Could putting people with adult-diagnosed diabetes into five clusters rather than two main types help to improve treatment and health?
Currently, most adults diagnosed with diabetes in adulthood are put into one of two main types: type 1 or type 2 diabetes. A new study poses the question of whether more categories of diabetes would help to improve treatment.
Type 2 represents the majority of adult diabetes diagnoses. In the UK, each person diagnosed with type 2 diabetes is treated in a similar way.
In their study, the researchers suggest that more precise treatment strategies could be used if diabetes, particularly type 2 diabetes, was divided into more groups.
The research showed that people with adult-diagnosed diabetes could be grouped into five distinct categories.
How the study was run
The study involved analysing data of around 9,000 newly-diagnosed adults with diabetes in Sweden.
The aim of the study was to see whether comparing and contrasting health information could lead to more refined and less broad categories of adult-diagnosed diabetes.
The researchers used six different factors to help towards dividing the participants into different categories:
- Age at diagnosis
- BMI
- HbA1c levels
- Insulin-producing ability
- Insulin resistance
- Presence of autoantibodies (glutamate decarboxylase antibodies or GADA)
Genetic risk factors were also reviewed, albeit not as part of the categorisation process.
Five clusters
The researchers found that they could group the people with adult-diagnosed diabetes into five different clusters.
Cluster 1 – severe autoimmune diabetes (SAID)
Characterised by:
- Early-onset
- Relatively low BMI
- High HbA1c
- Low insulin production
- Presence of at least one autoantibody
Ketoacidosis at diagnosis was most frequent within this cluster, occurring in about one in three people. As expected, metformin use was lowest in this cluster.
This was the smallest cluster and included people who would be diagnosed as having type 1 diabetes or LADA.
This cluster represented 6% of the participants.
Cluster 2 – severe insulin-deficient diabetes (SIDD)
Characterised by:
- Early-onset
- Relatively low BMI
- High HbA1c
- Low insulin production
- No presence of GADA autoantibody
Ketoacidosis at diagnosis was also common in this group, occurring in one in four people.
This cluster represented 18% of the participants.
Cluster 3 – severe insulin-resistant diabetes (SIRD)
Characterised by:
- High BMI
- Insulin resistance
This cluster had the highest prevalence of non-alcoholic fatty liver disease and the highest risk of developing chronic kidney disease.
Metformin use was low in this cluster despite the fact that the researchers noted that this cluster “would be expected to benefit the most from metformin”.
This cluster represented 15% of the participants.
Cluster 4 – mild obesity-related diabetes (MOD)
Characterised by:
- High BMI
- Lower HbA1c
- Low or no insulin resistance
This cluster represented 22% of the participants.
Cluster 5 – mild age-related diabetes (MARD)
Characterised by:
- Older onset
- Lower HbA1c
This cluster represented 39% of the participants.
Genetics
The researchers found further interesting results from the genetic data.
A variant of a gene associated with type 1 diabetes (rs2854275 variant in the HLA locus) was associated with the severe autoimmune diabetes cluster (cluster 1).
This genetic variant was not associated with the severe insulin-deficient diabetes (cluster 2) group. This further underlines the notion that this is a separate type to the autoimmune group.
A genetic variant that has previously been associated with non-alcoholic fatty liver disease (rs10401969 variant in the TM6SF2 gene) was associated with the severe insulin-resistant diabetes (cluster 3) group.
This variant was not associated with mild obesity-related diabetes (cluster 4), the other cluster characterised by high BMI. The researchers noted that this suggests that severe insulin resistant diabetes is unhealthier in terms of metabolic health than the mild obesity-related diabetes cluster.
A gene variant that has been linked with type 2 diabetes (rs7903146 variant in the TCF7L2 gene) was associated with the severe insulin-deficient, mild obesity-related and mild age-related clusters (clusters 2, 4 and 5). It is interesting that this gene variant was not associated with the severe insulin-resistant cluster (cluster 3).
Genetic risk score analysis
The researchers also reviewed for which clusters genetic risk scores for type 2 diabetes and insulin secretion matched up to.
Interestingly, the genetic risk score for type 2 diabetes was significantly associated with all of the clusters with the exception of the severe insulin resistant diabetes cluster (cluster 3).
The insulin secretion risk score used in the study was most strongly associated with the mild obesity-related and mild age-related clusters (clusters 4 and 5). The severe insulin-deficient group (cluster 2) showed a nominal association. No evidence of associated were found with the severe autoimmune or severe insulin-resistant clusters (clusters 1 and 3).
A new way of thinking about diabetes?
These results pose a number of questions related to how healthcare may be able to better diagnose and treat diabetes.
It is possible that some methods of screening and diagnosis may work better for some clusters than others. The findings also suggest that some clusters may benefit from treatments tailored to their cluster as opposed to the standard treatment algorithm that applies to type 2 diabetes and includes a number of these clusters.
Limitations
The researchers note that the way the study was ru, they cannot state that the different clusters all have separate causes. They acknowledge that the five clusters may not be the optimal way of classifying different subtypes of adult-diagnosed diabetes.
It is possible that someone might belong to one cluster initially and then be categorised within another cluster if re-analysed at a later date in their diabetes duration. The researchers cannot rule out the fact that this could take place.
The researchers analysed only two types of autoantibody. Had data been available on more autoantibodies, the results and clustering may have been different to some degree.
Furthermore, the researchers acknowledge that the classification of the five clusters was derived mainly from patients from northern Europe. It is possible that a different form of clustering may be found in other parts of the world.
All in all, the research provides a strong basis to begin evaluating whether adult-diagnosed patients could be better categorised into different subgroups.
The study, ‘Novel subgroups of adult-onset diabetes and their association with outcomes: a data-driven cluster analysis of six variables’, is published in The Lancet: Diabetes & Endocrinology journal.
