Olfactory Dysfunction Mediates Adiposity in Cognitive Impairment of Type 2 Diabetes: Insights From Clinical and Functional Neuroimaging Studies

The prevalence of being overweight or obese among people with type 2 diabetes was 87.1% in the U.S. from 2005 to 2010 (1) and 58.3% in China in 2011 (2). Obese individuals with diabetes are more susceptible to diabetes complications in the heart, brain, eyes, and kidneys (3,4). Cognitive decline and dementia are recognized as cerebral comorbidities associated with diabetes (5). Cognitive deficits, especially in executive function, impact self-care and medical compliance of these patients. Prevention, early diagnosis, and management of cognitive decrements present significant challenges.

Recently, neuroimaging techniques have attracted attention as they expand the research scope from animals to humans, allowing for visualizations of deep brain structures to assess early pathological changes, therapeutic targets, and neurofeedback to treatment (6). In structural and functional neuroimaging studies, people with diabetes have generally exhibited regional brain atrophy, small-vessel diseases, aberrant spontaneous brain activation, and disrupted functional connectivity underlying cognitive impairment (7). Particularly, obese individuals with type 2 diabetes have had pronounced cortical atrophy and white matter integrity disruption in MRI studies compared with normal-weight people with diabetes (8). However, the majority of functional neuroimaging studies in obese subjects with diabetes focused on eating behavior and energy homeostasis rather than cognitive function (9).

Olfactory dysfunction, characterized by increased odor thresholds and impaired odor discrimination and recognition, is an early risk sign of preclinical dementia and associated with the trajectories from normal cognition to mild cognitive impairment and late-life dementia (10,11). Epidemiological studies using olfactory tests found that patients with diabetes have decreased olfactory acuity (12). Furthermore, odor-induced functional MRI (fMRI) obtains clues about vulnerable brain regions before clinical measurable cognitive decline (13). Our previous study detected olfactory behavior and functional dysfunction in cognitively intact diabetes (14). However, associations between cognitive function, olfactory function, and olfactory central networks in obese individuals with diabetes, and how to improve olfaction associated with diabetes, remain unclear.

Glucagon-like peptide 1 receptors (GLP-1Rs) are found in central neural cells, particularly in the olfactory bulb and hippocampus (15). The antidiabetes GLP-1R agonists (GLP-1Ras) have been reported to partially reverse decreased brain glucose metabolism in Alzheimer disease (16). Little is known about their efficacy for olfactory and neurological disorders in patients with diabetes.

Therefore, this study evaluated the patterns and interrelationships of cognition, olfactory function, and odor-induced brain functional activations in obese patients with type 2 diabetes and investigated the therapeutic effects of GLP-1Ra treatment on cognition, olfaction, and olfactory brain function in these patients.

This study was conducted from January 2016 to August 2018 at Drum Tower Hospital, Nanjing University Medical School. The flowchart is presented in Supplementary Fig. 1. A total of 105 subjects including 35 obese and 35 nonobese participants with type 2 diabetes and 35 normal control subjects were enrolled and matched for age, sex, and education. All participants were right-handed and aged between 35 and 70 years and possessed >6 years of education. Demographic data and medication status of all participants were collected. Resting blood pressure, height, weight, and waist circumferences were measured. Type 2 diabetes diagnosis was based on the American Diabetes Association criteria (17). Standard BMI cutoffs were used to define normal weight (BMI <25 kg/m²), overweight (25 kg/m² ≤ BMI <30 kg/m²), and obesity (BMI ≥30 kg/m²) (18). Control subjects were included if they had normal glucose tolerance, were normal weight, and if their cognitive status was classified as normal by Montreal Cognitive Assessment (MoCA) (scores ≥26), while participants with diabetes were included with normal cognition or mild cognitive impairment assessed by MoCA (scores ≥19) (19). Exclusion criteria for all participants were a history of neurological and psychiatric disorders, abnormal thyroid, pancreatitis, cardiovascular or cerebrovascular disease, nasal pathologies, alcohol or substance abuse, steroid treatment, infections, or inability to undergo tests or MRI scanning.

This study was registered on ClinicalTrials.gov (NCT02738671) and was approved by the ethics committee of Drum Tower Hospital in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants.

