Hermanns Norbert, Ehrmann Dominic, Kulzer Bernhard, Klinker Laura, Haak Thomas, Schmitt Andreas
Research Institute of the Diabetes-Academy Mergentheim (FIDAM), Bad Mergentheim, Germany.
Department of Clinical Psychology and Psychotherapy, University of Bamberg, Bamberg, Germany.
Diabetes Obes Metab. 2025 Jan;27(1):61-70. doi: 10.1111/dom.15983. Epub 2024 Oct 7.
To analyse the potential drivers (glucose level, complications, diabetes type, gender, age and mental health) of diabetes symptoms using continuous glucose monitoring (CGM) and ecological momentary assessment.
Participants used a smartphone application to rate 25 diabetes symptoms in their daily lives over 8 days. These symptoms were grouped into four blocks so that each symptom was rated six times on 2 days (noon, afternoon and evening). The symptom ratings were associated with the glucose levels for the previous 2 hours, measured with CGM. Linear mixed-effects models were used, allowing for nested random effects and the conduct of N = 1 analysis of individual associations.
In total, 192 individuals with type 1 diabetes and 179 with type 2 diabetes completed 6380 app check-ins. Four symptoms showed a significant negative association with glucose values, indicating higher ratings at lower glucose (speech difficulties, P = .003; coordination problems, P = .00005; confusion, P = .049; and food cravings, P = .0003). Four symptoms showed a significant positive association with glucose values, indicating higher scores at higher glucose (thirst, P = .0001; urination, P = .0003; taste disturbances, P = .021; and itching, P = .0120). There were also significant positive associations between microangiopathy and eight symptoms. Elevated depression and diabetes distress were associated with higher symptom scores. N = 1 analysis showed highly idiosyncratic associations between symptom reports and glucose levels.
The N = 1 analysis facilitated the creation of personalized symptom profiles related to glucose levels with consideration of factors such as complications, gender, body mass index, depression and diabetes distress. This approach can enhance precision monitoring for diabetes symptoms in precision medicine.
使用连续血糖监测（CGM）和生态瞬时评估来分析糖尿病症状的潜在驱动因素（血糖水平、并发症、糖尿病类型、性别、年龄和心理健康）。
参与者使用智能手机应用程序在8天内对其日常生活中的25种糖尿病症状进行评分。这些症状被分为四个组块，以便每种症状在2天内（中午、下午和晚上）被评分6次。症状评分与前2小时用CGM测量的血糖水平相关。使用线性混合效应模型，允许嵌套随机效应并进行个体关联的N = 1分析。
共有192名1型糖尿病患者和179名2型糖尿病患者完成了6380次应用程序签到。四种症状与血糖值呈显著负相关，表明血糖水平较低时评分较高（言语困难，P = 0.003；协调问题，P = 0.00005；困惑，P = 0.049；以及食物渴望，P = 0.0003）。四种症状与血糖值呈显著正相关，表明血糖水平较高时得分较高（口渴，P = 0.0001；排尿，P = 0.0003；味觉障碍，P = 0.021；以及瘙痒，P = 0.0120）。微血管病变与八种症状之间也存在显著正相关。抑郁和糖尿病痛苦程度升高与较高的症状评分相关。N = 1分析显示症状报告与血糖水平之间存在高度特异的关联。
N = 1分析有助于创建与血糖水平相关的个性化症状档案，同时考虑并发症、性别、体重指数、抑郁和糖尿病痛苦程度等因素。这种方法可以提高精准医学中糖尿病症状的精准监测。
