Cognitive Dysfunction and Diabetes Mellitus

Abstract

The deleterious effects of diabetes mellitus on the retinal, renal, cardiovascular, and peripheral nervous systems are widely acknowledged. Less attention has been given to the effect of diabetes on cognitive function. Both type 1 and type 2 diabetes mellitus have been associated with reduced performance on numerous domains of cognitive function. The exact pathophysiology of cognitive dysfunction in diabetes is not completely understood, but it is likely that hyperglycemia, vascular disease, hypoglycemia, and insulin resistance play significant roles. Modalities to study the effect of diabetes on the brain have evolved over the years, including neurocognitive testing, evoked response potentials, and magnetic resonance imaging. Although much insightful research has examined cognitive dysfunction in patients with diabetes, more needs to be understood about the mechanisms and natural history of this complication in order to develop strategies for prevention and treatment.

• I. Introduction

• II. Cognitive Dysfunction in Patients with Diabetes

  • A. Type 1 diabetes
  • B. Type 2 diabetes
  • C. Hypoglycemia and cognitive dysfunction
  • D. Section summary
• III. Pathophysiology of Cognitive Dysfunction in Diabetes

  • A. The role of hyperglycemia
  • B. The role of vascular disease
  • C. The role of hypoglycemia
  • D. The role of insulin resistance and amyloid

• IV. Modalities for Assessment of Cognitive Dysfunction in Patients with Diabetes

• V. Future Directions

• VI. Conclusion

I. Introduction

DIABETES MELLITUS IS a complex metabolic disease that can have devastating effects on multiple organs in the body. Diabetes is the leading cause of end stage renal disease in the United States (1) and is also a common cause of vision loss, neuropathy, and cardiovascular disease. A less addressed and not as well recognized complication of diabetes is cognitive dysfunction. Patients with type 1 and type 2 diabetes mellitus have been found to have cognitive deficits that can be attributed to their disease. Both hypoglycemia and hyperglycemia have been implicated as causes of cognitive dysfunction, and many patients fear that recurrent hypoglycemia will impair their memory over time. Although much research has been done, the pathophysiology underlying this complication is not well understood, and the most appropriate methods to diagnose, treat, and prevent cognitive dysfunction in diabetes have not yet been defined. In this article, we will review the nature of cognitive dysfunction in type 1 and type 2 diabetes mellitus, the pathophysiology of cognitive dysfunction secondary to diabetes, methodologies used to assess cognitive deficits in patients with diabetes, and potential future directions of research that are needed to advance our understanding of this often overlooked complication of diabetes.

The purpose of this article is to present a comprehensive review of the literature regarding the subject of cognitive dysfunction in diabetes mellitus. To do this, we performed MEDLINE searches for such key words and terms as “diabetes mellitus,” “cognitive function,” “cognition,” “hypoglycemia,” “insulin resistance,” and “Alzheimer’s disease,” among others. We then pursued articles referenced in these sources. Although this is a comprehensive review, it is not exhaustive. In addition, it should be noted that the field of cognitive dysfunction in diabetes is still in its early stages. It must be remembered that although there have been many significant contributions regarding the association of diabetes and cognitive dysfunction and many hypotheses based on this association, the causative mechanisms of diabetes on cognitive dysfunction are still undergoing development.

II. Cognitive Dysfunction in Patients with Diabetes

A. Type 1 diabetes

Cognitive dysfunction in patients with diabetes mellitus was first noted in 1922, when patients with diabetes, who were “free from acidosis but usually not sugar free,” were noted to have impaired memory and attention on cognitive testing compared with controls (2). Since then, there have been many studies designed to better delineate the scope and magnitude of cognitive dysfunction in diabetes (Table 1). The most common cognitive deficits identified in patients with type 1 diabetes are slowing of information processing speed (3,4,5,6) and worsening psychomotor efficiency (3,4,7). However, other deficits have been noted, including deficits in motor speed (5,8,9,10), vocabulary (7,11,12,13), general intelligence (12,14), visuoconstruction (6,12), attention (6), somatosensory examination, motor strength (10), memory (7), and executive function (7,14). Glycemic control appears to play a role in cognitive performance in patients with type 1 diabetes. Functions such as psychomotor efficiency, motor speed (5,15), attention, verbal IQ scores (16,17,18), memory, and academic achievement (17) are improved with better glycemic control. Specifically, an 18-yr follow-up of the Diabetes Control and Complications Trial (DCCT) showed that those patients with type 1 diabetes mellitus with a time weighted mean glycated hemoglobin (HbA1c) less than 7.4% performed significantly better on tests of motor speed and psychomotor efficiency than those subjects whose time weighted mean HbAlc was greater than 8.8% (15). In addition, slowing of all cognitive function, an increased number of mental subtraction errors (19), loss of inhibition and focus (20), impaired speed of information processing, decreased attention, and impaired working memory (21) have all been noted during acute hyperglycemia in patients with type 1 and type 2 diabetes.

Table 1.

Summary of cognitive domains that have been found to be negatively affected by type 1 diabetes mellitus

Slowing of information processing*

Psychomotor efficiency*

Attention*

Memory

Learning

Problem solving

Motor speed

Vocabulary

General intelligence

Visuoconstruction*

Visual perception

Somatosensory examination

Motor strength

Mental flexibility*

Executive function

Domains marked by asterisks have particularly strong supporting data. 

A recent meta-analysis included 33 studies examining cognitive function in adult subjects with type 1 diabetes mellitus (22). It found that there were significant reductions in overall cognition, fluid and crystallized intelligence, speed of information processing, psychomotor efficiency, visual and sustained attention, mental flexibility, and visual perception in subjects with type 1 diabetes compared with controls. There was no difference in memory, motor speed, selective attention, and language. All studies included healthy matched control groups and used reliable testing measures at normal blood glucose values. Most studies included in the meta-analysis controlled for depression; however, similar findings were seen in those studies that did not control for depression. It is unclear whether any of these studies controlled for other chronic diseases that could affect cognitive function. Worse cognition was associated with increased diabetes complications, but not with glycemic control in these populations. However, this last finding was confounded by the heterogeneity of how different studies defined “well” vs. “poorly” controlled diabetes (22).

As was demonstrated in the work of Brands et al. (22), cognitive function may be worse in patients with type 1 diabetes who experience other diabetes complications. Deficits in fluid intelligence, information processing, attention, and concentration have been associated with the presence of background retinopathy (23). Proliferative retinopathy, macrovascular complications, hypertension, and duration of diabetes were associated with poorer performance on tests measuring psychomotor speed and visuoconstruction ability in patients with type 1 diabetes (4,5,6). Patients with distal symmetrical polyneuropathy displayed worse cognitive function on most cognitive domains except for memory (5). However, other studies were unable to identify a relationship between impaired cognitive function and diabetic complications (24). Future study will be necessary to determine whether there is a link between complications and alterations in cognition.

Although complications like retinopathy and nephropathy usually require years of diabetes before becoming clinically apparent, the onset of cognitive impairment has been found to occur early in patients with type 1 diabetes. Deficits in cognitive function have been detected as early as 2 yr after diagnosis in children with type 1 diabetes, and these patients experienced less positive changes than controls over time in general intelligence, vocabulary, block design, speed of processing, and learning (12). Six years after diagnosis, these same subjects had impaired IQ, attention, processing speed, long-term memory, and executive function compared with controls (14). The age of onset of type 1 diabetes may also contribute to the presence of cognitive dysfunction, because those who developed type 1 diabetes at less than 4 yr of age had impaired executive skills, attention, and processing speed when compared with those that were diagnosed after 4 yr of age (14). Of note, chronic disease and time away from school (secondary to illness, etc.) were not controlled for in these studies.

