Of the 232 studies surveyed, only 22 (9.5%) explicitly excluded individuals with T2DM(¹), and only 29 (12.5%) explicitly excluded individuals with hypertension(²). Thirteen studies—approximately 6%—excluded both T2DM and hypertension. Fourteen studies (6.0%) excluded individuals on antihypertensive medication(³), however few of these studies also clarified whether individuals were assessed for untreated hypertension and excluded, if present. Nineteen studies (8.2%) excluded medical illnesses, systemic illnesses, medical disorders or physical illnesses(⁴). This criterion implies that all medical conditions, including T2DM and hypertension, were excluded.

In contrast, almost half of the included studies (102; 44.0%) had exclusion criteria that made it difficult to determine whether T2DM was excluded(⁵), and 85 studies (36.6%) had exclusion criteria that made it difficult to determine whether hypertension and/or antihypertensive medications were excluded(⁶). These studies listed “major medical illnesses,” “significant medical conditions,” “serious systemic illnesses,” “conditions/medications interfering with cognitive and/or brain function,” “vascular disease,” “cardiovascular disease,” and/or “conditions/medications interfering with the fMRI signal” as exclusion criteria, or simply described their sample as “healthy.” There were also many studies that did not mention any selection criteria related to T2DM (80; 34.5%)(⁷) or hypertension (51; 22.0%)(⁸).

In addition, 26 studies (11.2%) included individuals with controlled hypertension(⁹), 8 studies (3.5%) included controlled and uncontrolled hypertension(¹⁰), 3 studies (1.3%) included individuals with controlled T2DM(¹¹), and 6 studies (2.5%) included individuals with controlled and uncontrolled T2DM in their healthy cohort(¹²). Figure 2 provides a visual depiction of these results.

The majority of studies (173; 75%) did not report how they conducted their medical screening(^a). Only 28 studies (12%) reported having screened subjects with physician-conducted medical examinations and/or laboratory testing(^b). Sixteen studies (7%) screened participants with telephone interviews, in-person clinical interviews, medical history, chart reviews, or a combination of these methods(^c). The remaining 15 studies (6%) used a self-report questionnaire to assess medical status(^d).

A survey of the 232 included studies found that just under one third (29%) acknowledged and/or addressed the potential influence of intersubject vascular variability on the reported results. Many excluded subjects with a high vascular burden by screening for white matter hyperintensities in the imaging data(^■). Others compared groups on vascular risk factors(⁺), compared outcome measures on hypertension status or antihypertensive treatment status(^♦), or attempted to control for health, blood pressure, and/or white matter hyperintensities in the reported associations (^❖). Several studies noted in their discussion the possibility that the reported results were influenced by vascular factors, or explained why they did not think this was an issue(^•). A few studies used the measured BOLD or PET signals to examine and account for individual differences in vascular health(^□); for example, by ensuring that groups were equated on BOLD signal variability, by comparing the temporal characteristics of the hemodynamic response curve across groups, with proportional scaling of the BOLD or PET signal, or by focusing on group by task interactions (instead of group main effects) or comparing within-subject task contrasts across individuals or groups to minimize any individual or group differences in vascular integrity.

There are rigorous ways to account for intersubject vascular variability, such as additional task data or an additional imaging contrast. Several studies included in the present review used arterial spin labeling (ASL) MRI (^▴) or PET (^▾) to measure resting cerebral blood flow and control for individual differences in perfusion. Three studies used a breath-hold task to index individual differences in cerebrovascular reactivity (^❍), and two studies included a low-level motor or baseline task to ensure that participants demonstrated an adequate hemodynamic response (^×).

Insulin is a peptide hormone that is critical for regulation of blood glucose levels. Binding of insulin to its receptors, found on nearly all cells throughout the body, facilitates the cellular uptake of glucose from the blood. When bound, insulin and insulin-like growth factor also activate complex intracellular signaling pathways that promote cell growth and survival, regulate glucose metabolism, and inhibit oxidative stress and apoptosis (for a review, see Nakae et al., 2001).

