Polygenic risk of type 2 diabetes is associated with Vascular Dementia but not with Alzheimer's Disease: a Prospective Cohort Study

doi:10.21203/rs.3.rs-1996829/v1

Download PDF

Research Article

Polygenic risk of type 2 diabetes is associated with Vascular Dementia but not with Alzheimer's Disease: a Prospective Cohort Study

https://doi.org/10.21203/rs.3.rs-1996829/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background and Objectives

Type 2 diabetes and dementia are associated, but their causal relationship is unclear. Genetic studies, i.e. using polygenic risk scores (PRS) or Mendelian Randomization (MR), and validated dementia endpoints with high diagnostic precision (instead of register-based diagnoses) may clarify causal links between these two entities.

Methods

We tested associations between PRS for type 2 diabetes, fasting glucose, fasting insulin and HbA_1c on the one hand and dementia on the other in 29139 adults (mean age 55) followed for 20–23 years. Dementia diagnoses were validated and included all-cause dementia, Alzheimer’s Disease (AD) and Vascular Dementia (VaD) as well as mixed dementia. We also tested causal associations between type 2 diabetes and dementia forms through 2-sample-MR analyses.

Results

Seven different PRSs (including SNPs with different significance thresholds) for type 2 diabetes were tested. A PRS including N = 4891 SNPs with p-value < 5e-04 showed strongest association with the different outcomes, including all-cause dementia (Hazard Ratio [HR] 1.11; Bonferroni corrected p = 3.6e-03), mixed dementia (HR 1.18; Bonferroni corrected p = 3.3e-04) and VaD cases (HR 1.28; Bonferroni corrected p = 9.6e-05). Associations were stronger for non-carriers of the AD risk gene APOE ε4. 2-sample-MR analyses could, however, not confirm a causal link between genetic risk markers of type 2 diabetes and dementia outcomes.

Conclusions

Polygenic risk of type 2 diabetes is associated with dementia risk, in particular VaD. Associations were strongest for non-carriers of APOE ε4. However, 2-sample-MR findings did not support a causal relationship between type 2 diabetes and VaD.

type 2 diabetes

dementia

Mendelian Randomization

Genetic

Prediction

Vascular Dementia

Alzheimer Dementia

Type 2 diabetes is associated with an increased risk of dementia in epidemiological studies with a risk ratio (RR) of 2.5 for Vascular Dementia (VaD) and 1.5 for Alzheimer’s disease (AD) (1). Interventional studies based on anti-diabetes drugs have, however, so far not been successful in preventing cognitive decline in people with type 2 diabetes (2) (3). It is unclear whether there is a causal link between type 2 diabetes and dementia or whether closely related traits to diabetes confound the association. As type 2 diabetes is a multifactorial disease, it has been a challenge to identify the specific mechanisms that are of greatest importance for this association. This may be needed to identify the best treatments to prevent cognitive decline associated with type 2 diabetes. Epidemiological studies have shown that vascular pathophysiological factors are likely to be the most predominant in causing cognitive decline (4), but that metabolic pathways also may play a role. For example, chronically elevated fasting glucose levels, HbA_1c levels, insulin resistance and Advanced Glycation End (AGE) products are some biomarkers that have found to be associated with cognitive decline in population-based studies(5).

A way to investigate a possible causal relationship between a trait (e.g. type 2 diabetes) and an outcome (e.g. dementia) is through Mendelian Randomization (MR), where genetic variants that are strongly associated with the exposure variable (the trait) are used as an instrumental variable analyzed in relation to the outcome. In this way the influence of confounding or reverse causation between the two entities of interest can be avoided (6). Some MR studies with type 2 diabetes as exposure and dementia as outcome have been carried out, but have not been able to identify a causal relationship between the two (7) (8) (9) (10). However, MR studies have found evidence of causal associations between closely related phenotypes to diabetes and dementia, for example between fasting plasma glucose and risk of AD(11) or unspecified dementia (12), between HbA_1c levels and impaired visual memory(8) and between insulin resistance and AD (11) (13).

While MR analyses provides the possibility to investigate causal relationships, it is also of value to study a wider set of genetic risk markers that are associated with a trait in relation to an outcome through use of polygenic risk scores (PRS) (14) which summarize an individual’s propensity to a trait. Associations between a PRS and an outcome enable the possibility to include a greater number of genetic variants with a weaker association to the trait, which can help to achieve higher statistical power in analyses than when using MR. Most PRS studies have so far not shown significant associations between type 2 diabetes and associated phenotypes such as insulin resistance on the one hand and dementia risk on the other (15) (16). One exception was a study in patients with major depressive disorder, MDD (but without type 2 diabetes) where an association between PRS for type 2 diabetes and impaired cognitive functioning was identified (17).

A limitation of the previous studies on type 2 diabetes and dementia is that they have not used validated dementia endpoints, but instead relied on register data which often lacks diagnostic precision on dementia, in particular sub-types such as AD and VaD (18). There is therefore a need of studies using diagnoses with higher reliability.

