In the Journals

CGM, machine learning reveal dysglycemia phenotypes in type 1 diabetes

Researchers have identified three distinct patterns of dysglycemia in adolescents with type 1 diabetes using a combination of blinded continuous glucose monitoring data and machine-learning techniques, offering clinicians an opportunity to better tailor therapy, according to findings published in Pediatric Diabetes.

Anna Kahkoska

“The study demonstrates that among adolescents with type 1 diabetes and elevated HbA1c, CGM data can be used to uncover subgroups for which glycemic control is challenged by different patterns in hypoglycemia, hyperglycemia and glycemic variability,” Anna Kahkoska, PhD, an MD/PhD candidate in the department of nutrition at the University of North Carolina at Chapel Hill, told Endocrine Today. “The analysis represents a novel use of CGM data toward broadening the concept of glycemic control from HbA1c to understanding a multifaceted profile that includes glycemic excursions and overall variability.”

Kahkoska and colleagues analyzed data from 234 adolescents aged 13 to 16 years with type 1 diabetes and an HbA1c of at least 8% who wore blinded CGM at baseline for 7 days as part of the FLEX randomized trial (49.8% girls; mean age, 15 years; 76.1% white; mean diabetes duration, 6.4 years; mean HbA1c, 9.6%). Participants had blood glucose readings for a median of 160 hours. Researchers grouped participants based on their placement on a self-organizing map, which is a machine-learning technique constructed from eight CGM metrics measuring hypoglycemia, hyperglycemia and glycemic variability. The researchers used mixed-effects models to characterize 18-month changes in HbA1c.

Using machine learning, researchers identified three distinct clusters of dysglycemia patterns based on the eight CGM metrics. Cluster 1, with 141 adolescents (60.3%), showed severe daytime hyperglycemia with low exposure to and incidence of hypoglycemia relative to other clusters. Cluster 1 also showed the lowest glycemic variability, with mean daytime and nighttime coefficients of variation of 35.5% and 35.7%, respectively.

Cluster 2, with 53 adolescents (22.7%), showed severe hyperglycemia overnight with moderate hypoglycemia (median daytime episodes, 4) and moderate variability.

Cluster 3, with 40 adolescents (17.1%), showed moderate hyperglycemia, with the highest measures of hypoglycemia exposure and incidence relative to the other clusters (median daytime episodes, 8). Cluster 3 also showed the highest glycemic variability in the daytime and overnight, with mean daytime and nighttime coefficients of variation of 4.1% and 51.7%, respectively.

In analyzing 18-month HbA1c measures, the researchers observed between-cluster differences that persisted after adjustment for study site and randomization (P for interaction = .006). Adolescents in cluster 3 experienced an increase in HbA1c over 18 months, rising from a mean of 8.7% to 9.6%, whereas HbA1c levels for clusters 1 and 2 remained stable during the study, according to researchers.

Compared with participants with a baseline HbA1c of less than 9%, researchers found that adolescents with a baseline HbA1c greater than 9% were less likely to use an insulin pump (P = .02), had higher insulin doses (P = .03), more frequent blood glucose monitoring (P = .004), lower motivation (P = .03) and poorer adherence to diabetes self-management (P = .003).

“The data reinforce the concept that adolescents with type 1 diabetes and elevated HbA1c do not show homogenous patterns in CGM measures of blood glucose dynamics,” Kahkoska said. “Interestingly, different patterns in dysglycemia are not explained by the individual sociodemographic, clinical or psychosocial characteristics that typically drive treatment recommendations with regard to HbA1c.”

Kahkoska said the observed heterogeneity in CGM measures of dysglycemia among adolescents with type 1 diabetes and elevated HbA1c warrants the development of tailored interventions targeted to specific dysglycemia phenotypes.

“Further understanding of these subgroups, including the patient-level drivers of significant CGM profiles, is crucial to pave the way for targeted interventions to optimize dysglycemia and the associated clinical outcomes in type 1 diabetes,” Kahkoska said. – by Regina Schaffer

For more information:

Anna Kahkoska, PhD, can be reached at the Gillings School of Global Public Health, University of North Carolina at Chapel Hill, 245 Rosenau Hall, Campus Box 7461, Chapel Hill, NC 27599; email: anna_kahkoska@med.unc.edu.

Disclosures: The National Institute of Diabetes and Digestive and Kidney Diseases and the Leona M. and Harry B. Helmsley Charitable Trust funded this research. One of the study authors reports he has consulted for Abbott, Eli Lilly and Sanofi.

