Diabetes and the elderly brain: sweet memories?

Abstract

Type 2 diabetes is common in older people and is associated with higher risk of both vascular dementia and Alzheimer’s disease. This review examines the evidence for increased risk of dementia and mild cognitive impairment in patients with diabetes and the role of potential confounders. The relationship of diabetes and impaired fasting glucose with brain structure is also reviewed, focusing on longitudinal studies in older people. The pathophysiology underlying cognitive change in type 2 diabetes is examined with reference to vascular disease, hypoglycaemia, inflammation and insulin levels. Implications for clinical care in older people with diabetes are discussed, with a recommendation for cognitive evaluation as a routine part of end-organ, diabetes complication review.

Introduction

The ability to think clearly and make independent and appropriate decisions concerning self and others independently is something we usually take for granted. Many patients with diabetes, as they enter early old age, become concerned about their cognitive health. They often resort to the consumption of various supplements purported to slow cognitive decline, and may take up pursuits such as bridge and mathematical games as cognitive exercises to bolster neural activity and agility. Cognitive health of patients with diabetes is as much a concern for clinicians, not only because of the impact on a patient’s quality of life, but also because benchmarked diabetes care places expectations on the ability for self-directed care and diabetes decision making. It is timely to review the evidence for the extent by which diabetes affects cognition and brain structure in older people.

Type 2 diabetes mellitus (diabetes) is a common disease in older people, affecting about 30% of people aged 65–85 [Steinman et al. 2012]. High diabetes rates are due to the environmental exposures of overnutrition and sedentariness, but also ageing of the population. In parallel, the prevalence of complex diseases such as dementia has also risen [Alzheimer’s Disease International, 2011]. Advances in healthcare in developed countries have translated to longer survival of people with diabetes due to effective treatment of infectious disease, cardiovascular disease, some cancers and diabetic complications. The classic microvascular diabetic complications of nephropathy, retinopathy and peripheral neuropathy are well known outcomes of hyperglycaemia. Other clinically relevant complications such as macrovascular disease, diabetic cardiomyopathy and dementia impact substantially on health outcomes. More critically, these consequences of diabetes may overtake or dictate future clinical care priorities.

As a consequence, the natural history of type 2 diabetes in the twenty-first century developed world appears to be changing as more of the population reaches advanced years with long-term, treated diabetes that has benefited from improved interventions and vascular risk factor management. It is relevant to consider the intersection of diabetes (and its glycaemic, vascular, metabolic and inflammatory milieu) with the brain in the elderly. There is abundant evidence that diabetes detrimentally affects brain structure and function, across a spectrum from mild cognitive decline and impairment, through to dementia. Diabetes is currently considered to account for 6–8% of all cases of dementia in older people [Kloppenborg et al. 2008]. Increasingly, there have been calls to recognize the brain as a target of end-organ damage in diabetes [Strachan, 2011] and for recognition of a central diabetic encephalopathy or neuropathy [Sima, 2010]. This review summarizes the evidence for changes in brain function and structure in type 2 diabetes mellitus.

Diabetes and dementia

Studies published in the 1980s suggested that diabetes was less prevalent in Alzheimer’s disease [Bucht et al. 1983; Wolf-Klein et al. 1988], now explained by the likelihood of survival bias. A number of prospective observational studies since then have consistently reported an association between type 2 diabetes mellitus and dementia, showing that diabetes increased dementia risk 1.3–3.4 fold [Biessels et al. 2006; Strachan et al. 2011], as reviewed systemically elsewhere [Biessels et al. 2006]. It is relevant to note that differences in dementia risk have been found in observational studies, depending on whether the cohort was in mid life or older and the type of dementia examined [Biessels et al. 2006].