The aim of this study was to investigate possible causal associations between genetic risk markers of type 2 diabetes, fasting glucose, fasting insulin and HbA_1c on the one hand and dementia diagnoses on the other, applying both PRS and Mendelian Randomization analyses. The dementia diagnoses were carefully validated and included the sub-types mixed dementia (AD in combination with significant cerebrovascular co-pathology), AD (without significant cerebrovascular co-pathology), and VaD, determined through careful examinations of medical records, neuro-imaging data and cerebrospinal fluid (CSF) analyses. We also tested the influence of the AD risk gene variant APOE ε4 burden on the associations, as this may further increase dementia risk in persons with type 2 diabetes (19).

Participants

We used data from the Malmö Diet and Cancer Study (MDCS) baseline examination carried out during 1991–1994, a prospective population-based cohort study of 30,446 participants from Malmö, Southern Sweden. The mean age at baseline was 55 years and the participation rate 41%. Reasons for non-participation and loss-to-follow-up have previously been described (20).

A self-administered questionnaire was distributed with questions on demographics, previous diseases, lifestyle habits and medications. A health examination was carried out in which blood pressure (mmHg), weight (kg) and height (m) was measured and fasting blood samples were drawn. Educational level was classified into elementary school (< 8 years of education), upper secondary school (9–12 years), higher education/university (≥ 12 years), or missing data. Smoking was classified as ‘never-smoker’, ‘past’, ‘current’ smoking or missing data. Physical activity level was calculated into a score based on data on number of minutes per week of activity, type of activity and an activity specific factor, and was then divided into tertiles with an added category for missing data. Alcohol was classified into ‘no consumption’, ‘consumption below risk level’ (less than 9 standard units of 12 g per week for women or 14 units for men), ‘consumption above risk level’ or missing data. A history of cardiovascular disease (CVD) was defined as reporting having had a coronary event or a stroke.

Genotyping was carried out when technically possible in 29,451 of the blood samples using the Illumina GSA v1 genotyping array. Due to missing data on age, blood sample number or a mismatch in ID and blood sample number, 72 participants were excluded, and a further 240 due to failing the genetic data quality control (QC) as described below. The final study population thus arrived at 29,139 participants.

Genetic QC

Standard QC steps were performed for SNP-level filter. High-quality variants (autosomal, non-monomorphic, bi-allelic variants with Hardy–Weinberg Equilibrium (HWE) P > 5 × 10^− 8 and with a call rate of > 99%) were used. Multi-allelic variants and SNPs with a data imputation score < 0.2 have been excluded as part of post-imputation QC and genotype calls with a posterior likelihood < 0.9 have been set to fail (i.e. hard-called). SNPs with a genotyping rate > 0.9 were retained. SNPs with Minor Allele Frequency (MAF) ≥ 5% were taken for the analysis. Supplementary sensitivity analysis using MAF ≥ 1% were also performed regarding PRS for type 2 diabetes and dementia outcomes, to check the possible impact of rare genetic variants on the results. Multi-dimensional scaling was done using PLINK 2 (21) to create principal components in genetic analyses to account for ancestry.

Polygenic Risk Score Calculation

Using an r² < 0.1 threshold over 1000 kb sliding windows (index threshold and clumped SNPs of p < 1), linkage-disequilibrium (LD) clumping was performed using PLINK’s clump function. LD clumping ensures that PRS/PGS calculation is not overloaded by large blocks of correlated SNP sets. The PRS / PGS was determined for each subject by summing up the effective number of alleles (0, 1, 2) of the SNPs weighted by the natural logarithm of their respective odds ratio (OR). The default formula for PRS calculation in PLINK is:

$${PRS}_{j}= \frac{{\sum }_{i}^{N}{S}_{i}*{G}_{ij}}{P* {M}_{j}}$$

where the effect size of SNP i is S_i; the number of effect alleles observed in sample j is G_ij; the ploidy of the sample is P (is generally 2 for humans); the total number of SNPs included in the PRS is N; and the number of non-missing SNPs observed in sample j is M_j. If the sample has a missing genotype for SNP i, then the population minor allele frequency multiplied by the ploidy (MAF_i ∗ P) is used instead of G_ij.

Publicly accessible summary statistics from reported GWAS studies (not overlapping with our dataset) of type II diabetes (22), HbA_1c (23), fasting glucose (24) and fasting insulin (25) was used to define PRS. We iterated over a variety of p-values (0.05 to 5 × 10^− 8) to evaluate the appropriate p-value threshold: p = 0.05 [PRS 1], p = 5 × 10^− 3 [PRS 2], p = 5 × 10^− 4 [PRS 3], p = 5 × 10^− 5 [PRS 4], p = 5 × 10^− 6 [PRS 5], p = 5 × 10^− 7 [PRS 6] and the GWAS-level significance thresholds of p = 5 × 10^− 8 [PRS 7] creating models named PRS 1–7.

Clinical type 2 diabetes

Type 2 diabetes at baseline was defined using combined data from the Diabetes 2000 Register, the National Diabetes Register (NDR), the HbA_1c registry, the Swedish Cause of Death Register, the Swedish National Patient Register (NPR), the Swedish Prescribed Drug Register and the Alla Nya Diabetiker I Skåne (ANDIS) Register, as well as cohort data from the Malmö Preventive Project and the present cohort including data from blood samples, self-reported diabetes and drug use. Type 2 diabetes was then classified as (a) having a diabetes diagnosis in any of the sources, (b) not having specified type 1 diabetes, gestational diabetes, secondary diabetes, MODY diabetes or other specified diabetes types, and (c) not having self-reported insulin treatment as only diabetes treatment at baseline. After exclusion criteria were applied, 1,119 participants were defined as having prevalent type 2 diabetes at baseline.

