The association of diabetes mellitus (DM), an established risk factor for dementia in the general population, with incident dementia in patients with atrial fibrillation (AF) has not been explored. We performed a cohort study where we identified subjects with incident AF in the Atherosclerosis Risk in Communities cohort (1987 to 2017) and determined their DM status, fasting blood glucose before AF diagnosis and hemoglobin A1c levels using information from the closest previous study visit. Incident dementia was expert adjudicated using information from cognitive assessments, informant interviews and hospitalization surveillance. We calculated hazard ratios (HRs) and 95% confidence intervals (CIs) of incident dementia for each level of exposure using Cox models and adjusting for potential confounders. We analyzed 3,020 patients with AF in the Atherosclerosis Risk in Communities cohort (808 with DM) and 530 had incident dementia after a mean follow-up of 5.3 years after AF diagnosis. After multivariable adjustment, patients with AF with prevalent DM had higher rates of dementia than those without DM, HR 1.45 (95% CI 1.16 to 1.80). A value of hemoglobin A1c ≥6.5% was associated with a HR 1.29 (95% CI 0.97 to 1.71) of dementia. However, fasting blood glucose was not associated with rates of dementia independent of DM status. In conclusion, DM was associated with higher rates of dementia in patients with AF. DM prevention and control could be a promising avenue for reducing risk of dementia in AF.
Dementia is characterized by the deterioration of memory, thinking, behavior, and the ability to perform day to day activities. It has been estimated that 6.2 million Americans >65 years of age have Alzheimer’s disease, the most common form of dementia, and it has been projected that these numbers may increase to 13.8 million by 2060. Diabetes mellitus (DM) is a major lifestyle-associated chronic disease which is approaching enormous proportions globally. The International Diabetes Federation has estimated that the prevalence of DM is 9.3% globally, with the number of patients with DM worldwide likely to increase from 463 million in 2019 to 700 million by 2045. Atrial fibrillation (AF), a common arrhythmia, is associated with increased risk of dementia, even in the absence of associated stroke. Although studies have been conducted demonstrating the association between midlife cardiovascular risk factors like DM with the development of dementia later in life, evidence of the role of DM as a risk factor for dementia in patients with AF is lacking. To address these knowledge gaps, we evaluated the association of DM, fasting blood glucose, and hemoglobin A1c (HbA1c), a marker of glycemic control, with the incidence of dementia among patients newly diagnosed with AF in a community-based cohort study.
Methods
The study population for this analysis was selected from the Atherosclerosis Risk in Communities Study (ARIC) cohort. ARIC is a prospective cohort study that recruited 15,792 men and women aged 45 to 64 years at baseline in 4 US communities (Forsyth County, North Carolina; Jackson, Mississippi; Minneapolis suburbs, Minnesota; Washington County, Maryland). The study had a total of 6 visits in addition to baseline (1987 to 1989): 1990 to 1992, 1993 to 1995, 1996 to 1998, 2011 to 2013, 2016 to 2017, and 2017 to 2019. Details about study design and methods have been published elsewhere. The ARIC study has been approved by institutional review boards of all participating institutions. Participants provided written informed consent at baseline and at each follow-up study visit.
We restricted our analyses to participants who developed incident AF during follow-up through 2017 or the latest available year and without dementia at the time of AF diagnosis. AF was ascertained in this cohort through 3 sources: study electrocardiogram (ECGs), hospital discharge codes, and death certificates. ECGs were performed during the study examinations using MAC PC Personal Cardiographs (Marquette Electronics, Milwaukee, Wisconsin) where a standard supine 12-lead ECG at rest was performed after 12-hour fast followed by a light snack and at least 1 hour after smoking tobacco or ingestion of caffeine. These ECGs were processed by the EPICARE center (Wake Forest University, Winston-Salem, North Carolina). Visual inspection of the ECGs was performed to assess the quality and look for technical errors. In addition, trained abstractors obtained and recorded all hospital discharge diagnoses using International Classification of Diseases, Ninth Revision, Clinical Modification or International Classification of Diseases, Tenth Revision, Clinical Modification codes. AF was defined as International Classification of Diseases, Ninth Revision, Clinical Modification codes 427.31 or 427.32 and, starting in October 2015, International Classification of Diseases, Tenth Revision, Clinical Modification codes I48x, not occurring in the context of open-heart surgery. A validation study showed a positive predictive value of 89% with a sensitivity of 84% and a specificity of 98% for this method of AF ascertainment. Our analysis excluded Asian, Native American, and Black participants from Minneapolis and Washington County because of very small numbers. Ultimately, the baseline population for our study consisted of 3,020 participants with AF .