Interestingly, several studies have shown patient gender to influence neurocognitive function in patients with type 1 diabetes mellitus. Skenazy and Bigler (10) found that men with type 1 diabetes had reduced performance on oscillation, strength grip, and somatosensory testing compared with male controls, and the magnitude of this difference was greater than that measured between women with type 1 diabetes and their gender-matched controls. In addition, a decline in verbal intelligence was seen in boys with type 1 diabetes between the ages of 7 and 16, which correlated with worse glycemic control. This was not seen in girls of similar ages (13). However, most human studies have not distinguished between genders when describing results of neurocognitive testing, and therefore more controlled analysis should be done before any conclusions are drawn.

Of note, the strength of these neurocognitive studies is variable. Covariates that could affect neurocognitive testing include age, education, sex, history of other chronic illnesses, psychiatric disorders, neurological disorders, substance abuse, absence from school, socioeconomic status, and hypoglycemia/hyperglycemia during testing. The reviewed reports controlled for at least some of these covariates, however most fail to control for all of them. For example, only two studies that have been mentioned (6,24) have reported controlling for hyperglycemia at the time of testing, which has been proven to affect cognitive function (see Section II.A). The cognitive domains that are affected by type 1 diabetes with the best evidence based on our review are indicated in Table 1 with an asterisk.

B. Type 2 diabetes

Patients with type 2 diabetes mellitus have also been found to have cognitive impairment (Table 2). Type 2 diabetes has been associated with decreases in psychomotor speed (25,26), frontal lobe/executive function (26,27,28), verbal memory (29), processing speed (29), complex motor functioning (26), working memory (27,28), immediate recall, delayed recall (30), verbal fluency (26,31), visual retention (32), and attention (33). The impact of these subtle neurocognitive deficits on the daily lives of patients with type 2 diabetes is not clear. Sinclair et al. (34) found that subjects with mini-mental status exam scores less than 23 fared worse on measures of self care and ability to perform activities of daily living. These subjects also displayed an increased need for personal care and increased rates of hospitalization when compared with controls. Patients with diabetes also have been found to have slower walking speed, lack of balance, and increased falls associated with type 2 diabetes, but whether the cerebral affects of diabetes contributed to these abnormalities is debatable (35). Complicating the impact of mild neurocognitive dysfunction secondary to diabetes on daily living is the observation that patients with diabetes are twice as likely to have depression (27,36), which will also negatively affect cognitive function and daily activities. Type 2 patients also have an increased incidence of Alzheimer’s disease (37,38,39,40,41,42,43,44) and increased incidence of vascular dementia (38,42,45). Recently, Bruce et al. (36) found that 17.5% of elderly patients with type 2 diabetes had moderate to severe deficits in activities of daily living, 11.3% had cognitive impairment, and 14.2% had depression.

Table 2.

Summary of cognitive domains that have been found to be negatively affected by type 2 diabetes mellitus

Memory*

Verbal memory

Visual retention

Working memory

Immediate recall

Delayed recall

Psychomotor speed*

Executive function*

Processing speed

Complex motor function

Verbal fluency

Attention

Depression

Domains marked by asterisks have particularly strong supporting data. 

Glycemic control appears to play a role in determining the degree of cognitive dysfunction detected in patients with type 2 diabetes, although this has not uniformly been observed (46). In a population of nearly 2000 postmenopausal women, Yaffe et al. (47) found that those with a HbA1c of more than 7.0% had a 4-fold increase in developing mild cognitive impairment. Grodstein et al. (30) found that elderly subjects who took oral diabetic medication, unlike those on insulin, had similar scores on general tests for cognition as subjects without diabetes. Other studies have demonstrated an inverse relationship between HbAlc and working memory (27,28), executive functioning (27), learning (26), and complex psychomotor performance (26,48) in patients with type 2 diabetes mellitus, supporting the hypothesis that worsening glucose control leads to worsening cognitive function much like with type 1 diabetes. Also similar to type 1 diabetes is the association between alterations in cognitive function in patients with type 2 diabetes and diabetes complications like peripheral neuropathy (28) and duration of type 2 diabetes (25,33).

Impaired glucose tolerance without diabetes is also a risk factor for cognitive dysfunction. Multiple investigations of patients with impaired glucose tolerance have shown them to have lower mini-mental status exam and long-term memory scores (49), impaired verbal fluency (31), increased Alzheimer’s dementia (39), and increased vascular dementia (38) compared with control subjects. These observations mirror the positive relationship found between hyperglycemia in patients without diabetes and cardiovascular disease (50,51,52). The pathophysiology of this relationship is unclear, and there is evidence that both hyperglycemia and other aspects of insulin resistance could contribute to this, which will be addressed later. Of note, however, not all studies found that patients with impaired glucose tolerance (33,53,54) or type 2 diabetes mellitus (54,55) perform worse than normoglycemic individuals.

However, like neurocognitive studies examining type 1 diabetes, the strength of these neurocognitive studies evaluating type 2 diabetes and impaired glucose tolerance is variable. Although most of these studies controlled for age, there was uneven control for other covariates including education, psychiatric disorders, neurological disorders, hyperglycemia and hypoglycemia during testing, and chronic illness. The cognitive domains that are affected by type 2 diabetes with the best evidence based on our review are indicated in Table 1 with an asterisk.

C. Hypoglycemia and cognitive dysfunction

Repetitive episodes of moderate to severe hypoglycemia have been implicated as one possible etiology of cognitive dysfunction in diabetes. This is significant because the risk of hypoglycemia increases as efforts to achieve the level of glycemia necessary to minimize the risk of developing the microvascular complications of diabetes are intensified (56,57,58). The reason for severe hypoglycemia secondary to intensive insulin management is complex and multifactorial, however the initial intelligence of patients with type 1 diabetes before intensive management does not predispose to more future hypoglycemia episodes, as shown by an analysis of data collected during the DCCT (59). During acute hypoglycemia episodes, it has been shown that performance on immediate verbal memory, immediate visual memory, working memory, delayed memory, visual-motor skills, visual-spatial skills, and global cognitive dysfunction are all impaired (60,61). In a more recent study, prospective memory (that is, “remembering to remember”) and immediate and delayed recall were both impaired secondary to hypoglycemia in patients with type 1 diabetes, suggesting that impairments with both recall and learning/consolidation occur during hypoglycemia (62). Interestingly, in some studies there was no difference in reaction time (63), memory (62), and overall cognitive performance (61) between hypoglycemia aware and unaware patients during hypoglycemic episodes, despite the fact that the glucose level at which the counterregulatory hormone response was elicited was higher in subjects with awareness of hypoglycemia.

