CHAPTER 24 DIABETES AND COGNITIVE IMPAIRMENT José A. Luchsinger, MD, MPH, Christopher Ryan, PhD, and Lenore J. Launer, PhD
Dr. José A. Luchsinger is Associate Professor of Medicine and Epidemiology, Columbia University Medical Center, New York, NY. Dr. Christopher Ryan is Director, Human Research Protection Program, University of California San Francisco, San Francisco, CA. Dr. Lenore J. Launer is Senior Investigator and Chief, Neuroepidemiology Section, Laboratory of Epidemiology and Population Sciences, National Institute on Aging, National Institutes of Health, Bethesda, MD.
SUMMARY
Cognitive impairment, ranging from mild cognitive impairment to dementia, is increasingly recognized as a potential complication of diabetes. The increase in the prevalence of diabetes along with the aging of the population may result in a large increase in the prevalence of cognitive impairment in persons with diabetes. Approximately one-third of the United States adult population has prediabetes or diabetes. However, about one-half of persons age ≥60 years, who are most at risk for cognitive impairment, have prediabetes or diabetes. The association of diabetes and cognitive impairment may reflect a direct effect on the brain of hyperglycemia or the effects of the diabetes-associated comorbidities, such as hypertension, dyslipidemia, or hyperinsulinemia. There is evidence for an effect of diabetes-related processes on both neurodegenerative and vascular processes, thus contributing to both Alzheimer’s-like cognitive impairment, i.e., amnestic, and cerebrovascular-type cognitive impairment characterized by impairment in executive cognitive function.
This chapter reviews epidemiologic studies of the association between diabetes and cognitive impairment in large clinical or community- based studies, emphasizing studies in the United States. Type 1 diabetes is, in general, related to modest decrements in cognitive abilities, primarily in executive-frontal abilities. Type 2 diabetes seems to be more strongly related to vascular and non-amnestic forms of cognitive impairment, including non-amnestic mild cognitive impairment and vascular dementia, but it also seems to have a less strong association with amnestic forms of cognitive impairment, including amnestic mild cognitive impairment and Alzheimer’s dementia.
The relation of diabetes to cognitive impairment has therapeutic implications. There is conflicting evidence from clinical trials of diabetes interventions examining the effects of better glucose control on cognition and the brain. Lifestyle interventions among persons with diabetes and prediabetes do not seem to be related to better cognitive outcomes in the long term.
Despite many studies documenting the association between diabetes and cognitive impairment, the causal nature and the mechanisms of this association have not been fully established and require further study.
INTRODUCTION
This chapter describes important DESCRIPTION OF COGNITIVE Performance in Cognitive Tests concepts in cognitive impairment and IMPAIRMENT The most frequently studied cognitive epidemiologic data on cognitive impair- To study cognitive impairment, persons domains broadly defined are memory ment in the general U.S. population. with and without diabetes are compared in and executive-frontal abilities. Memory Then, the relationship between diabetes performance of cognitive tests and by the can be defined as the ability to recall an and cognitive impairment is explored, presence or absence of cognitive dis orders, experience after it has ended (1). Memory including potential mechanisms and such as mild cognitive impairment (MCI) is tested by repetition of words usually epidemiologic data linking the two and dementia. A brief description of these delivered verbally as lists of several words conditions, as well as implications for types of cognitive impairment follows. or as a part of tests of global cognition, treatment and prevention. The review Importantly, in this chapter, Alzheimer’s such as the 5-minute recall of three of the epidemiologic data from diabetic disease (AD) is referred to as one items in the Mini Mental Status Exam (2). cohorts emphasizes large longitudinal pathologic process underlying cognitive Memory problems are probably the most studies in the United States and Canada impairment, characterized by the presence frequent cognitive complaints that bring and clinical trials when available. Both of amyloid plaques and neurofibrillary patients to medical attention. Several key type 1 diabetes and type 2 diabetes are tangles on brain autopsy, and not the processes are involved in the long-term covered in this chapter. manifestation of cognitive impairment, maintenance of memories. Consolidation, a frequent source of confusion. Dementia, primarily located in the hippocampal the most severe clinical manifestation of formation, enables the acquisition of new AD, is referred to as AD dementia. memories (1). Consolidation is the first
