糖尿病（diabetes mellitus, DM）是一种常见的慢性代谢性疾病，其流行率在我国达到了12.4%；DM的发生率随年龄增长，60岁及以上的中国老年人群中，DM的流行率达到了20%。DM的人口基数大且难以根治；但同时，该疾病是全球十大死亡原因之一，长期的高血糖状态将导致多种并发症，影响脑结构与功能，并将使失智症的患病风险提升1.5至2倍，严重威胁健康。
糖尿病前期（prediabetes, PreD）是介于血糖代谢正常（normal glucose metabolism, NGM）与DM之间的过渡状态。PreD是一个相当庞大的群体：在中国的成年人群中，其流行率超过了35%。PreD一方面是发展为DM的高危阶段，其发展为DM的风险约为血糖正常人群的2.5至3倍；另一方面，不同于DM的难以根治，PerD存在逆转的可能。关注疾病的前驱阶段，了解PreD是否与认知下降和脑结构异常有关能够为将PreD作为干预对象以预防DM相关认知障碍及脑病的合理性提供信息，这对于降低DM的负面影响非常重要。
目前，PreD对认知的影响尚不明确，甚少研究关注PreD脑结构损伤的分布模式。因此，本研究基于多认知域神经心理学测验探究了糖尿病前期的认知表现特征；基于多模态神经影像技术，从灰质萎缩、白质萎缩及白质微结构损伤三个角度揭示了糖尿病前期阶段脑结构的损伤模式并建立了血糖代谢状态、脑结构异常与认知表现的关联。
本研究所纳入的参与者均来自于“北京老年脑健康促进计划”社区临床队列研究数据库，包括78名NGM（30M/48F，年龄：65.462±6.722岁），92名PreD（34M/58F，年龄：66.337±6.091岁）和108名DM（45M/63F，年龄：65.528±6.537岁），三组参与者的性别、年龄、教育程度无显著组间差异。所有参与者均完成了人口学资料、疾病史、成套神经心理学测验及T1加权成像及弥散张量成像采集。所有非DM均测定了糖化血红蛋白（glycated hemoglobin, HbA1c）和空腹血糖。其中，神经心理学测验包括总体认知功能、注意与加工速度、记忆、视空间、言语和执行功能。张量指标使用各项分数异性（fractional anisotropy, FA）、平均弥散率（Mean Diffusivity, MD）、轴向弥散率L1、径向弥散率L23m。
分别使用卡方检验或单因素方差分析对人口学资料、疾病史及生化指标中的分类变量（性别与高血压、高脂血、脑血管病疾病史）和连续变量（其余变量）的组间差异进行检验。使用单因素协方差分析（Analysis of Covariance, ANCOVA）探究认知表现的组间差异。脑影像方面，使用ANCOVA作为统计模型，在全脑范围进行基于体素的分析（voxel-based analysis, VBA），得到灰质体积、白质体积及白质微结构存在显著组间差异的团块后提取团块内相应指标的平均值进行事后检验。ANCOVA以颅内总体积（仅在分析灰质与白质体积组间差异时纳入）、年龄、性别、教育程度、疾病史、BMI为协变量，事后检验均采用Bonferroni 校正（p<0.05）。中介分析用于讨论血糖代谢状态、脑结构特征和认知之间的关系。
本研究的研究结果显示，（1）认知表现上，注意与加工速度是PreD主要影响的认知域，而DM阶段的总体认知功能、注意与加工速度、视空间功能和执行功能相比血糖正常老年人和PreD均显著降低；血糖代谢状态由血糖正常到PreD再到DM，除言语功能外的各项认知功能均呈逐级下降的趋势。（2）灰质体积上，相比血糖正常老年人，PreD在右侧小脑、海马、扣带回中部（GRF，pvoxel<0.005，pcluster<0.05），左侧枕中回与扣带回中部，右侧舌回及双侧额叶（pvoxel<0.005，cluster size>50）的灰质体积显著降低。DM在上述区域亦表现出显著的萎缩；此外，相比PreD，DM的左侧小脑、海马、舌回以及右侧枕中回的灰质体积进一步降低。（3）白质体积上，相比血糖正常老年人，PreD的白质萎缩主要位于左侧枕中回及右侧小脑上脚（pvoxel<0.005，cluster size>50）；而在DM阶段，包括丘脑后辐射、前后放射冠、内囊、外囊、颞叶、中央前回等在内的广泛的白质区域均表现出显著的白质萎缩。（4）白质微结构完整性上，相比血糖正常老年人，PreD的右侧小脑上脚（GRF，pvoxel<0.005，pcluster<0.05）及右侧中央后回（pvoxel<0.005，cluster size>50）表现出显著的白质微结构损伤，该区域的MD、L23m或L1显著增加。DM组表现出更广泛区域的白质微结构损伤，受损区域包括双侧外囊，双侧胼胝体膝部、体部、压部，双侧尤其是左侧终纹与丘脑等。（5）在非DM人群中，枕中回白质体积的萎缩介导了PreD对加工速度的影响。此外，从血糖正常到PreD再到DM的整个血糖代谢范围内，弥散性的白质和白质微结构损伤而不是灰质萎缩介导了HbA1c值的持续升高对记忆、加工速度与执行功能的负面影响。 
综上，本研究发现在诊断为糖尿病之前，已经能够在糖尿病前期观察到注意与加工速度下降的趋势及局部的脑结构异常；这些脑结构的异常，尤其是白质异常，介导了血糖代谢异常对认知的影响。预防策略应尽早开始，避免脑结构进一步的损伤以及多认知域的认知下降。糖尿病前期的干预可能有助于预防脑病、失智症等糖尿病相关并发症的发生。

Diabetes mellitus (DM) is a common chronic metabolic disease, with a prevalence rate of 12.4% in China. The incidence of DM increases with age. Among the elderly population aged 60 and above in China, the prevalence rate of DM reaches 20%. DM has a large population base and is difficult to cure. At the same time,, it is one of the top ten causes of death worldwide. Prolonged exposure to hyperglycemia in DM can lead to various complications, affecting brain structure and function, and increasing the risk of dementia by 1.5 to 2 times, posing a serious threat to health.
