Objective

In China, one in every 10 adults aged 18 years and over has diabetes. In 2013, over 100 million adults in China had diabetes, corresponding to a 10% incidence of diabetes [1]. Diabetes affects brain function, the course of disease is long, and complications can aggravate the cognitive impairment associated with diabetes. Elderly patients with diabetes have a longer course of disease, and diabetes aggravates age-related cognitive decline; cognitive decline in elderly patients with diabetes is more common [2]. A number of conditions are potentially involved, including cardiovascular diseases, peripheral neuropathy, being overweight, osteoarthritis, visual deficits, skeletal muscle dysfunction, and cognitive impairment; these conditions are more prevalent in patients with diabetes [3]. Furthermore, a previous study suggests that hyperglycaemia induces attention and gait deficits in patients with diabetes [4]. Recently, diabetologists have agreed that cognitive impairment and dementia are new emerging complications of type 2 diabetes [5]. Similarly, individuals living with diabetes may face an increase in the rate of cognitive decline by as much as 1.5–2.0 times relative to their non-diabetic peers in later life [2]. Furthermore, this cognitive decline experienced by elderly individuals with diabetes incurs a tremendous cost to both individuals and society. Gaining a better understanding of the biological markers of this decline could help to reduce these costs.

Orexigenic and anorexigenic peptides, i.e. stomach-derived ghrelin and adipocyte-derived leptin, respectively, regulate energy homeostasis and food intake via the hypothalamus [6]. However, leptin and ghrelin receptors are expressed in many brain areas, including the hippocampus, where they promote learning and memory [7,8,9,10,11]. Although widely expressed in the temporal lobe of healthy humans, ghrelin is consistently reduced in patients with Alzheimer’s disease (AD) and is thus implicated in disease progression [12]. Similarly, the inhibition of ghrelin receptors in the brain impairs memory encoding in rats [13], and intracerebral administration of ghrelin affects memory formation in rodents [14,15,16,17]. In animal models of neurodegenerative diseases and age-related memory impairment, ghrelin appears to exert a neuroprotective effect [18,19,20,21]. In humans, however, its role in cognition is yet to be defined [22, 23], and studies in older adults are scarce. Only one study, conducted by Spitznagel et al. [24] with a sample of 35 older adults, has suggested that higher ghrelin levels are associated with poorer cognitive performance, which conflicts with conclusions from animal studies.

Conversely, leptin acts as an appetite inhibitor [25, 26]. Levels of circulating leptin have been shown to be lower in patients with AD versus age-matched controls [27,28,29]. However, prior studies on leptin have been primarily focused on its association with emotion rather than cognition; these have found both high and low levels of leptin in patients with depression, while others have shown no association [30,31,32,33,34,35]. Further, the association between depression and leptin appeared to be mediated by increased adiposity in patients with depression [36, 37].

Cortisol is a stress hormone produced by the adrenal cortex that may also affect cognition. Previous studies have reported a negative correlation between levels of serum cortisol and memory performance in elderly participants [38, 39]. Higher levels of urinary cortisol have also been associated with a greater risk of impairment in overall cognitive, verbal memory, and executive function [40]. Excess glucocorticoids stimulate hippocampal atrophy, leading to memory loss [38, 41].

The objective of this study was to assess the association of ghrelin, leptin, and cortisol levels with cognitive decline and short-term memory in elderly diabetic patients. Cognitive decline is common in elderly diabetic patients; however, it is not yet known whether the change in hormone levels in these patients is caused by diabetes or cognitive decline. To disentangle ageing effects from diabetes-related effects, we also investigated the relationship between cognitive decline and hormone levels in non-diabetic elderly people. Based on prior research, we hypothesized that individuals with lower cognitive functioning would exhibit lower levels of fasting ghrelin and leptin, higher levels of cortisol, and impaired short-term memory than those with normal cognition.

Methods

Participants

A cohort of 286 adults aged 65–85 years, with diabetes (n = 156; 81 men and 75 women) and without diabetes (n = 130; 67 men and 63 women), participated in the study. Participants without diabetes were community residents recruited through advertisements. The process of selecting study subjects without diabetes is shown in Fig. 1.

