Introduction
Diabetes is a chronic disease characterised by disorders of glucose metabolism. With the significant change of people’s diet structure and lifestyle, the incidence of diabetes has also been increasing annually, with a trend towards a younger age at diagnosis. In 2019, the number of patients with diabetes reached 116 million in China, which was a larger number than that in any other country in the world. Notably, the young and middle-aged populations accounted for more than 50% of the total number of cases.1 Diabetes causes a variety of comorbidities, including cognitive impairment (CI). It is a cognitive state between normal brain ageing and dementia, and is mainly characterised by a decline in cognitive abilities in memory, language, execution, attention and other cognitive domains.2 Cognitive decline and the ensuing dementia mean loss of function, placing huge burdens on individuals and society. Diabetes is a recognised risk factor for CI. Persistent hyperglycaemia and chronic inflammation in patients with diabetes can cause progressive damage to multiple organs (including the brain), affect cognitive function and contribute to CI.3 Moreover, diabetes accelerates the progression of CI to more severe stages, such as dementia.4 Patients with CI and diabetes have a 1.5–3.0 times higher risk of progressing to dementia.3
Obesity is not only a common comorbidity in patients with diabetes, but also has an impact on cognitive function. Results from previous studies suggest that obesity in midlife may impair cognitive function through metabolic syndrome or cardiovascular disease,5 while the relationship between obesity and cognitive performance in older adults is more complex and controversial.6 Many prospective studies have shown that a decline in body mass index (BMI) predicts the development of CI in older adults.6 The decrease in BMI in the elderly may be the result of early cognitive dysfunction. Only a few studies have investigated the relationship between obesity and cognitive function in elderly diabetic patients, suggesting that central obesity is a risk factor for cognitive deficits.7 Anyways, there are few research on cognition and obesity in middle-aged persons with diabetes, although adiposity is quite common in this population.
In recent years, increasing attention has been paid to global research on diabetes-associated cognitive dysfunction. However, to date, most of these studies are limited to older patients with type 2 diabetes (T2DM). Relatively few studies have been conducted on middle-aged patients with T2DM. More importantly, it has been reported that cognitive decline associated with diabetes is not limited to older individuals but may already begin to appear in middle-aged individuals.8 Some scholars have suggested that more attention should be paid to the prevention and control of CI in midlife rather than twilight years in T2DM patients.9 10 Therefore, identifying CI early in middle-aged individuals with diabetes has important implications for its prevention and development.
Considering that there are currently few effective methods to prevent or treat CI or dementia, determining modifiable risk factors for effective interventions is important. The present study aimed to explore the risk factors of middle-aged T2DM patients complicated by CI, particularly considering the influence of overall and abdominal obesity, in order to support strategies for the early diagnosis of CI in middle-aged T2DM patients.
Study population
Sample size calculations of this cross-section study were estimated using PASS V.11 software. Based on the results documented in most previous literature, we identified the prevalence of CI in T2DM patients as 46%.11 The allowable error was 4.6%, the confidence level was 0.95 and a total sample size of 471 persons was required. In order to minimise selection bias as much as possible, we recruited 524 consecutive patients with T2DM from among hospitalised patients at the Department of Endocrinology and Metabolism of the Tianjin Union Medical Center, from July 2018 to September 2021. T2DM was diagnosed according to the American Diabetes Association (ADA) criteria:12 fasting plasma glucose≥7.0 mmol/L and/or postprandial glucose≥11.1 mmol/L; or in a patient with classic symptoms of hyperglycaemia or hyperglycaemic crisis, a random plasma glucose level≥11.1 mmol/L or haemoglobin A1c (HbA1c) ≥ 6.5%. All patients were diagnosed with T2DM by oral glucose tolerance test, so that transient elevated blood glucose could be excluded.
