Abstract
Aims:
To determine the association between diabetes status, glycemia, and cognitive function among a national U.S. sample of older adults in the 2011–2014 National Health and Nutrition Examinations Surveys.
Methods:
Among 1,552 adults age ≥60 years, linear and multivariable logistic regressions were used to determine the association between diabetes status (diabetes, prediabetes, normoglycemia) and cognitive function [Consortium to Establish a Registry for Alzheimer’s Disease-Word Learning (CERAD W-L), Animal Fluency test, Digit Symbol Substitution Test (DSST)].
Results:
Overall, diabetes was associated with mild cognitive dysfunction. In age-adjusted models, adults with diabetes had significantly poorer performance on the delayed and total word recalls (CERAD W-L) compared to those with normoglycemia (5.8 vs. 6.8 words; p=0.002 and 24.5 vs. 27.6 words;p<0.001, respectively); the association was non-significant after adjusting for cardiovascular disease. Among all adults, cognitive function scores decreased with increasing HbA1c for all assessments, but remained significant in the fully adjusted model for the Animal Fluency and DSST [beta coefficient=−0.44;−1.11, p<0.05, respectively]. As measured by the DSST, the proportion with cognitive impairment was significantly higher for older adults with HbA1c≥8.0% (≥64mmol/mol) vs. HbA1c<7.0% (<53mmol/mol) (14.6% vs. 6.3%, p=0.04).
Conclusions:
Dysglycemia, as measured by HbA1c, was associated with poorer executive function and processing speed.
Keywords: Cognitive function, Glycemia, Hemoglobin HbA1c, NHANES, Older adults
1. Introduction
The prevalence of diagnosed and undiagnosed diabetes among adults in the U.S. has significantly increased over the past several decades from 9.8% in 1988–1994 to 12.4% in 2011–2012 and the burden is substantially higher among older adults (24.7% age ≥65 years) (1). In addition, older adults have a higher risk for cognitive impairment, including Alzheimer’s Disease, which is estimated at 11% of adults age ≥65 years (2). In diabetes, cognitive impairment may largely be due to vascular complications associated with hyperglycemia and diabetes, including cardiovascular disease (CVD) and stroke (3–6). Severe and repetitive hypoglycemic events have also been associated with initial cognitive impairment and accelerated decline among those with type 2 diabetes (7); however, no effect was found prospectively among those with type 1 diabetes(8). Thus, the relationship between glycemic status and cognition is complex(9).
Several epidemiological review studies have shown that cognitive dysfunction is a complication of hyperglycemia in diabetes (10, 11) and even prediabetes (12). A Framingham Heart Study showed that hyperglycemia was associated with impaired attention and memory among those with undiagnosed diabetes and prediabetes (13). In the prospective Atherosclerosis Risk in Communities (ARIC) Study, diabetes, higher HbA1c, and prediabetes in midlife was associated with a significantly greater cognitive decline compared to those without diabetes (14).
However, a specific and detailed assessment of the association of diabetes status, glycemic control, and cognitive function, while also evaluating other factors that may explain this association, has not been carried out in a nationally representative older U.S. sample. While a previous 1999–2002 NHANES study did assess fasting plasma glucose and the association with cognition, the focus was on metabolic syndrome and only one cognitive function assessment was implemented (15). Another 1999–2002 NHANES study focused on insulin resistance and cognition among older adults but, again, implemented only one cognitive assessment (16). Finally, a 1988–1994 NHANES study assessed the presence of diabetes and/or hypertension on cognition, but this study was among younger participants age 30–59 years (17). The more recent 2011–2014 NHANES includes additional measures of cognition as compared to previous NHANES cycles. Thus, to fill this void in the literature, we determined the association between diabetes status, glycemia, and cognitive function among older adults while also accounting for other factors that may explain these associations.
2. Subjects, Materials, and Methods
2.1. Study Participants
The National Health and Nutrition Examination Survey (NHANES) is a stratified multistage probability cluster survey conducted in the non-institutionalized civilian U.S. population (18) and includes an in-home interview and a physical examination at a mobile examination center (MEC) (19, 20). Written informed consent approved by the National Center for Health Statistics Institutional Review Board. Our analyses included adults age ≥60 years who completed cognitive functioning tests and had information on diabetes status in the 2011–2014 survey cycles (N=1,552). Participants self-reported age at interview, sex, race/ethnicity, education, and household income.
2.2. Diabetes Status
Diabetes status groups were defined as follows: diabetes--self-report of a diagnosis by a physician or health care professional, HbA1c ≥6.5% (≥48 mmol/mol), or fasting plasma glucose (FPG) ≥126 mg/dL (fasting 8– <24 hours) (n=208); prediabetes--HbA1c 5.7%–6.4% (39–46 mmol/mol) or FPG 100–125 mg/dL (n=345); normoglycemia--HbA1c <5.7% (<39 mmol/mol) and FPG <100 mg/dL (n=999). A phlebotomist obtained a blood sample during the MEC visit using a standardized protocol (21). HOMA-IR, a measure of insulin resistance, was calculated using the following formula: fasting serum insulin ( lU/mL) * fasting plasma glucose (mmol/L) / 22.5.