As diabetes is accompanied by increased vascular risk, it is relevant to focus on those longitudinal studies that have examined diabetes and dementia risk in older people but also included vascular risk factors as covariates. Representative studies that illustrate dementia risk after adjustment of important vascular risk factors known to increase dementia risk are summarized in Table 1. In evaluating the results of these studies as a whole, it is important to note the methods of case ascertainment for both diabetes and dementia. All studies assessed for dementia at baseline and follow up, in one instance by medical record examination [Hassing et al. 2002] and in others by initially screening for cognitive impairment and then performing a diagnostic assessment [Luchsinger et al. 2001; MacKnight et al. 2002; Peila et al. 2002; Xu et al. 2004]. In all studies, dementia diagnosis was confirmed by consensus within a central committee of experts. Consistent diagnostic criteria for dementia were applied from either the Diagnostic and Statistical Manual of Mental Disorders, 3rd edition [American Psychiatric Association, 1980] or the fourth edition [American Psychiatric Association, 1994]. A diagnosis of Alzheimer’s disease was based on the National Institute of Neurological and Communicative Diseases and Stroke and the Alzheimer’s Disease and Related Disorders Association [McKhann et al. 1984] and vascular dementia by the National Institute of Neurological and Communicative Diseases and Stroke and the Association Internationale pour la Recherche et L’Ensignment en Neurosciences criteria [Roman et al. 1993]. Diabetes was assessed by fasting glucose and oral glucose tolerance testing in only one study [Peila et al. 2002]; one study performed random glucose levels [Xu et al. 2004] and three studies relied on history and medications for diabetes ascertainment [Hassing et al. 2002; Luchsinger et al. 2001; MacKnight et al. 2002]. With undiagnosed diabetes estimated at 30% in older people [Thomas et al. 2005], misclassification of undiagnosed diabetes into the control group might underestimate the impact of diabetes on dementia risk. Further, dropout rates were substantial in these studies, being up to 39%. Even with these limitations, these studies consistently show a 2.0–3.4-fold increased risk of vascular dementia and a 1.8–2.0-fold increased risk of Alzheimer’s disease in older people with diabetes. For completeness, other longitudinal studies also show increased risk of dementia in diabetes, but they are not presented here because of a lack of vascular risk adjustment [Arvanitakis et al. 2004; Brayne et al. 1998; Ott et al. 1999; Yoshitake et al. 1995].

Table 1.

Type 2 diabetes and dementia risk in longitudinal observational studies, after adjusting for vascular risk factors.

Author	Source	Age at base (years)	Men (%)	N at FU (% diabetes)	FU	Diabetes method	Dropout rate (%)	All dementia risk* (95% CI)	Vascular dementia risk* (95% CI)	Alzheimer’s dementia risk* (95% CI)
Peila et al. [2002]	USA	77	100	2574 (35%)	2.9	FGL, OGTT	27	1.5 (1.0–2.2)	2.3 (1.1–5.0)	1.8 (1.1–2.9)
MacKnight et al. [2002]	Canada	76	39	5574 (9.1%)	5.0	History	39	1.3 (0.9–1.8)	2.0 (1.2–3.6)	1.3 (0.8–3.0)
Xu et al. [2004]	Sweden	81	25	1301 (8.7%)	4.7	RGL	0	1.5 (1.0–2.1)	2.6 (1.2–6.1)	1.3 (0.9–2.1)
Hassing et al. [2002]	Sweden	84	34	702 (15.4%)	6.0	History	0	1.2 (0.8–1.7)	2.5 (1.4–4.8)	0.8 (0.5–1.5)
Luchsinger et al. [2001]	USA	76	32	1138 (20.3%)	5.5	History	37	-	3.4 (1.7–6.9)	2.0 (1.4–2.9)

^*Risk following adjustment for vascular risk factors.

CI, confidence interval; FGL, fasting glucose level; FU, years of follow up; OGTT, oral glucose tolerance test.

Diabetes and mild cognitive impairment

Mild cognitive impairment (MCI) has been proposed as a predementia state that might identify individuals at increased risk for progression to dementia [Petersen et al. 2009]. Since first being proposed, there has been considerable debate and controversy about its definitional constraints and prognostic utility [Ritchie and Touchon, 2000; Whitehouse, 2007]. A number of studies suggest that diabetes is associated with a 1.5-fold increased risk of developing amnestic MCI or nonamnestic MCI [Luchsinger et al. 2007; Yaffe et al. 2006; Solfrizzi et al. 2004]. The difficulty in interpreting these studies with relation to diabetes is that MCI is likely to represent a number of heterogeneous disorders rather than a condition on a single disease-severity spectrum.