DEXA scans (Lunar iDXA software; GE Healthcare, Milwaukee, WI) were used to measure body fat distribution. Patients with diabetes underwent a 100-g standard meal test, and control subjects underwent a 75-g oral glucose tolerance test (OGTT). Plasma glucose, insulin, and C-peptide concentrations were measured at fasting and 2 h after a meal. Serum fasting total cholesterol, triglyceride, and HDL and LDL cholesterol concentrations were measured. Insulin resistance and β-cell function were estimated using the HOMA2 Calculator (HOMA2 v2.2.3; Diabetes Trials Unit, University of Oxford).

General cognition was evaluated by the Mini-Mental State Examination (MMSE) and MoCA (Beijing Version). Multiple cognitive subdomains including episodic memory, working memory, word fluency, executive function, and information processing speed were evaluated with the following tests: 16-word Philadelphia Verbal Learning Test and Wechsler Memory Scale, Trail Making Test Part A and Part B, Animal Naming Test, Boston Naming Test, Digit Span Test (forward and backward), and Stroop Color and Word Test (parts I, II, and III). Psychological status was evaluated by the Clinical Dementia Rating and the Hamilton Depression Rating Scale. All tests required ∼60 min to complete in a fixed order. Subject groupings were not known to examiners.

Olfactory testing was performed using Olfactory Function Assessment by Computerized Testing (OLFACT) (Osmic Enterprises, Inc.). Based on the University of Pennsylvania Smell Identification Test (UPSIT), OLFACT tests were computerized, standardized, and self-administered. Higher scores indicated better ability to detect, identify, and memorize odors. Threshold testing was performed by a series of binary dilutions of n-butanol solution in light mineral oil, and scores ranged from 1 to 13.5. Olfactory identification and memory tests included two tasks: task A with 10 different odors (menthol, clove, leather, strawberry, lilac, pineapple, smoke, soap, grape, and lemon), and task B with 20 odors (banana, garlic, cherry, baby powder, grass, tutti-frutti, peach, chocolate, dirt, orange, and the 10 odors in the task A). Each participant was asked to identify each odor from four pictures in tasks A and B and to indicate whether each was old or new in task B. There was a 10-min break between the two tasks.

The odor-induced fMRI task was used as previously described (Supplementary Fig. 2) (14). Each participant underwent a series of scans to measure temporal brain response to four increasing concentrations of lavender odor (0.032%, 0.10%, 0.32%, and 1.0%) diluted in 1,2-propanediol (Sigma-Aldrich, St. Louis, MO). The visual cues of “+” and “smell” were used for baseline and odor stimulation, respectively. Each concentration was assessed three times, with fresh air and scent occurring alternately. Participants were instructed to press a button once they smelled the lavender scent.

Brain MRI data were acquired using a 3.0T MR-scanner (Achieva TX; Philips Medical Systems, Eindhoven, the Netherlands) with an eight-channel head coil. fMRI was acquired with a gradient-echo planar imaging sequence scan (repetition time 2,000 ms, echo time 30 ms, field of view 192 mm × 192 mm × 140 mm, slice thickness 4 mm, gap 0 mm, flip angle 90°, and voxel size 3 mm × 3 mm × 4 mm), with 230 volumes for resting state and 222 volumes for task. Structural images were acquired with high-resolution T1-weighted three-dimensional fast field echo structural scans (repetition time 9.7 ms, echo time 4.6 ms, field of view 256 mm × 256 mm × 192 mm, flip angle 8°, and voxel size 1 mm × 1 mm × 1 mm).

Preprocessing of fMRI data was performed using Statistical Parametric Mapping 8 (SPM8) and Data Processing & Analysis for Brain Imaging (DPABI_V3.1_180801) (20) in the following stages: 1) removal of the first six time points in the task sequence and first 10 time points in the resting state sequence, 2) correction for head movement, 3) segmentation of high-resolution T1 image and alignment with fMRI image, 4) spatial normalization to Montreal Neurological Institute space template, 5) spatial smoothing, and 6) linear detrending and temporal bandpass filtering (0.01–0.08 Hz) for the resting state sequence. Participants with excessive head movement or image artifacts (>2.5° rotation or >2.5 mm shift) were excluded.