The defining characteristic of T2DM is peripheral insulin resistance, which occurs when cells in the body decrease their response to insulin stimulation. In the developing stage of this disease, the pancreas is able to produce enough insulin to overcome this resistance. This results in peripheral hyperinsulinemia, and blood glucose levels remain within the normal range. As the disease progresses, however, the pancreas can no longer keep up, and blood glucose levels begin to rise. When blood glucose levels are high even in the fasting state, T2DM is diagnosed.

Peripheral insulin resistance and hyperinsulinemia have a counterintuitive impact on insulin levels within the central nervous system. In the face of peripheral hyperinsulinemia, insulin transport across the blood brain barrier is effectively reduced, resulting in a brain hypo-insulinemic state (e.g., Heni et al., 2013). Low brain insulin levels and disrupted insulin signaling contribute to cognitive impairments directly, particularly in medial temporal lobe regions where insulin receptors are abundant (Convit, 2005; Craft, 2006). Indirectly, low brain insulin levels exacerbate amyloid beta (Aβ) and tau pathology, hallmarks of Alzheimer disease (AD). It is here that we see the link between T2DM and Alzheimer disease pathology: brain insulin deficiency results in the down-regulation of insulin degrading enzyme (IDE; Luchsinger, 2008), which also has a role in degrading Aβ (Carlsson, 2010). As a result, Aβ degradation is effectively reduced, contributing to its aggregation and amyloid plaque formation. Decreased brain insulin levels also suppress the enzymes involved in tau phosphorylation, contributing to the formation of neurofibrillary tangles (Akter et al., 2011). While the downstream impact of T2DM-mediated brain insulin deficiency and insulin resistance is more moderate than that associated with AD, the underlying pathogenic mechanisms are similar (Steen et al., 2005), and it has been proposed that AD is a form of diabetes mellitus that selectively affects the brain (T3DM; for discussion, see de la Monte and Wands, 2008). Given this, is not surprising that individuals with T2DM show a pattern of memory impairment, medial temporal lobe atrophy, and reduced hippocampal connectivity that is similar to the classic pattern of memory deficits, neurodegeneration, and network disruption in AD (e.g., Gold et al., 2007; Zhou et al., 2010; Baker et al., 2011; Cui et al., 2014).

When cells in the body become resistant to the effects of insulin, blood glucose levels rise, resulting in hyperglycemia. Endothelial cells are particularly vulnerable to the effects of hyperglycemia, because they are less efficient at reducing glucose uptake in the face of high blood glucose levels (Kaiser et al., 1993). Under such conditions, the resultant intracellular hyperglycemia induces an overproduction of reactive oxygen species in the mitochondria, which increases oxidative stress within the cell. This initiates a cascade of biochemical events that mediate much of the microvascular and macrovascular damage associated with T2DM including, but not limited to, increased intracellular formation of advanced glycation end-products (AGEs) and protein kinase C activation (Du et al., 2000; Nishikawa et al., 2000; Brownlee, 2005; Giacco and Brownlee, 2010; Johnson, 2012).

AGEs are formed during normal metabolism on proteins with slower rates of turnover, in almost all cells throughout the body. AGE accumulation over time is a major factor in normal aging; however, under hyperglycemic conditions, AGE production is exacerbated beyond normal levels. AGEs cause intracellular damage and induce apoptosis through a process called cross-linking (Shaikh and Nicholson, 2008). AGEs also contribute to oxidative stress, and themselves activate inflammatory signaling cascades (for a review, see Yan et al., 2008). Critically, under hyperglycemic conditions, the Aβ protein itself can act as an AGE (Granic et al., 2009), which enhances its own aggregation and further increases amyloid plaque formation.

Protein kinase C activation, on the other hand, affects a variety of changes in gene expression that culminate in vascular dysfunction. Production of nitric oxide (NO), a vasodilator, is decreased, and production of endothelin-1, a vasoconstrictor, is increased. As a result, blood vessels are less able to dilate to accommodate increased blood flow demand. Over time, chronic exposure to high concentrations of endothelin-1 and decreased concentrations of NO contribute to diminished vessel elasticity, and structural changes in the vessel wall that result in atherosclerotic plaque formation (Kalani, 2008).

