Letter. Novel subgroups of adult‐onset diabetes in a UK community setting

Abstract

Sir, A recent study by Ahlqvist et al. proposed that individuals with adult-onset diabetes could be grouped into five clusters based on six variables: i.e. age, BMI, HbA1c, presence of glutamic acid decarboxylase (GAD) antibodies, and homeostasis model assessment (HOMA) of insulin resistance.1 We investigated the applicability of this new classification among a multi-ethnic UK population. We analysed data from the electronic care record for participants attending community diabetes clinics in Sutton, an ethnically diverse London borough, between October 2016 and February 2017. We defined the diabetic clusters as close to the Swedish paper as possible. Thus, participants with evidence of autoimmune (GAD positive) insulin-requiring diabetes were classified as cluster 1. This cluster included individuals with latent autoimmune disease of adults (LADA). We categorised participants in cluster 2 as severely insulin-deficient if they were insulin requiring but GAD antibody negative with very low 2-hour post meal C-peptide concentrations <94pmol/L. In the absence of HOMA assessments, we defined severe insulin resistance (cluster 3) as participants requiring >1 unit/kg insulin.2 We included participants in cluster 4 if they required <1 unit/kg insulin or were only on oral medication. Participants classified as having mild age-related diabetes (cluster 5) were aged >60 years at diagnosis with modest metabolic derangement. Fisher's test was used to compare categorical variables and the level of significance was p<0.05. Mann-Whitney U test was used to compare non-parametric data. Compared with the Swedish dataset in which mild obesity-related diabetes (cluster 4) comprised 21.6% of participants, we classified 72.5% in this category (Figure 1). This discrepancy may have resulted from our definition of severe insulin resistance based on insulin dose/kg rather than on HOMA-IR assessment. The second largest subgroup were classified as having severe insulin-resistant diabetes (cluster 3) comprising 13.5% (15.3% in the Swedish dataset) and had the highest incidence of macroalbuminuria, chronic kidney disease (CKD) and non-alcoholic fatty liver disease (NAFLD); (Table 1). By contrast, cluster 3 participants in our study were younger compared with the Swedish subgroup and this may account for a low incidence of coronary artery disease and stroke in our group. We categorised 7% of our participants within the severe autoimmune insulin deficient cluster, a result similar to that reported by Ahlqvist et al. (6.4%).1 As expected, participants were younger, non-obese, had poor metabolic control, and had the shortest time to sustained insulin use. We observed the highest risk of retinopathy (79%) in this cluster. The severe insulin-deficient diabetes cluster 2 comprised 4.5% of our study group and at diagnosis were non-obese with median HbA1c of 67mmol/mol; range 57–93mmol/mol (8.3% [7.4–10.7%] in NGSP units). In accordance with the Swedish findings, the mild age-related diabetes cluster (2.5%) had only mild metabolic derangement with median HbA1c at diagnosis of 48mmol/mol (6.5% in NGSP units). Cluster 1 (n=14) Cluster 2 (n=9) Cluster 3 (n=27) Cluster 4 (n=145) Cluster 5 (n=5) Retinopathy at diagnosis: no. (%) Any retinopathy Maculopathy 0 0 0 0 3 (11) 0 16 (11) 3 (2) 0 0 Current retinopathy: no. (%) Any retinopathy Maculopathy 11 (79) 3 (21) 3 (33) 1 (11) 17 (63) 9 (33) 88 (61) 23 (16) 0 0 We further examined our data by ethnicity; Caucasians represented 74% of the participants, South Asians 22.5% and African/Afro-Caribbeans 3.5%. Eighty-one percent of non-Caucasians were classified to the mild obesity-related diabetes subgroup (cluster 4). Non-Caucasians within this cluster were younger, had lower BMI and a lower incidence of current retinopathy (46% vs 74%; p<0.001) compared with Caucasians. It is well recognised that diabetes occurs at younger ages and lower levels of BMI in South Asians.3, 4 Our data suggest that it may be possible to characterise a UK population with diabetes into the novel subgroups based on simple clinical criteria. A community registry-based study such as this has limitations. Some participant subgroups, like the mild age-related diabetes cluster, are largely managed by local general practitioners and not referred to community clinics, and are therefore under-represented in this work. Another limitation is that we had no data on HOMA-IR. Nevertheless, we identified a similar subgroup, who had severe insulin resistance and a significantly higher incidence of NAFLD when compared with those classified as mild insulin resistance (66% [cluster 3] vs 11% [cluster 4]; p<0.001). This may have implications for their management and risk of complications. Larger population-based studies are needed to confirm the generalisability of the Swedish classification in more diverse populations. There are no conflicts of interest declared.