Validated dementia endpoints

The validation process of dementia diagnoses in the MDC cohort has previously been described (26, 27). The validity was assessed using data from medical records by trained medical doctors at the Memory Clinic, Skåne University Hospital in Malmö. National Patient Register (NPR) diagnoses of dementia from the baseline examination until 31st December 2014 were obtained (n = 2,206), covering 99% of all inpatient medical diagnoses during the study period, and from 2001 also outpatient diagnoses (28), although not diagnoses from primary care facilities. The participants were thus followed for 20–23 years after the baseline examination in 1991–1994.

Out of 2,206 dementia cases, 1,697 (77%) were diagnosed by a specialist in cognitive disorders, 1,904 (86%) had data on symptoms, 1,894 (86%) on cognitive test results and 2,080 (94%) on Computer Tomography (CT), Magnetic Resonance Imaging (MRI), or both. CSF analyses were available for 839 (38%) of the participants. The NPR diagnoses were through this information re-classified in accordance with the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-V) (29).

The final number of dementia cases after validation was 2,039 out of which 578 were mixed dementia (AD + VaD), 510 were VaD, 598 were AD and the remaining 353 were other forms of dementia (not analyzed in this study).

Statistical analyses

Statistical analyses were performed using IBM SPSS version 25 for Mac OSX, R version 4.1.1, R studio version 1.4.1717 and Python version 3.9.6. Independent-samples t-tests and Chi-square tests were used to test for differences in cohort characteristics between participants with and without type 2 diabetes.

Associations of time to first dementia event between participants with and without clinical type 2 diabetes were compared through Cox regression analyses. In Model 1 we adjusted for age, gender and education. Model 2 was adjusted for the same factors and additionally for smoking, alcohol consumption, physical activity, SBP, BMI, ApoB/ApoA-ratio, history of CVD (stroke or coronary event), use of anti-hypertensive medication and use of lipid-lowering treatment. Age, sex, education and cardiovascular risk factors can all affect risk of both type 2 diabetes and dementia, and have traditionally been adjusted for in epidemiological studies on the association between the two diseases(30). Moreover, cardiovascular risk factors are seen as potential mediators of the association(31). An interaction model including APOE ε4 status (0 or 1–2 alleles), equivalent of Model 1, was also performed. The reason is that carriership of APOE ε4 is a strong risk factor of AD (32) which also can also affect prognosis of type 2 diabetes, and possibly also diabetes-associated dementia, through increasing the risk of atherosclerosis and cardiovascular disease (33) (34). The analyses were then carried out stratified for APOE ε4 status.

To examine the association between PRS 1–7 (as defined above) for type 2 diabetes as exposure and clinical type 2 diabetes at baseline as outcome in our study sample, binary logistic regression analyses were carried out.

We then performed Cox regression analyses to investigate associations between PRS 1–7 for (a) type 2 diabetes, (b) fasting glucose, (c) fasting insulin and (d) HbA_1c as exposure and time to first dementia event as outcome. The analyses were adjusted for age, gender and education in Model 1 and the same factors and additionally ApoE-ε2 and ApoE-ε4 genotype (0, 1 or 2 alleles) in Model 2. We also tested interactions between APOE ε4 status (0 or 1–2 alleles), and PRS 1–7 on the association with dementia. For analyses where interaction terms were significant, we then stratified for APOE ε4 status in secondary Cox regression analyses equivalent of Model 1. A sensitivity analysis using MAF ≥ 1% was also performed for the results of PRS for type 2 diabetes and dementia outcomes.

We also carried out 2-sample MR analyses with the same reference dataset(22) using the MR methods MR Egger, to test whether type 2 diabetes shared genetic effects with any of the dementia subtypes. The MR test was based on 254 independent variants (used as instrumental variables) of type 2 diabetes with a genome-wide significant association at p < 0.5*10^− 8. We could not perform MR analyses for the exposure variables fasting glucose, insulin and HbA_1c, as there were not enough SNPs with genome wide significance.

Data Availability

Data can be applied for from the Steering Committee of the Malmo Diet Cancer Cohort, with contact person Mr. Anders Dahlin, data manager (e-mail:[email protected]).

Standard Protocol Approvals, Registrations, and Patient Consents

The study was approved by the Ethical Committee of Lund University, Sweden (MDC baseline examination LU-51-90, MDC Re-examination Dnr.532/2006 and for genetic analyses in 2009 (Dnr 2009/682). Written informed consent was obtained from all participants.

Characteristics of the MDC cohort are presented in Table 1. Participants with type 2 diabetes were on average 3.6 years older, had a lower average educational level and a higher proportion of men than those without type 2 diabetes. Risk factors for cardiovascular disease were also more prevalent among these participants, as regards previous smoking, lipid status, BMI, blood pressure and previous disease history. Current smoking and alcohol intake was however lower for this group. There was no significant difference between the groups as regards APOE status. Out of people with type 2 diabetes, 11% were diagnosed with dementia during the follow-up period compared to 6.8% of the control group, for which all dementia subtypes were less prevalent.

Table 1. Characteristics of the MDC cohort and comparisons between participants with type 2 diabetes and controls.

