We performed a cross-sectional study of 257 older adults aged ≥70 years who participated in the Coyoacán Cohort, an observational study conducted in the Coyoacán borough located in Southern Mexico City conducted between 2008 and 2009. Complete methodological details for the design and protocols of this study have been published elsewhere, and questionnaire materials were similar for the full cohort [7, 12]. Briefly, recruited participants were non-institutionalized older adults with residence in Coyoacán who were registered at the comprehensive “Food Support, Medical Care and Free Drugs Program” (FMDP) government program. All participants underwent face-to-face interviews to collect self-reported socio-demographic and health data and a comprehensive geriatric assessment which included physical performance tests, cognitive, nutritional, and medical assessment. Baseline questionnaire data were collected between April–May 2008; clinical evaluation and biological sample collection were carried out between June 2008 and July 2009. In the original cohort, a sample of 1294 participants was calculated to ensure a sample size which could estimate a prevalence of ~ 14% of frailty with α = 0.05 and β = 0.20. Among contacted potential participants, acceptance rate was 86.9% and a total of 1124 participants completed the initial interview, which included individuals with and without T2D. For the present study, we included participants with T2D without previous clinical diagnosis of dementia, defined as self-report of previously diagnosed T2D and/or self-report of taking T2D medications (n = 236). To account for subjects who were not previously diagnosed with T2D, we included subjects with fasting glucose ≥126 mg/dL (n = 21), who had enough information for their dementia-risk stratification using the DSDRS (overall, n = 257).
Definitions for potential correlates with DSDRS
We used a modified definition of the one proposed by Fried et al. which was previously validated for this population [13, 14]. This modified definition uses data from questionnaires and self-report to define the following dominions, previously described as: a) Unintentional weight loss ≥5 kg in the last 12 months, b) Exhaustion, c) Low physical activity, d) Slowness, and e) Weakness. Participants were categorized as frail if they fulfilled ≥3 criteria, pre-frail if they fulfilled 1–2 criteria, and non-frail if none.
The presence of depressive symptoms was defined as a score > 5 in the 15-item version of the Geriatric Depression Scale (GDS).
Cognitive evaluation comprised an interview-based assessment, which included a questionnaire-based cognitive evaluation comprising Mini-Mental State Examination (MMSE) evaluation, verbal fluency abilities with the Isaacs Set Test (IST) where four semantic categories were successively used (cities, fruits, animals, and colors) and the Clock-drawing test to assess visuo-constructional abilities. Low cognitive performance was based on a modified definition by Blaum et al., defined as scores <25th percentile in both MMSE and the IST semantic verbal fluency test or clock-drawing test, adjusted for sex, age, and schooling based on normative cutoffs previously validated in the Coyoacán Cohort Study to account for inter-ethnic influences on MMSE scores which impact on the sentitivity of the test to detect cognitive impairment [12, 15].
Determined using Lawton-Brody Instrumental Activities of Daily Living (IADL) scale and Katz Index for the Activities of Daily Living (ADL). We defined ADL or IADL disability as having at least one impaired dominion in the Katz scale (ADL disability) or the Lawton-Brody scale (IADL disability) [16,17,18].
Risk of malnutrition
Assessed by the Mini-Nutritional Assessment (MNA) questionnaire, scores < 24 were indicative of at-risk of malnutrition.
Assessed using the self-administered generic instrument SF-36 health questionnaire in the translated and validated version for Mexican population. Items are formulated as statements to evaluate eight specific health scales including physical functioning, physical pain, role limitations due to physical health problems, role limitations due to personal or emotional problems, emotional well-being, social functioning, energy/fatigue and general health perceptions. Scales were classified in two physical (PCS) and mental component scores (MCS).
