Cluster analysis may help to predict diabetic complications
medwireNews: Allocating people with diabetes to one of five previously described data-driven clusters may help to identify those at high immediate risk for developing diabetic neuropathy or non-alcoholic fatty liver disease (NAFLD), say researchers.
Michael Roden (Heinrich Heine University, Düsseldorf, Germany) and colleagues used the methodology of the team that first proposed the diabetes clusters to classify 1105 people with newly diagnosed diabetes who participated in the German Diabetes Study.
They were classified as having mild age-related diabetes (MARD; 34.9%), mild obesity-related diabetes (MOD; 29.2%), severe insulin-resistant diabetes (SIRD; 10.9%), severe insulin-deficient diabetes (SIDD; 2.5%), or – for those with positive glutamic acid decarboxylase (GAD) antibodies – severe autoimmune diabetes (SAID; 22.3%).
The researchers who originally proposed these cluster-based diabetes definitions reported that the SIDD group was at the greatest risk for retinopathy and the SIRD group had a particularly high risk for developing kidney disease.
In the current analysis, published in The Lancet Diabetes & Endocrinology, the researchers were unable to confirm the association of retinopathy with the SIDD classification due to the very small number of cases in their cohort. But they did find that this group had a high prevalence of neuropathy, present in 36% compared with 5–17% of the other groups.
In addition, the team found that people in the SIRD group had the highest baseline hepatocellular lipid content (median 19% vs 1–7%) and fatty liver index, although these NAFLD markers increased in people in the MOD and MARD groups over the following 5 years.
The SIRD group also had more baseline markers of hepatic fibrosis, and by the end of follow-up 26% had developed hepatic fibrosis, compared with 0–15% of the other groups.
The analysis also confirmed the greater prevalence of diabetic nephropathy in the SIRD group relative to all others, both at baseline and after 5 years.
However, a previous analysis by a team from the University of Exeter, UK, although successfully replicating the original clusters, also suggested that routine clinical variables provided a more straightforward and as accurate means of predicting complications.
In a commentary accompanying the current study, Stefano Del Prato (University of Pisa, Italy) says the evidence overall suggests that “we might not yet have a reliable way to move forward towards a precision medicine approach for treatment of type 2 diabetes,” adding that use of routine clinical variables “could be informative enough, but not sufficiently precise.”
He also highlights that 23% of participants switched between clusters during the 5 years of follow-up, owing to changes in glucose and triglyceride levels and the fatty liver index, and six people developed GAD antibodies and moved into the SAID cluster.
“Precision medicine for type 2 diabetes, therefore, could require even more accurate profiling of individuals belonging to a given cluster, by integrating genetic and omics data, digital phenotyping, sensor-based behavioural monitoring, and pharmacogenomics,” concludes Del Prato.
medwireNews is an independent medical news service provided by Springer Healthcare. © 2019 Springer Healthcare part of the Springer Nature group