Although cognitive impairment may occur during hypoglycemia, the effect of repetitive hypoglycemia on subsequent cognitive function during euglycemia is less clear. Studies have shown impaired verbal IQ scores (14,64), full scale IQ scores (14,20,64), attention (20), verbal skills (11), short-term memory, verbal memory (17), vigilance (65), and visual-spatial memory (8,18) in patients with a history of type 1 diabetes and severe hypoglycemia (defined as being associated with seizures, coma, or the need for external assistance), compared with patients with type 1 diabetes without a history of severe hypoglycemia. However, it is possible that some of these “abnormalities” could be the result of slower, more deliberate completion of the tasks without loss of accuracy (66). More recently, no association between multiple severe episodes of hypoglycemia and impaired cognitive function in patients with type 1 diabetes mellitus was found in an 18-yr follow-up of the DCCT (15). Although the DCCT follow-up has been regarded as a landmark study that provides reassurance to diabetologists and patients, it is important to recognize its limitations, including the fact that it did not randomize patients into a “severe hypoglycemia” group and a control group (because such randomization would be not only logistically impossible but also unethical). A lack of association between severe hypoglycemia and cognitive dysfunction was confirmed by other studies (23,68,69,70,71), as well as with a meta-analysis, which showed no association between hypoglycemia and cognitive function (22). Of note, however, most data analyzing the effects of hypoglycemia look at young to middle-aged patients; data regarding the impact of hypoglycemia on older individuals is lacking.

One possible reason that some studies found an association between frequent hypoglycemia and cognitive dysfunction and others did not is that the positive investigations may have included subjects with diabetes onset earlier in life. Patients with type 1 diabetes diagnosed at less than 5 yr of age may have more severe (often with seizures) and frequent hypoglycemia episodes than those diagnosed at ages older than 5 yr; these younger patients have been found to have worse cognitive dysfunction (9,18,72,73). The severity of the hypoglycemia as well as the susceptibility of young brains to injury may explain the discrepancy (9,74). Another explanation for discrepancy between reports is that subjects with more hypoglycemia may have overall tighter glycemic control, which may offset the neurocognitive damage from hypoglycemia. This was most likely the case in the population studied by Kaufman et al. (17) in which children with more frequent episodes of hypoglycemia (<70 mg/dl) actually had increased memory and verbal scores, as well as overall better academic achievement when compared with less well-controlled children with diabetes.

D. Section summary

Clearly, much research has been done on cognitive dysfunction in patients with type 1 and type 2 diabetes mellitus. Although results are not consistent and many different deficits have been identified, some conclusions can be drawn. In patients with type 1 diabetes mellitus, deficits in speed of information processing, psychomotor efficiency, attention, mental flexibility, and visual perception seem to be present, whereas in patients with type 2 diabetes, an increase in memory deficits, a reduction in psychomotor speed, and reduced frontal lobe/executive function have been identified. Severe hypoglycemic episodes may contribute to cognitive dysfunction in the young; however, as patients age episodes seem to have less of an influence. Finally, improved diabetes control and decreased diabetic complications seem to be associated with less cognitive dysfunction, although this association is clearer in patients with type 2 diabetes than with type 1 diabetes.

However, some questions remained unanswered. First, it is not clear whether cognitive impairments seen in neurocognitive testing result in meaningful deficits either socially or professionally. Given the subjective nature of assessing professional and social activities, it will be difficult to address this question. Second, although the data suggest that hyperglycemia contributes to cognitive impairment, the magnitude of this contribution and how hyperglycemic one must be to experience the ill effects of hyperglycemia on cognition are not clear. Lastly, it is unknown whether mild neurocognitive impairments will progress to overt dementia. Large randomized controlled trials such as the Epidemiology of Diabetes Interventions and Complications (EDIC) study/DCCT and the ongoing Action to Control Cardiovascular Risk in Diabetes (ACCORD) study should hopefully continue to address these last two questions.

III. Pathophysiology of Cognitive Dysfunction in Diabetes

The pathophysiology underlying the development of cognitive dysfunction in patients with diabetes has not been completely elucidated. Many hypotheses with supporting evidence exist, including potential causative roles for hyperglycemia, vascular disease, hypoglycemia, insulin resistance, and amyloid deposition (Fig. 1). Although further research into each of these candidate mechanisms is necessary, it may be that the cause of cognitive dysfunction in patients with diabetes will turn out to be a combination of these factors, depending on the patient’s type of diabetes, comorbidities, age, and type of therapy.

Figure 1.

Summary of possible mechanistic contributors to cognitive dysfunction seen in diabetes mellitus. Not all mechanisms are present in every patient.

A. The role of hyperglycemia

As reviewed in Sections II.A and II.B, hyperglycemia appears to be related to abnormalities in cognitive function in patients with both type 1 and type 2 diabetes. However, the mechanisms through which hyperglycemia might mediate this effect are less than clear. In other organs, hyperglycemia alters function through a variety of mechanisms including polyol pathway activation, increased formation of advanced glycation end products (AGEs), diacylglycerol activation of protein kinase C, and increased glucose shunting in the hexosamine pathway (75,76,77,78). These same mechanisms may be operative in the brain and induce the changes in cognitive function that have been detected in patients with diabetes.

It has long been known that hyperglycemia increases flux through the polyol pathway in nervous tissue. In the streptozotocin-treated rat (glucose concentration 27.4 ± 0.3 mmol/liter vs. 5.9 ± 0.1 mmol/liter in control rats), an increase in sorbitol was measured in cranial nerves, sciatic nerve, cerebral cortex, and retina. This accumulation was reduced significantly when the animals were treated with the aldose reductase inhibitor tolerstat (79). Another study looking at streptozotocin-treated rats (HbA1c 7.9 ± 0.3 vs. 3.3 ± 0.0 in control rats) also found that administering an aldose reductase inhibitor, sorbinil, reduced the accumulation of brain tissue sorbitol and corrected the reduced cognitive function normally seen in rats with hyperglycemia (80). Whether this pathway contributes to neurocognitive dysfunction in humans with diabetes is unknown.

The role of AGEs and receptors for AGE (RAGEs) in the development of cerebral complications of diabetes also remains uncertain. Diabetic mice (32% HbA1c vs. 12% in control mice) with demonstrated cognitive impairment have been found to have increased expression of RAGEs in neurons and glial cells and damage to white matter and myelin (78), suggesting a possible role of RAGEs in the development of cerebral dysfunction (81). In humans, patients with diabetes and Alzheimer’s disease have been found to have greater N-carboxymethyllysine (a type of AGE) staining on brain slices obtained postmortem than patients with Alzheimer’s disease alone (82). However, a second autopsy study failed to find a difference in the quantity of AGE-like glycated protein rich neurofibrillary tangles and senile plaques, like those seen in patients with Alzheimer’s disease, between human subjects with diabetes and controls (83). Experiments performed in animal models provide limited evidence to support the hypothesis that AGE-induced brain injury may be a mechanism through which hyperglycemia and diabetes alter cerebral function. In vitro the addition of AGEs to bovine brain microvascular endothelial cells up-regulates both tissue factor mRNA (which induces blood coagulation) and reactive oxygen species through a mechanism that is reversed with treatment by the free radical scavenger edaravone (84). In addition, in a rat model of focal cerebral ischemia, the infusion of AGEs increased cerebral infarct size, whereas the coadministration of aminoguanidine, an inhibitor of AGE cross-linking, attenuated the infarct volume (85).

Few investigators have examined the role of diacylglycerol activation of protein kinase C and increased glucose shunting in the development of cognitive dysfunction in diabetes. Brain expression of protein kinase C-α was shown to be significantly increased in one untreated diabetic rat model (approximate blood glucose, 15 mmol/liter) compared with the treated diabetic rat and control rats (approximate blood glucose, 6 mmol/liter) (86), but another study found no increase in protein kinase C activity in diabetic rats (approximate blood glucose, 17 mmol/liter) compared with controls (approximate blood glucose, 6.4 mmol/liter) (87). Support for the possible role of the hexosamine pathway comes from the interesting observation that cerebral chitin, an N-acetyl-glucosamine polymer produced via the hexosamine pathway, is increased in human subjects with Alzheimer’s disease on autopsy (88). If hyperglycemia from diabetes shunts glucose toward the production of chitin, it is possible that the accumulation of this molecule could contribute to abnormalities in cognition.