	Total study sample (n=29,139)	No type 2 diabetes (n=28,020)	Type 2 diabetes (n=1,119)	P for difference between groups
BASELINE DATA
Age, mean (SD)	58.1 (7.61)	57.9 (7.61)	61.3 (6.69)	<0.001
Gender (% women/men)	60.4/39.6	60.9/39.1	46.7/53.3	<0.001
Education (%) Missing data <8 years 9-12 years ≥ 13 years	6.2 42.0 35.0 22.9	6.1 39.1 33.1 21.8	8.8 48.1 27.5 15.6	<0.001
Smoking (%) Missing data Never Former Current	6.0 26.5 31.9 35.6	5.9 26.7 31.7 35.7	8.5 21.6 37.8 32.1	<0.001
Alcohol consumption (%) Missing data No consumption Below risk level Above risk level	8.0 15.6 42.5 33.9	7.9 15.2 42.6 34.3	11.2 25.2 40.4 23.2	<0.001
Physical activity level (%) Missing data Low Medium High	6.9 31.0 31.0 31.0	6.7 30.9 31.3 31.1	9.7 35.7 25.6 29.0	<0.001
Systolic blood pressure, mean (SD)*	141.2 (20.0)	140.8 (20.0)	150.3 (19.7)	<0.001
BMI, mean (SD)*	25.8 (4.04)	25.7 (3.97)	28.6 (4.68)	<0.001
S-ApoB/ApoA-ratio, mean (SD)	0.71 (0.22)	0.70 (0.22)	0.82 (0.27)	<0.001
History of CVD (%)	3.0	2.8	8.1	<0.001
Use of anti-hypertensive treatment (%)	17.4	16.4	41.0	<0.001
Use of lipid lowering treatment (%)	3.0	2.8	10.2	<0.001
*APOE* e4 (%) Non-carrier Heterozygous Homozygous	69.9 27.4 2.7	69.8 27.5 2.7	69.9 27.4 2.7	0.422
*APOE* e2 (%) Non-carrier Heterozygous Homozygous	85.0 14.4 0.6	85.0 14.4 0.6	84.6 14.6 0.8	0.634
OUTCOMES
All-cause dementia, n (%)	2,039 (7.0)	1,914 (6.8)	125 (11.2)	<0.001
Mixed (AD+VaD), n (%)	578 (2.0)	541 (1.9)	37 (3.3)	0.001
VaD, n (%)	510 (1.8)	464 (1.7)	46 (4.1)	<0.001
AD, n (%)	598 (2.1)	571 (2.0)	27 (2.4)	0.386

* Valid data for n=29,091

Cox regression analyses of associations between clinical type 2 diabetes and time to first dementia event are presented in Table 2. Participants with prevalent type 2 diabetes at baseline had a 1.46 times higher HR of all-cause dementia, a 1.61 times higher HR for mixed dementia and a 1.84 times higher HR for VaD after full adjustment. The group did not have a significantly increased risk of AD compared to controls. When including an interaction term for the effect of APOE ε4 status (0 or 1–2 alleles) on type 2 diabetes for dementia outcome in Model 1, there was a significant negative interaction between APOE ε4 status and type 2 diabetes on the risk of dementia for all-cause dementia (HR for interaction term 0.62, p = 0.011), as well as AD (HR 0.46, p = 0.047), but not for mixed dementia or VaD. When stratifying for APOE ε4 status, associations between type 2 diabetes and all forms of dementia were stronger for non-carriers of APOE ε4. In APOE ε4 carriers there was only a significant association between type 2 diabetes and VaD in Model 1, whereas there were no other significant associations between type 2 diabetes and dementia endpoints.

Table 2

Multivariable Cox Proportional Hazards Modelling of years from birth to first dementia event with clinical type 2 diabetes as predictor. Analyses are presented for the total cohort and then stratified for ApoE-ε4-genotype (0 or 1–2 alleles).
	Model 1		Model 2
	HR (95% CI)	p	HR (95% CI)	p
Total cohort (n = 29,139)
All-cause dementia	1.61 (1.34–1.92)	< 0.001	1.46 (1.20–1.77)	< 0.001
Mixed dementia	1.83 (1.32–2.54)	< 0.001	1.61 (1.12–2.30)	0.010
VaD	2.25 (1.67–3.03)	< 0.001	1.84 (1.32–2.58)	< 0.001
AD	1.16 (0.79–1.70)	0.464	1.26 (0.84–1.89)	0.272
No APOE-ε4 (n = 20,359)
All-cause dementia	1.99 (1.58–2.52)	< 0.001	1.83 (1.42–2.36)	< 0.001
Mixed dementia	2.23 (1.42–3.48)	< 0.001	2.00 (1.23–3.27)	0.006
VaD	2.44 (1.67–3.56)	< 0.001	2.16 (1.43–3.26)	< 0.001
AD	1.85 (1.07–3.20)	0.027	1.82 (1.00-3.31)	0.049
APOE-ε4+ (n = 8,780)
All-cause dementia	1.26 (0.94–1.68)	0.117	1.16 (0.84–1.60)	0.369
Mixed dementia	1.45 (0.87–2.40)	0.152	1.22 (0.69–2.16)	0.488
VaD	1.96 (1.17–3.27)	0.010	1.59 (0.87–2.92)	0.136
AD	0.90 (0.52–1.56)	0.704	1.09 (0.62–1.91)	0.774
Model 1 is adjusted for age, sex and education.
Model 2 is adjusted for age, sex, education, smoking, alcohol consumption, physical activity level, SBP, BMI, blood pressure medication, lipid lowering treatment, ApoB/ApoA-ratio and history of CVD.
Significant p-values are highlighted in bold text.