The primary exposure of interest is prevalent DM (yes/no) at the time of AF diagnosis, using the variables from the visit before, or at the same time as, AF diagnosis. DM was defined in all visits as fasting blood glucose levels ≥126 mg/dl, nonfasting blood glucose levels ≥200 mg/dl, self-reported physician diagnosis of DM or self-reported use of antidiabetic medications.
For secondary analyses, we considered fasting blood glucose concentrations measured at all study visits and HbA1c measured at visits 2 (1990 to 1992) and 5 (2011 to 2013) (not available in other visits) as additional exposures. Serum glucose in the ARIC cohort was measured using the hexokinase method. HbA1c was measured in whole blood samples maintained at −80°C using high-performance liquid chromatography using instruments that were standardized to the Diabetes Control and Complications Trial assay. We used the most recent values of fasting blood glucose and HbA1c before AF diagnosis for the analysis.
The primary outcome of interest was incident dementia defined according to standard ARIC procedures after expert adjudication. There were several approaches used to ascertain dementia. First, ARIC participants taking part in visits 5 and 6 (2011 to 2013, 2016 to 217) underwent a detailed assessment of neurocognitive function. A subset of these participants was selected to receive a neurologic examination and a magnetic resonance imaging of the brain. Second, a validated phone-based cognitive assessment, the modified version of the Telephone Interview for Cognitive Status (TICSm), was administered to participants who were alive at the time of visit 5 but unable or unwilling to participate in an in-person examination. When the participants were deceased or unable to complete the TICSm by themselves, informants provided additional information using the Clinical Dementia Rating and Functional Activities Questionnaire. Finally, in the full cohort, hospitalization codes were used to identify incident dementia occurring from visit 1 to end of visit 6. For our analysis, we considered cases of dementia identified through any of these sources. The date of dementia diagnosis was defined depending on the source of dementia diagnosis. In participants identified through in-person cognitive evaluations, the date of assessment was used as the date of dementia diagnosis, with an exception of using the hospitalization dates in those with a previous dementia hospitalization. The earliest date from TICSm, informant interview, or hospitalization discharge, as applicable, was used for study participants with dementia diagnosis from other sources. To account for the lag in determining dementia identified by interviews, deaths, and hospitalization, 6 months were subtracted from the dates to identify the date of dementia onset. In study participants who were never diagnosed with dementia, the earliest of the date of visit 6 examination, date of loss to follow-up, December 31, 2017, or the date of death was used to calculate the follow-up time.
Covariates used in our analysis included participant demographics, co-morbidities, and use of certain medications ascertained at study visits or the time of AF diagnosis. The demographic information included self-reported age (at time of AF diagnosis), gender, race (White, Black), study site (Forsyth County, Jackson, Minneapolis suburbs, Washington County), education level (basic, intermediate, high), and smoking and alcohol drinker status (current, former, never, missing). Because visit center and the race of participants were correlated, we categorized participants jointly by race and center (White participants from Forsyth County, White participants from Minneapolis, White participants from Washington county, Black participants from Forsyth County, and Black participants from Jackson).
Heart failure, stroke and myocardial infarction incidence were defined according to criteria described elsewhere. Total cholesterol was measured at all visits using standard procedures. The use of blood pressure-lowering medications, anticoagulants, statins, and aspirin were ascertained by self-report at all visits. Systolic and diastolic blood pressures were measured 3 times and the mean of the second and third measurements were used for analysis. However, in visit 4, blood pressure was only measured twice and the mean of these 2 values were used. Genotyping for APOE polymorphisms in ARIC cohort were performed using the TaqMan assay, where variants on the codons 130 and 176 were assayed separately. The data obtained from these codons were then combined to generate the 6 APOE genotypes: 22, 23, 33 (used as reference), 24, 34, and 44.
SAS 9.4 (SAS Institute Inc., Cary, North Carolina) and Stata 16.1 (Stata Corp. LLC, College Station, Texas) were used for statistical analysis. DM status in study participants was defined with respect to the date of diagnosis of AF. Means and SDs were calculated for continuous variables, and frequencies and percentages for categoric variables by DM status. Time-to-event was calculated as the time from AF diagnosis to time of dementia diagnosis or censoring (death, lost to follow-up, visit 6 date, or December 31, 2017). We calculated crude incidence rates of dementia in participants with and without DM, and the corresponding incidence rate ratio (participants without DM as the reference). Cumulative incidence curves were generated for the association between DM and dementia before and after accounting for the competing risk of death.