Fig. 1
figure 1

Process of selecting the study non-diabetic elderly

Full size image

Diabetic subjects were selected from inpatients and fulfilled one or more of the following criteria: (a) had a fasting blood glucose > 7 mmol/L on two separate occasions, (b) a 2-hour blood glucose level > 11.10 mmol/L during a 75-g oral glucose tolerance test on two separate occasions, or (c) a prior diagnosis of type 2 diabetic mellitus (T2DM). First, we excluded diabetic patients who had large fluctuations in blood glucose levels and recurrent hypoglycaemia or hyperglycaemia. Other exclusion criteria were as follows: had been administrated a hormone (e.g. thyroid hormones, sex hormones, cortisol) in the last 3 months; low visual acuity or serious hearing problems; inability to speak Mandarin; significant, acute medical conditions (e.g. recent myocardial infarction and heart failure); a history of significant neurological or psychiatric disorders (e.g. AD, stroke, schizophrenia). Hospitalised diabetics who met the above criteria voluntarily participated in this study at the Jinhua Central Hospital between January 2015 and January 2017. The experiment conforms to The Code of Ethics of the World Medical Association (Declaration of Helsinki). The Human Investigations Committee of the Central Hospital of Jinhua approved this research. All participants provided written informed consent and all persons gave their informed consent prior to their inclusion in the study. Research funding to conduct this work was provided by the Jinhua Central Hospital.

Measures

Cognitive test

We measured overall cognitive performance using the Shanghai version of the MMSE and MoCA. The MMSE screens cognitive impairment [42]. The Chinese version of the MMSE determines the dementia boundary value according to the level of education, with the illiterate group, primary school group, and high school (or above) groups, scoring less than or equal to 17 points, less than or equal to 20 points, and less than or equal to 24 points, respectively. The score ranges from 0 to 30 points [43,44,45]. The MoCA was scored out of 30 points [46, 47]. This study used the Shanghai version of the MMSE as a screening tool for cognitive impairment, and the MoCA scale to check criterion-related validity.

Short-term memory test

To eliminate the confounding effect of education on short-term memory [48], this study used pictures of common objects to test short-term memory. Twenty pictures were presented on a computer screen for 30 s, and subjects were asked to verbally recall the pictures. This was scored according to the number of correct free recalls ranging from 0 to 20 points.

Evaluation of physiological indices

Following a 10-h overnight fast (i.e. participants ate nothing after 10:00 p.m.), blood samples were obtained at 8:00 a.m., at 10:00 a.m., at 4:00 p.m., and at 12 p.m. Further, 24-h urine samples were collected. Participants ate breakfast immediately after blood samples had been taken at 8:00 a.m.; this included steamed bread (70 g; 180 kCal), an egg (50 g; 75 kCal), soy milk (200 g; 45 kCal), and Chinese cabbage (100 g; 20kCal). Blood samples were stored at − 80 °C until further analysis. Leptin levels were analysed using leptin radioimmunoassay kits (Anhui Anke Biotechnology Ltd., by Share Ltd.), and total ghrelin plasma levels were determined using radioimmunoassay kits (Phoenix Pharmaceuticals, Inc., California, USA). Cortisol concentration was determined using chemiluminescence, using the commercially available LN CLIA Microparticles kits (Autobio Diagnostics Co. Ltd., Zhengzhou, China). The height and weight of each subject were measured to calculate the body mass index (BMI).

Analyses

Results are expressed as the mean ± standard deviation. Non-normal data were log-transformed. We compared the three groups using an analysis of variance (ANOVA). Cortisol data were analysed using a repeated-measures ANOVA, which included one within-participants variable (sampling times) and one between-group variable (groups: cognitively normal older adults without diabetes, cognitively impaired older adults without diabetes, cognitively normal older adults with diabetes, and cognitively impaired older adults with diabetes), the introduction of BMI as covariates. Multiple between-group comparisons of mean differences were checked using Bonferroni post hoc tests. Independent and paired-sample t tests were used, as appropriate.