In Table S1, Additional File 1, associations between PRS 1–7 for type 2 diabetes and clinical type 2 diabetes are presented. All associations were significant. The area under the curve (AUC) in Receiver Operating Curves (ROC) was around 0.65 for all PRS.

In Fig. 1 as well as Tables S2, S3, S4 and S5, Additional File 1, results of Cox regression analyses with PRS 1–7 for type 2 diabetes (standardized) as exposure and time to first dementia event as outcome are presented, adjusted for age, sex and education in Model 1 and for age, sex, education and APOE burden (APOE ε4 and ε2 count) in Model 2. There were no significant associations between any of the PRS scores for type 2 diabetes and the dementia endpoints using adjustment Model 1. In Model 2, however, PRS 1 (variants with p-value < 5e-02) and 2 (variants with p-value < 5e-03) were significantly associated with all-cause dementia (HR of 1.11 for both PRSs, Bonferroni corrected p-value 3.9e-03 and 3.6e-03 respectively, Table S2). PRS 1, 2, 3 and 4 were also significantly associated with Mixed Dementia with the strongest association for PRS 2 (HR1.18, Bonferroni corrected p 3.3e-04, Table S3). No significant associations were found between PRS for type 2 diabetes and AD (Table S4). All the type 2 diabetes PRSs were however significantly associated with risk of VaD, out of which PRS 2 showed the strongest association with a HR of 1.28 (Bonferroni corrected p-value 9.6e-05) (Table S5). When including an interaction term of standardized PRS-score multiplied by APOE ε4 status (zero or 1–2 alleles) into Model 2, the interaction term was significant for the outcome all-cause dementia for PRS-scores 1 and 2 (Table S2), and for VaD for all PRS-scores (Table S5), but not for other dementia types.

In Fig. 2 and Table S6, Additional File 1, results stratified for APOE ε4 status (0 or 1–2 alleles) are presented for Cox regression analyses with PRS 1–7 for type 2 diabetes (standardized) as exposure and time to first dementia event (all-cause dementia and VaD) as outcome. For non-carriers of APOE ε4, higher PRS score 1, 2, 3 and 7 significantly increased the risk of all-cause dementia and all PRS-scores increased the risk of VaD (average HR per SD of PRS-score 1.21, p < 0.002). For carriers of APOE ε4, none of these associations were significant.

Results of Cox regression analyses of PRS for exposure variables HbA_1c, fasting glucose and fasting insulin (standardized) on the one hand and dementia outcomes on the other are presented in Tables S7-S10 (HbA_1c), S11-S14 (fasting glucose) and S15-S18 (fasting insulin), Additional file 1, and results for fasting insulin are also visualized in Fig. 3. There were no significant associations between PRS scores for HbA_1c or fasting glucose and dementia outcomes. However, there was a significant negative association between PRS 3 for fasting insulin and mixed dementia (HR 0.88, Bonferroni corrected p-value 1.2e-02).Fasting insulin PRS 3 was also significantly negatively associated with all-cause dementia (HR = 0.94; uncorrected p-value = 4.3e-02) and PRS 7 with AD (HR = 0.86; uncorrected p-value = 1.6e-02), although not after Bonferroni correction.

In Supplementary Results S1 and Figure S1 and S2, Additional file 2, as well as Table S19, Additional file 1, sensitivity analyses with MAF ≥ 1% for the analyses of PRS for type 2 diabetes and dementia outcomes are shown and described more in detail. Using this level did not largely change the results compared with the main results using MAF ≥ 5%.

In Table S20, Additional file 1, 2-sample-MR analyses of genetic variants of type 2 diabetes (shared effects of all 254 genetic variants with genome wide significance) as exposure and dementia endpoints as outcome are presented using various MR methods (details in methodology). No significant causal associations between type 2 diabetes and dementia were found for any of the methods. We also tested the casual association of type 2 diabetes with APOE ε4 negative subgroup using 2-sample-MR. However, we did not find any significant causality effect between the two traits using different 2-sample-MR methods (Table S21, Additional file 1).

In this population-based study we investigated associations between genetic risk markers of type 2 diabetes and validated dementia endpoints. Polygenic risk of type 2 diabetes was associated with all-cause dementia and VaD, with stronger associations for non-carriers of APOE ε4. We also noted that PRS 2 (a PRS generated using weighted score of less significant variants [p < 0.005]) was most significant among all the PRSs in most of the models. This suggests that the combined estimated effect of these less significant variants might be responsible for a disease. Using this clumping and thresholding approach for generating different PRSs is a suitable way of choosing the best set of variants to predict a disease. In 2-sample-MR analyses, however, no causal associations were found between type 2 diabetes and dementia outcomes. Furthermore, there were negative associations between fasting insulin and the outcomes all-cause dementia and AD, although the associations did not remain after Bonferroni correction. PRS of fasting glucose and HbA_1c were however not significantly associated with dementia outcomes. Our findings also show that there are epidemiological associations between clinical type 2 diabetes and dementia which correspond to the findings in the genetic analyses.

