Diabetes and Elevated HbA1c levels are Associated with Brain Hypometabolism but not Amyloid Accumulation

Abstract

Dysfunctional insulin signaling may affect brain metabolism or amyloid deposition. We investigated the associations of type 2 diabetes with amyloid accumulation measured using ¹¹C-Pittsburgh Compound B (PiB) and brain hypometabolism measured using ¹⁸F-fluorodeoxyglucose (FDG) positron emission tomography (PET).

Methods

We studied a sample of non-demented participants from the population-based Mayo Clinic Study of Aging. All subjects underwent MRI, amyloid PET and FDG PET. Alzheimer’s disease (AD) signature and region of interest (ROI) measures for PiB retention ratio and FDG ratio were measured. Diabetes was assessed from the Rochester Epidemiology Project medical records-linkage system.

Results

Among 749 participants (median age 79.0 years; 56.5% male, 81.0% cognitively normal; 20.6% diabetics), FDG hypometabolism (FDG ratio < 1.31) in the AD signature meta-ROI was more common in diabetics (48.1%) than in non-diabetics (28.9%; p <0.001). The median FDG ratio was lower in diabetics vs. non-diabetics in the AD signature meta-ROI (1.32 vs. 1.40, p < 0.001), and in the angular (1.40 vs. 1.48, p < 0.001) and posterior cingulate gyri ROIs (1.63 vs. 1.72, p < 0.001). The odds ratio (OR [95% confidence interval]) for abnormal AD signature FDG hypometabolism was elevated (OR, 2.28 [1.56, 3.33]) in diabetics vs. non-diabetics after adjustment for age, sex, and education, and after additional adjustment for Apolipoprotein ε4 allele, glycemic level, and cognitive status (OR, 1.69 [1.10, 2.60]). However, AD signature PiB retention ratio was similar in diabetics vs. non-diabetics (OR, 1.03 [0.71, 1.51]; p = 0.87). In post-hoc analyses in non-diabetics, a 1% increase in HBA1c was associated with greater AD signature hypometabolism in cognitively normal subjects (OR, 1.93 [1.03, 3.62; p = 0.04]) and in the total cohort (OR 1.59 [0.92, 2.75; p = 0.10).

Conclusion

Diabetes and poor glycemic control in non-diabetics may enhance glucose hypometabolism in AD signature regions. These factors should be investigated in longitudinal studies for their role in detecting onset of symptoms in AD.

Introduction

Several studies have reported associations of type 2 diabetes with cognitive impairment, Alzheimer’s disease (AD), or vascular dementia (1–5). The exact mechanisms underlying the association of diabetes with AD remains incompletely understood, but studies suggest that ischemic cerebrovascular disease and hippocampal atrophy are involved (6–10). Other potential mechanisms including global and regional brain hypometabolism (6, 10) and amyloid deposition have been hypothesized (11), but require further investigation. Neuropathological studies, which provide the gold standard for AD pathology, have been inconsistent; some studies have reported greater amyloid-related pathology in diabetic brains (11) whereas others have not (3, 7, 12, 13). Given the dysfunctional insulin signaling in type 2 diabetes, we hypothesized abnormal cerebral glucose metabolism in diabetics may lead to neuronal injury and cognitive impairment (14). Positron emission tomography (PET) offers the opportunity to study in vivo markers of both abnormal brain metabolism and amyloid accumulation.

The objective of our study, therefore, was to specifically examine the association of type 2 diabetes with in vivo markers of amyloid accumulation and cerebral glucose metabolism measured using ¹¹C-Pittsburgh Compound B [PiB-PET]) and ¹⁸F-fluorodeoxyglucose [FDG-PET]), in a subset of participants from the population-based Mayo Clinic Study of Aging (MCSA).