Hyperglycemia has also been proposed to cause end organ damage through increases in reactive oxygen species, in particular superoxide, which could then lead to increased polyol pathway activation, increased formation of AGEs, activation of protein kinase C, and increased glucose shunting in the hexosamine pathway (76). Using streptozotocin to induce diabetes in rats (blood glucose, 20.72 ± 2.25 vs. 6.04 ± 0.64 mmol/liter in control rats), Aragno et al. (89) found that RAGEs, galectin-3 (a proatherogenic molecule), and the polyol pathway activation all were increased in rat brains, whereas activity of the glycolytic enzyme glyceradehyde-3-phosphate dehydrogenase was decreased, indicating elevated superoxide levels. Nuclear factor κB transcription factors, a proinflammatory gene marker up-regulated by AGEs, and S-100 protein, a marker for brain injury that can bind to RAGEs, were both up-regulated in the hippocampus in this animal model, although the effect in other regions was not assessed. These data suggest that oxidative stress may trigger a cascade of events that lead to neuronal damage. Interestingly, dehydroepiandrosterone, an adrenal androgen and antioxidant, significantly reduces these changes, suggesting a potential therapy worthy of more investigation.

In addition to hyperglycemic-induced end organ damage, altered neurotransmitter function has been observed in diabetic models and may also contribute to cognitive dysfunction. In diabetic rats (blood glucose 28.6 ± 1.1 vs. 6.3 ± 0.2 mmol/liter in control rats), there is an impairment of long-term potentiation, defined as activity-dependent prolonged enhancement of synaptic strength, in neurons rich in receptors for the neurotransmitter N-methyl-d-aspartate (NMDA), which could contribute to learning deficits (90). Other, neurochemical changes have been observed, including decreased acetylcholine (91), decreased serotonin turnover, decreased dopamine activity, and increased norepinephrine (86,92) in the brains of animals with diabetes. Interestingly, these changes were all reversed with insulin. One proposed hypothesis is that the alternating high and low glucose levels seen in patients with poorly controlled diabetes may worsen neurotransmitter function (92).

B. The role of vascular disease

Patients with diabetes have a 2- to 6-fold increased risk in thrombotic stroke (41,93), and vascular disease has long been hypothesized to contribute to abnormalities in cognition in such patients. Autopsy studies on patients with long-standing type 1 diabetes have shown changes related to vascular disease, including diffuse brain degeneration, pseudocalcinosis, demyelination of cranial nerves and spinal cord, and nerve fibrosis (94,95). Thickening of capillary basement membranes, the hallmark of diabetic microangiopathy, has been found in the brains of patients with diabetes (96). Patients with diabetes have also been found to have decreased global rates of cerebral blood flow as measured using xenon, and the magnitude of reduction correlates with the duration of the disease. However, blood glucose levels were not controlled during the experiment (range, 3.1–21.2; mean, 8.8 ± 4.74 mmol/liter) (97). Interestingly, the rate of cerebral blood flow in patients with diabetes is similar to that found in Alzheimer’s patients with dementia (92). These observations in humans with diabetes are supported by studies in streptozotocin-treated rats with chronic hyperglycemia (mean plasma glucose, approximately 29 mmol/liter) (98). One can speculate that the decrease in cerebral blood flow, coupled with the stimulation of the thromboxane A2 receptor known to occur in patients with diabetes (92), could contribute to the inability of cerebral vessels to adequately vasodilate, which may in turn increase the likelihood of ischemia. The coexistence of ischemia and hyperglycemia may be particularly detrimental to the brain. Even modestly elevated blood glucose levels (greater than 8.6 mmol/liter) in humans during a cerebrovascular event correlates with poorer clinical recovery (99). One potential mechanism through which hyperglycemia could potentiate ischemic damage is lactate accumulation. Hyperglycemia provides more substrate for lactate to form, causing cellular acidosis and worsening injury (93). Another mechanism is the accumulation of glutamate in the setting of hyperglycemia and ischemia (100). Glutamate, an excitatory amino acid neurotransmitter, has been shown to cause neuronal damage in the brain (101).

Although the exact mechanism is not known, the lack of C-peptide in patients with type 1 diabetes may by itself worsen cognitive impairment through its actions on the endothelium. Evidence for this is suggested by a rat model (blood glucose approximately 23 mmol/liter) in which replacement of C-peptide to normal levels normalized cognitive function and reduced hippocampal apoptosis (102). The relevance to humans is uncertain, however, because humans with type 1 diabetes do not have hippocampal atrophy (103).

C. The role of hypoglycemia

As mentioned in Section II.C, whether repeated episodes of hypoglycemia contribute to cognitive dysfunction is controversial and most likely depends on the age of the patient. However, there is no argument that if severe hypoglycemia lasts for a very long time, brain damage and death can occur (93,104,105,106). Most endocrinologists have personal experience with patients who have experienced severe hypoglycemia (<2 mm) with little or no permanent consequence. This most likely is secondary to inaccuracies of glucometer at low blood glucose levels (107), inadequate time with severe hypoglycemia, or variations in patients’ glycogen stores. That said, it has been shown in animal models that after 30–60 min of blood glucose levels between 0.12 and 1.36 mmol/liter, neuronal necrosis occurs with accompanying extracellular increases in aspartate, alkalemia, and neuronal energy failure, ultimately leading to a flat electroencephalograph (EEG) (105,106). The cortex, basal ganglia, and hippocampus appear to be most vulnerable to hypoglycemia, with laminar necrosis and gliosis found in these regions on autopsies performed in human patients who died of hypoglycemia (104). Other human autopsy studies done after death secondary to hypoglycemia have shown multifocal or diffuse necrosis of the cerebral cortex and chromatolysis of ganglion cells (108). In animal models, hypoglycemia-induced damage seems to be selective to neurons with sparing of astrocytes and oligodendrocytes (92). Although counterintuitive, the time to neuronal death may be asymmetric between hemispheres in severe, prolonged hypoglycemia, making the differentiation of hypoglycemic brain damage from ischemia difficult on a clinical basis (105). Some have hypothesized that hypoglycemia-induced neuronal damage occurs as a result of overactivation of a subtype of the excitatory neurotransmitter NMDA receptor (109). Interestingly, there exists an NMDA receptor antagonist that has been shown to prevent neuronal necrosis, suggesting a potential therapy for hypoglycemia-induced brain damage (110). Such a therapy may be helpful in young children with type 1 diabetes who seem to be particularly susceptible to cerebral complications of hypoglycemia. There may also be a relationship to hypoglycemia during early nocturnal sleep, a time in which consolidation of memories occurs, and cognitive dysfunction. Compared with test outcomes after a night of sleep in euglycemia, human control subjects and subjects with type 1 diabetes exhibited impaired declarative memory (memory of facts) after undergoing a short, relatively mild hypoglycemic clamp (2.2 mmol/liter) during early sleep (111). However, no neurocognitive deficits were seen in several other studies in which nocturnal hypoglycemia was induced later during the sleeping period (112,113).