The reason that type 2 diabetes and dementia are associated epidemiologically(1), but that causal associations cannot be found is yet unclear, but one possible explanation is that the heterogenicity in the pathogenesis of type 2 diabetes makes direct causal associations between the two diseases as entities hard to detect. It is instead possible that there are multiple clinical characteristics of type 2 diabetes that cause dementia, but that cannot be detected through 2-sample-MR analyses with a high p-value threshold of the included genetic variants for the disease as it is defined today. The fact that a PRS of type 2 diabetes was associated with all-cause dementia and VaD in this study shows that there instead may be other genetic risk markers that increase the risk of dementia through less strong or indirect pathways. This could partly be due to the complexity of both diseases. Findings from MRI studies and pre-clinical studies have previously shown that both vascular and neurodegenerative pathophysiological processes are likely of importance in causing diabetes-associated cognitive impairment (4). Although cardiovascular risk factors in people with type 2 diabetes is likely the predominant contributing cause through increasing the risk of VaD, studies have also shown possible negative effects on cognitive functioning of biomarkers such as glucose and Advanced Glycation End products (AGEs), as well as positive effects of insulin and incretin hormones such as glucagon-like peptide-1 (GLP-1) (5).

Another explanation for the strong epidemiological association between type 2 diabetes and dementia could be that there are causal pathways between type 2 diabetes and phenotypes close to dementia. For example, type 2 diabetes was causally associated with stroke and brain atrophy in a study(35). In other words, there are many pathways that need further investigation through studies including genetic risk markers of different phenotypes related to both type 2 diabetes and to dementia and cognitive impairment.

This study supports other MR studies where a causal link between type 2 diabetes and dementia have not been detected (7) (8) (9) (10). In contrast to other studies (15) (16), we have, however, identified a link between PRS for type 2 diabetes on the one hand and all-cause dementia and VaD on the other. One reason for this could be that we have well validated data on dementia diagnoses, in particular VaD, which is lacking in all other studies so far. These previous genetic studies have solely used register data on dementia, which is often less accurate. For example, a validation study from the UK Biobank comparing such data with validated diagnoses presented a positive predictive value of 82.5% for all-cause dementia, 71.4% for AD and 43.8% for VaD (18). Evidence from previous epidemiological studies showing strong links between diabetes and increased risk of in particular VaD supports these findings.

Our findings interestingly show a strong association between polygenic risk of type 2 diabetes, as well as clinical type 2 diabetes, and VaD in non-carriers of APOE ε4. This is in contrast to a meta-analysis that reported that carriers are more vulnerable to diabetes-associated dementia of all forms including VaD, although the study focused on clinical type 2 diabetes as exposure (19). It is possible that our results are partly influenced by less statistical power in the subgroup of APOE ε4 carriers. Another possibility is that carriers of APOE ε4 have a competing risk of AD and therefore exposed to a relatively lower risk of diabetes-associated dementia.

We found a protective effect of polygenic risk of higher fasting insulin on AD and all-cause dementia (although the estimates did not survive Bonferroni correction). We have not found any other study that examined this particular hypothesis, although epidemiological studies have shown an excess risk of dementia in people with lower serum insulin levels (36). Studies have also shown that higher insulin resistance is associated with greater dementia risk (11) (13), in line with this hypothesis. We could however not find a link between genetic risk of elevated HbA_1c (8) or fasting glucose (11) and dementia as some studies previously have reported.

The most important limitations of the study were the relatively low participation rate at baseline (41%) and the study sample size which is somewhat smaller than other studies investigating the same research questions. Follow-up for validated dementia diagnoses was restricted to the end of 2014 for logistical reasons. The study population, dominated by white Europeans, is derived from Skåne in Southern Sweden, why the findings may not be pertinent to broader populations at risk of type 2 diabetes and dementia, or other ethnicities. Furthermore, we could not perform external validation due to the non-availability of relevant data including validated dementia endpoints.

Further studies using both larger cohorts and validated dementia diagnoses are needed to confirm the results. More studies that investigate the role of genetic risk markers in non-carriers of APOE ε4 are also needed.

The results of this population-based study show that there is not a causal relationship between type 2 diabetes and dementia that can be detected by 2-sample-MR analysis. However, as type 2 diabetes is a heterogenous disease with a multifactorial etiology, it is possible that there are still factors that cause dementia among the contributing pathophysiological factors of type 2 diabetes, which could be identified in further studies. Our results showing that PRS of type 2 diabetes was associated with dementia indicate that weaker causal mechanisms could be present, although not detectable in this study.

Alzheimer’s Disease

AGE

Advanced glycation end products

ApoB/ApoA-ratio

Apolipoprotein B/Apolipoprotein A-ratio

APOE

Apolipoprotein E

BMI

Body Mass Index

CVD

Cardiovascular Disease

GLP-1

Glucagon-Like Peptide-1

GWAS

Genome Wide Association Study

HbA_1c

Hemoglobin A_1c

HWE

Hardy Weinburg Equilibrium

Linkage Desiquilibrium

MAF

Minor Allele Frequency

PRS

Polygenic Risk Score

Mendelian Randomization

SBP

Systolic Blood Pressure

SNP

Single Nucleotide Polymorphism

VaD

Vascular Dementia

Quality Control

Ethics approval and consent to participate

Consent for publication

Not applicable.