The inclusion criteria for the study were as follows: age 40–64 years, diagnosis of T2DM, capability to complete neuropsychological tests independently. The midlife period is now commonly considered to begin at ages 35–45 and end at age 64, while those aged 65 and older are classified as elderly.13 Thus, we selected T2DM patients aged between 40 and 64 years as the study population. The exclusion criteria were as follows: (1) inability to complete neuropsychological tests due to communication difficulties, physical disability, severe limitation of movement or severe vision, hearing, reading, language impairment or for any other reasons. (2) Anaemia (haemoglobin (Hb) < 90 g/L), cachexia, liver insufficiency (alanine transaminase (ALT) ≥ 120 U/L), renal insufficiency (creatinine (Cr) ≥ 265 µmol/L), severe cardiopulmonary insufficiency, thyroid dysfunction and severe infection. (3) Parkinson’s disease, epilepsy, brain trauma, encephalitis, brain tumours, schizophrenia, severe depression, diagnosed dementia, alcohol or drug addiction and long-term use of drugs affecting cognitive function.
Screening evaluation for CI
This study used standard screening tools (the Mini-Mental State Examination (MMSE) and the Montreal Cognitive Assessment (MoCA)) for early screening of CI in middle-aged patients with T2DM, rather than an informal diagnosis. According to the US Preventive Services Task Force, screening for CI and dementia does not equal a diagnosis.14 In view of cultural and linguistic differences, we used the Chinese version of MMSE and MoCA to assess cognition, which Katzman and Lu et al have demonstrated to be valid and reliable in the Chinese population.15 16 In a quiet environment, both cognitive tests were conducted face-to-face by qualified professionals in strict accordance with the guidelines.
The MMSE is currently the broadest test tool for screening for dementia in clinical and research settings, with a total score of 30.14 The questionnaire included simple tasks in several cognitive domains: orientation (time and place), memory and recall, attention, arithmetic and language.17 In previous studies, the most commonly used tangential value for the MMSE was 23/24 (score below 24 is considered dementia), with total sensitivity and specificity reaching 88.3% and 86.2%, respectively.14 The main focus of this study was CI, and dementia was not part of our study, so the MMSE was mainly used to exclude patients with possible dementia (score below 24).
The MoCA is an effective mild CI screening tool developed by Nasreddine et al, with a sensitivity of up to 90% and specificity of 87%,18 and is now widely used in clinical practice.16 19 The test has a total score of 30 and includes 12 items such as memory, visuospatial ability, animal naming, language, arithmetic, attention, abstract ability, orientation of time and place.18 Subjects with higher MoCA scores are associated with better cognitive function, and scores below 26 are considered likely to have CI.20 An additional point is added if the subject has less than 12 years of schooling.20 Individuals were grouped based on these scores, as follows: normal cognitive function (NCF) group: MMSE≥24 and MoCA≥26; CI group: MMSE≥24 and MoCA<26.
Data collection
Demographic characteristics and diabetes-related information were obtained through a questionnaire survey and hospitalisation medical records. The data collected included sex, age, marital status, education level, history of smoking and drinking, lifestyle, diabetic dietary pattern, age at diagnosis of diabetes, duration of diabetes, use of antidiabetic medications and lipid lowering agents, hypoglycaemic episodes in the recent 3 months, diabetic microangiopathy, low extremity atherosclerosis disease (LEASD), diabetic peripheral neuropathy (DPN), coronary heart disease (CHD), hypertension and cerebrovascular disease (CVD). The diabetic dietary pattern refers to a diversified dietary pattern based on cereals, high dietary fibre intake, low salt, low sugar and low fat intake and a reasonable and balanced intake of calories and various nutrients required by the body, such as increasing the intake of vegetables and dietary fibre, reducing the intake of alcohol, saturated fatty acids and simple sugars, rather than simply dieting or eating less.21 A sedentary lifestyle is defined as less than 150 min of moderate intensity (50%–70% of maximum heart rate) aerobic exercise per week.21 Diabetic microangiopathy included diabetic nephropathy and/or diabetic retinopathy. CVD included haemorrhagic and/or ischaemic strokes. Hypertension was defined as a blood pressure≥140/90 mm Hg or the use of antihypertensive medications.22
