Researchers have developed a novel diabetes classification system that calls for five distinct disease etiologies to better target early treatment to patients who would benefit the most, according to an analysis published in The Lancet Diabetes & Endocrinology.
“More accurately diagnosing diabetes could give us valuable insights into how it will develop over time, allowing us to predict and treat complications before they develop,” Leif Groop, PhD, professor in the department of clinical sciences at Lund University Diabetes Centre, Sweden, and the Institute for Molecular Medicine Finland, said in a press release. “Existing treatment guidelines are limited by the fact they respond to poor metabolic control when it has developed, but do not have the means to predict which patients will need intensified treatment. This study moves us towards a more clinically useful diagnosis and represents an important step toward precision medicine in diabetes.”
Groop and colleagues performed data-driven cluster analysis for 8,980 adults with newly diagnosed diabetes from the All New Diabetics in Scania (ANDIS) cohort in Sweden. Clusters were based on six variables: presence of glutamate decarboxylase antibodies (GADA), age at diabetes diagnosis, BMI, HbA1c, and homoeostatic model assessment 2 estimates of beta-cell function and insulin resistance. Variables were related to prospective data from patient records, including new diabetes drugs prescriptions. Researchers replicated the findings in three further independent cohorts: the Scania Diabetes Registry (n = 1,466), All New Diabetics in Uppsala (n = 844) and Diabetes Registry Vaasa (n = 3,485). Researchers used Cox regression and logistic regression analyses to compare time to medication, time to reaching the treatment goal and risk for diabetic complications across clusters, as well as genetic associations.
The researchers identified five clusters in both men (n = 5,334) and women (n = 3,646) in the ANDIS cohort. Cluster 1 (n = 577) was characterized by early-onset disease, low BMI, poor metabolic control, insulin deficiency and presence of GADA. Patients in cluster 2 (n = 1,575) had characteristics similar to cluster 1, but were GADA-negative. Cluster 3 (n = 1,373) was characterized by severe insulin resistance and high BMI, whereas cluster 4 (n = 1,942) was characterized by obesity, but not insulin resistance. Cluster 5 (n = 3,513) was described as mild, age-related diabetes with older patients but only modest metabolic derangements, according to researchers. Clustering was replicated across the remaining three cohorts.
In comparing disease progression, treatment and development of diabetes-related complications, the researchers observed differences depending on diabetes cluster. Clusters 1 and 2, for example, had a higher HbA1c at diagnosis, with poor control persisting during follow-up. Diabetic ketoacidosis was more frequent in clusters 1 and 2, whereas cluster 3 had the highest prevalence of nonalcoholic fatty liver disease.
Researchers found that the proportion of patients prescribed metformin was highest in cluster 2 and lowest in cluster 1, but the proportion of patients on metformin was also low in cluster 3, a group expected to benefit most from the therapy.
The metformin findings show that “traditional classification is unable to tailor treatment to the underlying pathogenic defects,” the researchers noted.
In ANDIS, patients in cluster 3 were most likely to develop chronic kidney disease during follow-up (mean, 3.9 years), whereas early signs of diabetic retinopathy were more common in cluster 2 vs. other clusters (OR = 1.6; 95% CI, 1.3-1.9), both findings that were replicated across the other cohorts.
Researchers did not observe between-cluster differences for age- and sex-adjusted coronary event and stroke risk, and no genetic variant was associated with all the clusters.
“The results of our study suggest that this new clustering of patients with adult-onset diabetes is superior to the classic diabetes classification because it identifies patients at high risk for diabetic complications at diagnosis and provides information about underlying disease mechanisms, thereby guiding choice of therapy,” the researchers wrote. “By contrast with previous attempts to dissect the heterogeneity of diabetes, we used variables reflective of key aspects of diabetes that are monitored in patients. Thus, this clustering can easily be applied to both existing diabetes cohorts (eg, from drug trials) and patients in diabetes clinics.”
The researchers noted that a web-based tool to assign patients to specific clusters, provided the appropriate variables have been measured, is under development. – by Regina Schaffer
Disclosures: The authors report no relevant financial disclosures.