A general linear model was used to estimate odor-induced brain activation. Contrasts between “fresh air > rest” and “scent > rest” for each participant were made. Bilateral parahippocampus, amygdala, piriform cortex, insula, orbitofrontal cortex, and hippocampus in Automated Anatomical Labeling templates and Brodmann areas 28 and 34 (entorhinal cortices) were extracted and merged as olfactory regions of interest (ROIs) (total cluster size 5,029 voxels) for further analyses (14).

Brain regions showing significantly different activations among control subjects and nonobese and obese patients with diabetes were selected as seed regions for functional connectivity analyses to determine correlations between the two olfactory brain regions.

After information collection, olfactory and cognitive assessments, and fMRI scans, obese subjects with diabetes with inadequate glycemic control (HbA_1c between 7% [53 mmol/mol] and 9% [75 mmol/mol]), a stable dose of 1,500 mg daily metformin monotherapy, and stable weight were included in the GLP-1Ra group. Patients receiving dipeptidyl peptidase 4 inhibitors and incretin-based or insulin therapy were excluded (Supplementary Fig. 1). Twenty obese individuals with type 2 diabetes were included, provided written informed consent before participation, and were randomized to liraglutide (Victoza) or exenatide (Byetta) treatment. Liraglutide was administered starting at 0.6 mg daily and, if well tolerated, increased to 1.2 mg and then 1.8 mg daily within 2 weeks. Exenatide (Byetta) was administered starting at 5 μg twice daily and increased to 10 μg twice daily within 2 weeks.

After 3-month GLP-1Ra treatment, these patients underwent anthropometric and biochemical measurements, standard meal tolerance tests, and cognitive, olfactory, and odor-induced fMRI evaluation. A MoCA alternative version was used to avoid duplication. Olfactory tests and odor-induced fMRI scans used the same procedures as used at baseline. One patient treated with liraglutide was excluded for image artifacts, and therefore 19 obese patients with diabetes were included for further analyses.

Analyses were performed using SPSS software (version 20.0; SPSS, Chicago, IL). One-way ANOVA and post hoc tests (Dunnett T3) were performed for continuous variables and Pearson χ² for dichotomous variables. Independent-samples t tests were performed to compare duration and body fat distribution in groups with diabetes. Paired t tests were used to investigate effects of 3-month GLP-1Ra treatment on anthropometric and metabolic parameters. Linear regression with vascular risk factors included was applied to compare cognitive and olfactory functions among the three groups. Mixed model regression including age, sex, education, and vascular risk factors showing significant differences after treatment was used to evaluate the effects of GLP-1Ras on olfaction and cognition in obese subjects with diabetes. P < 0.05 indicated statistical significance.

fMRI data were analyzed by DPABI software. One-way ANCOVA was performed to determine the differences in brain activation and seed-based functional connectivity among the three groups with correction for age, sex, education, and vascular risk factors (blood pressure and serum lipids). Paired t tests were conducted to investigate effects of GLP-1Ra treatment on olfactory functional alterations in obese patients with diabetes. The significance threshold correction was based on Gaussian random field (GRF) theory with a voxel level of P < 0.001 and a cluster level of P < 0.05.

For assessment of relationships between olfactory test scores and cognitive assessments, partial rank correlation analysis corrected for age, sex, and education was performed. Additionally, a Bonferroni correction was applied to reduce the inflation of type 1 error from multiple comparisons.

For determination of how metabolic parameters influenced cognition through olfaction, linear regression was performed for mediation analyses using the PROCESS SPSS macro (21). All analyses involved 5,000 bias-corrected bootstrap 95% CIs, and P < 0.05 indicated statistical significance.

Based on the brain activation changes in the olfactory regions of interest in the odor-induced task fMRI of our previous study (14), for achievement of 90% power and detection of differences among the means (SD 0.52) in a one-way ANOVA study at a 0.05 significance level, the sample size is no less than 32 individuals for each of the three groups. PASS (Power Analysis and Sample Size) software (version 11; NCSS, LLC, Kaysville, UT) was used to estimate the sample size.