While wearing light clothes and without shoes, participants’ height and weight were measured and recorded by professional nurses. BMI (kg/m2) was obtained by dividing the weight (kg) by the square of the height (m2). At the end of the patient’s expiration, waist circumference (WC (cm)) was measured horizontally with a graduated tape at the midpoint of the line from the lowest rib and the iliac crest. According to the WHO,23 18.5≤BMI<25 kg/m2, 25≤BMI < 30 kg/m2 and BMI≥30 kg/m2 were classified as normal weight, overweight and obese, respectively. In addition, there were no underweight (BMI<18.5 kg/m2) subjects in the study. Abdominal obesity (AO) is defined as a WC≥90 cm for men and ≥85 cm for women.24
After 10–12 hours of overnight fasting, venous blood was drawn by professional nurses in the early morning and samples were analysed by the central laboratory for the following indicators: white cell count (WCC), neutrophil count, lymphocyte count, red blood cell count, Hb haematocrit, platelet count (PLT), calcium, phosphorus, alanine transaminase (ALT), creatinine (Cr), urinary microprotein/creatinine (ACR), uric acid (UA), creatine kinase (CK), total cholesterol (TC), triglyceride (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), homocysteine (HCY), thyroid-stimulating hormone and HbA1c. Hyperhomocysteinaemia was defined as HCY≥15 µmol/L.25
Statistical analysis
In this study, SPSS software (V.25.0; IBM, Armonk, NY, USA) was used for statistical data analysis. The Kolmogorov-Smirnov test was used to test whether the measurement data were normally distributed. For descriptive statistics, the measurement data that conformed to the normal distribution were described as mean±SD, and those that did not fit an abnormal distribution were described by the median and IQR (P25, P75). Counting data are represented as number (percentage). In the statistical analysis, Student’s t-test was used to compare normally distributed data between the two groups, while the non-parametric Mann-Whitney U test was used to compare the non-normally distributed data. A χ2 test was used for comparing counting data. Multiple linear regression analysis was used to test whether there was multicollinearity between covariables. Multivariate logistic regression analysis and generalised linear model (GLM) were used to screen the risk factors of CI and explore the relationships between obesity parameters and the risk of CI. We also calculated the subjects’ relative risks (RR) of developing CI. Linear regression analysis was used to investigate the correlations between adiposity measures (BMI and WC) and cognitive scores (the MMSE and MoCA). Piecewise linear regression and interaction analysis were used to further study the relationships between obesity parameters and cognitive function under different categories. The diagnostic performances of individual or multiple influencing factors were assessed by the area under the curve (AUC) in the receiver operating characteristic (ROC) curve analysis. All tests were two-tailed, p<0.05 was regarded as indicating a significant difference.
Results
Online supplemental figure S1 shows the specific screening process of the participants and the reasons for their exclusion. Finally, 524 eligible participants with a median age of 58 years were enrolled, which included 285 males (54.4%) and 239 females (45.6%). They were divided into two groups according to their cognitive function assessment: 320 (61.1%) in the NCF group and 204 (38.9%) in the CI group.
Supplemental material
Characteristics of the study subjects
Demographic and clinical characteristics of the participants are presented in table 1. Compared with individuals in the NCF group, age (59.5 vs 57.0 years), WC (94.6 vs 91.9 cm) and the proportions of participants using insulin (55.4% vs 44.1%), with normal weight (45.0% vs 37.0%), with AO (78.3% vs 61.1%), with low extremity atherosclerosis disease (54.5% vs 44.7%) and with CVD (28.2% vs 13.5%) were significantly higher in patients with CI. In contrast, education attainment (9 vs 12 years), MMSE scores (29 vs 30), MoCA scores (24 vs 27) and the proportions of participants who were married (88.7% vs 97.2%), overweight (44.6% vs 48.4%) and obese (10.4% vs 14.6%) and who used diabetic dietary pattern (58.3% vs 71.2%) were significantly lower in the CI group. There were no significant differences in gender, BMI, current smoking, current drinking, sedentary lifestyle, duration of diabetes, age at diabetes diagnosis, the use of several antidiabetic drugs and statins, hypoglycaemic episode in 3 months, diabetic microangiopathy, DPN, CHD or hypertension (p>0.05). In addition, there were no significant differences in WCC, NEUT, LYM, NLR (NEUT/LYM), RBC, Hb, HCT, PLT, calcium, phosphorus, ALT, creatinine, ACR, UA, CK, TC, TG, HDL-C, LDL-C, THR (TG/HDL-C), HCY, hyperhomocysteinaemia, TSH or HbA1c (p>0.05).