D. The role of insulin resistance and amyloid

Although the role of insulin on cerebral metabolism and function is still evolving, fascinating research has given us more insight into this field over the last 20 yr. Historically, the brain was thought to be an insulin-independent organ; however, many recent discoveries have questioned that notion. Insulin receptors and mRNA expression have been found to be widely distributed in rat brain using immunohistochemistry and in situ hybridization (114,115), respectively, including in the olfactory bulb, hypothalamus, hippocampus, cerebellum, piriform cortex, cerebral cortex, and amygdala. The insulin-responsive glucose transporter 4 (GLUT4) has also been found in select regions of the rat brain, including the pituitary, hypothalamus, and medulla (116). GLUT8, also known as GLUTx1, is also found in the rat brain, specifically in the hippocampus, hypothalamus, cerebellum, and brainstem (117). GLUT8 has similar properties to GLUT4 and is up-regulated in response to insulin in some (118) but not all murine tissues, including the brain (119). Despite the presence of insulin receptors and insulin-sensitive glucose transporters, the effect of insulin on cerebral glucose metabolism is still uncertain. Many laboratories, including our own, have failed to demonstrate an effect of insulin on cerebral glucose metabolism in humans (120,121,122). However, other laboratories using fluorodeoxyglucose positron emission tomography (PET) have found a significant increase in brain glucose metabolism in the setting of hyperinsulinemia in humans (123), an effect that is reduced in subjects with peripheral insulin resistance (124). Despite the ongoing controversy of the effect of insulin on cerebral glucose metabolism, there is a large and growing body of evidence that insulin resistance, long recognized as a factor contributing to the onset of type 2 diabetes, may play a role in the pathogenesis of Alzheimer’s disease.

The clinical diagnosis of Alzheimer’s disease is made in the presence of a significant gradual and progressive decline in memory with at least one other cognitive, social, or occupational disturbance (125). The incidence of Alzheimer’s disease has been found to be approximately 1.2- to 1.7-fold higher in patients with type 2 diabetes and insulin resistance compared with a control population in most (37,38,39,40,41,42,43,47), but not all (126,127,128), investigations. The reason for the discrepancy could be the populations studied. Data that support a relationship between Alzheimer’s and insulin resistance all examined older subjects ascertained from the general population. The reports that failed to find an association collected data from a more narrowly defined population, such as those with a high incidence of the apolipoprotein E type 4 (APOE-ε4) allele (128) or a high percentage of early-onset Alzheimer’s disease (126). Interestingly, it appears that type 2 diabetes is also more common in populations with Alzheimer’s disease (129). Whether the association between Alzheimer’s disease and patients with type 2 diabetes reflects the impact of poor metabolic control on brain function or the actual effects of insulin and insulin resistance on the brain is unclear. Patients with Alzheimer’s disease and normal glucose tolerance have a more robust insulin secretory response to an oral glucose load than controls, suggesting that they may have increased insulin resistance (130,131). Some have suggested that the insulin resistance occurs in the brain itself and have hypothesized that the desensitization of neuronal insulin receptors plays an important role in the development of sporadic Alzheimer’s disease (132). Such a concept is supported by observations that patients with Alzheimer’s disease have an elevated concentration of insulin in their cerebral spinal fluid under fasting conditions (131), along with an increase in insulin receptor density in the occipital region and a decrease in tyrosine kinase activity (which is downstream to insulin binding) in the temporal and occipital lobes compared with controls (133). However, others have found that patients with Alzheimer’s disease have a decrease in cerebral spinal fluid insulin levels, suggesting that there may be impaired insulin transport across the blood brain barrier or increased insulin catabolism that accounts for the impaired central insulin action (134).

The mechanisms through which insulin resistance might alter cognitive function remain uncertain, but effects on neurotransmission and memory formation have been hypothesized. An impairment in central cholinergic activity is believed to contribute to the pathogenesis of Alzheimer’s disease (135), and interestingly, rats with streptozotocin- induced diabetes have a decrease in the production and release of acetylcholine compared with control rats (91). Mice models in which cholinergic activity is blocked by scopolamine experience amnesia and hyperactivity, a deficit that can be reversed by glucose administration (136,137). Glucose administration with a rise in endogenous insulin levels or insulin administration to patients with Alzheimer’s disease have also been shown to alter behavior, perhaps by enhancing cholinergic activity (138). In these studies, patients with Alzheimer’s demonstrated an improvement in declarative memory during either a hyperglycemic or a hyperinsulinemic euglycemic clamp (138,139). Furthermore, patients with memory impairment and Alzheimer’s disease had improved verbal memory acutely after intranasal insulin administration, which had no effect on peripheral glucose or insulin levels but had previously been shown to increase central nervous system insulin levels (140). In addition to affecting cholinergic activity, diabetes and insulin may affect long-term potentiation in opposing ways. Long-term potentiation is critical to the formation of memories and is induced by NMDA receptor activation, a process that is up-regulated in the presence of insulin (141). However, rats with diabetes, and presumed relative insulin deficiency, have decreased long-term potentiations in the hippocampus as measured by electrophysiology (90). As would be expected if long-term potentiation were reduced, rat hippocampal neurons exposed to insulin exhibited inhibition of spontaneous firing (142). Interestingly, patients with Alzheimer’s disease have a reduced cerebral glucose uptake as measured by PET (143,144,145) and have a reduced number of glucose transporters in the brain microvessels, frontal cortex, hippocampus, caudate nucleus, parietal, occipital, and temporal lobe compared with controls on autopsy studies (146,147). Perhaps the reduction in glucose uptake has a direct effect on how insulin regulates hippocampal function in these patients. Future experiments to identify the relative roles of glucose and insulin in human cognition are necessary to clarify these relationships.

The impact of insulin on cognitive function has also been examined in control subjects without Alzheimer’s disease or diabetes mellitus. In our laboratory, we found that inducing hyperinsulinemia using an insulin infusion in control subjects reduces parietal region P300 amplitude secondary to memory triggers (148). Other clamp studies found improved vigilance, memory, and selective attention in the setting of hyperinsulinemia (149,150), whereas intranasal insulin treatment for 8 wk improved delayed recall, enhanced mood, and self confidence and reduced anger in nondiabetic, nondementia subjects (151). Based on these studies, it is hypothesized that in Alzheimer’s disease, cerebral insulin resistance requires higher levels of insulin to facilitate memory (46,152). Although cerebral insulin is higher in these patients, it may not be enough to compensate for the insulin resistance. However, this does not necessarily prove that hyperinsulinemia directly improves cognitive function. Hyperinsulinemia can stimulate epinephrine release, and both insulin and epinephrine have been shown to increase lactate (153). Lactate can then in turn be used as a source of energy in brain metabolism (154,155), although the benefits of lactate therapy have not been proven to be beneficial yet in Alzheimer’s patients (156).

Insulin resistance and type 2 diabetes mellitus may contribute to cognitive dysfunction through three other indirect mechanisms. First, cognitive dysfunction in patients with type 2 diabetes has been correlated to inflammatory markers, and increased inflammation may contribute to the development of Alzheimer’s or macrovascular disease. In one investigation, patients with the metabolic syndrome, elevated C-reactive protein, and elevated IL-6 were found to have impaired cognitive function, whereas those patients with the metabolic syndrome and normal levels of these inflammatory markers had similar cognition to controls (47). Patients with type 2 diabetes are known to have higher levels of inflammatory markers including C-reactive protein, α- 1-antichymotrypsin, IL-6, and intercellular adhesion molecule 1 than control populations (157). These findings raise the possibility that insulin resistance and Alzheimer’s disease share a common pathophysiology, because patients with Alzheimer’s disease demonstrate increased inflammatory markers as well (158,159).