Dementia risk calculation
We evaluated dementia risk using the DSDRS . Self-reported variables included in the score considered duration of T2D from diagnosis in years, self-report of diabetic kidney disease (DKD) and diabetic retinopathy, history of insulin use or oral T2D treatment, previous diagnosis of diabetic foot or peripheral vascular disease, acute myocardial infarction, and stroke. Microvascular complications to estimate DSDRS considered the clustering of DKD, and/or diabetic retinopathy; this definition was also used in linear and logistic regression models. Acute metabolic event was defined as a previous episode of hyperglycemia which required hospitalization or hypoglycemia, defined by self-reported episodes of hyperglycemia or fasting glucose levels < 70 mg/dL as recommended by ADA guidelines. High dementia-risk was defined as an estimated 10-year dementia risk >75th age-specific percentile based on incremental 5-year age categories described by the DSDRS. The selection of this age-specific percentile cut-off was performed to provide fair comparisons of DSDRS to detect age-related phenotypes without a significant influence from age as considered in the DSDRS, whilst still allowing a detection of high-risk individuals within each age group.
Anthropometric and biochemical evaluation
We calculated the body mass index using anthropometric evaluation using the formula of weight in kilograms divided by height in m2. Blood samples were acquired after a 10–12 h fast to measure fasting glucose (Yellow Springs Instruments Co.); serum lipid concentrations assessed total cholesterol, triglycerides and HDL-C and were measured using colorimetric assays (Unicel DxC 600 Synchron Clinical System Beckman Coulter).
We compared groups according to the 75th age-adjusted percentile of DSDRS using Student’s t-test or Mann-Whitney U according to variable distribution. In all descriptive analyses, distribution of categorical variables is reported as frequencies which were compared between groups using chi-squared tests. A p-value < 0.05 was established as statistically significant.
Correlation between DSDRS, cognitive tests, frailty, and disability components
To investigate the association between dementia risk and the evaluated scores, we tested the correlation of DSDRS with the continuous MMSE, the IST, CDT, MNA, SF-36, Lawton, and Katz scores as well as the number of frailty components using Spearman’s correlation; 95% confidence intervals were estimated using 1000 bootstrap samples. To develop an explanatory model for DSDRS and identify independent predictors for dementia risk using these scores, we used step-wise multiple linear regression analyses adjusted for sex, years of schooling and years since diabetes diagnosis, with model selection carried out using Bayesian Information Criterion (BIC) minimization.
Logistic regression analyses
We developed an explanatory model for high-estimated 10-year dementia-risk to investigate the relation of subjects at higher risk with the investigated clinical phenotypes identified by the scores when transformed into categorical variables. For this purpose, we used logistic regression, treating high-dementia risk as the dependent variable and including as predictors frailty, ADL and IADL disability, risk of malnutrition and low cognitive performance; multiple logistic regression was carried out using step-wise models adjusted for years since diabetes diagnosis, years of schooling and sex. Model diagnostics were conducted using R2 and the Hosmer-Lemeshow test. Finally, we constructed ROC curves to estimate performance of DSDRS to identify phenotypes of frailty, low cognitive performance, ADL and IADL disability using probability estimates from regression modes; we also calculated sensitivity and specificity for each phenotype.
Contribution of DSDRS components to the observed associations
To investigate whether the association of DSDRS with cognition, disability, frailty and impaired QoL were driven by factors other than age, we fitted multiple linear regression models to evaluate which components of the DSDRS were primarily associated with the outcomes. Predictors included individual components of the DSDRS, including age, microvascular complications, depression, diabetic foot, acute metabolic events, cardiovascular and cerebrovascular disease. We included as dependent variables scores correlated to DSDRS, which included the frailty score, Lawton, Katz, MMSE, MNA and SF-36 PCS. Model diagnostics were conducted using R2 and BIC; multicollinearity was assessed using tolerance and variance inflation factor (VIF). Predictors were tested on homoscedasticity and linearity assumptions; model diagnostics were conducted evaluating normality of residuals. Model parameters are expressed using β-coefficients and 95%CI. All statistical analyses were performed using the SPSS software (Version 22.0), R (Version 3.6.1) and GraphPad Prism (Version 6.0).