2.3. Cognitive Function
The Consortium to Establish a Registry for Alzheimer’s Disease (CERAD W-L) assesses immediate and delayed learning ability for new verbal information (22, 23). The CERAD W-L consists of three consecutive learning trials and a delayed recall. For the three learning trials, participants were instructed to read aloud 10 unrelated words. Immediately following the presentation of the words, participants recalled as many words as possible. The delayed recall occurred approximately 10 minutes after the start of the word learning trials. The maximum score on each trial is 10; the maximum score for the total word list is 40 (sum of the three trials plus the delayed recall).
The Animal Fluency Test examines verbal category fluency, a component of executive function, as well as other functions such as semantic memory and processing speed (24). Participants were asked to name as many animals as possible in one minute with a point given for each named animal.
The Digit Symbol Substitution Test (DSST) is a global measure of brain health, relies on processing speed, visual scanning, sustained attention, and short-term memory (25). The test is conducted using a paper form with a key at the top containing 9 numbers paired with distinct symbols. Participants had two minutes to copy the corresponding symbols in the 133 boxes that adjoin the numbers.
2.4. Health Behaviors and Status
Participants self-reported smoking status (never, former, current), alcohol use (≥12 vs. <12 drinks in past year), and physical inactivity [<500 MET (metabolic equivalent) minutes per week of activity]. Height and weight were measured by trained technicians to determine body mass index (BMI, kg/m2). Three blood pressure (BP) readings were taken and averaged after participants were seated for 5 minutes. Hypertension was defined as BP ≥140/90mmHg or self-reported use of antihypertensive medication. Hyperlipidemia was defined as total cholesterol ≥200 mg/dL or self-reported use of cholesterol lowering medication. History of CVD was self-reported as a previous diagnosis of heart failure, coronary heart disease, angina, or heart attack; history of stroke was also self-reported. The Patient Health Questionnaire (PHQ9) was used to determine depression (PHQ9 score ≥10) (26).
2.5. Statistical Analysis
The Cohen’s d statistic was used to determine the unadjusted standardized effect sizes between those with diabetes or prediabetes and those with normoglycemia for each cognitive test (−0.2=small effect, −0.5=moderate effect, −0.8=large effect) (27).
Mean (95% confidence interval) cognitive test scores were determined by diabetes status and adjusted for age, race/ethnicity, sex, education, income, smoking status, alcohol consumption, exercise, BMI, hypertension, hyperlipidemia, history of CVD, depression, and history of stroke using linear regression. Each covariate was added individually to the previous model. For succinctness, the regression results are shown with the covariates grouped by the attributes they measure. The continuous association between HbA1c and cognitive test scores was assessed using linear regression (beta coefficients, 95% confidence intervals) and adjusted for the covariates mentioned above; a similar analysis was completed for log-transformed HOMA-IR in place of HbA1c. Finally, the proportion with cognitive impairment by diabetes status and HbA1c level was determined as ≤1.5 SD of the mean cognitive test score for the NHANES population using multivariable logistic regression (28). All analyses accounted for the cluster design and used sample weights that corrected for non-response, yielding representative estimates of the non-institutionalized U.S. population (SUDAAN User’s Manual, Release 9.2, 2008; Research Triangle Institute).
3. Results
3.1. Characteristics of Study Population by Diabetes Status
The age distribution was similar for those with diabetes, prediabetes, and normoglycemia (Table 1). There was a greater percentage of non-Hispanic black and Hispanic participants, and a lower percentage of women, in those with diabetes compared to those with normoglycemia. There was a higher prevalence of not exercising, obesity, hypertension, and CVD, and stroke among those with diabetes and prediabetes compared to those with normoglycemia.