Clinical characteristics of the participants grouped by cognitive function
Risk factors for CI complication in middle-aged T2DM patients
Binary logistic regression analysis and GLM were utilised to identify potential risk factors for CI complication in middle-aged T2DM patients. The dependent variable was coexisting CI. Eleven variables with p<0.1 in univariate analysis (age, marital status, education, diabetic dietary pattern, age at diabetes diagnosis, insulin use, BMI Cat, WC, low extremity atherosclerosis disease, CVD and neutrophil count-to-lymphocyte count ratio) and two variables that have been documented to be closely associated with the onset of CI (hypoglycaemic episode within past 3 months and hyperhomocysteinaemia)26 27 were introduced into the multivariate logistic regression model and GLM, respectively. Variables for inclusion were carefully chosen, given the number of events available, to ensure parsimony of the final model. In addition, the result of linear regression analysis showed that the variance inflation factors of all covariables were less than 10, and the tolerance was greater than 0.1 (online supplemental table S1), which meant that there was no multicollinearity between covariables.28 Table 2 shows the seven final factors contributing to CI as determined by logistic regression analysis. Age (OR=1.062, 95% CI: 1.014 to 1.113), WC (OR=1.125, 95% CI: 1.087 to 1.163), hypoglycaemic episode occurring within the past 3 months (OR=1.653, 95% CI: 1.134 to 2.822) and the presence of CVD (OR=2.510, 95% CI: 1.451 to 4.340) were independent risk factors, while the independent protective factors were having a better education (OR=0.835, 95% CI: 0.764 to 0.914), using diabetic dietary pattern (OR=0.487, 95% CI: 0.310 to 0.763), being overweight (OR=0.193, 95% CI: 0.107 to 0.347) and being obese (OR=0.020, 95% CI: 0.006 to 0.065). As an extension of the linear regression model, GLM is able to handle response variables that are not normally distributed, such as binary data, by specifying a link function. The flexible and powerful framework of GLM also allows for any forms of predictors.29 The same results were obtained using GLM analysis, with age, WC, hypoglycaemic episode within past 3 months and CVD as risk factors for CI, and education level, diabetic dietary pattern, overweight and obesity as protective factors (online supplemental table S2).
Multivariate logistic regression analysis to identify independent risk factors for cognitive impairment in middle-aged T2DM patients
The prevalence of CI in patients with T2DM in this study has exceeded 20%, and the ORs of some influencing factors have been greater than 2.5 or less than 0.5 (such as diabetic dietary pattern, overweight, obesity and CVD), in which case the OR may overestimate the true risk of disease.30 Therefore, we also calculated the RR to more accurately assess the impact of potential risk factors on developing CI. As expected, the RR for all categorical variables were closer to 1 than the OR (RR was still statistically significant), that is, the RRs of developing CI were smaller (table 2). In addition, because WC is a continuous variable, its RR calculation is grouped according to AO.