We assessed the association between DM diagnosis and incidence of dementia among ARIC participants with AF using Cox regression to calculate hazard ratios (HRs) and 95% confidence intervals (CIs). In model 1, we adjusted for demographics (age, gender, and race/center). In model 2, we additionally adjusted for education, smoking, drinking, anticoagulant use, aspirin use, antihypertensive use, statin use, myocardial infarction, stroke, prevalent heart failure, body mass index, total cholesterol, systolic blood pressure, diastolic blood pressure, and APOE genotype. Effect measure modification by age (<74, ≥74 years), gender (male, female), and race was assessed after adjusting for model 2 covariates. We repeated the analysis using a Fine-Gray subdistribution hazard model considering death as a competing risk and calculating subdistribution HR (SHR) and their 95% CIs.
Secondary analysis was performed using glucose tertile cut points as the exposure of interest. These cut points were created separately in participants with and without DM, and the lowest tertile among participants without DM were used as the reference category. We assessed the association between DM-specific glucose tertiles and diagnosis of dementia using Cox regression. An additional secondary analysis was performed using HbA1c as the exposure of interest, using a HbA1c value of 6.5% as the cut point. As with the primary analysis, we performed an initial analysis adjusting for demographic variables (model 1) followed by a model adjusting for multiple covariates (model 2). DM status was not included as a covariate in the model. We also explored effect measure modification by age, gender, and race/center as described previously.
Results
Of the 15,792 participants of the ARIC cohort, we included 3,020 eligible subjects in the final analysis. At the time of AF diagnosis, 808 participants (27%) had a diagnosis of DM. Use of antihypertensives, heart failure prevalence, and mean body mass index were higher in participants with DM compared with those without DM. The racial and gender distribution was similar between the 2 groups ( Table 1 ).
| Variable, N= 3020 | Diabetes mellitus | |
|---|---|---|
| Yes (n = 808) | No (n = 2212) | |
| Age (years) | 73 ± 8 | 74 ± 8 |
| Male | 407 (50%) | 1151 (52%) |
| Female | 401 (50%) | 1061 (48%) |
| Black | 212 (27%) | 316 (14%) |
| White | 596 (74%) | 1896 (86%) |
| Education level | ||
| None of the mentioned categories | 4 (0.5%) | 1 (0.1%) |
| Basic education or 0 years education | 253 (31%) | 504 (23%) |
| Intermediate education | 333 (41%) | 932 (42%) |
| Advanced education | 218 (27%) | 775 (35%) |
| Smoker | ||
| Current | 128 (16%) | 431 (20%) |
| Former | 388 (48%) | 1006 (46%) |
| Never | 273 (34%) | 742 (34%) |
| Unknown | 19 (2.4%) | 33 (1.5%) |
| Alcohol drinking status | ||
| Current | 302 (37%) | 1164 (53%) |
| Former | 302 (37%) | 636 (29%) |
| Never | 204 (25%) | 411 (19%) |
| Unknown | 0 (0%) | 1 (0.1%) |
| Aspirin use | 522 (65%) | 1390 (63%) |
| Antihypertensive use | 645 (80%) | 1245 (56%) |
| Anticoagulant use | 49 (6.1%) | 113 (5.1%) |
| Statin use | 237 (29%) | 394 (18%) |
| Total cholesterol (mmol/L) | 5.0 ± 1.2 | 5.1 ± 1.0 |
| Body mass index (kg/m 2 ) | 32.4 ± 6.3 | 28.7 ± 5.9 |
| Systolic blood pressure (mmHg) | 134 ± 23 | 131 ± 21 |
| Diastolic blood pressure (mmHg) | 69 ± 12 | 71 ± 12 |
| APOE ε4 allele | 211 (26%) | 605 (27%) |
| History of myocardial infarction | 176 (22%) | 265 (12%) |
| Prevalent CHD | 205 (25%) | 329 (15%) |
| Prevalent stroke | 42 (5.2%) | 73 (3.3%) |
| Prevalent heart failure | 327 (41%) | 613 (28%) |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