Received in final form February 12, 2016. 24–1 DIABETES IN AMERICA, 3rd Edition and most prominently affected process in considered large. Some sections of this This relative lack of research is particularly AD dementia and its antecedent, amnestic chapter make reference to Cohen’s d in relevant to persons with diabetes, who MCI. Another key memory process is the interpretation of effect sizes. likely have cerebrovascular disease and, retrieval (1), which is most affected thus, are at high risk of cognitive impair- by lesions disturbing the frontal lobe- Mild Cognitive Impairment ment involving executive-frontal functions. subcortical circuits in the brain. Retrieval MCI is a diagnostic category used to deficits manifest as difficulty in recalling capture the transitional state between Dementia words or objects that can be aided by normal cognitive function and dementia Dementia is a syndrome characterized by giving multiple choice options or cues (6,7) and has become a target for inter- impairment of memory and other cogni- (recognition). ventions (8). The feature that differentiates tive abilities, as well as behavior disorder MCI from dementia is functional impair- severe enough to impair the ability to live Executive-frontal abilities include those ment: individuals with MCI still have the independently (11). Dementia is the most skills that are used to plan and execute ability to function in daily activities (7), extreme form of cognitive impairment and complex tasks and comprise different while persons with dementia do not. In the one most often identified in research processes, such as attention, working MCI, general cognitive performance is well and clinical practice due to its clinical memory, and impulse control. Executive- preserved, but performance in one cogni- significance. The most common cause of frontal abilities are most often affected by tive domain falls below expectations for late-onset dementia is AD (12), comprising lesions that disrupt the frontal subcortical age and education. Typically, it is defined between 60% and 80% of cases. Vascular networks (3), the same network that following cutpoints of “pass/fail” results dementia is the second most common supports memory retrieval. A common on neuropsychological tests (e.g., 1.5 cause, comprising about 10% of cases, but type of lesion that can affect frontal standard deviations of established norms). approximately 50% of cases of dementia subcortical networks in diabetes is cere- This threshold is arbitrary, and the field have a vascular component (12). Memory brovascular disease. Commonly used of cognitive disorders has become inter- may be affected in vascular dementia, but tests of executive-frontal abilities include ested in even milder forms of cognitive the main cognitive domains affected are digits forward and backward (a test in impairment (e.g., using a threshold of generally related to executive-frontal abil- which persons are asked to repeat a 1.0 standard deviations rather than 1.5), ities. An important subgroup of dementia number series forward and backward), which some have termed pre-MCI (9). cases, the size of which likely differs the attention items of the Mini Mental across populations, has mixed pathology, Status Exam (2), and the Digit Symbol MCI can be classified as amnestic and having both neurodegenerative lesions Substitution Test (DSST) of the Wechsler non-amnestic MCI. Amnestic MCI involves typical of AD and cerebrovascular disease, Adult Intelligence Scale (4). memory impairment, while non-amnestic which is “typical” of vascular dementia MCI involves predominantly non-memory (13). This mixed dementia is particularly An important consideration about the domains, such as frontal-executive abilities. important when discussing cognitive evaluation of performance on cognitive Amnestic MCI is believed to be an early impairment in diabetes, because diabetes tests as outcomes is the judgment of the manifestation of AD and an antecedent is well known to be a risk factor for cere- clinical significance of differences found of AD dementia, while non-amnestic MCI, brovascular disease (14), but whether it between groups, for example, between such as executive MCI, may be a form of causes neurodegenerative disease is less persons with diabetes and those without vascular cognitive impairment (VCI) (6). clear (10). Other causes of dementia, such diabetes. Some studies will report that In VCI, cognitive impairment may present as frontotemporal dementia and Lewy persons with diabetes perform less well slowly and incrementally as subclinical body disease, are less common (15), are in a test than persons without diabetes cerebrovascular disease worsens (seen not known to be related to diabetes, and and provide a difference size or the score as white matter hyperintensities [WMH] are thus not discussed in this chapter. in performance in both groups. It can be on brain imaging, called mini-strokes concluded from such data that persons by clinicians, and “silent infarcts”), or it EPIDEMIOLOGY AND BURDEN with diabetes will perform less well may suddenly appear after an infarct. In OF COGNITIVE IMPAIRMENT IN than persons without diabetes, but the general, VCI affects executive-frontal abil- THE GENERAL U.S. POPULATION clinical importance of such a difference ities and memory retrieval and, similar to Frequency is uncertain. One way to overcome this amnestic MCI, may or may not progress to Major epidemiologic studies have shown problem is by providing standardized the syndrome of dementia. Some experts that dementia prevalence increases loga- differences. One common metric used believe that the cognition field has paid rithmically after age 70 years (16) and is Cohen’s d (5), which is calculated more attention to cognitive impairment may reach 50% in persons age ≥85 years as the difference between the means due to AD to the detriment of the iden- (17). According to the 2014 Alzheimer’s divided by the pooled standard deviation. tification of the other forms of cognitive Disease facts and figures report from A Cohen’s d of 0.2 is considered small, impairment described here, particularly the Alzheimer’s Association (12), 11% of 0.5 is considered medium, and 0.8 is those related to executive abilities (10). people age ≥65 years and one-third of
