The University of Edinburgh
Risk prediction models for dementia in people with Type 2 diabetes
Abstract
dc:description.abstractINTRODUCTION: Dementia is a complex, multifactorial disease, which results from the interplay of various genetic, lifestyle, and environmental factors. Type 2 Diabetes Mellitus (DM) is associated with a 70% greater risk for developing dementia. Over the last few years, the focus on treating dementia in later life has shifted to developing early treatment and preventive interventions to delay or even prevent the onset of dementia. One approach towards early identification can be achieved by the development of risk prediction models. The current dementia risks models for diabetes populations are limited and mainly based on traditional risk factors and/or co-morbid conditions. There is a huge scope to develop, update, improve and test risk models based on novel biomarkers and associated factors which are pertinent to diabetes-related pathological changes leading to dementia. AIMS: The study aim was to assess any improvement in the prediction ability of two base models reported in the literature in an Edinburgh diabetes cohort (ET2DS), based on the addition of i) important vascular risk factors reported in the scientific literature ii) inflammatory markers studied in the ET2DS and iii) Individual and combined cognitive tests measured at baseline in ET2DS. METHODS: Statistical analysis for this project was based on the Edinburgh Type 2 Diabetes Study (ET2DS), a population-based, diabetes-specific cohort of 1066 men and women living in Scotland, UK. The strength of E2TDS for this project includes its prospective design and availability of data on novel biomarkers to study various diabetes complications including dementia. For this PhD thesis project, dementia outcome assessment was performed using electronic data records including Information Services Division (ISD) database, TRAK data files (electronic patient management system, Lothian) and death certificates during 10 years of follow-up. Cognitive data was gathered by means of a comprehensive cognitive test battery applied at baseline, when data were also gathered on medical history, vascular events, circulating biomarkers, inflammatory markers (including Interlukin-6, Tumour necrosis factor alpha and C-reactive protein) and physiological parameters from physical examination. A systematic review (SR) was performed, and, among the published diabetes-specific risk models, two models were selected for testing in ET2DS. Based on the predictors of these models, base models were constructed in ET2DS, and their predictive performances were assessed. Another SR was performed to identify risk factors associated with dementia in type 2 diabetes populations. The factors identified in this SR and for which data was available in ET2DS, were identified as potential predictors, and were then assessed further (for ultimate inclusion in my models) according to biological significance, clinical significance, completeness of data and correlation between similar predictors. Cross validation with 10 folds employing logistic regression models was used to assess the predictive abilities of the base models, before and after the addition of the selected panels of risk factors. Area under the curve, along with sensitivity and specificity, were calculated to assess predictive performance. RESULTS: In ET2DS study, 103 people developed dementia out of 1065 participants followed over 10 years of the study period. The average age of the cohort at the start of the study was 67.90 ±4.20 years, around 547 (51.36%) were males and average duration of DM was 8.05±6.44 years. The two base models identified from the literature were based on a combination of age, gender, duration of DM (Diabetes Mellitus), BMI (Body Mass Index), fasting plasma glucose, HbA1c (Glycated haemoglobin), cerebrovascular disease and cardiovascular disease. Traditional and vascular predictors added to these base models included marital status, occupational status, serum cholesterol, neuropathy, alcohol usage, hypertension along with Scottish index for multiple deprivation score. The Area under the curve (AUC) for the base models was around 57%. Age as a predictor showed significant association with dementia (odds ratio (OR) 1.07; 95% CI 1.01-1.13) and females were less likely to develop dementia as compared to men (OR 0.62; 95% CI 0.38-0.99). With the addition of new traditional and vascular predictors, the AUC decreased to around 53%. Among the inflammatory markers tested, interlukin-6 showed a statistically significant association with dementia (OR 1.05; 95% CI 1.00-1.12). With the addition of inflammatory markers to the base models, the AUC reached 61%. All the cognitive tests showed a statistically significant association with dementia, moreover Trail-Making Test B showed AUC of 66% when assessed as the single predictor for dementia. CONCLUSIONS: Age, depression, male sex, inflammatory markers and cognitive measures were found to be associated with, and possible predictors of, dementia in older people with type 2 diabetes. Generally, trail-making Test B had improved predictive performance as compared to other markers combined together like combinations of multiple traditional, vascular and/or inflammatory markers. Cognitive tests as a single measure or in combination, administered in regular clinical practice, may prove more useful than risk models based on multiple traditional dementia risk factors to assess the early signs of dementia and predict risk of dementia onset in people attending diabetes clinics.
Degree
thesis:*- Grantor dc:publisher
- The University of Edinburgh
- Year dc:date.issued
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Khalid, Wardah
- Advisors dc:contributor.advisor
-
- Price, Jackie
- Muniz Terrera, Graciela
- Spiliopoulou, Athina
Subjects
dc:subject × 6Rights
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Identifier URI
- http://dx.doi.org/10.7488/era/5291