Table 1.
Participant characteristics [percentage (standard error)] by diabetes status among adults age ≥60 years, NHANES 2011–2014
| Diabetesa (N=208) | Prediabetes (N=345) | Normoglycemia (N=999) | |
|---|---|---|---|
| Age | |||
| 60–69 years | 50.7 (2.5) | 51.8 (3.6) | 61.6 (6.0) |
| ≥70 years | 49.3 (2.5) | 48.2 (3.6) | 38.4 (6.0) |
| Race/ethnicity | |||
| Non-Hispanic white | 65.8 (3.0)* | 76.5 (3.1)* | 86.3 (3.8) |
| Non-Hispanic black | 14.5 (2.3)* | 8.9 (1.3) | 6.1 (2.2) |
| Non-Hispanic Asian | 5.1 (1.0) | 4.9 (1.0)* | 2.5 (1.2) |
| Mexican-American | 7.0 (1.8)* | 4.0 (1.3) | 3.1 (1.4) |
| Other Hispanic | 4.8 (0.9)* | 3.7 (1.1) | 1.8 (0.6) |
| Female | 49.5 (2.7)* | 54.4 (3.1) | 58.9 (2.8) |
| Education | |||
| <high school | 27.4 (2.5)* | 19.1 (3.0) | 13.4 (3.1) |
| High school | 24.9 (2.1)* | 25.3 (2.7) | 15.9 (3.6) |
| >high school | 47.7 (2.8)* | 55.6 (3.5)* | 70.7 (5.3) |
| Household Income <$20,000 | 23.1 (2.1)* | 16.1 (2.4) | 12.1 (3.8) |
| Smoking status | |||
| Never | 46.3 (2.2) | 48.6 (4.2) | 50.5 (4.1) |
| Former | 43.6 (1.8) | 40.8 (4.6) | 38.2 (3.1) |
| Current | 10.1 (1.4) | 10.6 (1.6) | 11.3 (2.9) |
| Consume alcohol | 66.7 (2.2) | 70.8 (4.0) | 70.4 (3.8) |
| Does not exercise | 59.8 (2.1)* | 52.0 (2.7)* | 41.4 (2.3) |
| Body mass index | |||
| <25 kg/m2 | 15.1 (1.6)* | 26.1 (3.1)* | 42.8 (3.3) |
| 25–29.9 kg/m2 | 28.0 (1.9) | 39.0 (2.5) | 36.9 (4.7) |
| ≥30 kg/m2 | 57.0 (2.2)* | 34.9 (3.5)* | 20.3 (3.0) |
| Hyperlipidemia | 50.5 (2.6) | 41.1 (3.8) | 42.7 (3.5) |
| Hypertension | 79.1 (2.2)* | 68.9 (4.0)* | 49.2 (3.1) |
| History of CVD | 36.9 (2.4)* | 25.6 (2.9)* | 12.6 (3.0) |
| Depression | 12.4 (1.8) | 10.3 (1.9) | 7.7 (3.0) |
| History of Stroke | 13.3 (1.4)* | 7.6 (0.9) | 5.1 (1.9) |
p<0.05 compared to those with normoglycemia using two-tailed large sample z-tests
Includes both diagnosed and undiagnosed diabetes
3.2. Standardized Effect Sizes by Diabetes Status
Overall, adults with diabetes had significantly worse cognitive performance compared to those with normoglycemia (Figure 1). There was a moderate effect size for the total word recall score and DSST (Cohen’s d = −0.51 for both) and smaller effect sizes for the delayed word recall (Cohen’s d = −0.44) and the Animal Fluency test (Cohen’s d= −0.33) when comparing those with diabetes to those with normoglycemia.
Figure 1. Standardized effect sizes (Cohen’s d and 95% confidence interval) by cognitive assessment score compared to those with normal glucose.
Standardized effect sizes (Cohen’s d and 95% confidence interval) by cognitive assessment tests in those with diabetes or prediabetes compared to those with normoglycemia. Light gray shading indicates a small effect size, medium gray indicates a moderate effect size, and dark gray indicates a large effect size.
3.3. Cognitive Function Scores by Diabetes Status
Adults with diabetes recalled significantly fewer words in the delayed word recall compared to those with normoglycemia in the unadjusted, age-adjusted, and many partially adjusted models, but the association was non-significant in the fully adjusted model (6.0 vs. 6.6 words; p=0.064) (Table 2). A similar relationship was found with the total word recall test. Those with diabetes performed significantly worse on the DSST compared to those with normoglycemia in the unadjusted model and after further adjustment for age, race/ethnicity, and sex (47.1 vs. 52.3, p=0.024), but there was no significant difference after adjusting for education. The percent with cognitive impairment, for any test, was similar by diabetes status in unadjusted and adjusted models (data not shown).
Table 2.
Mean (95% confidence interval) cognitive assessment score by diabetes status among adults age ≥60 years, NHANES 2011–2014
| Diabetesa (N=200) | Prediabetes (N=331) | Normoglycemia (N=864) | p-value (Diabetes vs. Normoglycemia) | p-value (Prediabetes vs. Normoglycemia) | |
|---|---|---|---|---|---|
| Delayed Word Recall (CERAD W-L) | |||||
| Unadjusted | 5.8 (5.5–6.0) * | 6.5 (6.0–7.0) | 6.8 (6.3–7.2) | <0.001 | 0.478 |
| Age-adjusted | 5.8 (5.5–6.0) * | 6.6 (6.1–7.1) | 6.7 (6.3–7.1) | 0.002 | 0.687 |
| Model 1 | 5.8 (5.5–6.1) * | 6.5 (6.1–7.0) | 6.6 (6.2–7.0) | 0.010 | 0.818 |