24–2 Diabetes and Cognitive Impairment people age ≥85 years have late-onset AD It is important to put the epidemiology prevention (30) have received the most dementia. Given the longer life expectancy of cognitive impairment in the context attention in studies of causes, prevention, of the population, cases of AD dementia of the epidemiology of diabetes in the and treatment of AD. in persons age ≥65 years are expected to United States. Over 100 million U.S. increase by 40% from current numbers adults have diabetes or prediabetes: In 2010, an expert panel from the by 2025 (12). Estimates from the 2010 25 million have diabetes and 79 million National Institutes of Health concluded Census revealed that in 2010, there were have prediabetes (22), comprising that there was insufficient evidence to 4.7 million cases of AD dementia, of one-third of the U.S. population. This recommend treatment or prevention for whom 0.7 million were age 65–74 years, high prevalence is accounted for in AD dementia (31). The genetic causes 2.3 million were 75–84 years, and 1.8 large part by the age group most at risk of early-onset, familial AD dementia are million were ≥85 years (18). Based on for cognitive impairment, persons age known (32), but this form represents a these estimates, the projected number of ≥60 years in whom the prevalence of minority of cases and is not the focus of AD dementia cases in 2050 will be 13.8 pre diabetes and diabetes is >50% (>60% this chapter. APOE-ε4, encoding a form million, with 7.0 million age ≥85 years. in minorities) (22,23). A large proportion of apolipoprotein E, is the only robust AD dementia is the sixth leading cause of of this prevalence is accounted for by genetic susceptibility risk factor for late- death in the United States and the fifth non-Hispanic blacks and Hispanics onset AD dementia, occurring in up to leading cause for those age ≥65 years (22,24,25), who have about twice the 50% of cases. However, it has not proved (15). As previously mentioned, vascular prevalence of diabetes compared to useful for diagnosis, prevention, or treat- dementia is the second most common non-Hispanic whites (22,24). ment (33,34,35). In addition, APOE-ε4 cause, comprising about 10% of cases, has lower predictive value in non-His- but approximately 50% of all cases of An interesting aspect of dementia is that panic blacks and Hispanics compared dementia have a vascular component (12). it has ethnic differences similar to those to non-Hispanic whites (27,33). Among found in type 2 diabetes. In a study from nongenetic risk factors, older age, lower The prevalence of amnestic MCI varies Northern Manhattan in New York City, education, and diabetes (36) are the most from 3% to 20% depending on the criteria Hispanics and non-Hispanic blacks had robust. Studies of hypertension in relation applied (19) and the population studied. approximately double the prevalence to dementia have shown that mid-life Estimates increase from about 1% in of dementia compared to non-Hispanic exposure to risk factors may be a better persons age 60 years to 25% at age 85 whites (26,27). In the same cohort from indicator of future risk for dementia than years (20). The Aging, Demographics, and Northern Manhattan, it was estimated measures obtained close to the time that Memory Study estimated the prevalence that the higher prevalence of diabetes in cognitive impairment is detected (37,38). of cognitive impairment without dementia, minorities could account for some of the Other factors, such as smoking, dyslip- similar to MCI, in a representative sample ethnic disparities in cognitive impairment, idemia, diet, and physical activity, have of 856 persons in the United States (21). including MCI and dementia, among been studied, but results are inconsistent Twenty-two percent of persons age ≥71 persons age ≥65 years (28). (39). Risk factors for VCI (40) are mainly years met the study criteria for cognitive those associated with cerebrovascular impairment without dementia. Among Causes, Risk Factors, Prevention, and disease, including diabetes and other those, 8% died and 12% progressed to Treatment of Cognitive Impairment vascular risk factors, such as hyperten- dementia on an annual basis. Persons The pathologic hallmark of AD is the sion and dyslipidemia (41). with amnestic MCI may progress to AD presence of amyloid plaques and neuro- dementia at the rate of nearly 10%–15% fibrillary tangles identified in brains that per year (6) compared to 1%–2% in elderly have come to autopsy (29). To date, persons with normal cognition (7). the accumulation of amyloid and its
POTENTIAL MECHANISMS LINKING DIABETES AND COGNITIVE IMPAIRMENT
Diabetes and its related conditions, potential mechanisms linking diabetes The mechanisms for the association with adiposity and hyperinsulinemia, and with cognitive impairment are classified as AD dementia are not clear. However, the their clustering with other vascular cerebrovascular and non-cerebrovascular. presence of infarcts in persons with AD risk factors, such as hypertension and neuropathology may result in an increase dyslipidemia (42), are known to increase CEREBROVASCULAR SUBSTRATES in brain injury and increase the likelihood of the risk for cerebrovascular disease Infarcts, ascertained by clinical history reaching a clinical dementia endpoint in the (14,41,43,44,45,46). In addition, insulin as strokes causing neurologic deficits presence of such neuropathology (50,51). or diabetes-related byproducts may affect (48) or identified as focal lesions on brain the amyloid cascade that is thought imaging (49), are related to a higher risk to be responsible for AD (47). Thus, of AD dementia and vascular dementia.
