Background Liraglutide, a drug used for the management of obesity, has many known side effects. In this study, we developed a predictive model for the occurrence of liraglutide-related side effects using data from electronic medical records (EMRs).
Methods This study included 237 patients from Seoul St. Mary's Hospital and Eunpyeong St. Mary's Hospital who were prescribed liraglutide. An endocrinologist obtained medical data through an EMR chart review. Model performance was evaluated using the mean of the area under the receiver operating characteristic curve (AUROC) with a 95% confidence interval (CI).
Results A predictive model was developed for patients who were prescribed liraglutide. However, 37.1% to 75.5% of many variables were missing, and the AUROC of the developed predictive model was 0.630 (95% CI, 0.551–0.708). Patients who had previously taken antiobesity medication had significantly fewer side effects than those without previous antiobesity medication use (20.7% vs. 41.4%, P<0.003). The risk of side effect occurrence was significantly higher in patients with diabetes than in patients without diabetes by 2.389 times (odds ratio, 2.389; 95% CI, 1.115–5.174).
Conclusions This study did not successfully develop a predictive model for liraglutide-related side effects, primarily due to issues related to missing data. When prescribing antiobesity drugs, detailed records and basic blood tests are expected to be essential. Further large-scale studies on liraglutide-related side effects are needed after obtaining high-quality data.
Citations
Citations to this article as recorded by
The effects and side effects of liraglutide as a treatment for obesity Jeonghoon Ha, Jin Yu, Joonyub Lee, Hun-Sung Kim Cardiovascular Prevention and Pharmacotherapy.2022; 4(4): 142. CrossRef
Background Statin-associated muscle symptoms are one of the side effects that physicians should consider when prescribing statins. In this study, creatine kinase (CK) levels were measured following statin prescription, and various factors affecting the CK levels were determined using machine learning.
Methods Changes in the CK were observed every 3 months for a 12-month period in patients who received statins for the first time at Seoul St. Mary's Hospital. For each visit, we developed four basic models based on changes in the CK levels. Extreme gradient boosting, a scalable end-to-end tree boosting algorithm, which employs a decision-tree-based ensemble machine learning algorithm, was used for the prediction of changes in the CK.
Results A total of 23,860 patients were included. Among them, 19 patients (0.08%) had increased CK levels of 2,000 IU·L−1 or more 3 months after statin prescription, and 65 patients (0.27%) exhibited CK levels of over 2,000 IU·L−1 at least once during the 12-month study period. The area under the receiver operator characteristic of each model for each visit was 0.709–0.769, and the accuracy was 0.700–0.803. In each of the models, the variables that had the strongest influence on changes in the CK were sex and previous CK value.
Conclusions Through machine learning, factors influencing changes in the CK were identified. These results will provide the basis for future research, through which the optimal parameters of the CK prediction model can be found and the model can be used in clinical applications.
Background Glycated hemoglobin (HbA1c), which reflects the patient's blood sugar level, can only be measured in a hospital setting. Therefore, we developed a model predicting HbA1c using personal information and self-monitoring of blood glucose (SMBG) data solely obtained by a patient.
Methods Leave-one-out cross-validation (LOOCV) was performed at two university hospitals. After measuring the baseline HbA1c level before SMBG (Pre_HbA1c), the SMBG was recorded over a 3-month period. Based on these data, an HbA1c prediction model was developed, and the actual HbA1c value was measured after 3 months. The HbA1c values of the prediction model and actual HbA1c values were compared. Personal information was used in addition to SMBG data to develop the HbA1c predictive model.
Results Thirty model training sessions and evaluations were conducted using LOOCV. The average mean absolute error of the 30 models was 0.659 (range, 0.005–2.654). Pre_HbA1c had the greatest influence on HbA1c prediction after 3 months, followed by post-breakfast blood glucose level, oral hypoglycemic agent use, fasting glucose level, height, and weight, while insulin use had a limited effect on HbA1c values.
Conclusions The patient's SMBG data and personal information strongly influenced the HbA1c predictive model. In the future, it will be necessary to develop sophisticated predictive models using large samples for stable SMBG in patients.