Abstract: Background　Diabetes significantly increases the risk of cognitive impairment. Olfactory impairment is a precursor manifestation of cognitive disorders, making olfactory assessment a potential tool for identifying individuals with diabetes at high risk of cognitive impairment. However, the optimal cutoff value of olfactory tests for predicting the risk of mild cognitive impairment (MCI) in patients with type 2 diabetes mellitus (T2DM) remains unclear. Objective　To investigate the predictive value and optimal cutoff of the olfactory function score for MCI development in T2DM patients. Method　This study included a cross-sectional analysis and a longitudinal cohort validation. In the cross-sectional component, we enrolled T2DM patients hospitalized in the Department of Endocrinology, Nanjing Drum Tower Hospital, between January 2024 and September 2025. Demographic data, laboratory parameters, cognitive function, and olfactory performance were collected. Participants were categorized into normal cognition and MCI groups based on established diagnostic criteria. Multivariate logistic regression was used to assess the association between reduced olfactory function and MCI risk. Restricted cubic spline (RCS) analysis was performed to evaluate potential non-linear relationships and identify the optimal cutoff of the total olfactory score. Subgroup analyses were conducted to assess the robustness of the findings. For the longitudinal cohort, 111 T2DM patients with normal baseline cognition were followed up. Kaplan–Meier survival curves and multivariate Cox proportional hazards regression models were used to validate the predictive performance of the identified olfactory cutoff for incident MCI. Results　The cross-sectional analysis included 1,063 T2DM patients (424 in the normal cognition group and 639 in the MCI group). The MCI group had significantly lower total olfactory scores than the normal cognition group (P<0.05). Multivariate logistic regression confirmed that impaired olfactory function was independently associated with higher MCI risk(OR=0.609, 95%CI=0.527-0.705, P<0.001). RCS analysis revealed a non-linear relationship between olfactory score and MCI risk (P for non-linearity = 0.045) and identified an optimal cutoff of 42 points. Patients with scores ≤ 42 had a 114% higher risk of MCI (OR=2.14, 95%CI=1.66-2.76, P<0.05). Subgroup analyses showed consistent associations across age, sex, education level, smoking status, and comorbidities such as hypertension or dyslipidemia (all interaction P>0.05). In the longitudinal cohort, 39 of 111 participants developed MCI during a median follow-up of 9 months. kaplan survial curve analysis showed that the low-score group (olfactory score ≤ 42) had a significantly higher cumulative incidence of MCI than the high-score group (P=0.003), with a median time to MCI onset that was 6 months earlier. Multivariate Cox regression showed that the low-score group had a 2.61-fold higher risk of developing MCI (HR=2.61, 95%CI=1.29-5.28,P<0.05). Conclusions　Impaired olfactory function is an independent risk factor for MCI in T2DM patients. Olfactory assessment has predictive value for identifying high-risk individuals, with an optimal cutoff of 42 points for the total olfactory score.

Key words: Diabetes mellitus, type 2, Cognitive dysfunction, Olfaction disorders, Forecasting, Restricted cubic spline

摘要： 背景　糖尿病显著增加了认知障碍的发生风险。嗅觉损害是认知障碍的前驱表现，因此，嗅觉评估可作为识别糖尿病认知障碍高危人群的重要工具。然而，目前关于嗅觉评估预测2型糖尿病（T2DM）人群轻度认知障碍（MCI）发生风险的最佳切点尚不明确。目的　探究嗅觉功能评分对T2DM患者发生MCI的预测价值及切点。方法　本研究采用横断面研究和纵向队列研究。横断面研究纳入2024年1月—2025年9月在南京鼓楼医院内分泌科住院的T2DM患者，收集患者的一般资料及实验室检查数据，并进行认知功能评估及嗅觉行为学测试。根据MCI诊断标准将患者分为正常认知组和合并MCI组，采用多因素Logistic回归分析探究T2DM患者嗅觉功能降低与MCI发生风险之间的关系；采用限制性立方样条分析（RCS）评估潜在的非线性联系，探究嗅觉功能评分预测T2DM患者MCI发生风险的最佳截断值；通过亚组分析进一步验证结果的稳健性。纵向队列研究共纳入111名基线认知正常的T2DM患者，采用Kaplan-Meier生存曲线及多因素Cox比例风险回归模型验证嗅觉测评总分对T2DM患者MCI发生风险的预测效能。结果　横断面研究共纳入1 063例T2DM患者，其中正常认知组424例，合并MCI组639例。与正常认知组相比，合并MCI组患者嗅觉测评总分更低（P<0.05）。多因素Logistic回归分析显示，嗅觉功能减退与T2DM患者MCI发生风险增加相关（OR=0.609，95%CI=0.527~0.705，P<0.001）。RCS曲线分析显示嗅觉测评总分与MCI发生风险存在非线性关系（P非线性=0.045），嗅觉测评总分预测T2DM患者MCI发生风险的最佳截断值为42分，低于该最佳截断值的T2DM人群发生MCI的风险增加114%（OR=2.14，95%CI=1.66~2.76，P<0.05）。亚组分析显示，嗅觉功能与认知障碍的关联强度不受年龄、性别、受教育年限、吸烟史、合并高血压或高脂血症等因素的影响（P>0.05）。在纵向队列研究，中位随访时间为9个月，共有39例研究对象发生MCI。Kaplan Meier生存曲线分析显示，低分组（嗅觉测评总分≤42分）累积MCI发生率高于高分组（嗅觉测评总分>42分）（P=0.003），其中位MCI发生时间较高分组提前6个月。多因素Cox回归分析显示，嗅觉测评总分低分组患者MCI发生风险高于高分组（HR=2.61，95%CI=1.29~5.28，P<0.05）。结论　嗅觉功能下降是T2DM患者发生MCI的独立危险因素，嗅觉评估在识别MCI高危人群方面具有一定预测价值，嗅觉测评总分对T2DM患者MCI发生风险的最佳截断值为42分。

关键词: 糖尿病, 2 型；认知功能障碍；嗅觉障碍；预测；限制性立方样条