| Model 2 | 5.9 (5.6–6.2) * | 6.5 (6.1–7.0) | 6.6 (6.2–7.0) | 0.015 | 0.852 |
| Model 3 | 5.9 (5.6–6.2) * | 6.6 (6.1–7.1) | 6.6 (6.2–7.0) | 0.028 | 0.986 |
| Model 4 | 6.0 (5.7–6.3) * | 6.6 (6.1–7.0) | 6.6 (6.2–6.9) | 0.018 | 0.989 |
| Model 5 | 5.9 (5.7–6.2) * | 6.6 (6.1–7.1) | 6.6 (6.3–7.0) | 0.006 | 0.899 |
| Model 6 | 6.0 (5.7–6.3) | 6.6 (6.1–7.1) | 6.6 (6.2–7.0) | 0.059 | 0.929 |
| Model 7 | 6.0 (5.7–6.3) | 6.6 (6.1–7.1) | 6.6 (6.2–7.0) | 0.064 | 0.934 |
| Total Word Recall (CERAD W-L) | |||||
| Unadjusted | 24.5 (23.6–25.3) * | 26.7 (25.5–28.0) | 27.8 (26.6–29.1) | <0.001 | 0.240 |
| Age-adjusted | 24.5 (23.7–25.3) * | 26.9 (25.6–28.2) | 27.6 (26.3–28.9) | <0.001 | 0.416 |
| Model 1 | 24.7 (23.9–25.6) * | 26.8 (25.9–28.1) | 27.4 (26.0–28.6) | 0.003 | 0.541 |
| Model 2 | 24.8 (24.0–25.7) * | 26.8 (25.6–28.0) | 27.3 (26.0–28.6) | 0.006 | 0.582 |
| Model 3 | 25.1 (24.3–25.8) * | 26.9 (25.7–28.1) | 27.1 (25.9–28.3) | 0.012 | 0.813 |
| Model 4 | 25.3 (24.5–26.0) * | 26.9 (25.8–28.0) | 27.1 (26.–28.2) | 0.011 | 0.769 |
| Model 5 | 25.2 (24.5–25.9) * | 27.0 (25.8–28.1) | 27.3 (26.1–28.5) | 0.005 | 0.665 |
| Model 6 | 25.5 (24.7–26.3) | 26.9 (25.7–28.1) | 27.1 (25.8–28.3) | 0.052 | 0.847 |
| Model 7 | 25.5 (24.7–26.3) | 26.9 (25.7–28.1) | 27.1 (25.8–28.3) | 0.054 | 0.845 |
| Animal Fluency | |||||
| Unadjusted | 16.8 (16.2–17.4) * | 17.7 (16.8–18.6) | 18.7 (17.2–20.2) | 0.033 | 0.233 |
| Age-adjusted | 16.9 (16.3–17.4) | 17.8 (17.0–18.6) | 18.5 (17.1–19.9) | 0.056 | 0.384 |
| Model 1 | 17.2 (16.6–17.7) | 17.8 (17.0–18.5) | 18.1 (16.8–19.4) | 0.211 | 0.595 |
| Model 2 | 17.1 (16.6–17.7) | 17.7 (17.0–18.5) | 18.2 (16.9–19.5) | 0.189 | 0.572 |
| Model 3 | 17.4 (16.9–17.9) | 17.8 (17.1–18.6) | 18.0 (16.9–19.0) | 0.443 | 0.857 |
| Model 4 | 17.6 (17.1–18.1) | 17.8 (17.1–18.6) | 18.0 (17.1–18.9) | 0.553 | 0.787 |
| Model 5 | 17.6 (17.1–18.1) | 17.9 (17.1–18.6) | 18.2 (17.2–19.1) | 0.309 | 0.609 |
| Model 6 | 17.5 (17.0–18.0) | 17.9 (17.2–18.6) | 18.2 (17.3–19.1) | 0.225 | 0.563 |
| Model 7 | 17.5 (17.0–18.0) | 17.9 (17.2–18.6) | 18.2 (17.3–19.1) | 0.181 | 0.556 |
| Digit Symbol Substitution Test (DSST) | |||||
| Unadjusted | 45.6 (43.5–47.8) * | 50.7 (48.1–53.3) | 54.4 (50.4–58.5) | <0.001 | 0.147 |
| Age-adjusted | 45.8 (43.9–47.8) * | 51.1 (48.3–53.9) | 53.7 (49.7–57.7) | 0.002 | 0.314 |
| Model 1 | 46.9 (45.2–48.7) * | 50.9 (48.4–53.5) | 52.5 (48.7–56.2) | 0.019 | 0.510 |
| Model 2 | 47.1 (45.5–48.8) * | 50.9 (48.4–53.4) | 52.3 (48.5–56.1) | 0.024 | 0.538 |
| Model 3 | 48.3 (46.7–49.8) | 51.2 (48.8–53.7) | 51.3 (47.9–54.7) | 0.143 | 0.970 |
| Model 4 | 48.9 (47.3–50.4) | 51.4 (49.4–53.5) | 51.2 (47.9–54.4) | 0.230 | 0.904 |
| Model 5 | 48.6 (46.9–50.3) | 51.5 (49.6–53.5) | 51.8 (48.8–54.8) | 0.080 | 0.874 |
| Model 6 | 48.8 (47.0–50.6) | 51.5 (49.5–53.4) | 51.2 (48.3–54.0) | 0.209 | 0.856 |
| Model 7 | 48.8 (47.0–50.6) | 51.4 (49.5–53.5) | 51.2 (48.4–54.1) | 0.183 | 0.866 |
p<0.05 compared to those with normoglycemia
Includes both diagnosed and undiagnosed diabetes
Score ranges: Delayed word recall: 0–10, total word recall: 0–40, Animal Fluency: 0-no limit, Digit Symbol Substitution Test: 0–133
Covariates were added individually, but are shown grouped by attribute for succinctness
Covariates were added individually, but are shown grouped by attribute for succinctness
Model 1: Adjusted for age, race/ethnicity
Model 2: Model 1 + sex
Model 3: Model 2 + education, income
Model 4: Model 3 + smoking status, alcohol consumption, exercise
Model 5: Model 4 + body mass index
Model 6: Model 5 + hypertension, hyperlipidemia, history of CVD
Model 7: Model 6 + depression, history of stroke
3.4. Cognitive Function Scores by HbA1c and HOMA-IR
In the total population, cognitive function scores decreased with increasing HbA1c for all assessments (Table 3). For the DSST, every unit increase in HbA1c resulted in 1.11 fewer matched symbols in the fully adjusted model (p=0.029). The percent with cognitive impairment, as assessed by the DSST, was significantly higher for those with HbA1c≥8.0% (≥64 mmol/mol) vs. HbA1c<7.0% (<53 mmol/mol) (14.6% vs. 6.3%, p=0.041) (Table 4). In a separate analysis replacing HbA1c with log-transformed HOMA-IR, HOMA-IR was not associated with any cognitive function test (data not shown).
