Carsten Wridt Stoltenberg, Stine Hangaard, Ole Hejlesen, Thomas Kronborg, Peter Vestergaard, Morten Hasselstrøm Jensen
{"title":"预测 2 型糖尿病患者在接受速效胰岛素治疗后 HbA1c 达不到目标值的情况:在临床试验数据中使用机器学习框架。","authors":"Carsten Wridt Stoltenberg, Stine Hangaard, Ole Hejlesen, Thomas Kronborg, Peter Vestergaard, Morten Hasselstrøm Jensen","doi":"10.1177/19322968241280096","DOIUrl":null,"url":null,"abstract":"<p><strong>Background and aims: </strong>Glycemic control is crucial for people with type 2 diabetes. However, only about half achieve the advocated HbA1c target of ≤7%. Identifying those who will probably struggle to reach this target may be valuable as they require additional support. Thus, the aim of this study was to develop a model to predict people with type 2 diabetes not achieving HbA1c target after initiating fast-acting insulin.</p><p><strong>Methods: </strong>Data from a randomized controlled trial (NCT01819129) of participants with type 2 diabetes initiating fast-acting insulin were used. Data included demographics, clinical laboratory values, self-monitored blood glucose (SMBG), health-related quality of life (SF-36), and body measurements. A logistic regression was developed to predict HbA1c target nonachievers. A potential of 196 features was input for a forward feature selection. To assess the performance, a 20-repeated stratified 5-fold cross-validation with area under the receiver operating characteristics curve (AUROC) was used.</p><p><strong>Results: </strong>Out of the 467 included participants, 98 (21%) did not achieve HbA1c target of ≤7%. The forward selection identified 7 features: baseline HbA1c (%), mean postprandial SMBG at all meals 3 consecutive days before baseline (mmol/L), sex, no ketones in urine, baseline albumin (g/dL), baseline low-density lipoprotein cholesterol (mmol/L), and traces of protein in urine. The model had an AUROC of 0.745 [95% CI = 0.734, 0.756].</p><p><strong>Conclusions: </strong>The model was able to predict those who did not achieve HbA1c target with promising performance, potentially enabling early identification of people with type 2 diabetes who require additional support to reach glycemic control.</p>","PeriodicalId":15475,"journal":{"name":"Journal of Diabetes Science and Technology","volume":" ","pages":"19322968241280096"},"PeriodicalIF":4.1000,"publicationDate":"2024-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11571615/pdf/","citationCount":"0","resultStr":"{\"title\":\"Prediction of People With Type 2 Diabetes Not Achieving HbA1c Target After Initiation of Fast-Acting Insulin Therapy: Using Machine Learning Framework on Clinical Trial Data.\",\"authors\":\"Carsten Wridt Stoltenberg, Stine Hangaard, Ole Hejlesen, Thomas Kronborg, Peter Vestergaard, Morten Hasselstrøm Jensen\",\"doi\":\"10.1177/19322968241280096\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><strong>Background and aims: </strong>Glycemic control is crucial for people with type 2 diabetes. However, only about half achieve the advocated HbA1c target of ≤7%. Identifying those who will probably struggle to reach this target may be valuable as they require additional support. Thus, the aim of this study was to develop a model to predict people with type 2 diabetes not achieving HbA1c target after initiating fast-acting insulin.</p><p><strong>Methods: </strong>Data from a randomized controlled trial (NCT01819129) of participants with type 2 diabetes initiating fast-acting insulin were used. Data included demographics, clinical laboratory values, self-monitored blood glucose (SMBG), health-related quality of life (SF-36), and body measurements. A logistic regression was developed to predict HbA1c target nonachievers. A potential of 196 features was input for a forward feature selection. To assess the performance, a 20-repeated stratified 5-fold cross-validation with area under the receiver operating characteristics curve (AUROC) was used.</p><p><strong>Results: </strong>Out of the 467 included participants, 98 (21%) did not achieve HbA1c target of ≤7%. The forward selection identified 7 features: baseline HbA1c (%), mean postprandial SMBG at all meals 3 consecutive days before baseline (mmol/L), sex, no ketones in urine, baseline albumin (g/dL), baseline low-density lipoprotein cholesterol (mmol/L), and traces of protein in urine. The model had an AUROC of 0.745 [95% CI = 0.734, 0.756].</p><p><strong>Conclusions: </strong>The model was able to predict those who did not achieve HbA1c target with promising performance, potentially enabling early identification of people with type 2 diabetes who require additional support to reach glycemic control.</p>\",\"PeriodicalId\":15475,\"journal\":{\"name\":\"Journal of Diabetes Science and Technology\",\"volume\":\" \",\"pages\":\"19322968241280096\"},\"PeriodicalIF\":4.1000,\"publicationDate\":\"2024-09-20\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11571615/pdf/\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Journal of Diabetes Science and Technology\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1177/19322968241280096\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"ENDOCRINOLOGY & METABOLISM\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Diabetes Science and Technology","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1177/19322968241280096","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"ENDOCRINOLOGY & METABOLISM","Score":null,"Total":0}
Prediction of People With Type 2 Diabetes Not Achieving HbA1c Target After Initiation of Fast-Acting Insulin Therapy: Using Machine Learning Framework on Clinical Trial Data.
Background and aims: Glycemic control is crucial for people with type 2 diabetes. However, only about half achieve the advocated HbA1c target of ≤7%. Identifying those who will probably struggle to reach this target may be valuable as they require additional support. Thus, the aim of this study was to develop a model to predict people with type 2 diabetes not achieving HbA1c target after initiating fast-acting insulin.
Methods: Data from a randomized controlled trial (NCT01819129) of participants with type 2 diabetes initiating fast-acting insulin were used. Data included demographics, clinical laboratory values, self-monitored blood glucose (SMBG), health-related quality of life (SF-36), and body measurements. A logistic regression was developed to predict HbA1c target nonachievers. A potential of 196 features was input for a forward feature selection. To assess the performance, a 20-repeated stratified 5-fold cross-validation with area under the receiver operating characteristics curve (AUROC) was used.
Results: Out of the 467 included participants, 98 (21%) did not achieve HbA1c target of ≤7%. The forward selection identified 7 features: baseline HbA1c (%), mean postprandial SMBG at all meals 3 consecutive days before baseline (mmol/L), sex, no ketones in urine, baseline albumin (g/dL), baseline low-density lipoprotein cholesterol (mmol/L), and traces of protein in urine. The model had an AUROC of 0.745 [95% CI = 0.734, 0.756].
Conclusions: The model was able to predict those who did not achieve HbA1c target with promising performance, potentially enabling early identification of people with type 2 diabetes who require additional support to reach glycemic control.
期刊介绍:
The Journal of Diabetes Science and Technology (JDST) is a bi-monthly, peer-reviewed scientific journal published by the Diabetes Technology Society. JDST covers scientific and clinical aspects of diabetes technology including glucose monitoring, insulin and metabolic peptide delivery, the artificial pancreas, digital health, precision medicine, social media, cybersecurity, software for modeling, physiologic monitoring, technology for managing obesity, and diagnostic tests of glycation. The journal also covers the development and use of mobile applications and wireless communication, as well as bioengineered tools such as MEMS, new biomaterials, and nanotechnology to develop new sensors. Articles in JDST cover both basic research and clinical applications of technologies being developed to help people with diabetes.