24–3 DIABETES IN AMERICA, 3rd Edition
FIGURE 24.1. Relation of Glycemia and Insulin Resistance to Risk of Alzheimer’s Disease Neuropathology, the Hisayama Study, 1998–2003
Model 1 Model 2 A B C 6 6 * 6 * 5 5 5 4 4 4 † * 3 * 3 * 3 2 2 2 Odds ratio Odds ratio Odds ratio 1 1 1 (n) (45) (46) (44) (n) (37) (52) (45) (n) (45) (46) (43) 5.9 5.9 to .6 .6 4 4 to 6 6 .97 .97 to 1.77 1.77 2-hour Post-load Plasma Glucose (mmol/L) Fasting Insulin (μU/mL) HOMA-IR
Odds ratios for each tertile of (A) glucose, (B) insulin, and (C) HOMA-IR versus the lowest tertile for the presence of neuritic plaques. Model 1: adjusted for age and sex. Model 2: adjusted for age, sex, systolic blood pressure, total cholesterol, body mass index, current smoking, regular exercise, and cerebrovascular disease. Conversions for glucose and insulin values are provided in Diabetes in America Appendix 1 Conversions. HOMA-IR, homeostasis model assessment of insulin resistance. * p<0.05 versus the lowest tertile † p<0.10 versus the lowest tertile SOURCE: Reference 87, copyright © 2010 Wolters Kluwer, reprinted with permission
White matter disease, called mini-strokes in important brain structures, including their receptors (RAGE) (70,71,72), and by clinicians, is ascertained as WMH, also the hippocampus and entorhinal cortex, AGEs are known to be related to the tradi- known as leukoaraiosis, on brain imaging are affected early in AD (62). Insulin and tional microvascular complications of type (40). WMH may represent microvascular amyloid are competing substrates for 2 diabetes (73,74,75,76,77,78). Increased disease in the brain or demyelination. insulin-degrading enzyme, which has been expression of RAGE is also observed in AD However, the nature of WMH is still not linked to clearance of amyloid in the brain dementia (79,80,81), and expression of well understood. WMH are thought to be (63). Insulin in the brain can increase the RAGE is enhanced in blood vessels near ischemic in origin in the same way that deposition of amyloid and Tau protein amyloid deposits in AD brains (79,82). infarcts are (52) and, thus, have been phosphorylation, which are central to the proposed as surrogate markers of cerebro- pathogenesis of AD (60). The pathways Several studies have examined the vascular disease (52). However, WMH are relating insulin in the periphery with relation between diabetes and dementia common in AD dementia, and some may amyloid clearance in the brain are multiple mechanisms, particularly surrogate be related to cerebral amyloid angiopathy and complex (64). One potential pathway markers of AD, and found conflicting (53,54,55,56), a correlate of AD, inflam- is that peripheral hyperinsulinemia down- results. These markers have included mation, or other demyelinating processes. regulates insulin uptake in the blood brain neuropathology studies, brain imaging WMH are common correlates of cognitive barrier due to saturation over physiologic studies, and studies of other biomarkers. impairment in diabetes (57), but it is levels (65). This may result in reduction Neuropathology studies from 233 partic- unclear whether these WMH are markers of insulin levels in the brain, as well as ipants in the Religious Orders Study, a of microvascular injury or if they represent downregulation of the expression of insu- study of deceased older Catholic nuns, a process related to amyloid deposition or lin-degrading enzyme (66) and reduction priests, and brothers from across the another neurodegenerative process. in its related amyloid clearance function United States (83), and from a study (63). This complex observation has been of 385 nursing home residents who NON-CEREBROVASCULAR used to support the seemingly paradoxical underwent autopsy (84) found that a MECHANISMS use of rosiglitazone, an insulin sensitizer history of diabetes was not related to AD Hyperinsulinemia caused by insulin resis- (67,68), and intranasal insulin (69) in the neuropathology, similar to a study in a tance is an important antecedent and treatment of AD dementia. Finnish population (85). However, a study companion of type 2 diabetes (58,59). of Japanese Americans in Hawaii found Hyperinsulinemia is a plausible risk factor Advanced glycation endproducts (AGEs) that diabetes was associated with higher for AD dementia, independent of cere- are closely linked with glycemia and prevalence of AD neuropathology (86). brovascular disease for several reasons. diabetes and have also been related to The most compelling neuropathology Insulin can cross the blood brain barrier AD. Glycated hemoglobin (A1c), used by study to date found, in a Japanese sample, (60), and peripheral insulin infusions in clinicians to track diabetes control, is the that increased glycemia and insulin the elderly may affect amyloid levels in most common example of an AGE. With resistance measured 10 years before cerebrospinal fluid (61), an indirect marker hyperglycemia, diabetic animal and autopsy were related to increased risk of of AD risk. In addition, insulin receptors human tissues increase both AGEs and AD neuropathology (Figure 24.1) (87). One
24–4 Diabetes and Cognitive Impairment brain imaging study in the United States FIGURE 24.2. Patterns of Brain Metabolism and Amyloid Brain Accumulation by Positron found that insulin resistance was related Emission Tomography, by Diabetes Status and Cognitive Impairment, Mayo Clinic Study of to a profile of brain metabolism on brain Aging, 2006–2012 positron emission tomography (PET) that is consistent with AD (88). Similarly, 2. 2. another brain imaging study from the United States among 749 persons with a mean age of 79 years found that diabetes 1.5 1.5 was associated with a brain metabo-
