Status of Glycosylated Hemoglobin and Prediction of Glycemic Control Among Patients with Insulin-Treated Type 2 Diabetes in North China: A Multicenter Observational Study
Type 2 diabetes mellitus (T2DM) is a global health concern, with its prevalence expected to rise significantly in the coming decades. In China, the situation is particularly alarming, as the country has the largest number of individuals living with diabetes worldwide. Effective blood glucose control is crucial for managing T2DM and preventing complications. This study aimed to investigate the status of glycemic control among insulin-treated T2DM patients in North China and explore the application of machine learning algorithms combined with elastic network (EN) regression for predicting glycemic control.
The study was conducted across 27 centers in six cities in North China, including Tianjin, Tangshan, Datong, Qinhuangdao, Cangzhou, and Taiyuan. Data were collected from 2787 consecutive participants between January 2016 and December 2017. The inclusion criteria were a diagnosis of T2DM, age of 18 years or older, and basal insulin use for at least three months. Participants were excluded if they refused to sign informed consent, had a history of drug allergy, were pregnant or lactating, or had psychiatric conditions.
Data collection included basic information such as sex, age, smoking status, alcohol consumption, and marital status, as well as diabetes-related information such as disease duration, exercise habits, diet, oral medications, complications, hypoglycemia events, and insulin dosage. Physical examinations measured height, weight, waist circumference, hip circumference, and blood pressure. Laboratory tests included fasting blood glucose (FBG), 1-hour blood glucose (1HBG), 2-hour blood glucose (2HBG), glycosylated hemoglobin (HbA1c), total cholesterol (TC), triglycerides (TGs), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C).
The study defined glycemic control as HbA1c levels below 7.0%, following the guidelines for the prevention and treatment of T2DM in China. The results showed that only 45.82% of participants achieved this target, indicating that more than half of the patients had poor glycemic control. Factors associated with better glycemic control included a history of hypertension, atherosclerotic cardiovascular disease (ASCVD), regular exercise, and higher total cholesterol levels. Conversely, central adiposity, family history of diabetes, longer duration of T2DM, complications, higher insulin doses, elevated blood pressure, and hypertension were identified as risk factors for poor glycemic control.
To address the issue of multicollinearity among the variables, the study employed an elastic network (EN) regression, which combines the least absolute shrinkage and selection operator (LASSO) and ridge regression methods. The EN regression was used to reduce the dimensionality of the data, selecting 19 out of 42 initial variables. The selected variables included protective factors such as hypertensive history, ASCVD history, nocturnal hypoglycemia, exercise, and TC, as well as risk factors such as central adiposity, family history, duration of T2DM, typical disease characteristics, complications, insulin dose, oral hypoglycemic drugs (OHA), FBG, 2HBG, blood pressure, HDL-C, LDL-C, and hypertension.
Following the dimensionality reduction, three machine learning algorithms—random forest (RF), support vector machine (SVM), and back propagation artificial neural network (BP-ANN)—were used to predict glycemic control. The data were divided into training and testing sets in a 7:3 ratio, and the models’ performance was evaluated using sensitivity, specificity, accuracy, and the area under the receiver operating characteristics curve (AUC). The results showed that the RF algorithm performed best among the three, with sensitivity, specificity, accuracy, and AUC values of 0.79, 0.73, 0.75, and 0.75, respectively, after dimensionality reduction. The SVM algorithm showed the most significant improvement after dimensionality reduction, with sensitivity increasing by 37.70%, specificity by 7.94%, accuracy by 17.74%, and AUC by 18.03%.
Compared to traditional logistic regression models, the machine learning models combined with EN regression demonstrated higher sensitivity and accuracy. The logistic regression model had sensitivity and accuracy values of 0.52 and 0.56, respectively, while the RF, SVM, and BP-ANN models achieved values of 0.79, 0.84, and 0.78 for sensitivity and 0.70, 0.73, and 0.73 for accuracy, respectively. This indicates that the EN and machine learning models are superior alternatives to traditional logistic regression for predicting glycemic control in T2DM patients.
The study also highlighted the importance of lifestyle factors in glycemic control. Regular exercise and dietary adjustments, particularly in the consumption of vegetable oil, were associated with better glycemic control. Conversely, central adiposity, a sedentary lifestyle, and poor dietary habits were linked to elevated HbA1c levels. These findings underscore the need for comprehensive lifestyle interventions in the management of T2DM.
In conclusion, this study provides valuable insights into the status of glycemic control among insulin-treated T2DM patients in North China. The findings reveal that more than half of the patients had poor glycemic control, putting them at a higher risk of developing diabetic complications. The study also demonstrates the potential of combining elastic network regression with machine learning algorithms to improve the prediction of glycemic control. These methods offer a more accurate and sensitive approach compared to traditional logistic regression models, providing a valuable tool for healthcare providers in managing T2DM.
The study’s limitations include the lack of rigorous random sampling, which may introduce bias, and the focus on outpatients, which may not fully represent the broader population of T2DM patients in China. Future research should aim to include a more diverse sample and explore the causal relationships between the identified factors and glycemic control. Despite these limitations, the study’s findings have significant implications for the management of T2DM in China and highlight the need for continued efforts to improve glycemic control and prevent complications.
doi.org/10.1097/CM9.0000000000000585
Was this helpful?
0 / 0