MATERIALS AND METHODS

Study Participants

The study design and methodology have previously been published (15). Briefly, participants were Olmsted County, MN residents aged 70 – 89 on October 1, 2004, who were randomly selected from an enumeration of the population using the Rochester Epidemiology medical records linkage system, and invited to participate in the MCSA (16). Starting in 2008, additional subjects were recruited continually to maintain the sample size. Participants were invited to undergo magnetic resonance imaging (MRI) of the brain in 2005 and PET imaging in 2006. All subjects signed a written informed consent that included the Health Insurance Portability and Accountability Act authorization to use and disclose protected health information. The Mayo Clinic and Olmsted Medical Center Institutional Review Boards approved all study protocols.

Assessment of Cognitive Status

Each participant underwent a nurse/study coordinator interview, a physician evaluation, and cognitive testing by a psychometrist. The interview included questions about memory (participant) and the Clinical Dementia Rating Scale (CDR) (17) and Functional Activities Questionnaire (FAQ) (18) (informant). The physician evaluation included the Short Test of Mental Status (19) and a neurological examination. The cognitive testing used 9 tests to assess performance in memory, executive function, language, and visuospatial skills domains. Cognitive domain scores were computed from age-adjusted test scores that were scaled (mean, 10; SD, 3), then summed within domains, and scaled (15, 20, 21).

MCI criteria were defined as: cognitive concern (subject, informant, nurse, or physician); impairment in ≥ 1 cognitive domains; essentially normal functional activities; and no dementia (15, 21, 22). Dementia was based on the Diagnostic and Statistical Manual of Mental Disorders, 4^th edition [DSM-IV] criteria. Subjects who performed within the range for cognitively normal subjects from the population and did not meet criteria for MCI or dementia were considered cognitively normal (CN) (15, 21, 22).

Ascertainment of Type 2 Diabetes

Type 2 diabetes was abstracted from the medical records (16) as: treatment for diabetes (oral anti-diabetic agents, insulin), or a fasting blood glucose ≥ 126 mg/dl reported two or more times, or a physician diagnosis (23).

Other Covariates

Date of birth and years of education were assessed by interview. HBA1c level at enrollment was assessed from the medical record (16). APOE genotype was performed at baseline, and blood glucose was measured just prior to PET imaging.

FDG-PET and PiB-PET Acquisition

MRI was performed at 3T with a 3-dimensional volumetric T1-D-MPRAGE (magnetization-prepared rapid gradient–echo) sequence previously described (24). PET images were acquired using a PET/CT scanner operating in 3-dimensional mode (25). A CT image was obtained for attenuation correction. The ¹¹C PiB-PET scan consisted of four 5-minute dynamic frames acquired 40–60 minutes after injection (26). Subjects were injected with ¹⁸Fluorodeoxyglucose (18F-FDG) 1 hour after the PiB scan and imaged after 30–38 minutes, for an 8-minute image acquisition of 4, 2-minute dynamic frames (26, 27). An in-house fully automated image processing pipeline was used for quantitative image analysis for PiB and FDG. Each subject’s PET images were registered to his/her MRI which had been labeled with our parcellation atlas (26). An AD signature cortical PiB-PET meta-region of interest (meta-ROI) retention ratio was calculated from the median uptake over voxels in the prefrontal, orbitofrontal, parietal, temporal, anterior cingulate, and posterior cingulate/precuneus ROIs normalized to the cerebellar gray matter ROI of the atlas (25). An AD-signature FDG-PET ratio was calculated in a similar manner based on glucose metabolic rates from an “Alzheimer signature” meta-ROI and consisted of the average bilateral angular gyri, posterior cingulate/precuneus, and inferior temporal cortical ROIs from both hemispheres normalized to pons uptake (28–30).