lism pattern similar to AD, but not with FDG ratio 1. 1. increased amyloid brain accumulation, as ascertained with brain PET with an amyloid ligand (Figure 24.2) (89). A study .5 P = . 1 .5 P = .1 from the Baltimore Longitudinal Study No DM DM No DM DM of Aging (BLSA) (90) found among 197 persons with autopsy at a mean age 3.5 3.5 of 88 years that hyperglycemia, insulin resistance, and diabetes assessed 3. 3. approximately at age 66 years were 2.5 2.5 not associated with increased brain AD
pathology. The same study had data PIB ratio 2. 2. on 53 subjects with brain amyloid PET acquired at a mean age of 79 years and 1.5 1.5 also found no evidence of a relation with 1. P = . 1. P = . brain amyloid. No DM DM No DM DM Cognitively normal MCI The conflicting results of these studies do not help resolve the question of Box plots for 18F-FDG and 11C-PiB retention ratio in Alzheimer’s disease signature regions by diabetes mellitus (DM) whether diabetes is related to brain in cognitively normal subjects and those with mild cognitive impairment (MCI). = women; = men. amyloid accumulation. The limitations SOURCE: Reference 89, copyright © 2014 Society of Nuclear Medicine and Molecular Imaging, reprinted with permission of the neuropathology studies include survival bias and selection bias and the that found that hyperglycemia and insulin from survival and selection biases (i.e., lack of proper ascertainment of diabetes resistance were related to AD pathology persons with diabetes may have a higher status compared with the previously (87). The limitations of the few amyloid mortality risk that could result in survival mentioned Japanese study with detailed brain imaging studies include old age and bias and affect the study of dementia and data on glycemia and insulin resistance small clinic samples, which also suffer neuropathology outcomes).
TYPE 1 DIABETES AND COGNITIVE IMPAIRMENT
Historically, most research on the rela- defined in terms of performance differ- among those children and adolescents tionship between type 1 diabetes and ences between demographically similar with diabetes. Effects were largest for cognitive impairment was conducted on subjects with and without diabetes. Most measures of “crystalized intelligence”— relatively small “convenience” samples, recent studies have expressed these as factual information about the world, verbal and was restricted to children or young standardized differences (e.g., z score learning, and visual and motor integration. and middle-aged adults. Cognitive status differences; Cohen’s d) (93). With very Because a number of studies have consis- was typically measured with a battery of few exceptions (94,95), researchers have tently indicated that children with an “early neuropsychological tests that assessed not made formal ratings of clinically onset” of diabetes—usually diagnosis multiple cognitive domains, including significant impairment like dementia or before age 6 or 7 years—perform more attention/concentration, learning and MCI, largely because it is quite unusual to poorly than those with a later onset of memory, mental and motor speed, find clinical evidence of serious cognitive diabetes (97), the meta-analysis was visual perceptual skills, and reasoning impairments in this patient population. repeated, but comparing early-onset or “executive function” (91,92). Because Drawing on results from 15 studies with later-onset subjects. The magnitude the focus was on identifying mild to comprising 1,029 subjects with diabetes of the cognitive differences was found moderate decrements in cognitive func- and 751 control subjects, Gaudieri et al. to be somewhat larger in persons with tioning, dysfunction was operationally (96) found modestly poorer performance early onset of diabetes compared with
24–5 DIABETES IN AMERICA, 3rd Edition later onset, with more domains equal to FIGURE 24.3. Meta-Analysis of Cognitive Performance Among Adults With Type 1 Diabetes or greater than an effect size (Cohen’s d) of 0.15, considered a small difference as Overall cognition ( ) p . 1 compared with controls. Taken together, Intelligence these results support the view that in Crystalized intelligence (27 ) p . 1 children and adolescents, diabetes is Fluid intelligence (1 ) p . 1 associated with modest decrements on a Learning and memory range of cognitive skills, with these effects Wor ing memory (244) being greatest in those youngsters who Visual learning and immediate memory (1 7) were diagnosed with diabetes early in life. Verbal learning and immediate memory (2 4) Another meta-analysis, adding several Visual delayed memory (15 ) newer studies (24 studies published Verbal delayed memory (15 ) between 1980 and 2005), identified a Psychomotor activity and speed of information processing generally similar pattern of results (98), Psychomotor efficiency (3 ) p . 5 with modest differences between diabetic Speed of information processing ( ) p . 5 and nondiabetic samples (Cohen’s d Motor speed (14 ) 0.15–0.30) evident on tests of verbal and Attention performance intelligence, as well as on Visual attention (195) p . 1 measures of visuospatial ability, motor Sustained attention (217) p . 1 Cognitive domains (number of patients included) speed, attention, and reading and writing. Divided attention (5 ) Selective attention ( ) In terms of cognition in adults with type 1 Cognitive flexibility (3 4) p . 