Table 3.
Beta coefficients for the association between HbA1c and cognitive assessment score among all adults age ≥60 years, NHANES 2011–2014
| Delayed Word Recall (CERAD W-L) | Total Word Recall (CERAD W-L) | Animal Fluency | Digit Symbol Substitution Test (DSST) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Beta | 95% CI | p-value | Beta | 95% CI | p-value | Beta | 95% CI | p-value | Beta | 95% CI | p-value | |
| Unadjusted | −0.19 | −0.34– −0.03 | 0.019 | −0.51 | −1.02– −0.01 | 0.048 | −0.64 | −1.25– −0.03 | 0.040 | −2.55 | −3.99– −1.10 | 0.001 |
| Age adjusted | −0.19 | −0.35– −0.03 | 0.022 | −0.52 | −1.07– 0.03 | 0.062 | −0.65 | −1.24– −0.06 | 0.031 | −2.53 | −3.92– −1.14 | 0.001 |
| Model 1 | −0.15 | −0.32– 0.01 | 0.066 | −0.38 | −0.95– 0.18 | 0.175 | −0.50 | −1.09– 0.10 | 0.100 | −1.74 | −3.12– −0.35 | 0.016 |
| Model 2 | −0.14 | −0.30– 0.02 | 0.082 | −0.33 | −0.87– 0.21 | 0.220 | −0.51 | −1.12– 0.10 | 0.100 | −1.61 | −2.88– −0.33 | 0.015 |
| Model 3 | −0.12 | −0.29– 0.05 | 0.153 | −0.27 | −0.79– 0.25 | 0.302 | −0.42 | −0.99– 0.15 | 0.146 | −1.32 | −2.43– −0.21 | 0.021 |
| Model 4 | −0.10 | −0.26– 0.05 | 0.189 | −0.21 | −0.71– 0.29 | 0.395 | −0.38 | −0.90– 0.13 | 0.139 | −1.17 | −2.16– −0.18 | 0.023 |
| Model 5 | −0.13 | −0.30– 0.03 | 0.097 | −0.26 | −0.76– 0.24 | 0.305 | −0.47 | −0.94– −0.01 | 0.046 | −1.36 | −2.33– −0.39 | 0.007 |
| Model 6 | −0.12 | −0.29– 0.05 | 0.165 | −0.22 | −0.73– 0.29 | 0.387 | −0.44 | −0.86– −0.02 | 0.042 | −1.12 | −2.11– −0.14 | 0.026 |
| Model 7 | −0.12 | −0.29– 0.05 | 0.168 | −0.22 | −0.72– 0.29 | 0.394 | −0.44 | −0.86– −0.01 | 0.044 | −1.11 | −2.11– −0.12 | 0.029 |
Covariates were added individually, but are shown grouped by attribute for succinctness Model 1: Adjusted for age, race/ethnicity
Model 2: Model 1 + sex
Model 3: Model 2 + education, income
Model 4: Model 3 + smoking status, alcohol consumption, exercise
Model 5: Model 4 + body mass index
Model 6: Model 5 + hypertension, hyperlipidemia, history of CVD
Model 7: Model 6 + depression, history of stroke
Table 4.
The proportion (%, SE) with cognitive impairment in the total population by A1c level among adults age ≥60 years, NHANES 2011–2014
| A1c< 7.0% (<53 mmol/mol) | A1c 7.0– <8.0% (53– <64 mmol/mol) | A1c ≥8.0% (≥64 mmol/mol) | |
|---|---|---|---|
| CERAD WL Delayed | |||
| Model 1 | 4.5 (0.70) | 3.5 (1.56) | 9.7 (3.73) |
| Model 2 | 4.5 (0.70) | 3.5 (1.55) | 9.6 (3.81) |
| Model 3 | 4.5 (0.70) | 2.5 (1.54) | 9.8 (3.98) |
| Model 4 | 4.0 (0.70) | 2.3 (1.62) | 8.9 (3.72) |
| Model 5 | 3.9 (0.69) | 3.2 (1.67) | 9.6 (3.71) |
| Model 6 | 4.0 (0.73) | 2.0 (1.32) | 9.4 (3.70) |
| Model 7 | 4.0 (0.72) | 2.1 (1.38) | 9.3 (3.67) |
| CERAD WL Total | |||
| Model 1 | 5.4 (0.85) | 3.0 (1.56) | 11.4 (4.59) |
| Model 2 | 5.4 (0.85) | 2.9 (1.57) | 11.2 (4.56) |
| Model 3 | 5.5 (0.86) | 1.9 (1.59) | 11.5 (4.77) |
| Model 4 | 5.2 (0.86) | 1.5 (1.70) | 10.4 (4.52) |
| Model 5 | 5.1 (0.87) | 2.4 (1.91) | 11.1 (4.38) |
| Model 6 | 5.1 (0.91) | 1.3 (1.43) | 11.0 (4.58) |
| Model 7 | 5.1 (0.91) | 1.2 (1.49) | 11.1 (4.61) |
| Animal Fluency | |||
| Model 1 | 5.5 (1.02) | 6.5 (2.35) | 6.6 (2.75) |
| Model 2 | 5.5 (1.01) | 6.5 (2.34) | 6.6 (2.84) |
| Model 3 | 5.1 (1.01) | 4.8 (2.28) | 6.9 (2.93) |
| Model 4 | 4.7 (0.94) | 3.8 (2.83) | 6.6 (2.94) |
| Model 5 | 4.4 (0.84) | 3.5 (2.68) | 7.0 (3.13) |
| Model 6 | 4.6 (0.87) | 3.2 (2.94) | 6.7 (3.11) |
| Model 7 | 4.6 (0.87) | 2.7 (2.78) | 7.2 (3.09) |
| DSST | |||
| Model 1 | 7.4 (1.16) | 5.7 (1.36) | 14.8 (4.20) |
| Model 2 | 7.4 (1.16) | 5.7 (1.35) | 14.9 (4.18) |
| Model 3 | 7.0 (1.09) | 4.0 (1.63) | 14.5 (3.97) |
| Model 4 | 6.5 (1.02) | 3.4 (1.85) | 14.5 (4.14) * |
| Model 5 | 6.2 (0.94) | 3.2 (2.10) | 14.9 (4.10) * |
| Model 6 | 6.3 (0.97) | 2.3 (1.82) | 14.7 (4.17) * |
| Model 7 | 6.3 (0.98) | 2.3 (1.72) | 14.6 (4.26) * |
p<0.05 vs. A1c <7.0%
Cognitive impairment defined as ≤1.5 SD of the mean cognitive test score for the NHANES population
Model 1: Adjusted for age, race/ethnicity
Model 2: Model 1 + sex