Statistical Analyses

We defined an abnormal AD signature meta-ROI PiB retention ratio (standardized uptake value ratio) as >1.50 and an abnormal AD signature meta-ROI FDG ratio as <1.31, corresponding to the 90% sensitivity in AD dementia (27, 31). For PiB-PET, where higher biomarker values are worse, the cutpoint corresponding to 90% sensitivity was at the 10th percentile of the AD distribution. For FDG-PET where lower biomarker values are worse, the cutpoint was at the 90th percentile of the AD distribution. We examined the association of diabetes with abnormal AD signature PiB-PET ratio and FDG-PET ratio using logistic regression models adjusted for age and sex. Given that FDG ratio varies with cognitive status, we performed analyses for the total sample, MCI, and cognitively normal [CN]. (29). We examined potential confounding and effect modification (interaction) by APOE ε4 allele, blood glucose at PET imaging, and cognitive status. We selected comparison non-AD regions as the primary sensorimotor regions and global whole brain (based on 35 ROIs), computed unadjusted and glucose-adjusted means of FDG ratio, and examined their associations with diabetes using linear regression models. We conducted post-hoc analyses in non-diabetes to determine whether glycemic control (HBA1c) was also associated with hypometabolism. All analyses were conducted using SAS v.9.3. (SAS, Cary, North Carolina). P-values were considered significant at an alpha < 0.05 (two-tailed tests).

RESULTS

Table 1 shows the characteristics of 749 participants without dementia who underwent PET imaging (2006–2012). The frequency of an abnormal AD signature meta-ROI FDG ratio was significantly higher among diabetics; the median was lower in diabetics vs. non-diabetics in both CN and MCI subjects (Fig 1, top panel) and in the total sample (not shown). FDG hypometabolism was found in all three subregions of the AD signature meta-ROI. Unadjusted means for AD signature FDG ratio for sensorimotor regions and whole brain and were significantly different in diabetics vs. non-diabetics, but glucose-adjusted means did not differ (Supplemental Table 1).

TABLE 1.

Characteristics of participants by type 2 diabetes

Characteristic	Type 2 diabetes, n = 154	No diabetes, n = 595	p
Age, yr	80.3 (75.8, 84.1)	78.9 (75.0, 83.5)	0.18
Male Sex,	96 (62.3)	327 (55.0)	0.10
Education, yr	13 (12, 16)	14 (12, 16)	0.04
APOE ε4 carriera	40 (26.0)	160 (26.9)	0.81
PET studies
AD signature meta-ROI FDG < 1.31b	74 (48.1)	172 (28.9)	<0.001
AD signature meta-ROI FDG ratio	1.32 (1.25,1.46)	1.40 (1.29, 1.49)	0.0001
Posterior cingulate	1.63 (1.53, 1.75)	1.72 (1.59, 1.83)	<0.001
Temporal gyrus	1.10 (1.00, 1.23)	1.15 (1.04, 1.26)	0.05
Angular gyrus	1.40 (1.30, 1.53)	1.48 (1.37, 1.60)	<0.001
Whole Brain FDG meta-ROI ratio	1.38 (1.31, 1.44)	1.43 (1.36, 1.49)	<0.001
AD signature meta-ROI PiB >1.50	56 (36.4)	209 (35.1)	0.77
AD signature meta-ROI PiBc ratio	1.39 (1.33, 1.75)	1.39 (1.31, 1.75)	0.30
Anterior cingulate	1.49 (1.41, 1.90)	1.45 (1.37, 1.83)	0.01
Posterior cingulate/precuneus	1.48 (1.39, 1.94)	1.45 (1.36, 1.87)	0.07
Whole Brain PiB meta-ROI ratio	1.39 (1.34, 1.67)	1.39 (1.32, 1.66)	0.14
Fasting blood glucose, mg/dl	114.5 (101, 131)	97 (90, 105)	<0.001
Years in study	4.04 (2.53, 6.35)	4.04 (2.56, 6.49)	0.36
MCI, n (%)	40 (29.9)	102 (19.2)	0.007

Estimates are n (%) or median (interquartile range).

Abbreviations: AD = Alzheimer’s disease; APOE = apolipoprotein E; MCI = mild cognitive impairment.

^aOne subject in no diabetes group had missing data.

^bAD signature FDG ratio.

^cAD signature PiB ratio.

FIGURE 1.