Diabetes and cognitive function

Consistent with observational studies showing increased risk of cognitive impairment, longitudinal studies have also shown that type 2 diabetes is associated with cognitive decline in some [Christman et al. 2011; Elias et al. 1997; Espeland et al. 2011; Fischer et al. 2009; Gregg et al. 2000; Grodstein et al. 2001; Hassing et al. 2004; Okereke et al. 2008; Yaffe et al. 2004, 2012] but not all studies [van den Berg et al. 2010]. A number of studies evaluating specific cognitive domains prospectively have demonstrated performance declines in important domains that might impact on independent diabetes self care, specifically memory [Elias et al. 1997; Grodstein et al. 2001; Hassing et al. 2004; Okereke et al. 2008], executive function [Christman et al. 2011; Fischer et al. 2009; Grodstein et al. 2001; Okereke et al. 2008; Yaffe et al. 2004], language [Grodstein et al. 2001, Okereke et al. 2008] information processing speed [Christman et al. 2011, Fischer et al. 2009; Gregg et al. 2000; Hassing et al. 2002], verbal memory [Espeland et al. 2011], and abstract reasoning. [Elias et al. 1997].

As a whole, these studies have substantial variations in cohort size, selection biases, the rigour of diabetes ascertainment and, as for dementia, possibly under represent the full impact of diabetes on cognition. Another significant limitation is the frequent lack of consideration of important covariates. For example, only selected studies adjusted for hypertension [Elias et al. 1997; Fischer et al. 2009; Gregg et al. 2000; Grodstein et al. 2001; Okereke et al. 2008], education [Elias et al. 1997; Gregg et al. 2000; Grodstein et al. 2001; Hassing et al. 2004; Okereke et al. 2008; Yaffe et al. 2004], lipids [Okereke et al. 2008], vascular disease [Elias et al. 1997; Gregg et al. 2000], and mood [van den Berg et al. 2010; Yaffe et al. 2004] and no published study has included all these important covariates into the analyses thus far.

Diabetes and brain structure

Structural brain changes are reported in observational studies of older people with type 2 diabetes. As might be expected with increased vascular risk, cross-sectional studies report more frequent structural brain lesions [Manschot et al. 2007; Tiehuis et al. 2008], greater cortical atrophy [Manschot et al. 2006; Schmidt et al. 2004; Tiehuis et al. 2008] and more numerate white matter hyperintensities [Manschot et al. 2006; Tiehuis et al. 2008] in people with diabetes compared with normoglycaemic controls. These studies have found relationships between these structural abnormalities and underlying hypertension and vascular disease [Manschot et al. 2007], but also diabetes-specific parameters such as diabetes duration and fasting glucose levels [Tiehuis et al. 2008]. Longitudinal observation studies are sparse and show a greater increase in lateral ventricular volume in diabetes at 4 years compared with controls [de Bresser et al. 2010]. Studies of brain volume have focused on total brain volume or surrogates; as yet, there are no published data on the potential effects of diabetes on brain regions associated with memory and executive function.

Impaired fasting glucose

The impact of prediabetic glucose disorders [impaired fasting glucose (IFG) and impaired glucose tolerance] on cognition is not clear. IFG in particular is common in older people. There are inconsistent data from longitudinal studies reporting declines in processing speed, executive function and verbal memory [Euser et al. 2010; Yaffe et al. 2004, 2007]. One study has examined the impact of fasting glucose over 4 years on hippocampal and amygdala volumes, restricted to blood glucose levels under 6.1 mmol/liter [Cherbuin et al. 2012]. High fasting glucose levels (in the range considered ‘impaired’) was associated with atrophy of these regions, accounting for up to 10% of the change, after accounting for the effects of a number of important covariates. Further studies are awaited to clarify whether IFG in older people adversely affects brain structure and function.

Pathogenesis

The specific mechanisms that mediate cognitive decline in diabetes are not clear. Macrovascular pathways may link hyperglycaemia, hyperlipidaemia, hypertension and low-grade systemic inflammation to structural changes, brain volume loss and cognitive decline, as has been proposed elsewhere [Kloppenborg et al. 2008; Reijmer et al. 2010; Strachan et al. 2011].

