0.013 (−0.002~0.026) |
0.06 |
| Processing speed | −0.017 (−0.022~−0.014) |
−0.030 (−0.043~−0.019) |
−0.013 (−0.024~−0.001) |
0.02 | −0.014 (−0.025~−0.001) |
0.02 |
| Language | −0.013 (−0.018~−0.009) |
−0.017 (−0.029~−0.005) |
−0.004 (−0.015~0.009) |
0.58 | −0.004 (−0.017~0.011) |
0.52 |
| Executive function | −0.008 (−0.013~−0.005) |
−0.020 (−0.035~−0.007) |
−0.012 (−0.026~0.002) |
0.09 | −0.012 (−0.027~0.001) |
0.09 |
AAnnual changes in scores presented as points/year. BUnadjusted P value; Dadjusted P value. CAdjusted for age, female sex, chronic kidney disease, cardiovascular disease, systolic blood pressure, diastolic blood pressure, aspirin use, cholesterol, whether in intensive treatment group, smoking status, body mass index, education level, and self-report history of alcohol abuse. AF, atrial fibrillation; CI, confidence interval.
We also explored associations between AF and both TBV and WML volume. In this analysis, there were 458 participants in total, 32 of whom had AF. The median follow-up time was 3.99 years. There were no significant differences in annual changes in TBV and WML volume between participants with and without AF (Supplementary Table 5).
The AF group was then divided into 2 groups based on anticoagulation treatment. Approximately 28% (70/252) of prevalent AF participants were on anticoagulation treatment. Of these, 56 were using warfarin, and the rest were on dabigatran. There was no significant association between anticoagulation treatment and cognitive decline in terms of either global function or individual cognitive domains (Supplementary Table 6).
In this study, the rate of decline in global cognitive function was greater in participants with than without AF, but the difference did not reach statistical significance. With regard to domain-specific cognitive function, AF was associated with impaired processing speed; there was no significant difference between patients with and without AF in memory, executive function, and language. Processing speed is the most prominent cognitive domain affected by AF, and may be a potential marker in the early screening for cognitive dysfunction in the AF population.
In parallel with previous studies,6 we found a faster rate of decline in global function in AF patients. The reason why the difference between AF and non-AF participants failed to reach statistical significance in the present study could be due to differences in assessment strategies. In SPRINT, global function was assessed by the comprehensive assessment of various domains and was reflected by scores from component measures instead of a single test, such as MoCA. In addition, it is possible that less affected domains could compensate for those cognitive domains that were impaired, leading to confounding effects on AF-related cognitive impairment of global function.
The association between AF and an accelerated decline in processing speed revealed in the present study is in accordance with the results of the ARIC cohort study. In the ARIC study, incident AF was associated with a speedier decline in the Digit Symbol Substitution test, which measures executive function and processing speed without specificity.13 Along with the ARIC study, the present study provides powerful evidence for the relationship between AF and impaired processing speed cohort because the processing speed was measured by careful combination of the Digit Symbol Coding test and the Trail Making Test: B minus A for better efficacy and effectiveness. Furthermore, we adjusted for skewness, as well as floor and ceiling effects, by using a robust linear mixed effect model.15
Processing speed is the cognitive ability to mentally manage pieces of information and is generally defined as the time to perceive, understand, and respond to information.21 It is a particularly important cognitive ability. Processing speed is the most impaired domain of functioning various coding tasks that are strongly correlated with daily life. In addition, it tends to be the strongest predictor of overall cognitive performance.21,22 The present study suggests that AF possibly causes an extra decline in processing speed apart from age-related degeneration. From a clinical perspective, processing speed may be useful for screening for early cognitive deterioration among AF patients.