Prediabetes (PreD) is a state between normal glucose metabolism (NGM) and DM. There is a considerable number of individuals in the PreD stage, with a prevalence rate exceeding 35% among adults in China. On one hand, PreD is a high-risk stage for developing DM, with a 2.5 to 3 times higher risk compared to individuals with normal glucose metabolism. On the other hand, unlike DM, PreD has the potential for reversal. By focusing on the preclinical stage of the disease and exploring the association between PreD, cognitive decline, and brain structural abnormalities, valuable information can be obtained to justify the rationale for targeting PreD as the intervention subjects to prevent DM-related cognitive impairments and brain disorders. This is crucial for mitigating the negative impact of DM.
Currently, the impact of PreD on cognition remains unclear, and there is limited research focusing on the regional distribution of brain structural damage in PreD. Therefore, this study explores the cognition performance of PreD based on neuropsychological tests assessing multiple cognitive domain. Additionally, this study reveals the patterns of gray matter atrophy, white matter atrophy, and white matter microstructural damage in the PreD stage by using multimodal neuroimaging techniques. This study also establishes the association between glucose metabolism status, abnormal brain structure, and cognitive performance.
Participants in the current study were recruited from the Beijing Aging Brain Rejuvenation Initiative (BABRI) community clinical cohort, including 78 NGM individuals (30M/48F, age: 65.462 ± 6.722 years), 92 PreD individuals (34M/58F, age: 66.337±6.091 years), and 108 DM individuals (45M/63F, age: 65.528±6.537 years). There were no significant differences in gender, age, or education level among the three groups of participants. All participants underwent the collection of demographic information, history of disease, a battery of neuropsychological tests, T1-weighted and diffusion tensor imaging scans. glycated hemoglobin(HbA1c) and fasting blood glucose were measured for all non-DM participants. Neuropsychological tests included general cognitive function, attention and processing speed, memory, visuospatial, language and executive function. Tensor indexs used in this study were fractional anisotropy (FA), mean diffusivity (MD), L1 (i.e., axial diffusivity), and L23m (i.e., radial diffusivity).