Additional microvascular pathways are also supported. Increasing severity of diabetic retinopathy has been cross-sectionally associated with lower verbal fluency, mental flexibility and processing speed in men but not women [Ding et al. 2010]. Retinopathy has also been associated with the 10-year decline in verbal fluency and processing speed, though this relationship was not restricted to diabetic retinopathy, but was also found with hypertensive retinopathy [Lesage et al. 2009]. The findings suggest that cerebral microangiopathy, as reflected in retinopathy, contributes to structural brain changes and ensuing cognitive decline.

The association of hypoglycaemia with cognitive decline and dementia risk has been examined, with inconsistent results. Severe hypoglycaemia was not associated with cognitive decline, but individuals with dementia appeared to be at risk of future severe hypoglycaemia in one study [Bruce et al. 2009]. In contrast, another study suggested hypoglycaemia frequency and severity appeared to increase risk of dementia [Whitmer et al. 2009]. These discordant results from two large type 2 diabetes cohorts indicate the care required in evaluating cohorts prospectively, since it might be expected that older individuals with baseline cognitive dysfunction may already be on a dementia trajectory and are more likely to make decisions that increase hypoglycaemia risk, such as missing a meal or making an inappropriate medication adjustment.

Genetic factors might impact or mediate the effect of diabetes on cognition. The apolipoprotein E ϵ4 genotype, known to increase the risk of Alzheimer’s disease, has been evaluated in several studies, with a diabetes–genotype interaction for cognitive decline suggested in some [Blair et al. 2005; Haan et al. 1999], but not consistently [Kanaya et al. 2004].

Inflammatory mediators are also implicated. Diabetes and obesity are associated with low-grade systemic inflammation. Inflammation increases risk of vascular disease, but may also have a direct effect on the pathogenesis of cognitive decline in diabetes, as suggested by increased inflammation in the brain in dementia [Halliday et al. 2000]. Circulating inflammatory cytokines have been implicated in cognitive decline [Schram et al. 2007]; however, few studies have examined the impact of inflammatory cytokines on cognitive decline in diabetes. One study reported that higher levels of C-reactive protein, interleukin 6 and tumour necrosis factor were associated with lower cognitive performance cross sectionally [Marioni et al. 2010]. Prospective studies examining the associations between circulating inflammatory markers and cognition in diabetes are awaited.

Other hormonal factors of interest in the pathogenesis of cognitive decline in diabetes include relative insulin deficiency [Sima, 2010; Craft and Watson, 2004] and glucocorticoid excess [Strachan et al. 2011]. Insulin is anabolic in most tissues. Early studies have shown that chronic hyperinsulinism downregulates insulin receptors at the blood–brain barrier, with reduced brain insulin transport [Wallum et al. 1987]. In rodents, insulin receptors are found in the hippocampus [Singh et al. 1997] and insulin enhances memory and learning [Park et al. 2000]. Readers are referred to comprehensive reviews of the role of insulin in the brain [Sima, 2010; Craft and Watson, 2004]. Perturbations in brain insulin levels, signal transduction and action in the brain may link with cerebral vascular dysfunction, but also with inflammation, cellular oxidative stress and, critically, impaired neuronal repair. Further research in this area is likely to inform the area of cognitive decline in diabetes, but will need to rely on animal studies.

Translation of the evidence to clinical practice

Diabetes care focuses on informed independent patient self care to achieve glycaemic targets and minimize vascular risk. Obviously, intact cognition is required to direct diabetes self care in the multiple comorbidities that often accompany diabetes. Annual screening for complications, the standard for diabetes care, does not evaluate cognition, even in older people in whom cognitive decline might be expected. The evidence presented above purports the question: should we consider another end-organ in diabetes complication assessment, the ‘brain’ or a ‘central neuropathy’? The evidence presented above indicates the brain is adversely affected in diabetes, not only by vascular disease and atherothrombotic events, but also by loss of brain volume longitudinally, associated with loss of cognitive abilities in executive function, information processing and memory.