Model 3: Model 2 + education, income
Model 4: Model 3 + smoking status, alcohol consumption, exercise
Model 5: Model 4 + body mass index
Model 6: Model 5 + hypertension, hyperlipidemia, history of CVD
Model 7: Model 6 + depression, history of stroke
4. Discussion
In this study of older adults in the United States, we found that those with diabetes, but not prediabetes, had mild cognitive dysfunction compared to those with normoglycemia. These results were found for the CERAD W-L tests after adjusting for demographic characteristics, health behaviors, and BMI, and were marginally significant after adjusting for comorbidities, including CVD. This suggests that immediate and delayed verbal learning ability is compromised among those with diabetes. While dysglycemia was not strongly associated with verbal learning ability, higher HbA1c was associated with worse executive function and processing speed, as measured by the DSST, and was independent of demographic characteristics and other cardiometabolic related indices.
Older adults with diabetes showed mild cognitive dysfunction as measured by the DSST, which is considered a sensitive measure of global cognitive function, compared to those with normoglycemia. This result aligns with a previous longitudinal study, ARIC, that found significantly greater cognitive decline among adults with diabetes compared to those without diabetes after adjustment for sociodemographics, CVD risk factors, and CVD (14). Regression analysis in our study showed that the association became non-significant after adjusting for education. Previous research has shown that level of education modifies the relationship between Alzheimer’s Disease pathology and cognitive function in older adults (29). Nevertheless, in the total U.S. population, increasing HbA1c, a measure of average blood glucose levels over an approximate three to four month period, was significantly associated with lower DSST scores even after accounting for multiple covariates, including education. In addition, we found that a significantly higher proportion of older adults with HbA1c≥8.0% (≥64 mmol/mol) had cognitive impairment, as measured by the DSST, compared to their counterparts with HbA1c<7.0% (<53 mmol/mol). Finally, we found few associations between HbA1c and cognitive impairment when the results were stratified by diabetes status (data not shown). For those with diabetes, current HbA1c control may not have as strong an effect on cognition as the duration of diabetes or the amount of time that diabetes is uncontrolled; in addition, sample size may have been too small to detect a difference. However, there was a significant inverse association between HbA1c and the DSST among the total population and this association remained after full adjustment for all covariates. This may suggest that dysglycemia, as measured by HbA1c, is strongly associated with a measure of global cognitive function regardless of other factors that may affect cognition.
Results from the CERAD W-L measure, a test of word learning and short-term memory, suggest that decreased sensitivity to glucose may affect memory structures in the brain, including the hippocampus (4). The longitudinal Health and Retirement Study found that among adults age ≥50 years, diabetes was associated with a 10% faster rate of memory decline, measured by immediate and delayed word list recalls (30). In multivariable analysis, adults with diabetes in our study reported significantly fewer words in the total and delayed recall tests compared to those with normoglycemia after adjustment for age, sociodemographic characteristics, lifestyle factors, and obesity, suggesting that verbal learning ability is compromised among those with diabetes, regardless of these factors. The association was attenuated after adjusting for cardiovascular disease; therefore, it appears that cardiovascular disease may moderate the association between diabetes and the CERAD W-L as previously reported(3).