The relationships between adiposity (BMI and WC) and cognitive function in middle-aged T2DM patients
Adjusted ORs of adiposity measures for CI
After adjusting for potential confounders (age, education, diabetic dietary pattern, hypoglycaemic episode within the past 3 months and CVD), the results of multiple logistic regression analysis show that BMI (OR=0.937, p=0.023) and obesity (OR=0.432, p=0.010) were protective factors against CI, whereas WC (OR=1.022, p=0.015) and AO (OR=2.217, p<0.001) were risk factors (table 3). In order to assess the independent associations of general and central adiposity with CI, both the WC and BMI variables were simultaneously included in the model. After further adjustments of AO or WC, BMI (OR=0.809 or 0.572, p<0.001), overweight (OR=0.241 or 0.191, p<0.001) and obesity (OR=0.125 or 0.018, p<0.001) showed more significant inverse associations with CI. Similarly, WC (OR=1.125 or 1.194, p<0.001) and AO (OR=6.635 or 5.670, p<0.001) also showed more significant positive associations with CI while accounting for BMI Cat or BMI (table 3).
Adjusted ORs of adiposity measures for cognitive impairment
We also repeated the above analysis process using GLMs to verify that obesity parameters are independent influences of CI, and we got exactly the same results.
Linear regression analysis of adiposity measures and cognitive scores
As shown in table 4, after adjusting for potential confounders (age, education, diabetic dietary pattern, hypoglycaemic episode within the past 3 months and CVD), there was no significant correlation between BMI and cognitive MMSE score (β=0.019, p=0.146). However, after further adjustment of AO or WC, a significant positive correlation between BMI and the MMSE score emerged (β=0.034 or 0.084, p=0.039 or 0.001). Similarly, WC and the MMSE score did not present a statistical association after adjusting for other confounders (β=−0.002, p=0.671). The negative association between WC and the MMSE score became significant after further adjustment for BMI Cat or BMI (β=−0.014 or −0.024, p=0.035 or 0.002). BMI was positively correlated with cognitive MoCA score with adjustment of confounders (β=0.081, p=0.006). Further adjustment for AO or WC resulted in a more significant positive association between BMI and the MoCA score (β=0.197 or 0.392, p<0.001). In contrast, there was no correlation between WC and the MoCA score after adjustment (β=−0.009, p=0.327). However, when BMI Cat or BMI was further included in the multiple linear regression model, a significant negative correlation between WC and the MoCA score appeared (β=−0.080 or −0.114, p<0.001).
The association of adiposity measures with cognitive scores
Piecewise linear regression can fit multiple linear models over the entire range of variables of interest, showing more clearly the non-linear association between the exposure and outcome.31 Thus, we tested for differences in the associations between adiposity measures and cognitive scores across the categories by using piecewise regression models and interaction analysis. Piecewise linear regression analysis showed that there was no significant correlation between BMI and the MMSE score in different BMI categories. However, in both normal and overweight cases, BMI was positively correlated with the MoCA score after adjusting for WC (β=0.289 and 0.264). In the case of obesity, there was no significant correlation between BMI and the MoCA score (online supplemental table S3). On the other hand, in the case of non-AO, WC was negatively correlated with the MMSE score after adjusting BMI Cat or BMI (β=−0.041 or −0.057). In the case of AO, there was no significant correlation between WC and the MMSE score. In addition, regardless of the presence or absence of AO, WC was negatively associated with the MoCA score after adjusting for BMI Cat or BMI (online supplemental table S3).
Interaction analysis showed that there was no interaction between BMI, AO and cognitive scores after adjusting for confounders (age, education, diabetic dietary pattern, hypoglycaemic episode in 3 months and CVD). Regardless of the presence or absence of AO, BMI was associated with higher MMSE (P for interaction=0.616) and MoCA scores (P for interaction=0.122). Patients with AO, whether normal weight or overweight, had lower MMSE (P for interaction=0.616) and MoCA scores (P for interaction=0.122) than patients with normal WC (online supplemental figure S2).
ROC curve analysis of individual or multiple factors for CI in middle-aged T2DM patients
ROC curve analysis was conducted to understand the diagnostic value of the single and combined detection capabilities of adiposity measures (BMI and WC) for CI. The results of ROC curve analysis (figure 1A and table 5) suggested that the predictive value of WC (AUC=0.604) was higher than that of BMI (AUC=0.538). The combination of WC and BMI testing greatly improved the predictive performance for CI (AUC=0.754). These findings implied that the combination of BMI and WC might have more diagnostic value for cognitive decline than either measure alone in middle-aged T2DM patients.