Control subjects and nonobese and obese participants with diabetes were matched for age, sex, educational years, and alcohol and smoking habits. Patients with diabetes had higher HbA_1c, fasting and 2-h plasma glucose, HOMA2 of insulin resistance (HOMA2-IR), and resting systolic blood pressure and lower HDL levels compared with control subjects. Obese individuals with diabetes had higher fasting and 2-h C-peptide levels, HOMA2-IR, HOMA2 of β-cell function (HOMA2-B), and fat ratio in the trunk and android region compared with nonobese subjects with diabetes (Table 1).

Compared with nonobese subjects with diabetes, obese subjects with diabetes had significantly lower general cognitive scores on the MMSE (29.0 ± 1.3 vs. 28.4 ± 1.2, P = 0.014) after adjustment for blood pressure and HDL level. No statistically significant differences were observed between the two diabetes groups in other cognitive subdomains except for working memory (Table 1).

Compared with that in nonobese participants with diabetes, the olfactory threshold score was lower in obese subjects with diabetes (9.7 ± 2.4 vs. 8.3 ± 2.8, P = 0.028), indicating that obese patients with diabetes exhibited impaired ability to detect odors. There were no significant differences in olfactory identification or memory tests scores among the three groups (Table 1). Meanwhile, subgroup analyses in control subjects and subjects with diabetes with intact cognition (MoCA ≥26) showed that subjects with diabetes likewise had a lower olfactory threshold score than control subjects (Supplementary Table 1).

Decreased activation in left hippocampus was observed in patients with diabetes compared with control subjects after correction for age, sex, education, and vascular risk factors. These decreases were more pronounced in obese subjects with diabetes than in nonobese subjects with diabetes (with GRF correction, voxel level P < 0.001, cluster level P < 0.05, and cluster size threshold 14 voxels) (Fig. 1A and Supplementary Table 2). Moreover, subgroup analyses in cognitively intact subjects with diabetes and control subjects also demonstrated significant decreased left hippocampal activation in subjects with diabetes (Supplementary Fig. 3).

Further, olfactory brain regions that showed significantly different activations among the three groups during the task in the left hippocampus and surrounding areas were used as seed regions for resting state functional connectivity analyses. Obese patients with diabetes exhibited decreased functional connectivity between seed regions and right insula compared with nonobese subjects with diabetes and control subjects (with GRF correction, voxel level P < 0.001, cluster level P < 0.05, and cluster size threshold 23 voxels) (Fig. 1B and Supplementary Table 2). These results indicated that obese patients with diabetes had more pronounced olfactory function impairment, decreased olfactory brain activation, and disrupted seed-based functional connectivity compared with nonobese subjects with diabetes and control subjects.

After adjustment for age, sex, and education with a Bonferroni correction with a rigorous P < 0.0018, both olfactory threshold and olfactory test total score showed significant positive associations with MoCA and negative associations with executive function time in patients with diabetes (Supplementary Table 3). Meanwhile, left hippocampus (seed regions) activation showed positive correlation with olfactory (r = 0.376, P = 0.002) and MoCA (r = 0.322, P = 0.008) test scores after adjustment for age, sex, and education (Supplementary Fig. 4A and B). Mediation analysis revealed that left hippocampal activation impacted general cognition through olfactory function (b = 0.567 [95% bootstrap CI 0.140, 1.282]) (Fig. 2A).

Next, we determined associations of metabolic parameters in diabetes with cognitive and olfactory assessments and odor-induced brain activation after adjustment for age, sex, and education. Negative associations were observed between BMI and left hippocampal activation (r = −0.328, P = 0.007), seed-based functional connectivity (r = −0.321, P = 0.008), and olfactory test total score (r = −0.254, P = 0.038) (Supplementary Fig. 4C–E). Interestingly, android region fat ratio also showed negative correlation with olfactory test total score (r = −0.314, P = 0.016) and episodic memory (r = −0.300, P = 0.022) (Supplementary Fig. 4F and G). Meanwhile, fasting insulin release positively correlated with left hippocampal activation (r = 0.302, P = 0.013) and negatively correlated with processing speed test time (r = −0.319, P = 0.009) (Supplementary Fig. 4H and I).