Regression models

After controlling for age, sex, BMI, diabetes (1: with diabetes; 0: without diabetes) and education, a multiple stepwise regression model was used to analyse the correlation between short-term memory and the levels of ghrelin, leptin, and cortisol. The multiple stepwise regression model included sex, age, BMI, years of education, diabetes, fasting ghrelin, fasting leptin, plasma cortisol at 8:00 a.m., plasma cortisol at 4:00 p.m., plasma cortisol at 12:00 p.m., and 24-h urinary cortisol for the prediction of short-term memory. The multiple stepwise regression for short-term memory was created, where by the independent variable with the highest explained variance was entered first, followed by the next highest, and so on, until adding an independent variable no longer led to significant change in explained variance.

Results

The Pearson’s correlation coefficient between the MMSE and MoCA scores was 0.793 and 0.683, respectively, between the MMSE and short-term memory scores, and 0.655 between the MoCA and short-term memory scores. For both non-diabetic and diabetic patients, the MMSE, MoCA, and short-term memory scores of the cognitively impaired participants were significantly lower than the normal cognition group (p < 0.01, Table 1). These results show that the cognitive tests had a good calibration validity.

Table 1 Comparison of the four groups
Full size table

As seen in Table 1, baseline ghrelin level was significantly lower in the cognitively impaired group (9.07 ± 1.13 ng/mL in non-diabetic subjects and 7.76 ± 1.34 ng/mL in diabetic subjects) compared with the cognitively normal group (10.94 ± 1.53 ng/mL in non-diabetic subjects and 9.93 ± 1.76 ng/mL in diabetic subjects). Short-term memory was significantly lower in the cognitively impaired group (5.99 ± 2.90 in non-diabetic subjects and 4.71 ± 2.14 in diabetic subjects) compared with the cognitively normal group (8.14 ± 2.23 in non-diabetic subjects and 7.82 ± 3.37 in diabetic subjects). Plasma cortisol levels at 8:00 a.m., 4:00 p.m., and 12:00 p.m. were significantly higher in the cognitively impaired group compared with the cognitively normal group. Years of education was significantly lower in the cognitively impaired group (6.50 ± 3.53 in non-diabetic subjects and 4.56 ± 3.67 in diabetic subjects) compared with the cognitively normal group (9.43 ± 3.42 in non-diabetic subjects and 6.74 ± 3.49 in diabetic subjects). No significant between-group differences in plasma levels of fasting leptin were observed.

As seen in Table 1, 24-h urine cortisol value and cortisol plasma levels at 8:00 a.m. and 4:00 p.m. of diabetic patients with normal cognition were significantly higher than those of non-diabetic patients with normal cognition. Years of education were significantly lower in diabetic patients with normal cognition than in non-diabetic patients with normal cognition. Compared with non-diabetic patients with cognitive decline, diabetic patients with cognitive decline were nearly 4 years ahead (78.25 ± 3.98 in non-diabetic subjects and 74.51 ± 3.69 in diabetic subjects), the 8:00 a.m. and 12:00 p.m. cortisol plasma levels were significantly higher, and the level of fasting ghrelin (7.76 ± 1.34 ng/mL) in the diabetic group with cognitive decline was significantly lower than that in non-diabetic group with cognitive decline (9.07 ± 1.13 ng/mL). There was significant difference in ghrelin change after meals between the cognitive decline group with diabetes (0.64 ± 1.03 ng/mL) and the non-diabetic cognitive decline group (0.07 ± 1.18 ng/mL). The number of years of education was significantly lower in the diabetic group (4.56 ± 3.67) than in the non-diabetic group with cognitive decline (6.50 ± 3.53). Short-term memory scores (4.71 ± 2.14) in the diabetic group with cognitive decline were significantly lower than those in non-diabetic cognitive decline group (5.99 ± 2.90).

As shown in Table 1, the diabetic group with cognitive decline had the highest level of 24-hour urinary cortisol and plasma cortisol levels at three sampling times compared to the other three groups, with the lowest fasting ghrelin levels (7.76 ± 1.34 ng/mL), worst decrease in ghrelin levels (0.64 ± 1.03 ng/mL), the lowest schooling years (4.56 ± 3.67), and the cognitive test scores, especially particularly low short-term memory scores (4.71 ± 2.14).