1 diabetes, Brands et al. (99) published a Visual perception (2 2) p . 1 meta-analysis of 33 studies published Language (144) between 1980 and 2004 that included -1.6 -1.4 -1.2 -1. - . - .6 - .4 - .2 .2 .4 .6 adults age >18 years and a comparison Standardized effect sizes (Cohen’s d) and 95% confidence intervals for the cognitive domains in patients with type 1 group of subjects without diabetes. In diabetes compared with nondiabetic control subjects. Number of patients included in each domain is listed between virtually all studies, subjects were age parentheses. Nonsignificant p-values are not shown. Points to the left of 0 indicate worse cognitive performance in <50 years at the time of the cognitive persons with type 1 diabetes compared to controls without diabetes. SOURCE: Reference 99, copyright © 2005 American Diabetes Association, reprinted with permission from The assessment. As illustrated in Figure American Diabetes Association. References for individual studies are available from Reference 99. 24.3 (99), adults with type 1 diabetes performed consistently more poorly than of children and adolescents: intelligence, severe hypoglycemia and because earlier controls on measures of overall cognition, as measured by formal IQ tests, is studies had demonstrated that severe intelligence, psychomotor efficiency, consistently lower—by about 3 to 5 hypoglycemia could adversely affect the speed of information processing, points on average (the majority of the structural and functional integrity of the cognitive flexibility, attention, and visual general population is between 85 and brain, the DCCT study group developed a perceptual skills. The largest effect 115 points of IQ)—and performance comprehensive battery of neurocognitive sizes (Cohen’s d 0.5–0.7) were found is slower—by 5%–15%—on tasks that tests that was administered at study entry on measures of crystalized intelligence, require attention and rapid responding. and repeated at years 2, 5, 7, and/or study fluid intelligence (e.g., problem-solving, end. Data from 1,441 subjects, followed reasoning), psychomotor efficiency, and The only large-scale study to compre- for an average of 6.5 years, demonstrated cognitive flexibility, which fall under the hensively assess adolescents and adults that intensive insulin therapy did not umbrella of non-amnestic cognitive with type 1 diabetes over time is the differentially disrupt functioning on any function. It is noteworthy that with the Diabetes Control and Complications Trial of the eight cognitive domains studied. exception of measures of crystalized (DCCT), conducted in 29 sites in the Despite high rates of hypoglycemia in the intelligence, the other cognitive domains United States and Canada (100). This study group, there was no relationship that were most affected by diabetes clinical trial, which compared the effects between moderately severe hypoglycemia tended to be those that required rapid of intensive insulin therapy (multiple daily and performance on any cognitive test responding or were otherwise time- injections) with conventional therapy (one (95,101). Use of an innovative clinical limited. Despite the relatively small or two daily injections), was designed to rating procedure (102) showed very low sample sizes of the studies included in test the hypothesis that intensive therapy rates of clinically significant worsening this analysis, as well as the wide range delayed or prevented the development of over time (by year 5, only 23 of 1,383 of measures used by the different microvascular and neuropathic compli- subjects showed significant cognitive researchers, these results are remarkably cations. Because intensive therapy was decline), and these effects were unrelated consistent with those reported in studies known to increase the incidence of to hypoglycemia.
24–6 Diabetes and Cognitive Impairment
The Epidemiology of Diabetes FIGURE 24.4. Effects of DCCT Treatment Group, Severe Hypoglycemia, and Glycated Interventions and Complications (EDIC) Hemoglobin on Changes in Cognition, From Entry Into DCCT to Year 12 in the EDIC Study study, a follow-up assessment of the DCCT cohort (N=1,144), studied a mean A. Original treatment assignment Intensive treatment Conventional treatment of 18 years following study entry, also .3 showed no between-group differences on any cognitive domain nor any rela- .2 tionship between recurrent episodes of .1 hypoglycemia and change in cognitive functioning over time (Figure 24.4) (103). However, poorer metabolic control, oper- - .1 ationalized as a time-weighted A1c value Change in z score - .2 of ≥8.2% (≥66 mmol/mol), was associated with consistently poorer performance over - .3 time on two cognitive domains: psycho- - .4 motor efficiency and motor speed. 1 2 3 4 5 6 7 Cognitive Domain Subsequent analyses, focusing on predictors of cognitive changes over time, B. Number of severe hypoglycemic events demonstrated that after controlling for age No episodes 1–5 episodes 5 episodes .3 and education, proliferative retinopathy, renal complications, and time-weighted .2 A1c values were each independently and .1 significantly associated with declining performance on measures of psycho- motor efficiency 104( ). This work is - .1 consistent with earlier longitudinal studies (105) that link the development of micro- Change in z score - .2 vascular complications to psychomotor - .3 slowing in adults with type 1 diabetes. - .4 1 2 3 4 5 6 7 Cognitive Domain
C. Degree of metabolic control Glycated hemoglobin 7.4 Glycated hemoglobin ≥7.4% and ≤8.8% .4 Glycated hemoglobin . p= . 1 .3 .2 .1 - .1 - .2 Change in z score - .3 p . 1 - .4 - .5 1 2 3 4 5 6 7 Cognitive Domain