Availability of data and materials

The data that support the findings of this study are available from the Steering Committee of the Malmo Diet and Cancer Cohort, but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of Mr. Anders Dahlin, data manager (e-mail:[email protected]).

Competing interests

The authors declare that they have no competing interests.

Funding

The study was funded by the Research Council of Sweden (Grant K2011-65X-20752-04-6), the Anders Pålsson Foundation, the Ernhold Lundström Foundation, the Regional Agreement on Medical Training and Clinical Research (ALF) between Skåne County Council and Lund University, the Swedish Alzheimer’s Foundation (Alzheimerfonden), the Diabetes Fund (Diabetesfonden), and research grants from Region Skåne. The study funders were not involved in the design of the study; the collection, analysis, and interpretation of data; writing the report; and did not impose any restrictions regarding the publication of the report.

Authors' contributions

ED and AK analyzed and interpreted the data. OH, OM, NMC, PMN, GE, and KN participated in planning the data collection, analyses and interpretation of findings, and also in text editing. All authors read and approved the final manuscript.

Acknowledgements

The authors wish to thank Margaretha Persson and Gerd Östling for co-ordinating the data collection at baseline, Anders Dahlin for data administration, Anna-Märta Gustavsson for data preparation and all medical doctors at the Memory Clinic, Skåne University Hospital in Malmö, who carried out the dementia validation process.