The results of the present study suggest a potential pattern of cognitive decline related to AF that may provide clues about the underlying mechanisms. Vascular-related mechanisms, such as cerebral infarction, cerebral hypoperfusion, and cerebral microbleeds,23,24 could explain the association of AF with the decline in processing speed. Inflammation could also play a role.24 Participants with AF have higher concentrations of the inflammatory factor interleukin-6, which is associated with deteriorated processing speed.25–27
In a longitudinal analysis of 2,682 participants from the Framingham cohort, AF was associated with a faster deterioration in executive function.14 However, in the present study, the difference in the rate of decline in executive function between participants with and without AF was of marginal statistical significance. Differences in assessment strategies could explain the apparent discrepancies in results. Executive function has usually been confounded with other frontal lobe functions, such as working memory and processing speed, because the measurements of these cognitive domains all require “frontal lobe” tasks.22 The Trail Making Test Part B requires robust executive function, and was used in both the Framingham study and SPRINT.14 In addition, SPRINT used the digit span test to measure executive function, and this test also reflects working memory.22 Thus, the estimated rate of decline in executive function may be confounded by other frontal lobe functions. Furthermore, the limited follow-up period and attrition during follow-up may have resulted in inadequate statistical power. The differences between participants with and without AF for memory and language declines were not statistically significant, which is in line with results from previous prospective studies.13,14
Anticoagulation treatment is associated with a lower risk of dementia, whether vascular dementia or Alzheimer’s disease.3,28 However, the Birmingham Atrial Fibrillation Treatment of the Aged (BAFTA) randomized clinical trial showed that, over a 33-month follow-up, there was a non-significant difference of 0.56 points in the Mini-Mental State Examination score between people assigned to warfarin or aspirin.29 Anticoagulation was not associated with improved cognitive performance in the present study. This result has limited generalizability because the proportion of patients with AF on oral anticoagulants was smaller in the present study compared with the general trend of anticoagulant use.30 In addition to the small sample size, there was a lack of monitoring of time in the therapeutic range (TTR) in SPRINT. Because 80% of the anticoagulation group was on warfarin, the difference in cognitive decline may be weakened by inadequate anticoagulation with unknown TTR. Several randomized clinical trials are underway to examine the effects of anticoagulation treatment on cognitive decline in patients with AF.
The present study has several major strengths, including the elaborate planning of the SPRINT, a relatively large sample size, and long follow-up period, as well as the comprehensive battery of efficient neuropsychological tests, which ensure quality assessment of cognitive status. However, the study has several limitations. First, because AF was ascertained based on previous history or biennial ECG examinations rather than long-term ECG monitoring, some cases of paroxysmal AF may not have been detected and may have been misclassified into the no-AF group. Second, SPRINT excluded people with previous stroke, diabetes, advanced kidney disease, and symptomatic heart failure; thus, the generalizability of the findings may be limited. Third, because SPRINT was terminated early, loss to follow-up during the extended follow-up period may have affected the difference in cognitive decline. Fourth, the type and burden of AF may be related to cognitive decline.31 Unfortunately, SPRINT did not collect information about the type or burden of AF. Thus, we could not further investigate the relationship between cognitive decline and AF type or burden based on current data.
In this post hoc analysis of SPRINT, processing speed was the most prominent cognitive domain affected by AF, which may be useful for early screening of cognition dysfunction.
The authors thank the SPRINT Investigators for making the study data available for public use through the National Heart, Lung, and Blood Institute BioLINCC Biologic Specimen and Data Repository.
This work was supported by the National Key Research and Development Program of China (Grant no. 2020YFC2004803), the National Natural Science Foundation of China (Grant no. 82100326 and 81900449), the Beijing Municipal Science and Technology Commission (Grant no. D171100006817001), and the Beijing Municipal Education Commission (Grant no. KM202210025012).
C.M. has received honoraria for presentations from AstraZeneca, Bayer Healthcare, Boehringer Ingelheim, Bristol-Myers Squibb, Johnson & Johnson, and Pfizer. J.D. has received honoraria for presentations from Johnson & Johnson. C.S.A. reports receiving grants from the National Health and Medical Research Council (NHMRC) of Australia and Takeda, paid to his institution, and advisory board feeds from Amgen and Boehringer Ingelheim. The remaining authors have no conflicts of interest to declare.
The present study was approved by the Wake Forest University Health Sciences Institutional Review Board (Reference no. IRB00014304).
SPRINT data for this study are publicly available at the National Heart, Lung, and Blood Institute BioLINCC data repository and can be accessed online (https://biolincc.nhlbi.nih.gov/studies/sprint/).
Please find supplementary file(s);
https://doi.org/10.1253/circj.CJ-22-0224