The bars show the changes within cognitive domains between cognitive testing at baseline in DCCT and follow-up testing (a mean of 18 years after baseline) expressed as (A) changes in z scores for intensive versus conventional treatment, (B) frequency of episodes of severe hypoglycemia (coma or seizure), and (C) mean glycated hemoglobin values. Across the three groups, higher levels of glycated hemoglobin were associated with moderate declines in psychomotor efficiency (p<0.001) and motor speed (p=0.001), but no other cognitive domain was affected significantly. Cognitive domains are numbered as follows: 1, problem solving; 2, learning; 3, immediate memory; 4, delayed recall; 5, spatial information; 6, attention; 7, psychomotor efficiency; and 8, motor speed. Conversions for glycated hemoglobin values are provided in Diabetes in America Appendix 1 Conversions. DCCT, Diabetes Control and Complications Trial; EDIC, Epidemiology of Diabetes Interventions and Complications study. SOURCE: Reference 103, copyright © 2007 Massachusetts Medical Society, reprinted with permission
24–7 DIABETES IN AMERICA, 3rd Edition
TYPE 2 DIABETES AND COGNITIVE IMPAIRMENT
DIABETES, COGNITIVE DECLINE, Self-reported diabetes was related to DIABETES AND DEMENTIA AND MILD COGNITIVE IMPAIRMENT a twofold higher risk of developing Numerous studies conducted in a wide The difference in cognitive function MCI among 7,027 osteoporotic post- range of ethnic groups have examined the between those with and without diabetes menopausal women with a mean age relation between diabetes and dementia is relatively robust across cohort studies, of 66 years participating in a clinical (112,113). Table 24.1 shows the results of although results vary in terms of which trial of the osteoporosis medication representative prospective studies in the specific functions differ between the two raloxifene in 25 countries, including United States. In general, the association groups. Over 4 years, a cohort of 999 the United States (108). A multiethnic between diabetes and dementia seems to individuals age 44–89 years was followed study of 1,088 subjects age ≥65 years be strongest for vascular dementia or AD in the Rancho Bernardo Study, a popula- in New York City found that diabetes dementia with vascular comorbidity, paral- tion-based study conducted in suburban was related to a higher risk of cognitive leling the findings for studies examining southern California. Diabetes, ascertained impairment-without dementia with cognitive performance and non-dementia by glucose tolerance testing, was associ- stroke, but not of cognitive impair- cognitive impairment as outcomes. One of ated with faster decline in a verbal fluency ment-without dementia without stroke, the most compelling reports documenting test (a test of executive-frontal abilities), after adjusting for demographic vari- the association between diabetes and but not a decline in the Mini Mental ables and the presence of the APOE-ε4 dementia came from a study of 2,067 Status Exam (a test of global cognition) allele (26). A Canadian prospective subjects age ≥65 years from Washington or Trail Making Tests (another test of study in 5,574 persons age ≥65 years State with repeated measures of glycemia executive-frontal abilities) (106). The similarly found that diabetes was related (114). In this study, increased glucose Atherosclerosis Risk in Communities Study only to VCI-without dementia (109). levels were related to a higher risk of (ARIC) was a prospective epidemiologic Another study in 918 subjects age dementia both among persons with and study of 10,963 persons age 47–70 years ≥65 years in New York City found that without diabetes. For example, for an in four U.S. communities (Washington diabetes was related to a higher risk of average glucose level of 190 mg/dL (10.55 County, Maryland; Forsyth County, North both amnestic and non-amnestic MCI, mmol/L) compared with 160 mg/dL (8.88 Carolina; Jackson, Mississippi; and highlighting the importance of diabetes mmol/L), there was a 40% increased risk Minneapolis, Minnesota). ARIC showed for both AD and VCI (110). A study of of dementia (Figure 24.5) (114). Some that diabetes, ascertained by self-report or 1,640 subjects age 70–80 years from studies have also reported a greater risk fasting glucose, was a strong predictor of Olmstead County, Minnesota, reported of dementia among persons with diabetes decline in the DSST of the Wechsler Adult that longer diabetes duration and treat- and the APOE-ε4 allele compared to Intelligence scale and the word fluency ment with insulin, a surrogate marker of persons with one of these risk factors or test, both tests of executive-frontal abili- diabetes severity, were related to higher others (86,115), suggestive of gene-envi- ties, after 6 years of follow-up (107). MCI risk (111). ronment interaction.
TABLE 24.1. Summary of Epidemiologic Studies From North America Relating Type 2 Diabetes and Dementia
YEARS (REF.) POPULATION SAMPLE STUDY CHARACTERISTICS STATISTICS 1970–1984 (153) Historical cohort in 1,455 persons Dementia ascertained by Persons with diabetes exhibited significantly increased risk of all Rochester, Minnesota with diabetes; medical record abstraction. dementia (RR 1.66, 95% CI 1.34–2.05) and Alzheimer’s disease age ≥45 years Standardized morbidity ratios (AD) dementia (for men, RR 2.27, 95% CI 1.55–3.31; for women, calculated using expected RR 1.37, 95% CI 0.94–2.01). age- and sex-specific dementia rates.