Further, mediation analyses with correction for age, sex and education showed that olfactory test total score mediated the correlation between android region fat ratio and episodic memory (b = −0.884 [95% bootstrap CI −1.987, −0.1168]) (Fig. 2B). Meanwhile, left hippocampus activation also mediated the association between BMI and olfactory function (b = −0.174 [95% bootstrap CI −0.437, −0.024]) (Fig. 2C) and that between fasting insulin release and processing speed time (b = −0.531 [95% bootstrap CI −1.346, −0.047]) (Fig. 2D). Taken together, both olfactory function and brain network showed mediating effects on relationships of adiposity and insulin release with cognitive function in diabetes (Fig. 2E).

Significantly, GLP-1Ra treatment for 3 months resulted in an average of 6.1 kg weight loss (89.4 ± 14.4 vs. 83.3 ± 13.6 kg, P < 0.001) and a 1.6% (16.4 mmol/mol) HbA_1c decrease (63 ± 9.8 vs. 46 ± 10.9 mmol/mol, P < 0.001) in obese patients with diabetes (Supplementary Table 4). In addition, 2-h postprandial insulin, C-peptide, and HDL cholesterol levels were increased while systolic blood pressure was decreased when compared with baseline (P < 0.05).

Of note, general cognition (MoCA) (26.6 ± 2.4 vs. 27.9 ± 1.9, P = 0.014) and olfactory function (48.5 ± 8.4 vs. 50.0 ± 8.3, P = 0.008) were increased after treatment with adjustment for age, sex, education, duration, and blood pressure. Particularly, the cognitive subdomains of recall memory (2.8 ± 1.4 vs. 3.7 ± 1.2, P = 0.005) and olfactory identification ability (15.6 ± 3.0 vs. 16.6 ± 2.9, P = 0.002) were improved (Supplementary Table 4). In addition, significantly increased right parahippocampus activation was observed in obese patients with diabetes compared with baseline (with GRF correction, voxel level P < 0.001, cluster level P < 0.05, and cluster size threshold 13 voxels) (Fig. 3 and Supplementary Table 5). No differences were found after treatment in other cognitive subdomains. No cluster was found after GRF correction in the seed-based functional connectivity analysis.

Importantly, reduced significance of GLP-1Ra effects on MoCA, recall memory, olfactory identification, and olfactory test total score was observed when BMI or HbA_1c or both were included in the mixed models (Supplementary Table 6). In addition, no difference of functional activation in the right parahippocampus was observed when neither HbA_1c nor BMI was included as a covariate (with GRF correction, voxel level P < 0.001, and cluster level P < 0.05; no cluster was found). There were no significant differences in metabolic characteristics, cognitive assessment, olfactory test scores, or odor-induced brain activation between liraglutide and exenatide treatments (Supplementary Table 7).

This study found that obese individuals with type 2 diabetes exhibited pronounced impaired general cognition, olfactory threshold score, odor-induced brain activation, and seed-based functional connectivity relative to nonobese subjects with diabetes. Adiposity negatively correlated, while insulin release positively correlated, with cognitive function in patients with diabetes. Olfactory function and functional network performed mediating roles in these correlations. GLP-1Ra treatment significantly improved MoCA and olfactory test total score and increased odor-induced brain activation in obese subjects with diabetes.

Cognitive dysfunction in patients with diabetes has attracted much attention in research and clinical care (22). Diabetes-associated cognitive decrements are reported to develop ∼50% faster than those in normal cognitive aging (5). Moreover, ample scientific and epidemiological research has shown that obesity and its accompanied adverse health conditions are related to cognitive impairment (23). Obese subjects with diabetes, especially those with central obesity, are reported to have worse cognitive function and higher risk of dementia than normal-weight patients with type 2 diabetes (8,24). This study also showed pronounced decrements in general cognition of obese subjects with diabetes compared with nonobese subjects with diabetes (Table 1).

Olfactory dysfunction is considered an early predictor of neurodegeneration and is associated with late-life cognitive impairment (25), and olfactory testing is a useful utility for screening cognitive decline progression due to its independence from cognitive confounds such as intelligence quotient and education. Lower scores on the olfactory function test were observed in patients with type 2 diabetes and were related to poor cognitive performance (26). This study and our previous work (14) also demonstrated lower olfactory threshold scores in patients with diabetes than in control subjects and positive associations of olfactory function with general cognition and executive function with use of a Bonferroni correction with a rigorous P < 0.0018 (Supplementary Table 3).