Box plots for FDG and PiB retention ratio in AD signature regions by diabetes mellitus (DM) in cognitively normal and MCI subjects. Women (filled circles) and men (open circles).

In the total sample, AD signature meta-ROI FDG ratio was significantly lower in younger diabetics vs. non-diabetics (Fig 2A and 2B), in men and in women (Fig 2C and 2D), and in APOE ε4 non-carriers but not in carriers (Fig 2E and 2F); the findings were consistent for CN and MCI (not shown). In representative scans, AD signature FDG hypometabolism was absent in subjects with no diabetes and no MCI (Fig 3A), minimal in non-diabetics with MCI (Fig 3B), and worst in diabetics with MCI (Fig 3C and 3D).

FIGURE 2.

Box plots of FDG ratio in AD signature regions in the total sample by diabetes (DM) status stratified by age (A: 70–79 years, B: 80 – 89 years), sex (C: women, D: men), and apolipoprotein ε4 allele (E: ε4 carrier, F: ε4 non-carrier). Women (filled circle), men (open circles).

FIGURE 3.

FDG scans for participants without diabetes (A, cognitively normal; B, MCI) and with diabetes (C and D; both with MCI). Each image set represents 3-D SSP regional hypometabolism normalized to age appropriate normal subjects and displayed as a Z-score; upper boundary as 7.0 (red) and the lower boundary as 0.0 (black). A Z-score of 1.5 (medium blue) or greater would be considered significant.

By contrast, the proportion of diabetics versus non-diabetics with an abnormal AD signature meta-ROI PiB ratio did not differ (Table 1). The median AD signature meta-ROI PiB ratio was similar for diabetics vs. non-diabetics (Table 1) among CN and MCI (Fig 1, bottom panel).

Association of Diabetes with AD Signature FDG Uptake

Diabetes was significantly associated with AD signature meta-ROI FDG hypometabolism of the range seen in AD dementia (Table 2). The OR decreased slightly after adjustment for glucose, but remained significant for the total sample and MCI cases for the fully-adjusted models. Diabetes was associated with decreased FDG ratio in the sensorimotor ROIs and the whole brain and but the associations were not significant after adjustment for glucose (Supplemental Table 2). There were no significant interactions of diabetes with age, sex, or APOE ε4; the OR for diabetes and hypometabolism was significant for subjects with glucose levels < 126 mg/dl (OR, 1.72 95%CI, 1.11, 2.69; p = 0.02) (Supplemental Table 3)

TABLE 2.

Association of type 2 diabetes with abnormal FDG-PET ratio^a in AD signature meta-ROI

	Odds ratios (95% CI): column headings represent covariates included in each model
	Age and Sex	Education	APOE ε4	Glucose	Cognition	Full model 1b	Full model 2b
Total sample
OR (95% CI)	2.24 (1.54, 3.27)	2.28 (1.56, 3.33)	2.26 (1.55, 3.30)	1.73 (1.13, 2.63)	2.17 (1.48, 3.18)	1.78 1.17, 2.72)	1.69 (1.10, 2.60)
P value	<0.001	<0.001	<0.001	.01	<0.001	0.008	0.02
Cognitively normal
OR (95% CI)	2.21 (1.42, 3.43)	2.22 (1.43, 3.45)	2.20 (1.42, 3.41)	1.50 (0.91, 2.49)	N/A	1.50 (0.91, 2.49)	N/A
P value	0.0004	0.0004	0.0005	0.12		0.12
MCI
OR (95% CI)	2.02 (0.93, 4.41)	2.22 (1.00, 4.92)	2.31 (1.03, 5.18)	1.96 (0.85, 4.51)	N/A	2.62 (1.08, 6.38)	N/A
P value	0.08	0.05	0.04	0.11		0.03

Abbreviations: AD = Alzheimer’s disease; MCI = mild cognitive impairment.

^aAD signature meta-ROI FDG ratio < 1.31.