The multiple stepwise regression final model included age, years of education, plasma cortisol (12:00 p.m.) level, and fasting ghrelin level for prediction of short-term memory, total variance explained by the following four independent variables: \( R^{2} = 0.43 \), included age: \( R^{2} = 0.13 \); years of education: \( R^{2} \) change = 0.17; plasma cortisol (12:00 p.m.): \( R^{2} \) change = 0.08; and fasting ghrelin: \( R^{2} \) change = 0.05. As shown in Table 2, a worse short-term memory was associated with older age (p < 0.001), higher nocturnal zero plasma cortisol concentration (p < 0.001), fewer years of education (p < 0.001), and lower fasting ghrelin level (p < 0.01). Sex, BMI, diabetes, fasting leptin, plasma cortisol at 8:00 a.m. and 4:00 p.m., and 24-h urinary cortisol were not significantly correlated with short-term memory.

Table 2 Multiple stepwise regression models for all elderly subjects
Full size table

A significant main effect of group were observed on plasma cortisol concentrations (F[3, 281] = 21.08, p < 0.001). The cortisol physiological rhythm curve of adults with cognitive impairment floated higher than that of the normal groups (Fig. 2). An interaction between sampling times and group was found (F[3, 281] = 6.79, p < 0.01); the cortisol levels of all participants but the cognitively impaired group with diabetes differed significantly over the three sampling times (all p < 0.001), decreasing from 8:00 a.m. to 4:00 p.m., and reaching its lowest levels at 12 p.m. In the impaired cognition group with diabetes, cortisol levels at the 4:00 p.m. and 12:00 p.m. fell flat on the steep slope and did not differ significantly over the two sampling times (p = 0.28).

Fig. 2
figure 2

Cortisol circadian rhythm of the four groups

Full size image

Discussion

The level of ghrelin was negatively correlated with short-term memory ability, whereby lower fasting ghrelin levels were associated with a greater memory loss. Our results also confirmed the positive correlation between cortisol level and short-term memory observed in prior research. Leptin levels were not associated with short-term memory. Taken together, these findings lead to a more complete picture of biomarkers associated with the cognitive ageing process.

The fasting ghrelin level in the normal cognition group (in both diabetic/non-diabetic patients) was significantly higher than that of the cognitive impairment group, and lower ghrelin was associated with lower short-term memory. This illustrates a possible long-term effect of ghrelin on neural protection and cognitive function. This supports previous studies that have shown decreased ghrelin levels in patients with neurodegenerative diseases such as AD and Parkinson’s disease [49, 50]. Furthermore, the inhibition of ghrelin signalling decreases the number of spine synapses in the striatum, followed by impaired memory performance [51]. Our study is also consistent with previous animal studies in which ghrelin appears to exert a neuroprotective effect in animal models of neurodegenerative diseases [18,19,20,21]. While Kunath et al. [52] did not consider ghrelin to act as a short-term cognitive enhancer in humans, they concluded that ghrelin may have potential neuroprotective effects in long-term or pathological models. Based on ours as well as previous results, we can propose that ghrelin has neuroprotective effects in the elderly.

Tschop et al. [11] state that the plasma ghrelin concentration in healthy people increases in fasted states and decreases in fed states; its levels are dynamically decreased after a meal. Wittekind and Kluge [53] showed no decrease in plasma ghrelin levels following a meal in some psychiatric disorders. Similarly, we found no decrease in ghrelin levels after eating in any of the four groups. Ghrelin levels increased most after a meal in the cognitively impaired with diabetes group, who showed the greatest impairment in the normal physiological function of ghrelin. Our study demonstrates that ghrelin is likely to be involved in the pathophysiology of varying forms of diabetes and ageing.