Neuroimaging techniques MRI and fMRI provide early diagnostic insights for neurological diseases by detecting deep brain structures and providing real-time neurofeedback. MRI studies showed that obese people with type 2 diabetes had reduced cortical thickness compared with normal-weight people with diabetes (8). Similarly, we also observed smaller intracranial volume in obese individuals with diabetes (Supplementary Table 8). Moreover, fMRI bridges the gap between brain structures and behavioral function by tracking neural reactions to various external stimuli and tasks (7). Previously, most task-based fMRI studies in obese patients with diabetes focused on eating behavior and neural activations in reward and motivation systems and evaluated therapeutic effects of antidiabetes drugs or bariatric surgery on brain reactions to food cues (6,27,28). Our previous work first provided new functional neuroimaging insights into the cognition-associated olfactory network in patients with type 2 diabetes and reported that these patients with intact cognition exhibited decreased olfactory activation and disrupted olfactory functional connectivity (14).

Both our subgroup analyses in cognitively intact individuals (Supplementary Table 1 and Supplementary Fig. 3) and our previous study (14) showed impaired olfactory performance and disrupted olfactory functional alterations in subjects with diabetes compared with control subjects, supporting the idea that such olfactory abnormalities were specifically related to type 2 diabetes. It is noteworthy that obese patients with diabetes in this study showed impaired general cognition, olfactory threshold, decreased odor-induced left hippocampus activation, and disrupted seed-based functional connectivity compared with nonobese subjects with diabetes (Table 1 and Fig. 1), indicating that obese people with diabetes suffered more pronounced cognitive and olfactory dysfunction than nonobese subjects with diabetes.

Further, we investigated the inherent link of olfaction with adiposity and cognition in diabetes. We found a negative correlation of the android region fat ratio with cognitive function in diabetes (Supplementary Fig. 4G). Meanwhile, olfactory function and olfactory cortex activation served as mediators between adiposity and cognitive function after adjustment for age, sex, and education (Fig. 2B and C). Indeed, the brain olfactory network not only serves as a nutritional sensor in the reward circuit (29) but also has direct connections to the hippocampus, a brain region vital to memory and primarily affected in dementia and that functions in odor memory processing (30). In mammals, brain olfactory structures of olfactory epithelium, olfactory bulb, piriform cortex, and hippocampus are reported to express a high density of metabolic hormone receptors (31). Previous studies showed that adipokines secreted from adipose tissues, such as leptin and adiponectin, bind to receptors on olfactory sensory neurons and enhance synaptic plasticity (32). Metabolic disorders disturb these pathways in olfactory areas, causing loss of olfactory neurons in rats (33) and resulting in chronic olfactory impairment and aberrant brain activation patterns in response to odors (34), which may result in dysfunctional odor memory processing.

Extensive efforts have been spent exploring lifestyle or drug interventions to delay dementia (22). Some antidiabetes drugs exhibit therapeutic effects on neurological diseases (35), particularly GLP-1Ras. In addition to inducing weight loss and improving glucose homeostasis and cardiovascular function, GLP-1Ras are shown to cross the blood-brain barrier and exert neuroprotective effects (36). GLP-1Rs are distributed in mammalian brains (15). In mouse models, glucagon-like peptide 1 was reported to increase activity of mitral cells that encode olfactory information in the olfactory bulb by decreasing activity of the voltage-dependent potassium channels Kv1.3 (37). GLP-1Ras were shown to improve synaptic plasticity and recognition memory by stimulating neurogenesis, reduce neurotoxic amyloid oligomer levels, and alleviate neuroinflammation and oxidative stress in rats (36). Human studies indicated that GLP-1Ra therapies reduced decline of brain glucose metabolism and cognition in Alzheimer disease patients (16) and reduced brain activation response to food stimulation in patients with type 2 diabetes (27). This study provided new evidence of the therapeutic effects of GLP-1Ras on cognition, olfactory function, and olfactory brain activity in obese subjects with diabetes. In this study, GLP-1Ra treatment for 3 months significantly improved MoCA, olfactory test total score, and olfactory brain activation (Fig. 3 and Supplementary Tables 4 and 5). Such neuroprotective effects of GLP-1Ras might benefit from reducing visceral fat and are potentially relevant to glucose homeostasis or its own effects on neurogenesis. Nevertheless, the significant effects of GLP-1Ras on MoCA, recall memory, olfactory test total score, and odor-induced functional activation were weakened or disappeared when BMI or HbA_1c was included as a covariate (Supplementary Table 6), indicating that such GLP-1Ra–related improvements may have occurred partly through glycemic control and weight loss. Long-term evaluations are necessary to determine the cognitive outcomes of GLP-1Ra therapies in diabetes.