The intactness of these important functional domains is critical for the ongoing safety of independent self care expected in people with type 2 diabetes. We expect our patients to make informed and immediate responses to blood glucose elevations and hypoglycaemia, often with decisions about medication or insulin dose adjustment. For example, ‘I have a virus and my sugars are high, I need to increase my medication or insulin dose.’ ‘I am low, I have checked my sugar to confirm this and need to eat 6 jelly beans, and then retest my sugar.’ We expect our patients to not only operate glucose monitoring equipment (‘where do I switch it on again?’) but problem solve when it malfunctions or readings appear inaccurate: ‘Was there enough blood on the strip?’ ‘Are the strips in date?’ ‘Is this a battery malfunction?’ We expect our patients to negotiate complex dietary instructions, discern increasingly opaque food labels and marketing claims: ‘It says low fat, so it must be ok.’ ‘It is low GI, so I can eat it freely.’ ‘This fruit juice says “No Added Sugar”, so it is fine to drink it all.’ These internal dialogues are commonly reported by patients and indicate that the terrain people with diabetes negotiate is fraught with decision-making pitfalls. Obviously sound and consistent diabetes education helps inform these decisions and may promote and maintain appropriate decision making in an assumed paradigm in which learning and decision making are retained. With the increased risk of cognitive impairment, can we assume that, even when self-care skill sets have been appropriate for some years, our patients will continue to execute these with equal accuracy as they age? Should we be assessing our patients’ diabetes care decision-making more formally? Should cognitive assessment form a part of the annual complication screening in people over 70 years? If so, is the Mini Mental State Examination [Folstein et al. 1975] or a similar screening instrument sufficient, or should more formal assessment be undertaken of cognitive domains involved in diabetes self care, such as executive function and memory? Depression, which is frequent in diabetes, requires active exclusion, but what is the optimal tool to be applied in clinical practice in the elderly?

If cognitive impairment or decline is detected, what intervention is required? Clearly optimizing vascular risk factor management is imperative, but so is reviewing the safety of self-directed diabetes care. This may be examined by at-consultation examination of decision making and establishing safe pathways, supported by education but perhaps supplemented with written instructions and visual supports. As yet, there are no randomized studies to show these strategies are effective or prevent errors in decision making, but they seem ‘common sense’. Medication regimens in diabetes might also need to be simplified to minimize medication-related hypoglycaemia risk and reduce the complexity associated with diet and equipment. Further, additional ‘one-to-one’ support services may be necessary, along with ad hoc and perhaps more frequent routine reviews. Family and patient advocates may also benefit from needs-specific diabetes education.

Randomized interventions of intensive glucose lowering have reported changes in cognition and brain structure over 40 months (ACCORD-MIND). While tight glycaemic control was associated with greater total brain volume at study end, there was no functional benefit, as assessed by measures of cognition [Launer et al. 2011]. Coupled with ACCORD’s unexpected findings of increased mortality with tight glycaemic control, it would not appear that tight glucose control to merely preserve brain volume (particularly without cognitive preservation) is justified.

Whether the loss of brain tissue and cognitive function are explained by direct toxic effects of glucose or is mediated by other metabolic, lipid or inflammatory markers that are altered in diabetes requires further study, to understand pathogenesis and how these effects might be attenuated. Until then, it is the responsibility of clinicians involved in the care of older people with diabetes to detect cognitive decline and make appropriate treatment regimen alterations to minimise harm.

Conclusion

Type 2 diabetes impacts adversely on cognition, at least in part due to macrovascular disease, microvascular changes, hyperlipidaemia, alterations in insulin homeostasis and inflammation. Diabetes care in older people requires monitoring for cognitive decline and modulation of diabetes self care and treatment targets as is appropriate for each individual’s situation. Our benchmark of care should ensure that treatment regimens minimize hypoglycaemia and vascular risk. The self-care regimen should be appropriately tailored to each individual’s cognitive strengths and sufficient clinical support should be offered. Clarification of the pathogenesis of cognitive decline in older people with type 2 diabetes will provide opportunities for minimization of the cognitive decline in this group.