Cheng G, Huang C, Deng H, Wang H. Diabetes as a risk factor for dementia and mild cognitive impairment: a meta-analysis of longitudinal studies. Intern Med J. 2012;42(5):484–91.
Biessels GJ, Verhagen C, Janssen J, van den Berg E, Zinman B, Rosenstock J, et al. Effect of Linagliptin on Cognitive Performance in Patients With Type 2 Diabetes and Cardiorenal Comorbidities: The CARMELINA Randomized Trial. Diabetes Care. 2019;42(10):1930–8.
Murray AM, Hsu FC, Williamson JD, Bryan RN, Gerstein HC, Sullivan MD, et al. ACCORDION MIND: results of the observational extension of the ACCORD MIND randomised trial. Diabetologia. 2017;60(1):69–80.
Biessels GJ, Despa F. Cognitive decline and dementia in diabetes mellitus: mechanisms and clinical implications. Nat Rev Endocrinol. 2018;14(10):591–604.
Dybjer E, Engström G, Helmer C, Nägga K, Rorsman P, Nilsson PM. Incretin hormones, insulin, glucagon and advanced glycation end products in relation to cognitive function in older people with and without diabetes, a population-based study. Diabet Med. 2020;37(7):1157–66.
Zheng J, Baird D, Borges MC, Bowden J, Hemani G, Haycock P, et al. Recent Developments in Mendelian Randomization Studies. Curr Epidemiol Rep. 2017;4(4):330–45.
Thomassen JQ, Benn M, Tolstrup JS, Frikke-Schmidt R. Type 2 diabetes and risk of alzheimer'S and vascular dementia: Observational and Mendelian randomization studies in > 1 million individuals. Atherosclerosis. 2020;315:e68-e9.
Garfield V, Farmaki AE, Fatemifar G, Eastwood SV, Mathur R, Rentsch CT, et al. The Relationship Between Glycaemia, Cognitive Function, Structural Brain Outcomes and Dementia: A Mendelian Randomisation Study in the UK Biobank. Diabetes. 2021.
Østergaard SD, Mukherjee S, Sharp SJ, Proitsi P, Lotta LA, Day F, et al. Associations between Potentially Modifiable Risk Factors and Alzheimer Disease: A Mendelian Randomization Study. PLoS medicine. 2015;12(6):e1001841-e.
Walter S, Marden JR, Kubzansky LD, Mayeda ER, Crane PK, Chang SC, et al. Diabetic phenotypes and late-life dementia risk: A mechanism-specific mendelian randomization study. Alzheimer Disease and Associated Disorders. 2016;30(1):15–20.
Pan Y, Chen W, Yan H, Wang M, Xiang X. Glycemic traits and Alzheimer's disease: a Mendelian randomization study. Aging. 2020;12(22):22688–99.
Benn M, Nordestgaard BG, Tybjærg-Hansen A, Frikke-Schmidt R. Impact of glucose on risk of dementia: Mendelian randomisation studies in 115,875 individuals. Diabetologia. 2020;63(6):1151–61.
Walter S, Marden JR, Kubzansky LD, Mayeda ER, Crane PK, Chang SC, et al. Diabetic Phenotypes and Late-Life Dementia Risk: A Mechanism-specific Mendelian Randomization Study. Alzheimer Dis Assoc Disord. 2016;30(1):15–20.
Lewis CM, Vassos E. Polygenic risk scores: from research tools to clinical instruments. Genome Medicine. 2020;12(1):44.
Andrews SJ, Fulton-Howard B, O'Reilly P, Marcora E, Goate AM, Farrer LA, et al. Causal Associations Between Modifiable Risk Factors and the Alzheimer's Phenome. Annals of Neurology. 2020.
James SN, Wong A, Tillin T, Hardy R, Chaturvedi N, Richards M. The effect of mid-life insulin resistance and type 2 diabetes on older-age cognitive state: the explanatory role of early-life advantage. Diabetologia. 2019;62(10):1891–900.
Repple J, König A, Opel N, Mauritz M, de Lange SC, Kircher T, et al. Genetic Risk for Type 2 Diabetes is Associated With Impaired Structural Brain Connectivity and Worse Cognitive Performance in Nondiabetic Patients With Major Depressive Disorder. Biological Psychiatry. 2020;87(9):S398-S9.
Wilkinson T, Schnier C, Bush K, Rannikmäe K, Henshall DE, Lerpiniere C, et al. Identifying dementia outcomes in UK Biobank: a validation study of primary care, hospital admissions and mortality data. Eur J Epidemiol. 2019;34(6):557–65.
Li L, Cavuoto M, Biddiscombe K, Pike KE. Diabetes Mellitus Increases Risk of Incident Dementia in APOEɛ4 Carriers: A Meta-Analysis. J Alzheimers Dis. 2020;74(4):1295–308.
Manjer J, Carlsson S, Elmståhl S, Gullberg B, Janzon L, Lindström M, et al. The Malmö Diet and Cancer Study: representativity, cancer incidence and mortality in participants and non-participants. Eur J Cancer Prev. 2001;10(6):489–99.
Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, Lee JJ. Second-generation PLINK: rising to the challenge of larger and richer datasets. Gigascience. 2015;4:7.
Mahajan A, Taliun D, Thurner M, Robertson NR, Torres JM, Rayner NW, et al. Fine-mapping type 2 diabetes loci to single-variant resolution using high-density imputation and islet-specific epigenome maps. Nature Genetics. 2018;50(11):1505–13.
Chen J, Spracklen CN, Marenne G, Varshney A, Corbin LJ, Luan Ja, et al. The trans-ancestral genomic architecture of glycemic traits. Nature Genetics. 2021;53(6):840–60.
Manning AK, Hivert MF, Scott RA, Grimsby JL, Bouatia-Naji N, Chen H, et al. A genome-wide approach accounting for body mass index identifies genetic variants influencing fasting glycemic traits and insulin resistance. Nat Genet. 2012;44(6):659–69.
Lagou V, Mägi R, Hottenga J-J, Grallert H, Perry JRB, Bouatia-Naji N, et al. Sex-dimorphic genetic effects and novel loci for fasting glucose and insulin variability. Nature Communications. 2021;12(1):24.
Gustavsson AM, van Westen D, Stomrud E, Engström G, Nägga K, Hansson O. Midlife Atherosclerosis and Development of Alzheimer or Vascular Dementia. Ann Neurol. 2020;87(1):52–62.
K Nägga VB, E Stomrud, O Melander, P M Nilsson, A-M Gustavsson, O Hansson. Prevalence and ascertainment of dementia cases in the Malmö Diet and Cancer study. JAD Reports. 2022.
Ludvigsson JF, Andersson E, Ekbom A, Feychting M, Kim JL, Reuterwall C, et al. External review and validation of the Swedish national inpatient register. BMC Public Health. 2011;11:450.
Association AP. Diagnostic and statistical manual of mental disorders (DSM-5®): American Psychiatric Pub; 2013.
Biessels GJ, Staekenborg S, Brunner E, Brayne C, Scheltens P. Risk of dementia in diabetes mellitus: a systematic review. Lancet Neurol. 2006;5(1):64–74.
Biessels GJ, Strachan MW, Visseren FL, Kappelle LJ, Whitmer RA. Dementia and cognitive decline in type 2 diabetes and prediabetic stages: towards targeted interventions. Lancet Diabetes Endocrinol. 2014;2(3):246–55.
Safieh M, Korczyn AD, Michaelson DM. ApoE4: an emerging therapeutic target for Alzheimer's disease. BMC Med. 2019;17(1):64.
Mahley RW. Apolipoprotein E: from cardiovascular disease to neurodegenerative disorders. J Mol Med (Berl). 2016;94(7):739–46.
Koren-Iton A, Salomon-Zimri S, Smolar A, Shavit-Stein E, Dori A, Chapman J, et al. Central and Peripheral Mechanisms in ApoE4-Driven Diabetic Pathology. Int J Mol Sci. 2020;21(4).
Georgakis MK, Harshfield EL, Malik R, Franceschini N, Langenberg C, Wareham NJ, et al. Diabetes Mellitus, Glycemic Traits, and Cerebrovascular Disease: A Mendelian Randomization Study. Neurology. 2021.
Mehlig K, Lapidus L, Thelle DS, Waern M, Zetterberg H, Björkelund C, et al. Low fasting serum insulin and dementia in nondiabetic women followed for 34 years. Neurology. 2018;91(5):e427.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Polygenic risk of type 2 diabetes is associated with Vascular Dementia but not with Alzheimer's Disease: a Prospective Cohort Study

Status:

Version 1

Abstract

Background and Objectives

Methods

Results

Conclusions

Figures

Introduction

Methods

Participants

Genetic QC

Polygenic Risk Score Calculation

Clinical type 2 diabetes

Validated dementia endpoints

Statistical analyses

Data Availability

Standard Protocol Approvals, Registrations, and Patient Consents

Results

Discussion

Conclusions

Abbreviations

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1