In the brain, hyperglycemia-mediated macro- and microvascular damage reduces the delivery of nutrients and oxygen required to meet metabolic demands. Altered cerebral autoregulation has been observed in middle-aged adults with T2DM (Brown et al., 2008), and may be an early manifestation of microvascular disease (Kim et al., 2008). Older adults with T2DM show decreased blood flow velocity, increased cerebrovascular resistance, and impaired vasoreactivity (Novak et al., 2006). Over time, declines in cerebrovascular health and reduced perfusion of brain tissue lead to structural atrophy and altered brain function.

The cognitive profile of individuals with T2DM includes deficits in attention, processing speed, learning and memory, and executive function (e.g., Reaven et al., 1990; Brands et al., 2007; Yeung et al., 2009; Whitehead et al., 2011). Moreover, these individuals, and individuals with pre-diabetes (impaired glucose tolerance), show an accelerated trajectory of cognitive decline relative to that associated with healthy aging (Gregg et al., 2000; Fontbonne et al., 2001; Arvanitakis et al., 2004; Yaffe et al., 2004; Fischer et al., 2009; Nooyens et al., 2010; Espeland et al., 2011; for conflicting results, see van den Berg et al., 2010).

Cognitive deficits in T2DM have been linked to multiple disease-related processes, including: (i) poor glucose control (i.e., hemoglobin A1c [HbA1c]; Ryan and Geckle, 2000; Kanaya et al., 2004; Cukierman-Yaffe et al., 2009; Maggi et al., 2009; Luchsinger et al., 2011; Tuligenga et al., 2014; for conflicting results, see Christman et al., 2011), (ii) glucose intolerance (Rizzo et al., 2010; Zhong et al., 2012b), (iii) high peripheral AGE levels (Yaffe et al., 2011), (iv) high levels of inflammatory cytokines (Marioni et al., 2010), and (v) peripheral hyperinsulinemia and insulin resistance (Bruehl et al., 2010; Zhong et al., 2012a). Even in non-diabetic adults, poorer glucoregulation has been associated with deficits and/or declines in verbal memory, working memory, processing speed, and executive function (Dahle et al., 2009; Bruehl et al., 2010; Messier et al., 2010, 2011; Ravona-Springer et al., 2012).

The link between cognitive impairment and poor metabolic control may be largely mediated by the structural and functional brain changes that occur in the presence of chronic insulin dysregulation and hyperglycemia. Associations between glucoregulation, hypoperfusion in temporal regions, hippocampal atrophy, and memory impairment have been observed in T2DM (Gold et al., 2007; Last et al., 2007), and in non-diabetic adults with decreased peripheral glucose regulation (Convit et al., 2003), or high fasting plasma glucose levels within the normal range (Cherbuin et al., 2012; Kerti et al., 2013). In other studies of T2DM, cognitive deficits and structural brain atrophy were linked to cerebral hypoperfusion and altered vascular reactivity (Last et al., 2007; Brundel et al., 2012), and disrupted default-mode network connectivity was associated with peripheral hyperinsulinemia, insulin resistance, and white matter integrity (Musen et al., 2012; Hoogenboom et al., 2014). Regardless of the underlying cause, brain atrophy in T2DM is associated with poor cognition (Moran et al., 2013), and cognitive declines have been associated with progression of brain atrophy over time (van Elderen et al., 2010; Reijmer et al., 2011). Some studies suggest that structural changes may occur early in the course of T2DM; enlarged lateral ventricles, particularly within the frontal horns, have been observed less than a year after diagnosis (Lee et al., 2013), and middle-aged, as well as older adults with T2DM, show reduced prefrontal volumes (Bruehl et al., 2009) and generalized global atrophy (de Bresser et al., 2010; Kamiyama et al., 2010; Espeland et al., 2013).