Poor cognition can have detrimental effects on diabetes management and care (31), which may subsequently cause additional comorbidities and complications, including hypoglycemia (32, 33). In supplemental analysis, we found no association between cognitive impairment and currently implementing these recommended health-related behaviors (ORs range from 0.84 to 1.13 in fully adjusted models, data not shown). In addition, there was no association between cognitive impairment and diabetes self-care behaviors (checking blood glucose levels and feet for sores) among those with diagnosed diabetes (ORs range from 0.79 to 1.27). Future analyses are needed with more comprehensive measures of self-care behaviors.
We found no difference in cognitive function among those with prediabetes compared to those with normoglycemia. Since prediabetes probably represents a lower magnitude and shorter duration of hyperglycemia relative to diabetes, a significant association may not have been detectable.
A major strength of this study is the nationally representative sample of the U.S. non-institutionalized population. In addition, we were able to assess cognitive function by levels of glycemia using laboratory measures for FPG and HbA1c. The NHANES included three different cognitive tests, which provided more depth than previous NHANES surveys; however, there were many domains of cognitive function that we not assessed. The cross-sectional study design limits our ability to make any causal statements about the association between diabetes and cognitive function. Information on repetitive episodes of hypoglycemia, which has also been associated with cognitive impairment in some studies, was not available in NHANES(7, 9). Finally, we cannot rule out that adjustment is incomplete and that those with higher HbA1c levels may be more frail compared to those with lower HbA1c levels. Those with high HbA1c levels may also be more likely to have episodes of hypoglycemia which makes it difficult to fully elucidate the effect of glycemia(33, 34).
In a national sample of older adults, we found that persons with diabetes had moderate decrements in cognition compared to those with normoglycemia and that hyperglycemia was associated with poorer executive function and processing speed, regardless of diabetes status. Further research is needed to understand factors leading to impaired cognition among patients with diabetes so that interventions to prevent cognitive decline can be developed.
Acknowledgements
The authors would like to acknowledge Keith F. Rust, PhD for his statistical support.
This work was funded by the National Institutes of Diabetes and Digestive and Kidney Diseases (GS-10F-0381L). The sponsor was involved with the study design and data collection only.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
No potential conflicts of interests relevant to this article were reported.
References
- 1.Menke A, Casagrande S, Geiss L, Cowie CC. Prevalence of and Trends in Diabetes Among Adults in the United States, 1988–2012. JAMA. 2015;314(10):1021–9. [DOI] [PubMed] [Google Scholar]
- 2.2014 Alzheimer’s disease facts and figures. Alzheimers Dement. 2014;10(2):e47–92. [DOI] [PubMed] [Google Scholar]
- 3.Snyder HM, Corriveau RA, Craft S, Faber JE, Greenberg SM, Knopman D, et al. Vascular contributions to cognitive impairment and dementia including Alzheimer’s disease. Alzheimers Dement. 2015;11(6):710–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Small SA, Perera GM, DeLaPaz R, Mayeux R, Stern Y. Differential regional dysfunction of the hippocampal formation among elderly with memory decline and Alzheimer’s disease. Ann Neurol. 1999;45(4):466–72. [DOI] [PubMed] [Google Scholar]
- 5.Watson GS, Peskind ER, Asthana S, Purganan K, Wait C, Chapman D, et al. Insulin increases CSF Abeta42 levels in normal older adults. Neurology. 2003;60(12):1899–903. [DOI] [PubMed] [Google Scholar]
- 6.Zlokovic BV, Gottesman RF, Bernstein KE, Seshadri S, McKee A, Snyder H, et al. Vascular contributions to cognitive impairment and dementia (VCID): A report from the 2018 National Heart, Lung, and Blood Institute and National Institute of Neurological Disorders and Stroke Workshop. Alzheimers Dement. 2020;16(12):1714–33. [DOI] [PubMed] [Google Scholar]
- 7.Feinkohl I, Aung PP, Keller M, Robertson CM, Morling JR, McLachlan S, et al. Severe hypoglycemia and cognitive decline in older people with type 2 diabetes: the Edinburgh type 2 diabetes study. Diabetes Care. 2014;37(2):507–15. [DOI] [PubMed] [Google Scholar]
- 8.Jacobson AM, Musen G, Ryan CM, Silvers N, Cleary P, Waberski B, et al. Long-term effect of diabetes and its treatment on cognitive function. N Engl J Med. 2007;356(18):1842–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Kodl CT, Seaquist ER. Cognitive dysfunction and diabetes mellitus. Endocr Rev. 2008;29(4):494–511. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Ryan CM, van Duinkerken E, Rosano C. Neurocognitive consequences of diabetes. Am Psychol. 2016;71(7):563–76. [DOI] [PubMed] [Google Scholar]
- 11.Saedi E, Gheini MR, Faiz F, Arami MA. Diabetes mellitus and cognitive impairments. World J Diabetes. 2016;7(17):412–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Buysschaert M, Medina JL, Bergman M, Shah A, Lonier J. Prediabetes and associated disorders. Endocrine. 2015;48(2):371–93. [DOI] [PubMed] [Google Scholar]