A second potential mechanism through which insulin resistance and type 2 diabetes could contribute to cognitive dysfunction is through the disruption of the hypothalamic-pituitary adrenal axis. Both animals (160) and humans (161) with type 2 diabetes have an up-regulation of the hypothalamic-pituitary-adrenal axis, with increased serum cortisol compared with controls. In other research, hypercortisolemia has been found to cause cognitive dysfunction. Healthy humans treated with dexamethasone (162), corticosterone (163), and hydrocortisone (164) to mimic stress conditions all performed worse on memory testing. In a study of healthy elderly patients, those with higher serum cortisol levels performed more poorly on memory and attention testing (165). In addition, patients with Cushing’s disease have been found to have worse performance on memory, attention, reasoning, and concept formation testing compared with controls (166), which may be attributed to a significant reduction in cerebral glucose metabolism found on PET scan in those patients with Cushing’s disease (167). Supporting these findings are the animal studies in which glucocorticoids cause structural damage and reduce function of neurons in the hippocampus (168,169,170,171,172). Based on the facts that type 2 diabetes causes an up-regulation of the hypothalamic-pituitary-adrenal axis and hypercortisolemia can cause cognitive dysfunction, it can be hypothesized that the increase in cortisol levels seen in patients with type 2 diabetes may contribute to cognitive dysfunction.

The third potential mechanism through which insulin resistance may indirectly contribute to cognitive dysfunction is by promoting the formation of senile plaques found in Alzheimer’s disease. Intracellular neurofibrillary tangles and extracellular senile plaques composed of β-amyloid are the pathological hallmarks of Alzheimer’s disease (173,174,175). β-Amyloid is formed from the cleavage of amyloid precursor protein (APP), produced in neurons (176), by the enzymes β- and γ-secretase (177). β-Amyloid is eventually degraded by the insulin-degrading enzyme (178,179). Amyloid β- peptides can by themselves bind to RAGEs and bring about microglial and neuronal dysfunction and oxidative stress (81). Interestingly, amyloid β-peptides, AGEs, and RAGEs have all been colocalized in astrocytes using immunohistochemistry in human brain slices (180). In addition, there is a growing body of evidence that insulin and insulin resistance can affect the metabolism of APP and β-amyloid, thus potentially increasing the burden of cerebral senile plaques. The role of insulin resistance in the metabolism of APP and β-amyloid was further clarified by Craft et al. (181). In their experiment, plasma levels of APP, the precursor to β-amyloid, were lower in those subjects with insulin resistance and Alzheimer’s disease when undergoing a hyperinsulinemic-euglycemic clamp. This corresponded with improved memory testing. One potential explanation of this observation is that insulin resistance may cause decreased APP degradation that can be overcome by the elevating serum and presumably tissue insulin levels (181). Similar findings have come from experiments in rat hippocampal neurons, where insulin was found to up-regulate insulin degrading enzyme, thereby increasing β-amyloid degradation (182). However, not all studies have agreed with this hypothesis. In a study using neuroblastoma cell lines by Gasparini et al. (178), insulin was found to decrease intracellular β-amyloid and increase extracellular levels of β-amyloid by both promoting its secretion and inhibiting its degradation via the insulin-degrading enzyme. This would contradict the majority of the evidence that insulin has a protective effect against memory loss. However, whereas it is widely believed that extracellular accumulation of β-amyloid plays a critical role in the development of Alzheimer’s, other evidence is suggesting there is a pathogenic role for intracellular β-amyloid (183,184,185). More research is needed concerning the pathophysiology of β-amyloid and insulin before conclusions can be drawn.

Of interest is the observation that the pancreatic islets in patients with type 2 diabetes mellitus are characterized by β-cell loss and deposition of islet amyloid (186), which is reminiscent of the neuronal loss and β-amyloid deposition seen in Alzheimer’s disease (187). The constitutions of islet and neural β-amyloid are similar (129,188), and both are toxic to islet and neurons, respectively (189,190). In a series of 29 patients in whom both brain and pancreas autopsy specimens were available, all had amyloid detected to some degree in both the brain and pancreas (187). In another study, islet amyloid was more abundantly present on autopsy in patients with Alzheimer’s disease than in those without Alzheimer’s disease (129). Based on the similarity between islet and neural β-amyloid, some have speculated that a shared pathogenesis may be present in patients with type 2 diabetes and patients with Alzheimer’s disease, possibly involving a defect in a chaperone protein (191) that helps intracellular protein trafficking (129). Rat models of type 2 diabetes (BBZDR/Wor), even more so than type 1 (BB/Wor) diabetes, demonstrate an increase in Alzheimer’s pathology, including increased APP, β-amyloid, and β-secretase and loss of neurons (192). Despite this compelling evidence linking insulin resistance and type 2 diabetes mellitus to Alzheimer’s disease and pathology, several autopsy studies performed in humans have failed to identify an increase in senile plaques or neurofibrillary tangles in subjects with diabetes compared with age-matched controls (83,129), although the duration of diabetes did correlate with the density of senile plaques in one of these studies (129).

The relationship between insulin resistance and cognitive dysfunction in Alzheimer’s disease appears to depend on the presence or absence of the APOE-ε4 allele. Curiously, although the presence of the APOE-ε4 allele is associated with an increased incidence of Alzheimer’s disease (193), it seems that insulin resistance is only a significant risk factor for Alzheimer’s disease in those patients without the APOE-ε4 allele (39,134). Patients with Alzheimer’s disease and no APOE-ε4 allele have been observed to have lower glucose disposal rates during a hyperinsulinemic-euglycemic clamp than subjects with Alzheimer’s disease and the APOE-ε4 allele, as well as those subjects without Alzheimer’s disease or the APOE-ε4 allele. Subjects with Alzheimer’s disease without the APOE-ε4 allele also had improved memory scores in the setting of hyperinsulinemia, which was not the case in APOE-ε4 allele-positive subjects (149,181). Based on this information, it seems that insulin resistance/type 2 diabetes and APOE-ε4 allele positivity are distinct and separate risk factors for the development of Alzheimer’s disease, a hypothesis that is supported by the fact that those with diabetes had a low incidence of the APOE-ε4 allele (128). However, there is again a contradiction in the literature; in the Honolulu-Asia Aging Study, those subjects with both type 2 diabetes and the APOE-ε4 allele had an additive increased risk of dementia and Alzheimer’s pathology (43). This study was specific for elderly Japanese-American men, so it seems that additional multiethnic studies are needed to understand this discrepancy better.

The association between insulin resistance and Alzheimer’s disease has been sufficiently compelling for investigators to examine whether peroxisome proliferator-activated receptor-γ (PPAR-γ) agonists can treat Alzheimer’s disease in the absence of diabetes. To date, two trials have demonstrated rosiglitazone to have a beneficial effect on memory in patients with Alzheimer’s disease. In a small randomized study published by Watson et al. (194) in 2005, patients with mild Alzheimer’s disease treated for 6 months with rosiglitazone had better memory and selective attention than controls. A much larger study published in 2006 found that patients with Alzheimer’s disease without the APOE-ε4 allele had improvements in cognitive testing after 6 months of rosiglitazone, whereas those Alzheimer’s disease patients with the APOE-ε4 allele did not have improvements (195). Multiple mechanisms have been proposed to address how PPAR-γ agonists may affect the pathophysiology responsible for Alzheimer’s disease, including reducing serum glucocorticoids (196), decreasing glial inflammation (197,198), protecting against β-amyloid-induced neurodegeneration (199), decreasing β-amyloid production (197), increasing β-amyloid degradation (196,200), and decreasing phosphorylation of tau proteins, the mechanism by which neurofibrillary tangles are formed (201). Interestingly, despite all of these data, it is still not totally clear how rosiglitazone benefited patients with Alzheimer’s disease in clinical trials because it has been shown not to cross the blood-brain barrier (196). These data are compelling enough to warrant further, longer-term studies on the benefits of PPAR-γ agonists in the treatment of Alzheimer’s disease. However, clinicians must weight the benefits against the newly documented cardiovascular risks of these treatments (202,203).