Researchers have identified three distinct patterns of dysglycemia in adolescents with type 1 diabetes using a combination of blinded continuous glucose monitoring data and machine-learning techniques, offering clinicians an opportunity to better tailor therapy, according to findings published in Pediatric Diabetes.

Anna Kahkoska

“The study demonstrates that among adolescents with type 1 diabetes and elevated HbA1c, CGM data can be used to uncover subgroups for which glycemic control is challenged by different patterns in hypoglycemia, hyperglycemia and glycemic variability,” Anna Kahkoska, PhD, an MD/PhD candidate in the department of nutrition at the University of North Carolina at Chapel Hill, told Endocrine Today. “The analysis represents a novel use of CGM data toward broadening the concept of glycemic control from HbA1c to understanding a multifaceted profile that includes glycemic excursions and overall variability.”

Kahkoska and colleagues analyzed data from 234 adolescents aged 13 to 16 years with type 1 diabetes and an HbA1c of at least 8% who wore blinded CGM at baseline for 7 days as part of the FLEX randomized trial (49.8% girls; mean age, 15 years; 76.1% white; mean diabetes duration, 6.4 years; mean HbA1c, 9.6%). Participants had blood glucose readings for a median of 160 hours. Researchers grouped participants based on their placement on a self-organizing map, which is a machine-learning technique constructed from eight CGM metrics measuring hypoglycemia, hyperglycemia and glycemic variability. The researchers used mixed-effects models to characterize 18-month changes in HbA1c.

Using machine learning, researchers identified three distinct clusters of dysglycemia patterns based on the eight CGM metrics. Cluster 1, with 141 adolescents (60.3%), showed severe daytime hyperglycemia with low exposure to and incidence of hypoglycemia relative to other clusters. Cluster 1 also showed the lowest glycemic variability, with mean daytime and nighttime coefficients of variation of 35.5% and 35.7%, respectively.

Cluster 2, with 53 adolescents (22.7%), showed severe hyperglycemia overnight with moderate hypoglycemia (median daytime episodes, 4) and moderate variability.

Cluster 3, with 40 adolescents (17.1%), showed moderate hyperglycemia, with the highest measures of hypoglycemia exposure and incidence relative to the other clusters (median daytime episodes, 8). Cluster 3 also showed the highest glycemic variability in the daytime and overnight, with mean daytime and nighttime coefficients of variation of 4.1% and 51.7%, respectively.

In analyzing 18-month HbA1c measures, the researchers observed between-cluster differences that persisted after adjustment for study site and randomization (P for interaction = .006). Adolescents in cluster 3 experienced an increase in HbA1c over 18 months, rising from a mean of 8.7% to 9.6%, whereas HbA1c levels for clusters 1 and 2 remained stable during the study, according to researchers.

PAGE BREAK

Compared with participants with a baseline HbA1c of less than 9%, researchers found that adolescents with a baseline HbA1c greater than 9% were less likely to use an insulin pump (P = .02), had higher insulin doses (P = .03), more frequent blood glucose monitoring (P = .004), lower motivation (P = .03) and poorer adherence to diabetes self-management (P = .003).

“The data reinforce the concept that adolescents with type 1 diabetes and elevated HbA1c do not show homogenous patterns in CGM measures of blood glucose dynamics,” Kahkoska said. “Interestingly, different patterns in dysglycemia are not explained by the individual sociodemographic, clinical or psychosocial characteristics that typically drive treatment recommendations with regard to HbA1c.”

Kahkoska said the observed heterogeneity in CGM measures of dysglycemia among adolescents with type 1 diabetes and elevated HbA1c warrants the development of tailored interventions targeted to specific dysglycemia phenotypes.

“Further understanding of these subgroups, including the patient-level drivers of significant CGM profiles, is crucial to pave the way for targeted interventions to optimize dysglycemia and the associated clinical outcomes in type 1 diabetes,” Kahkoska said. – by Regina Schaffer

For more information:

Anna Kahkoska, PhD, can be reached at the Gillings School of Global Public Health, University of North Carolina at Chapel Hill, 245 Rosenau Hall, Campus Box 7461, Chapel Hill, NC 27599; email: anna_kahkoska@med.unc.edu.

Disclosures: The National Institute of Diabetes and Digestive and Kidney Diseases and the Leona M. and Harry B. Helmsley Charitable Trust funded this research. One of the study authors reports he has consulted for Abbott, Eli Lilly and Sanofi.