- 13.Weinstein G, Maillard P, Himali JJ, Beiser AS, Au R, Wolf PA, et al. Glucose indices are associated with cognitive and structural brain measures in young adults. Neurology. 2015;84(23):2329–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Rawlings AM, Sharrett AR, Schneider AL, Coresh J, Albert M, Couper D, et al. Diabetes in midlife and cognitive change over 20 years: a cohort study. Ann Intern Med. 2014;161(11):785–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Tsai CK, Kao TW, Lee JT, Wu CJ, Hueng DY, Liang CS, et al. Increased risk of cognitive impairment in patients with components of metabolic syndrome. Medicine (Baltimore). 2016;95(36):e4791. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Sherzai AZ, Shaheen M, Yu JJ, Talbot K, Sherzai D. Insulin resistance and cognitive test performance in elderly adults: National health and nutrition examination survey (NHANES). J Neurol Sci. 2018;388:97–102. [DOI] [PubMed] [Google Scholar]
- 17.Pavlik VN, Hyman DJ, Doody R. Cardiovascular risk factors and cognitive function in adults 30–59 years of age (NHANES III). Neuroepidemiology. 2005;24(1–2):42–50. [DOI] [PubMed] [Google Scholar]
- 18.Fox K, Borer JS, Camm AJ, Danchin N, Ferrari R, Lopez Sendon JL, et al. Resting heart rate in cardiovascular disease. J Am Coll Cardiol. 2007;50(9):823–30. [DOI] [PubMed] [Google Scholar]
- 19.Centers for Disease Control and Prevention (CDC). National Center for Health Statistics (NCHS). National Health and Nutrition Examination Survey Examination Protocol. Hyattsville, MD: Department of Health and Human Services, Centers for Disease Control and Prevention; 2008. [Google Scholar]
- 20.Centers for Disease Control and Prevention (CDC). National Center for Health Statistics (NCHS). National Health and Nutrition Examination Survey Laboratory Protocol: Cholesterol and Triglycerides. Hyattsville, MD: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention; 2008. [Google Scholar]
- 21.Centers for Disease Control and Prevention. National Health and Nutrition Examination Survey Laboratory Protocol. In: U.S. Department of Health and Human Services, ed. Hyattsville, MD; 2011. [Google Scholar]
- 22.Fillenbaum GG, van Belle G, Morris JC, Mohs RC, Mirra SS, Davis PC, et al. Consortium to Establish a Registry for Alzheimer’s Disease (CERAD): the first twenty years. Alzheimers Dement. 2008;4(2):96–109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Morris JC, Heyman A, Mohs RC, Hughes JP, van Belle G, Fillenbaum G, et al. The Consortium to Establish a Registry for Alzheimer’s Disease (CERAD). Part I. Clinical and neuropsychological assessment of Alzheimer’s disease. Neurology. 1989;39(9):1159–65. [DOI] [PubMed] [Google Scholar]
- 24.Clark LJ, Gatz M, Zheng L, Chen YL, McCleary C, Mack WJ. Longitudinal verbal fluency in normal aging, preclinical, and prevalent Alzheimer’s disease. Am J Alzheimers Dis Other Demen. 2009;24(6):461–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Wechsler DS. WAIS Manual – Third Edition. In: Psychological Corporation, ed. New York, ; 1997. [Google Scholar]
- 26.Kroenke K, Spitzer RL, Williams JB. The PHQ-9: validity of a brief depression severity measure. J Gen Intern Med. 2001;16(9):606–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Cohen J A power primer. Psychol Bull. 1992;112(1):155–9. [DOI] [PubMed] [Google Scholar]
- 28.Nunley KA, Rosano C, Ryan CM, Jennings JR, Aizenstein HJ, Zgibor JC, et al. Clinically Relevant Cognitive Impairment in Middle-Aged Adults With Childhood-Onset Type 1 Diabetes. Diabetes Care. 2015;38(9):1768–76. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Bennett DA, Wilson RS, Schneider JA, Evans DA, Mendes de Leon CF, Arnold SE, et al. Education modifies the relation of AD pathology to level of cognitive function in older persons. Neurology. 2003;60(12):1909–15. [DOI] [PubMed] [Google Scholar]
- 30.Marden JR, Mayeda ER, Tchetgen Tchetgen EJ, Kawachi I, Glymour MM. High Hemoglobin A1c and Diabetes Predict Memory Decline in the Health and Retirement Study. Alzheimer Dis Assoc Disord. 2017;31(1):48–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Tomlin A, Sinclair A. The influence of cognition on self-management of type 2 diabetes in older people. Psychol Res Behav Manag. 2016;9:7–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Bruderer SG, Bodmer M, Jick SS, Bader G, Schlienger RG, Meier CR. Incidence of and risk factors for severe hypoglycaemia in treated type 2 diabetes mellitus patients in the UK--a nested case-control analysis. Diabetes Obes Metab. 2014;16(9):801–11. [DOI] [PubMed] [Google Scholar]
- 33.Lee AK, Lee CJ, Huang ES, Sharrett AR, Coresh J, Selvin E. Risk Factors for Severe Hypoglycemia in Black and White Adults With Diabetes: The Atherosclerosis Risk in Communities (ARIC) Study. Diabetes Care. 2017;40(12):1661–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Miller ME, Bonds DE, Gerstein HC, Seaquist ER, Bergenstal RM, Calles-Escandon J, et al. The effects of baseline characteristics, glycaemia treatment approach, and glycated haemoglobin concentration on the risk of severe hypoglycaemia: post hoc epidemiological analysis of the ACCORD study. BMJ. 2010;340:b5444. [DOI] [PMC free article] [PubMed] [Google Scholar]