^bAll models include adjustment for age and sex; each column shows the variable added separately to a model with age and sex. Full model 1 is adjusted for age, sex, education, APOE ε4 allele, and glucose level. Full model 2 includes variables in full model 1 and cognition (MCI or normal cognition).

Association of Diabetes with AD Signature PiB

In contrast to unadjusted estimates, the OR for abnormal AD signature meta-ROI PiB was significantly reduced in subjects with MCI after adjustment for age, sex, and education, but not after adjustment for APOE e4 allele (Table 3).

TABLE 3.

Associations of type 2 diabetes with abnormal PiB-PET retention ratio^a in AD signature meta-ROI

	OR (95% CI): column headings represent covariates included in each modelb
	Age and Sex	Education	APOE ε4	Full model
Total sample
OR (95% CI)	1.02 (0.70, 1.49)	1.03 (0.71, 1.51)	1.05 (0.70, 1.57)	1.06 (0.71, 1.59)
P value	0.91	0.87	0.82	0.78
Cognitively normal
OR (95% CI)	1.26 (0.82, 1.95)	1.28 (0.83, 1.97)	1.26 (0.80, 1.99)	1.28 (0.81, 2.02)
P value	0.30	0.27	0.32	0.29
MCI
OR (95% CI)	0.39 (0.17, 0.88)	0.39 (0.17, 0.90)	0.46 (0.18, 1.16)	0.46 (0.18, 1.16)
P value	0.02	0.03	0.10	0.10

Abbreviations: AD = Alzheimer’s disease; MCI = mild cognitive impairment.

^aAD signature PiB ratio > 1.50.

^bAll models include adjustment for age and sex; each column shows the variable added separately to a model with age and sex. Full model is adjusted for age, sex, education, and APOE ε4 allele.

HbA1c and AD signature FDG hypometabolism

In CN non-diabetics, the OR for hypometabolism in the AD signature meta-ROI increased with each 1% increase in HBA1c after adjustment for age, sex, and education (OR, 1.93 [1.03, 3.62]; p = 0.04]), and after additional adjustment for APOE ε4 allele (OR, 1.89 [1.00, 3.57]; p = .049).

DISCUSSION

In this elderly persons sample, cohort, diabetes was associated with FDG hypometabolism. This association persisted in the AD signature region for the total sample and MCI cases even after controlling for serum glucose, but not in sensorimotor cortex. By contrast, diabetes was not associated with increased PiB retention ratio (amyloid accumulation) in AD signature meta- ROIs for the total sample, or for CN subjects, but the OR was significantly reduced in subjects with MCI in models that did not include APOE e4.

Glucose hypometabolism in diabetics suggests that diabetes impacts cerebral energy metabolism independently of amyloid deposition, possibly through mechanisms that cause neuronal injury (32). This hypometabolism in CN diabetics suggests that neuronal injury is present even before symptomatic evidence of cognitive impairment. The loss of significance in the CN group after adjustment for glucose level, however, suggests that the effect of diabetes may be small in CN or below the threshold for sustained effect, such that a larger sample of CN is required to detect a significant difference. Alternately, the findings in CN may be due to a more generalized “diabetes effect” rather than a specific AD meta-ROI effect. The persistence of significant associations for the total sample and for MCI cases, even after adjustment for glucose, suggests the effect may be real. The stronger association for MCI subjects suggests that hypometabolism may be in the causal pathway for cognitive impairment in diabetics. Finally, the absence of an effect in multivariable models for the sensorimotor cortex suggests that the associations are not entirely explained by a competitive inhibition of labeled glucose uptake by elevated serum glucose in diabetics.

Our subgroup analyses showed stronger associations of diabetes with hypometabolism for APOE ε4 non-carriers, younger subjects, and men. The stronger OR for APOE ε4 non-carriers suggests that APOE ε4 allele effects might share some mechanistic commonalities with diabetes, and that are independent of APOE ε4-related amyloidosis (33). The stronger association in 70 – 79 year olds may relate to earlier diabetes onset, age-related differences in diabetes effects, non-participation in older subjects, or to potential survival bias. The stronger OR in men suggests etiologic differences as observed for amnestic MCI in our previous studies (21, 34). Follow-up of the cohort will provide greater power to detect significant interactions.