IV. Modalities for Assessment of Cognitive Dysfunction in Patients with Diabetes

Although progress is being made, the difficulty of detecting neurocognitive dysfunction in patients with diabetes in the clinical setting may explain in part why the field of cognitive dysfunction in diabetes has not advanced similarly to other fields dealing with hyperglycemia-associated end organ damage. Neurocognitive testing in which an examiner administers a battery of tests to assess different aspects of cerebral function has long been the gold standard for the assessment of neurocognitive function. Although cumbersome to administer and score, it has been very useful in assessing neurocognition in a variety of disease states, including diabetes, as was demonstrated in Section II. However, such tests have a relatively high rate of intrasubject variability that reduces their ability to identify mild deficits or preclinical disease. Also, many studies examining the effect of diabetes on brain function use multiple neurocognitive tests that assess the same psychological process. When the results of these different tests don’t agree, determining which results to base conclusions on can be confusing (204). In addition, not all neurocognitive tests are created equal. Although many neurocognitive tests are well validated in a diverse population to distinguish between “normal” and “abnormal” results, other tests do not have adequate reliability data, are based on unacceptably small norms, are administered inappropriately, or do not properly distinguish between two or more diagnostic groups (205). Finally, neurocognitive testing is unable to provide specific information about the neural structures responsible for any dysfunction identified. For example, although it appears that white matter function such as processing speed, attention, and visual-spatial processing are particularly affected by diabetes (4), localization of this dysfunction to white or gray matter is not possible using the battery of tests available to assess neurocognition.

Because of the limitations in neurocognitive testing, a number of modalities have been used to assess cognitive function in patients with diabetes (Table 3). One of the oldest modalities has been to measure electrical activity such as evoked response potentials in the brain after the administration of different stimuli. Abnormal evoked response potentials can reveal subclinical sensory nerve conduction deficits that may not otherwise be apparent (206). For example, flash electroretinography has shown decreased potentials from the retina in diabetic subjects before ophthalmoscopic signs of retinopathy were seen (108). In addition, pattern electroretinogram, which looks at the pattern of retinal stimuli originating from the ganglion cells, is also decreased in patients with diabetes (108). The evoked response of nerves involved in sensing auditory stimuli is also abnormal; brainstem auditory-evoked potentials demonstrated acoustic pathway impairment in patients with diabetes (207,208,209). Evoked potentials related to memory may also be affected because auditory P300 event-related potentials had significantly longer latencies in patients with type 2 diabetes compared with controls, which could relate to attention and short-term memory defects (210). Central somatosensory-evoked potentials were found to be prolonged in patients with diabetes as well (211). In another study looking at both type 1 and type 2 diabetes, slowed latency in visual and somatosensory-evoked potentials was observed in patients with type 1 diabetes, whereas patients with type 2 diabetes had slowed latency of visual, somatosensory, and brainstem auditory-evoked potentials (212). In this investigation, increasing HbA1c was related to reduced cognitive performance. Event-related potentials have also helped define brain adaptations to hypoglycemia. During hypoglycemia, normal subjects do not experience a delay in initial perception and precognitive processing, but they do have a delay in central processes such as stimulus selection and selective central motor activation (213). Of note, although all except one of these studies controlled for hypoglycemia during testing (211), none of the studies adequately controlled for hyperglycemia during testing, although Kurita et al. (210) found no difference between those subjects with high and relatively normal blood glucose values.

Table 3.

Summary of modalities for assessment of cognitive dysfunction in patients with diabetes

Neurocognitive testing

Evoked response potentials

EEG

MRI

fMRI

SPECT

PET

EEG can also assess spontaneous cerebral electrical activity and has been used in patients with type 1 and type 2 diabetes. Patients with type 2 diabetes have been found to have slowing in the EEG frequency band analysis over the central cortex area and reduction of alpha activity over the parietal area. These findings correlated with reduced visual retention on neurocognitive testing but were not simply related to hyperglycemia because making nondiabetic subjects with hyperglycemia did not reproduce these findings (32). Subjects with type 1 diabetes have also been found to have abnormal EEG results compared with controls, with those patients with a history of having severe hypoglycemia having the most abnormalities (65,214,215).

Magnetic resonance imaging (MRI) has been used in a number of studies to examine cerebral structure in patients with type 1 and type 2 diabetes and has pretty consistently found the brains of such subjects to have leukoariosis, which are hyperintense white matter lesions (207,216). One study found that 69% of middle-aged adults with long-standing type 1 diabetes had abnormal MRI scans, compared with 12% of healthy, aged-matched volunteers, with an increased number of larger, high-signal lesions in the cerebrum, cerebellum, and brain stem being the primary abnormality identified (207). However, this was not confirmed by a more recent published study in which relatively young patients (25–40 yr old) with type 1 diabetes for more than 15 yr did not have a significant difference in white matter hyperintensities compared with healthy controls. In addition, white matter hyperintensities did not correlate with depressive history, retinopathy, severe hypoglycemia, glycemia control, and most neurocognitive tests (with the exception of delayed memory) (7). This is in agreement with previous studies (3,217). The reason for the discrepancy may have been that subjects in the former study had more severe microvascular complications and that differences in cardiovascular risk factors between subjects with diabetes and controls were not controlled for. In patients with type 2 diabetes, these white matter hyperintensities have been noted to correlate with reduced performance on tests of attention, executive function, information processing speed, and memory (216,218). The nature of these white matter lesions is uncertain, but investigators have hypothesized that they could represent demyelination, increased water content, angionecrosis, cystic infarcts, or gliosis (i.e. brain tissue scarring) (92).

MRI has also demonstrated that subjects with type 2 diabetes have hippocampal and amygdala atrophy relative to control subjects (219). The hippocampus and amygdala are responsible for such functions as memory and behavior and, interestingly, are also found to be atrophied in Alzheimer’s patients (219). However, a similar study in subjects with type 1 diabetes failed to identify reductions in hippocampal and amygdala volume, although these subjects did have an increase in cerebrospinal fluid on MRI, suggesting mild global cerebral atrophy (103). Another study compared MRI findings and neuropsychometric testing in patients with an early age of onset of type 1 diabetes (younger than age 7) and a later age of onset (7–17 yr old). Subjects with early onset disease had larger ventricular volumes and more prevalent ventricular atrophy than those with later onset, which corresponded to poorer intellectual and information processing ability (220). Atrophy in subcortical and periventricular areas has also been associated with reduced performance on memory tasks in patients with type 2 diabetes (216). A history of hypoglycemia appears to be related to an increase in cerebral atrophy (70).