1991–1996 (26) Multiethnic cohort from 1,262 subjects Diabetes diagnosed by self- The adjusted relative risk of AD dementia among persons with New York City recruited without report or use of diabetes diabetes compared to those without diabetes was 1.3 (95% CI between 1992 and dementia at medications. Dementia and 0.8–1.9). The adjusted relative risk of stroke-associated dementia 1994 baseline; mean dementia subtype diagnosed in persons with diabetes was 3.4 (95% CI 1.7–6.9). age 75.6 years by standard research criteria.
1991–1992 (109) Canadian Study of 5,574 subjects; Diabetes diagnosed by self- Diabetes at baseline was associated with incident vascular dementia Health and Aging mean age report. Dementia diagnosed (RR 2.03, 95% CI 1.15–3.57), but not with mixed AD/vascular 74±6.4 years by research criteria. dementia (RR 0.87, 95% CI 0.34–2.21), or AD dementia (RR 1.30, 95% CI 0.83–2.03), or all dementias (RR 1.26, 95% CI 0.90–1.76). 1991 and 1994 Population-based 2,574 subjects; Diabetes ascertained by Diabetes was associated with total dementia (RR 1.5, 95% CI 1.01– (86) cohort of Japanese mean age 76.9 interview and direct glucose 2.2), AD dementia (RR 1.8, 95% CI 1.1–2.9), and vascular dementia American men enrolled years testing. Dementia assessed (RR 2.3, 95% CI 1.1–5.0). Individuals with both type 2 diabetes and in the Honolulu-Asia by research criteria. the APOE-ε4 allele had a relative risk of 5.5 (95% CI 2.2–13.7) for Aging Study AD dementia compared with those with neither risk factor.
Table 24.1 continues on the next page. 24–8 Diabetes and Cognitive Impairment
TABLE 24.1. (continued)
YEARS (REF.) POPULATION SAMPLE STUDY CHARACTERISTICS STATISTICS 1994–2003 (154) Religious Orders Study 824 nuns, Diabetes ascertained by self- Diabetes was related to higher risk of AD dementia (HR 1.65, 95% across the United priests, and report or use of medications. CI 1.10–2.47). States brothers; mean AD ascertained by research age 75 years criteria. 1964–1973 and Participants of a 8,845 subjects Diabetes ascertained when Diabetes was related to a higher risk of dementia (HR 1.46, 95% CI 1994–2003 (155) health maintenance participants were between 1.19–1.79). organization in ages 40 and 44 years. Northern California Dementia ascertained when participants were between ages 70 and 74 years. Both diabetes and dementia ascertained from medical records. 1992–2000 (115) Cardiovascular Health 2,547 subjects; Diabetes ascertained by Diabetes was only related to dementia (HR 1.44, 95% CI 1.03–2.01), Study mean age 74.67 fasting blood glucose AD dementia (HR 1.62, 95% CI 0.98–2.67), and mixed AD (HR 1.75, years or clinical history. AD 95% CI 1.00–3.04), but not vascular dementia (HR 0.80, 95% CI ascertained by research 0.30–2.09). Compared with those who had neither diabetes nor criteria. APOE-ε4, those with both factors had a significantly higher risk of AD dementia (HR 4.58, 95% CI 2.18–9.65) and mixed AD (HR 3.89, 95% CI 1.46–10.40). 1999–2007 (156) Cohort from New York 1,488 subjects; Diabetes ascertained by self- Diabetes associated with a higher risk of AD dementia (HR 1.6, 95% City recruited between mean age 76.0 report. Dementia ascertained CI 1.0–2.6) and vascular dementia (HR 5.4, 95% CI 0.3–95.0). 1999 and 2001 years by research criteria. 1994–2004 (114) Cohort from the Group 2,067 subjects; Glycemia measured Among participants without diabetes, higher average glucose levels Health Cooperative in mean age 76 with glucose or glycated within 5 years before baseline were related to an increased risk of Washington State years hemoglobin (A1c). dementia (p=0.01). For a glucose level of 115 mg/dL compared with 100 mg/dL, the adjusted hazard ratio for dementia was 1.18 (95% CI 1.04–1.33). Among participants with diabetes, higher average glucose levels were also related to an increased risk of dementia (p=0.002); for a glucose level of 190 mg/dL compared with 160 mg/dL, the adjusted hazard ratio was 1.40 (95% CI 1.12–1.76). Conversions for glucose values are provided in Diabetes in America Appendix 1 Conversions. AD, Alzheimer’s disease; CI, confidence interval; HR, hazard ratio; RR, relative risk. SOURCE: References are listed within the table.
FIGURE 24.5. Risk of Incident Dementia Associated With the Average Glucose Level During the Preceding 5 Years, According to the Presence or Absence of Diabetes, Adult Changes in Thought Study, 1994–2004
A. Participants without diabetes B. Participants with diabetes 2. 2. 1. 1. 1.6 1.6 1.4 1.4
1.2 1.2
1. 1. Hazard ratio Hazard raatio . .