Our findings for HBA1c in non-diabetics support our hypothesis that altered glycemic control is associated with neuronal injury, and are consistent with other studies. In CN adults with newly diagnosed pre-diabetes or diabetes, increasing insulin resistance was associated with greater FDG hypometabolism in AD regions (35), and higher fasting glucose levels CN in non-diabetics were associated with FDG hypometabolism (36). These findings suggest that markers of pre-diabetes should be examined for early detection of brain hypometabolism.

The analyses reported here did not use partial volume correction. When we applied partial volume correction, our findings changed somewhat. Significant associations persisted with adjustment for age- and sex-adjusted models in AD signature regions. However, glucose-adjusted models were not significant. This difference in results may mean that the FDG changes we observed may in part, reflect loss of brain volume due either to diabetes or to other etiology. Alternately, we recognize that if diabetes and related hyperglycemia contributes to structural as well as metabolic changes, controlling for blood glucose may result in over-controlling.

The absence of an association of diabetes with PiB retention ratio in AD signature regions in CN subjects is consistent with findings from several studies (3, 7, 12, 13, 37). In one study, diabetes was associated with cerebral infarctions but not with AD pathology (7); another study found no difference in AD pathology among diabetics and non-diabetics (3). A third study reported that diabetic APOE ε4 carriers had a lower amyloid beta load compared to non-diabetics (12). In contrast, one study reported that diabetic APOE ε4 carriers had greater AD pathology (11), and another reported greater plaque density with longer duration of diabetes (3).

Absence of an association of diabetes with AD signature PiB ratio and significantly AD signature PiB in diabetic MCI subjects, are inconsistent with the hypothesis that diabetes contributes to increased amyloid deposition (38, 39). However, they are consistent with other studies (40). In the Baltimore Longitudinal Study of Aging, diabetes-related parameters were not associated with AD pathology.(40). It is possible that diabetes may have greater effects on neuronal injury and brain atrophy than on amyloid accumulation in elderly persons (8, 9). The null association may also be due to interactions of timing of onset of vascular pathology with amyloid deposition in diabetics. By contrast, subjects with diabetes and PiB deposition occurring independently at earlier ages may progress more rapidly from MCI to dementia, and may be less likely to be captured in our cohort since we did not initially enroll demented subjects to imaging studies. The absence of a significant difference in AD signature PiB ratio for MCI diabetics and non-diabetics (Figure 1, lower panel) vs. the significant association observed in multivariable models suggests possible confounding; the association was significant after adjustment for age, sex, and education, but not APOE ε4 allele. This indicates the need to account for potential confounders when comparing FDG ratio across regions.

One potential limitation of our study is that we cannot assess causality due to the cross-sectional design. The absence of associations of diabetes with AD signature FDG ratio in APOE ε4 carriers and with AD signature PiB retention ratio could be due to under-representation of diabetics, lower participation of diabetics with more severe or longer disease duration, or to lower frequency of APOE ε4 allele in this older age-group. HBA1c was assessed at enrollment and not at the time of imaging; however, the findings support our hypothesis that impaired glucose metabolism impacts brain metabolism. Ongoing recruitment should increase the number of diabetics in our sample. Finally, over 98% of our cohort is of northern European ancestry, thus our findings should be confirmed in other ethnicities.

Our study has several important strengths. Participants were randomly selected from the population; thus, our findings are less subject to selection or volunteer bias. The diagnosis of diabetes was reliably ascertained using the medical records linkage system and not subject to recall bias as occurs with self-report. We had a large sample size of subjects with PET imaging that provides more reliable estimates than studies of small sample size.

CONCLUSION

In elderly persons, diabetes and impaired glucose metabolism in CN, might contribute to neuronal injury. These factors should be examined in future longitudinal studies for their potential role in detecting onset of AD symptoms.