Hypertension places enormous stress on the cerebral circulation (for a comprehensive review, see Pires et al., 2013). A hallmark of chronic hypertension is increased vascular resistance, particularly in the small blood vessels that perfuse the brain. Vascular resistance increases as vessel walls thicken. This remodeling is an adaptive response required to maintain chronically increased blood pressure, but it decreases the interior space of the blood vessels (the lumen). Vascular resistance also increases as the number of blood vessels decrease. Rat models of hypertension have shown both of these effects: reductions in lumen diameter and in the number of capillaries making up capillary beds in the cerebral vasculature (Sokolova et al., 1985).

Blood flow is reduced when vascular resistance is high, and chronic hypertension-mediated hypoperfusion has been linked to white matter degradation, gray matter atrophy, and cognitive deficits. Studies of older adults with hypertension show reduced blood flow, particularly in occipito-temporal, prefrontal, and medial temporal lobe regions (Beason-Held et al., 2007), positive correlations between blood pressure and white matter burden (White et al., 2011; Raji et al., 2012), and negative correlations between blood pressure and total brain volume (Nagai et al., 2008). Blood vessel function is also impacted by hypertension. Cerebral autoregulation (i.e., the ability to maintain a constant perfusion rate over a range of arterial pressures) is impaired, as is cerebrovascular reactivity, the ability of blood vessels to dilate to accommodate increased blood flow demand (Last et al., 2007; Hajjar et al., 2010).

The cognitive profile of older adults with hypertension includes poorer performance on tests of executive function, including verbal fluency, Trails B-A switching score, Stroop interference scores (Bucur and Madden, 2010), slowed processing speed (Dahle et al., 2009), and deficits in attention and memory (see Gifford et al., 2013 for a meta-analysis). Prospective cohort studies show that midlife cardiovascular risk factors like hypertension predict cognitive impairment in later life (e.g., Virta et al., 2013), and, similarly, cross-sectional studies show a relation between higher systolic blood pressure and poorer cognitive performance, even within the normotensive range, a relation that is particularly strong in midlife (e.g., Knecht et al., 2008, 2009). Hypertension is associated with decreases in cognitive reserve (Giordano et al., 2012), and older adults with MCI and cardiovascular risk factors like hypertension are twice as likely to develop dementia compared to those without such risk factors (Johnson et al., 2010; Ettorre et al., 2012). Moreover, cognitive declines may be faster in those with MCI and hypertension, compared to those without hypertension (Li et al., 2011; Goldstein et al., 2013).

The association between hypertension and cognitive decline appears to be strongest in executive and processing speed domains, and weakest in memory and language domains. Hypertension increased the risk of non-amnestic MCI, but not amnestic MCI, regardless of APOEε 4 genotype or hypertensive medication status (Reitz et al., 2007), and predicted progression to dementia in non-amnestic MCI, but not amnestic or multi-domain MCI (Oveisgharan and Hachinski, 2010). The impact of hypertension on executive and processing speed domains is consistent with studies that show a positive relation between hypertension and white matter changes (Kennedy and Raz, 2009; Raz et al., 2012), and between white matter changes and deficits in processing speed, executive function, and attention, but not memory (e.g., Debette et al., 2011).

Cognitive deficits in hypertensive adults are linked to various indicators of vascular and brain health. There are correlations between white matter integrity and performance on tests of executive function and attention (Hannesdottir et al., 2009), and between decreased flow-mediated dilation and poorer executive function (Smith et al., 2011). Deficits in attention and psychomotor speed in late middle-aged adults with hypertension are associated with reductions in global brain perfusion, reductions that were not fully ameliorated following 6-months of antihypertensive treatment (Efimova et al., 2008). Global cognitive decline has been linked to reduced cerebral blood flow in the face of white matter lesions and lacunar infarcts (Kitagawa et al., 2009), to higher pulse pressure and arterial stiffness (Scuteri et al., 2007; Waldstein et al., 2008; Triantafyllidi et al., 2009), and to hypertension-mediated deep-brain vascular pathology (Yakushiji et al., 2012). In another large study of patients with MCI, those with hypertension and deep white matter lesions were at higher risk of dementia (Clerici et al., 2012).