ROC curve analysis for cognitive impairment. Model 1: including age, education, diabetic dietary pattern, hypoglycaemic episode in 3 months, cerebrovascular disease and BMI. Model 2: including age, education, diabetic dietary pattern, hypoglycaemic episode in 3 months, cerebrovascular disease and WC. Model 3: including age, education, diabetic dietary pattern, hypoglycaemic episode in 3 months, cerebrovascular disease, BMI and WC. BMI, body mass index; ROC, receiver operating characteristic; WC, waist circumference.
ROC curve analysis for cognitive impairment
In addition, we examined the diagnostic performance of different models incorporating other influencing factors (figure 1B and table 5). All three models include age, education, diabetic dietary pattern, hypoglycaemic episode in 3 months and CVD. Models 1 and 2 included additional BMI and WC, respectively. ROC curve analysis showed that the diagnostic performance of model 2 (AUC=0.708) was slightly higher than that of model 1 (AUC=0.705). Model 3 had the highest diagnostic performance (AUC=0.808) with all seven influencing factors included.
Discussion
Results of this cross-sectional study based on hospitalised middle-aged T2DM patients with a median age of 58 years suggest the three main findings. First, we identified age, WC, a hypoglycaemic episode within the past 3 months, and the presence of CVD as independent risk factors for CI in hospitalised middle-aged T2DM patients. In contrast, higher education level, diabetic dietary pattern, overweight and obesity were independent protective factors against CI in these patients. Second, general and central obesity had distinctly different relationships with cognitive function. Results of regression analysis indicated that a higher BMI was associated with better cognitive performance within certain limits, while those with higher WC had poorer cognitive performance. Third, the combination of WC and BMI was more valuable for the diagnosis of CI than either measure alone, and it might be a meaningful indicator for early screening of CI in hospitalised middle-aged patients with T2DM.
In our study, the prevalence of CI among hospitalised middle-aged T2DM patients was 38.9%, which was much higher than that in the general population (3%–22%).32 Our findings of the CI prevalence were consistent with a recent meta-analysis that showed that the prevalence of MCI in middle-aged people with T2DM ranged from 33% to 59%.11 Metabolic disorders induce a state of hyperglycaemia accompanied by chronic inflammation in T2DM patients, resulting in damage and dysfunction of blood vessels, nerves, the brain, as well as other tissues and organs.3 4 In view of the high prevalence of CI in middle-aged diabetic individuals, this study is warranted.
The results of our present study showed that T2DM patients with CI were older and less educated than those with NCF. Age-related brain atrophy and neuronal structural changes (including a decrease in the dendritic number and length, loss of dendritic spines, a decrease in the number of axons and a significant loss of synapses) could lead to cognitive decline.33 Moreover, we found that age remained an independent risk factor for CI in this middle-aged diabetic population, even after correcting for other confounding factors, suggesting that ageing-related cognitive decline is already occurring in the middle-aged diabetic population, not just in the older population. It is well documented that significant cognitive decline begins from as early as 45 years.34 However, most of the current studies have focused on cognitive dysfunction and dementia in the elderly. More attention should be paid to the cognitive ability of middle-aged adults in the future, especially in a high-risk population such as combined diabetes mellitus. We found that as education increased, patients were at less risk of CI. Cognitive reserve (CR) theory may explain this result. CR is built through a series of life experiences, including educational level, professional complexity and cognitive activities.35 This reserve can maintain the integrity of cognitive function and prevent the decline in cognitive ability by compensating for nerve damage and using alternative neural networks.36 Our findings regarding education and CI suggest the necessity for widespread compulsory education and the continuous improvement of the general public’s knowledge. Relevant policies and projects are needed to move forward on this.