Chi-square tests or one-way analysis of variance (ANOVA) were used to test the group differences in categorical variables (gender, the history of hypertension, dyslipidemia, and cerebrovascular disease) and continuous variables (the others) in demographic information, history of disease, and clinical indexs. One-way analysis of covariance (ANCOVA) was used to explore group differences in cognitive performance. For brain imaging, ANCOVA was used as the statistical model for whole-brain voxel-based analysis (VBA) to identify clusters with significant group differences in gray matter volume, white matter volume, or white matter microstructure. The mean value of the corresponding index in each cluster was extracted for post-hoc tests. Total intracranial volume (only included in analyzing gray and white matter volume differences), age, gender, education level, history of disease and BMI were included as covariates in ANCOVA, and bonferroni correction was used for post-hoc tests (p<0.05). Mediation analysis was used to investigate the relationships among glucose metabolism, brain structural characteristics and cognitive performance.
The main findings of this study are as follows: (1). In terms of cognitive performance, attention and processing speed is the main cognitive domain affected in PreD. Compared to individuals with normal glucose metabolism and PreD, those in the DM stage showed significant declines in general cognitive function, attention and processing speed, visuospatial, and executive function. There was a gradual decline in various cognitive functions from normal glucose metabolism to PreD and then to DM, except for language function. (2). Regarding gray matter volume, PreD participants exhibited significantly reduced gray matter volume in the right cerebellum, hippocampus, middle cingulate gyrus(GRF corrected，pvoxel<0.005，pcluster<0.05), left middle occipital cortex and middle cingulate gyrus, right lingual gyrus, and bilateral frontal lobes (pvoxel<0.005，cluster size>50) compared to individuals with normal glucose metabolism. DM participants also showed significant atrophy in these regions. Furthermore, compared to PreD, DM participants exhibited further reductions in gray matter volume in the left cerebellum, hippocampus, lingual gyrus, and right middle occipital cortex. (3). In terms of white matter volume, PreD participants showed white matter atrophy mainly in the left middle occipital cortex and right cerebellar peduncle compared to individuals with normal glucose metabolism (pvoxel<0.005, cluster size>50). In the DM stage, widespread white matter atrophy was observed in regions including the posterior thalamic radiation, anterior and posterior corona radiata, internal capsule, external capsule, temporal lobe, and precentral gyrus. (4). Regarding white matter microstructural integrity, PreD participants showed significant white matter microstructural damage in the right cerebellar peduncle(GRF，pvoxel<0.005，pcluster<0.05) and right postcentral gyrus(pvoxel<0.005，cluster size>50), characterized by increased MD、L23m or L1. DM participants exhibited more extensive white matter microstructural damage, including bilateral external capsules, bilateral corpus callosum (genu, body and splenium), and bilateral stria terminalis and thalamus. (5). The mediation analysis results showed that in the non-DM group, the atrophy of the white matter volume of middle occipital cortex mediated the impact of PreD on processing speed. Additionally, within the entire range of glucose metabolism from normal to PreD to DM, diffuse white matter and white matter microstructural damage, rather than gray matter atrophy, mediated the negative effects of increasing HbA1c levels on memory, processing speed, and executive function.
In conclusion, this study found that before the diagnosis of diabetes, a decline in attention and processing speed and localized brain structural abnormalities can already be observed in the PreD stage. These brain structural abnormalities, especially white matter abnormalities, mediate the impact of abnormal glucose metabolism on cognition. Early prevention strategies should be implemented to avoid further brain structural damage and cognitive decline in multiple domains. Intervention during the PreD stage may help prevent the occurrence of diabetes-related complications such as brain diseases and dementia.