Of note, this study had two major strengths: First, distinguished from conventional olfactory assessment tools such as the UPSIT and Open Essence test (12), the olfactory function test in this study was designed based on the UPSIT and conducted by computerized processes that regulated time, duration, and odor concentrations and automatically recorded and scored performance to ensure data accuracy and reliability. Second, regarding olfactory fMRI techniques, we verified the reproducibility of odor-induced fMRI signals (38) and applied rigorous thresholds and covariate corrections to minimize inflated false positive rates resulting from multiple comparisons in fMRI data (39).

The limitations of the current study were as follows. First, relationships among adiposity, olfactory function, and cognition in diabetes were demonstrated in this observational study; however, the potential causal role of obesity with diabetes on olfaction and cognition should be evaluated in the future. Moreover, severely or morbidly obese patients with diabetes were not included in this study due to the MRI space constraint. Additionally, although this study demonstrated that the preclinical cognitive and olfactory dysfunction in obese patients with diabetes was improved by GLP-1Ra treatment, whether GLP-1Ras show ubiquitous neuroprotective effects in all people with diabetes cannot be determined, as nonobese subjects with diabetes in this study did not receive GLP-1Ra treatment. Further randomized controlled trials are also warranted to clarify whether the risk of cognitive outcome events can be reduced by GLP-1Ra therapy in patients with diabetes with clinical cognitive impairment.

In conclusion, this pilot study demonstrated that obese patients with diabetes had more pronounced impairment in cognitive and olfactory function and more disrupted odor-induced brain functional activity than nonobese patients with diabetes. Olfaction mediated adiposity in cognition of patients with diabetes. Of note, cognitive and olfactory abnormalities can be ameliorated by GLP-1Ra treatment. This study characterized cognition and related brain functional changes in obese patients with diabetes, enhanced the significance of olfactory dysfunction as an early diagnostic signal for cognitive decline in these patients, and provided corresponding potential therapeutic approaches.

Acknowledgments. The authors thank all volunteers for their participation in this study and thank medical personnel from Department of Endocrinology and Department of Radiology, Drum Tower Hospital, for their valuable assistance.

Funding. This study was supported by grants from the National Natural Science Foundation of China (81570737, 81570736, 81770819, 81500612, 81400832, 81600637, 81600632, 81703294, 81800752, and 81800719), the National Key Research and Development Program of China (2016YFC1304804 and 2017YFC1309605), Jiangsu Provincial Key Medical Discipline (ZDXKB2016012), Jiangsu Provincial Medical Talent (ZDRCA2016062), the Key Project of Nanjing Clinical Medical Science, the Key Research and Development Program of Jiangsu Province of China (BE2015604 and BE2016606), the Six Talent Peaks project of Jiangsu Province of China (WSN-165 and SWYY-091), the Fundamental Research Funds for the Central Universities (021414380208, 021414380142, and 021414380160), the Nanjing Science and Technology Development Project (201605019), and the Medical Scientific Research Foundation of Jiangsu Province of China (Q2017006).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Z.Z. contributed to data collection and statistical analyses and wrote the manuscript. B.Z. designed the protocol and reviewed the manuscript. X.W. contributed to data collection and MRI analysis and wrote the manuscript. X.Z. and Q.X.Y. designed the protocol. Z.Q. contributed to MRI analysis. W.Z. contributed to data collection. D.Z. conceived and designed the study. Y.B. designed the study and oversaw all clinical aspects of study conduct and manuscript preparation. Y.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 12th International Diabetes Federation Western Pacific Region Congress/10th Scientific Meeting of Asian Association for the Study of Diabetes, Kuala Lumpur, Malaysia, 22–25 November 2018.