Our study found that diabetic dietary pattern is a protective factor against cognitive dysfunction. Patients with controlled diabetes diets typically consume more dietary fibre and antioxidant vitamins and fewer sugars or fats (usually saturated fatty acids) than those without controlled diets. Excessive fat and sugar intake appears to have a direct effect on brain by reducing neural production and synaptic plasticity, which in turn lead to cognitive dysfunction.37 In contrast, antioxidants inhibit neuroinflammatory processes and neuronal apoptosis by reducing the production of free radicals and cytokines in activated microglia.38 However, we did not find a significant correlation between physical activity and cognitive function. The reason may be that this association is weaker in the diabetic subjects.39 Besides, our study population is relatively young and might have a more active lifestyle. The proportion of sedentary patients was about 30% in the present study.
We found that episodes of hypoglycaemia within the past 3 months was a risk factor for CI in middle-aged T2DM patients. The brain’s energy metabolism places a high demand for glucose. Hypoglycaemia interferes with energy metabolism in the brain, leading to neuronal degeneration and hippocampal atrophy, which in turn cause chronic CI.25 A prospective study reported that, in older people with undiagnosed diabetes, whose FBG<4.7 mmol/L and ≥6.3 mmol/L, the risk of developing CI within 5 years was increased by about 50%.40 This result suggests that both long-term hyperglycaemia and hypoglycaemia can promote the occurrence and development of CI. Possible pathophysiological mechanisms of hypoglycaemia-induced CI include hypoglycaemia-related neuronal damage, inflammatory processes, coagulation deficiency, endothelial cell abnormalities and synaptic dysfunction of hippocampal neurons during hypoglycaemia.41
CVD was an independent risk factor for CI in the present study. Diabetic stroke is caused by extracranial carotid artery disease and large and small intracranial vascular lesions. The clinical symptoms range from asymptomatic cerebral microvascular disease to transient ischaemic attack, and haemorrhagic or ischaemic stroke.42 Theoretically, multiple CVDs can cause vascular CI.43 A larger infarct volume and more regional, small subcortical infarcts have been related to cognitive decline and an increased risk of dementia, particularly when they exist in areas involved in cognitive function.44
Interestingly, we found that central obesity reflected by WC was a risk factor for CI in our study subjects, whereas overweight and obesity based on BMI were protective factors. In fact, AO, defined by WC, has been shown to contribute to cognitive deficits in patients with T2DM.45 One potential explanation is the strong association of central obesity with cardiovascular disease, stroke, diabetes and hypertension.5 However, AO remained a significant risk factor for CI after adjustment for CVD (univariate analysis showed that cardiovascular disease and hypertension were not confounders, p>0.05), suggesting that something inherent to central adiposity may increase the risk of cognitive dysfunction. Inflammatory responses and elevated circulating levels of saturated fatty acids induced by increased WC are associated with cognitive decline in diabetic patients.46 Additionally, obesity-related changes in brain structure and volume also mediate later cognitive deficits.47
The present study found that overweight and obesity based on BMI were protective factors against CI in middle-aged patients with diabetes. But many previous studies have shown that BMI is a risk factor for cognition in middle-aged adults.48 This inconsistency of results may be due to the heterogeneity of study population. Evidence from cohort studies and meta-analyses suggests that the phenomenon of the ‘obesity paradox’ (the protective effect of obesity on health) may occur in this particular population of diabetes.49 50 This large prospective cohort study over a period of 10 years showed that overweight or obese diabetic patients had lower all-cause mortality than normal-weight patients.49 A recent retrospective study showed that a higher BMI was associated with better cognitive function in patients with T2DM.51 If obesity impairs cognition largely through diabetes (including glycaemic control, insulin resistance, microvascular damage, etc),52 then the presence of diabetes is likely to attenuate the negative cognitive effects of obesity, which needs to be confirmed by more studies. In addition, BMI is only a rough indicator of obesity and cannot distinguish between specific body compositions, meanly including visceral fat, subcutaneous fat and lean body mass.53 When WC and BMI were entered in the same multivariate model, BMI mainly reflected lean body mass and WC mainly reflected abdominal fat.54 This explains why the combination of BMI and WC increased the predictive power of each other in our study. Previous studies have reported that elderly with low lean mass had worse cognitive performance.55 Loss of lean mass and cognitive decline were also mediated by osteopenia.56 There are biological evidences to support these results. Neurotrophins (including insulin-like growth factor-1 and brain-derived neurotrophic factor) produced during skeletal muscle contraction can modulate brain synapses and improve brain function.57 Finally, it is worth noting that the results of piecewise linear regression suggest that the positive association between BMI and the MoCA score might exist only in patients with normal weight or overweight, and not in patients with obesity. In fact, this is also consistent with the reality. Although BMI is not an accurate representation of obesity, when it exceeds a certain range, the cognitive damage of obesity is shown (eg, in this study, all obese patients with BMI≥30 kg/m2 were associated with AO.