There was no significant difference in fasting ghrelin levels and the changed amount of postprandial ghrelin between cognitive normal elderly patients with and without diabetes. In other words, diabetes did not cause a significant decrease in fasting ghrelin levels compared with non-diabetic elderly people, and no significant loss of postprandial ghrelin function was observed. Only when diabetes and cognitive decline coexisted did fasting ghrelin levels show a significant decline and the function of postprandial ghrelin showed significant damage. Thus, fasting ghrelin and the changes in postprandial ghrelin levels reflect the inner mind–body balance of the subjects.

Although low fasting plasma leptin levels have been linked to cognitive decline in patients with AD [27,28,29], our findings indicated no significant correlation between leptin levels and memory loss. Further, there were no significant group differences in leptin levels. This may be due to no significant difference in BMI between different cognitive levels of non-diabetes or diabetes patients, while leptin is adipocyte-derived.

We found a significant increase in cortisol secretion in older adults with cognitive impairment, and higher 12:00 p.m. cortisol levels were associated with a lower short-term memory. Vasiliadis et al. [54] reported elevated salivary cortisol in elderly individuals who self-reported more daily hassles. Previous studies have indicated that long-term stress may result in depression and a hyperactive hypothalamic–pituitary–adrenal (HPA) axis [43, 55, 56]. Depression in patients at morning and nocturnal cortisol levels is higher [57, 58]. In the current study, memory loss and ageing could have brought about a long-term stress-induced increase in plasma cortisol levels, especially in the diabetic group with cognitive impairment. This theory is supported by previous studies that have demonstrated a negative correlation between cortisol levels and memory performance in older adults [38, 39]. Higher levels of urinary cortisol have also been associated with a higher risk of overall cognitive, verbal memory, and executive function impairments [40]. Moreover, excess glucocorticoids have been found to stimulate hippocampal atrophy, leading to memory loss [38, 41].

The circadian rhythm curve of older adults with cognitive impairment floated higher on top, compared with the cognitively normal groups. In diabetic elderly patients with cognitive impairment, the cortisol levels at 4:00 p.m. and 12:00 p.m. fell flat on the steep slope and did not differ significantly over the two sampling times which demonstrates their physiology was most impaired of all the groups. This supports previous finding on the physiological rhythm of cortisol in elderly individuals with depression [59]. The cortisol rhythm represents the activity pattern of the HPA stress axis [60]. From Fig. 1, we can speculate that memory loss in the elderly may has encountered a considerable number of daily life problems, causing long-term emotional state of high stress.

Our results suggest that education may be a protective factor against diabetes and cognitive decline; the number of years of education was significantly lower in the cognitively impaired group (including both diabetic and non-diabetic patients) than in their peers. Similarly, the number of years of education in the diabetic group (including both cognitively normal and impaired patients) was significantly lower than their respective peers. A lower level of education may result in lower self-management knowledge, treatment compliance, and poor long-term glycaemic control, which could in turn result in a greater degree of cognitive impairment. Effective educational programs should therefore be formulated for patients with lower educational levels.

Some limitations of the current study should be noted. First, the sample size was not very bigger; further studies should verify our results with a larger sample of patients. Second, this was a cross-sectional study that evaluated subjects at one time-point. Thus, while markers correlated with memory loss can be identified, this study did not determine whether ghrelin and cortisol can increase memory loss risk. Further, since this was a descriptive study, so it remains to be seen in further appropriately designed studies whether this is a therapeutic option. In the future, we will do a longitudinal study where we will followed elderly subjects with or without diabetes, establishing a baseline for the metabolic and the cognitive status, and later asses these elderly again during the follow-up (i.e. 12 months and 24 months later). In this case, we would be able to corroborate whether ghrelin and cortisol are really predictors of cognitive decline. Third, follow-up studies could be conducted on physiological, behavioural, and cognitive interventions that may change the baseline ghrelin and plasma cortisol levels, with the aim of reducing or delaying memory loss in older adults.

Overall, irrespective of the presence of diabetes, older adults with cognitive impairment showed decreased levels of ghrelin and increased levels of cortisol compared to the cognitively normal group. The leptin levels, however, were conserved in all groups. Thus, these findings suggest that ghrelin and cortisol may be markers of memory loss, raising the possibility that adjusting their levels could have therapeutic effects.