Recently, voxel-based morphometry, in which differences in local characteristics of tissue are measured using MRI, has been used to evaluate both the gray and white matter of patients with type 1 diabetes. Musen et al. (221) found that compared with controls, patients with type 1 diabetes had lower gray matter density in certain areas of the temporal lobe, frontal lobe, and thalamus. They also found that higher HbA1c levels correlated with lower gray matter density in areas important for language, memory, and attention, whereas a history of severe hypoglycemia correlated with less gray matter density in the left cerebellar posterior lobe (221). Wessels et al. (222) performed a similar experiment but compared patients with type 1 diabetes and retinopathy, patients with type 1 diabetes and no retinopathy, and controls. They found that patients with type 1 diabetes and proliferative retinopathy had decreased gray matter density in the right inferior gyrus and right occipital lobe compared with those patients with diabetes and no retinopathy, as well as to controls. More recently, Wessels and colleagues applied voxel-based morphometry to white matter volumes. They found that those subjects with type 1 diabetes and proliferative retinopathy had significantly smaller white matter volume than subjects without diabetes, and that smaller white matter volume correlated with worse performances on attention, speed of information processing, and executive function. This was not seen in patients with type 1 diabetes who had no proliferative retinopathy, suggesting a possible common mechanism in the development of retinopathy and cerebral dysfunction (6). Preliminary findings by Perantie et al. (223) have found that a history of severe hypoglycemia in children is associated with decreased gray matter in the left superior temporal region, whereas increased HbAlc levels are associated with reductions of gray matter in the right cuneus and precuneus regions, reductions of white matter in the right posterior parietal region, and increased gray matter in the right prefrontal cortex.

Functional MRI (fMRI) has also been used to assess cerebral function in patients with diabetes. fMRI is based on the fact that increases in cerebral blood flow and metabolism during stimulus-induced neuronal activation are accompanied by a relative reduction in deoxyhemoglobin content of the activated tissue relative to the adjacent unactivated brain. Because deoxyhemoglobin is a paramagnetic molecule, it can be visualized by MRI. Rosenthal et al. (224) applied fMRI to the study of cerebral function during standard neurocognitive testing in subjects with type 1 diabetes subjected to both euglycemia and hypoglycemia. They found that the effect of acute hypoglycemia on cerebral blood flow is task and region specific. For example, during hypoglycemia, the slower finger tapping corresponded to decreased activation of the right premotor cortex, supplementary motor area, and left hippocampus and with increased activation in the left cerebellum and right frontal pole. In addition, during hypoglycemia deterioration of four-choice reaction time correlated with reduced activation in the motor and visual systems but with increased activation of the part of the parietal cortex involved in planning (224). More recently, Wessels et al. (24) applied this methodology to determine whether subjects with type 1 diabetes with known hyperglycemia-induced end organ damage, specifically retinopathy, had differing areas of cerebral activation with cognitive stimuli and hypoglycemia compared with patients with type 1 diabetes and no microvascular complications. Although there was no difference seen in cognitive ability between the two groups, there was an overall increase in activation and less appropriate deactivation of certain brain regions during hypoglycemia in the diabetic group with retinopathy. The investigators hypothesized that regional alterations in activation were secondary to hyperglycemia-induced end organ damage in the central nervous system, causing altered neurovascular coupling or functional microvascular alterations (24). Preliminary findings by Musen et al. (225) have found decreased activation in the superior temporal gyrus and the parahippocampal gyrus in response to lexical and recognition stimuli, respectively, in a group composed of patients with long-standing type 1 diabetes compared with controls.

Other imaging modalities such as single photon emission computed tomography (SPECT) and PET have been used to assess cerebral function in patients with diabetes mellitus. SPECT is particularly good at assessing cerebral perfusion and has demonstrated in an uncontrolled study that patients with type 2 diabetes and dementia have a high incidence of hypoperfusion in at least one area of the brain (226). However, SPECT has also demonstrated that patients with type 1 diabetes have hyperperfusion in the prefrontal and frontal brain regions compared with controls (227). Other investigators found that when the SPECT data are corrected for the increase in cerebral atrophy seen in patients with diabetes, cerebral blood flow and glucose metabolism values were within normal range (228). PET with fluorodeoxyglucose is a technique that can be used to assess glucose metabolism because the compound is taken up and trapped in the cell by phosphorylation. When this method was used in patients with type 1 diabetes and a history of severe hypoglycemia and hypoglycemia unawareness, no differences in glucose metabolism were found relative to controls, although neuropsychological testing was also not significantly different (69). Two pilot studies, one by our own laboratory, have used diffusion tensor imaging, a type of MRI that measures white matter integrity quantitatively by fractional anisotropy (229,230), in patients with diabetes. Preliminary findings show a reduction in white matter integrity in patients with type 1 diabetes that was associated with severity of hyperglycemia (231) and poorer performance on certain neurocognitive tests (67).

In general, assessment modalities to detect cognitive dysfunction associated with diabetes have been disappointing. Neurocognitive testing is cumbersome and lacks pathophysiological insight. Evoked response potentials and EEG seem to be good at detecting sensory/perception deficits but are also cumbersome and do not provide as much information about more complex cognitive functions. Although promising, there have been conflicting results with regard to MRI studies, specifically in patients with type 1 diabetes. The utility of SPECT, PET, and diffusion tensor imaging in monitoring or detecting changes in patients with cognitive dysfunction and diabetes remains to be determined.

V. Future Directions

Although much research has been done on the impact of diabetes mellitus on cognitive function, many questions still remain. It is clear that patients with type 1 and type 2 diabetes have been found to have abnormalities in neurocognitive function, although the natural history and clinical significance of these findings have not yet been clearly defined. For example, we know that there is an increased incidence of dementia in patients with type 2 diabetes (37,38,40,41,42,43,44), but we do not know whether the more subtle changes in memory and in other measures on neurocognitive testing are a precursor to true dementia or represent another process. We also do not know whether the incidence of dementia is increased in patients with type 1 diabetes compared with the rest of the population. In prior decades, this was not an issue because patients with type 1 diabetes died at relatively young ages from other complications of the disease. However, now that patients with type 1 diabetes are living longer and better with the disease, this must be assessed. Finally, it is not clear whether the subtle cognitive deficits identified in many studies truly impact the lives of patients living with diabetes (34). Future studies, perhaps longitudinal in nature or involving better serum/imaging biomarkers, will be of tremendous benefit in providing better understanding of the natural history of this complication.

Future study will also be important in understanding the pathogenesis of cognitive dysfunction secondary to diabetes. Although it seems that hyperglycemia and hyperglycemia-induced end organ damage contribute to this problem, the actual mechanisms through which hyperglycemia alters cerebral structure and function are not clear. Improved glycemic control is likely of therapeutic benefit, as has been suggested by many retrospective studies (5,15,16,17,18,26,27,28,48), but a prospective study is needed to determine whether this is true. In addition, identification of the mechanisms through which hyperglycemia may impair cognitive function in patients with diabetes will stimulate new research into ways to prevent and treat all of the hyperglycemia-associated complications of diabetes.

VI. Conclusion

In conclusion, there have been significant gains in our understanding of the effect of diabetes on cognitive dysfunction. Evidence from neurocognitive testing suggests that cognitive dysfunction should be listed as one of the many complications of diabetes, along with retinopathy, neuropathy, nephropathy, and cardiovascular disease. The pathogenesis of cognitive dysfunction is only partially understood. Although many studies suggest that changes in cerebral structure and function in diabetes are related to hyperglycemia-induced end organ damage, macrovascular disease, hypoglycemia, insulin resistance, and amyloid lesions may play a role in some patients. Greater understanding of the natural history of this diabetes complication and the mechanisms responsible for its development will continue to advance as biochemical and imaging modalities continue to evolve. As new knowledge is gained, it can be applied to develop new and improved ways to prevent and treat all of the hyperglycemia-related complications of diabetes.