Current studies on cognition mainly focus on the elderly population, and few studies have focused on CI in middle-aged patients with T2DM. However, there is substantial evidence that cognitive decline begins in midlife,34 especially when diabetes itself accelerates cognitive deterioration. Given the high prevalence of CI in middle-aged diabetic population (38.9% in our study) and the lack of effective treatments for dementia and cognitive deficits, it is important to identify modifiable risk factors for diabetes in midlife and provide early intervention. In particular, we investigated the joint effects of global and AO on CI in hospitalised middle-aged patients with T2DM, and found that BMI and WC might have opposite effects on cognitive function. The results of ROC curve analysis showed that combination of BMI and WC had a diagnostic significance for CI in hospitalised middle-aged patients with T2DM.
This study had several limitations. First, it was a cross-sectional study. Due to the restriction of the research design, we could not determine a causal relationship between CI and the identified factors. Second, although we adjusted for some traditional demographic and health-related characteristics, we failed to adjust for other unmeasured potential confounding factors, such as β-Amyloid, hyper-phosphorylated τ, apolipoprotein ε4 gene genotype, C-reactive protein and depression. Third, our study population was middle-aged hospitalised patients with T2DM who had poor glycaemic control, with an average glycation level of 9.06%. The population also had a higher incidence of comorbidities than the general T2DM population:58 59 microvascular complications, macrovascular complications and hypertension were 83.5% vs 53.5%, 39.2% vs 27.2% and 61.0% vs 54.0%, respectively. Therefore, our findings may not apply to other people with T2DM, such as those with good glycaemic control or those without comorbidities. Fourth, the subject of this study is the analysis of risk factors in middle-aged patients with T2DM. Although we used T2DM patients without CI as substitute controls, we lacked non-diabetic healthy people as controls, so we could not fully understand the distribution of these risk factors in normal people, which may bring bias to our analysis. Finally, BMI cannot distinguish between fat and muscle; however, the effects of adipose tissue and muscle mass on cognitive function are entirely different. Therefore, more researches are needed to understand the impact of different body components on cognitive function.
In summary, we found that age, occurrence of a hypoglycaemic episode within the past 3 months, WC and CVD were independent risk factors for CI in middle-aged T2DM patients. On the other hand, independent protective factors were education, use of diabetic dietary pattern, overweight and obesity. Within a certain range, a higher BMI was associated with a higher MoCA score, while a higher WC was associated with lower cognitive scores (the MMSE and MoCA). Our findings may contribute to a more complete understanding of modifiable risk factors for CI in hospitalised middle-aged type 2 diabetic inpatients. In addition, we found that the combination of BMI and WC might have more diagnostic value for cognitive decline than either measure alone in middle-aged T2DM patients. These two simple and readily available anthropometric measures may serve as potential predictive markers for CI in hospitalised middle-aged T2DM patients. It is important to emphasise the possible different effects of different obesity parameters on cognition and the comprehensive consideration of BMI and WC in the early screening of CI in middle-aged diabetic patients, not just BMI. Nevertheless, more prospective studies are still needed to support their relationships.
This post was originally published on https://bmjopen.bmj.com