Classification of type 1 diabetes: A pathogenic and treatment-based classification

AIM

To improve the diagnosis and classification of patients who fail to satisfy current type 1 diabetes diagnostic criteria.

METHODS

Review of the literature and current diagnostic guidelines.

DISCUSSION

We propose a novel, clinically useful classification based on islet autoantibody status and non-fasting C-peptide levels. Notably, we discuss the subgroup of latent autoimmune diabetes in the young and propose a new subgroup classification of autoantibody negative type 1 diabetes in remission.

CONCLUSION

A novel classification system is proposed. Further work is needed to accurately diagnose and manage minority type 1 diabetes subgroups.

Secondary analysis of newly diagnosed type 2 diabetes subgroups and treatment responses in the MARCH cohort

OBJECTIVE

To incorporate new clusters in the MARCH (Metformin and AcaRbose in Chinese patients as the initial Hypoglycemic treatment) cohort of newly diagnosed type 2 diabetes (T2D) patients and compare the anti-glycemic effects of metformin and acarbose across different clusters.

METHODS

K-means cluster analysis was performed based on six clinical indicators. The diabetic clusters in the MARCH cohort were retrospectively associated with the response to metformin and acarbose.

RESULTS

A total of 590 newly diagnosed T2D patients were classified by data-driven clusters into the MARD (mild obesity-related diabetes) (34.1 %), MOD (mild obesity-related diabetes) (34.1 %), SIDD (severe insulin-deficient diabetes) (20.3 %) and SIRD (severe insulin-resistant diabetes) (11.5 %) subgroups at baseline. At 24 and 48 weeks, 346 participants had finished the follow-up. After the adjustment of age, gender, weight, baseline HbA1c, baseline fasting glucose and 2-h postprandial blood glucose (2hPG), metformin mainly decreased the fasting glucose (0.07 ± 0.89 vs -0.26 ± 0.83, P = 0.043) in the MARD subgroup presented with OGTT (oral glucose tolerance test) results compared with acarbose group at 24 weeks. Acarbose led to a greater decrease in 2hPG in the MOD subgroup compared with metformin group (0.08 ± 0.86 vs -0.24 ± 0.92, P = 0.037) at 24 weeks. There was a also significant interaction between cluster and treatment efficacy in HbA1c (glycated hemoglobin) reduction in metformin and acarbose groups at 24 and 48 weeks (pinteraction<0.001).

CONCLUSIONS

Metformin and acarbose affected different metabolic variables depending on the diabetes subtype.

The relationship between islet autoantibody status and the genetic risk of type 1 diabetes in adult-onset type 1 diabetes

AIMS/HYPOTHESIS

The reason for the observed lower rate of islet autoantibody positivity in clinician-diagnosed adult-onset vs childhood-onset type 1 diabetes is not known. We aimed to explore this by assessing the genetic risk of type 1 diabetes in autoantibody-negative and -positive children and adults.

METHODS

We analysed GAD autoantibodies, insulinoma-2 antigen autoantibodies and zinc transporter-8 autoantibodies (ZnT8A) and measured type 1 diabetes genetic risk by genotyping 30 type 1 diabetes-associated variants at diagnosis in 1814 individuals with clinician-diagnosed type 1 diabetes (1112 adult-onset, 702 childhood-onset). We compared the overall type 1 diabetes genetic risk score (T1DGRS) and non-HLA and HLA (DR3-DQ2, DR4-DQ8 and DR15-DQ6) components with autoantibody status in those with adult-onset and childhood-onset diabetes. We also measured the T1DGRS in 1924 individuals with type 2 diabetes from the Wellcome Trust Case Control Consortium to represent non-autoimmune diabetes control participants.

RESULTS

The T1DGRS was similar in autoantibody-negative and autoantibody-positive clinician-diagnosed childhood-onset type 1 diabetes (mean [SD] 0.274 [0.034] vs 0.277 [0.026], p=0.4). In contrast, the T1DGRS in autoantibody-negative adult-onset type 1 diabetes was lower than that in autoantibody-positive adult-onset type 1 diabetes (mean [SD] 0.243 [0.036] vs 0.271 [0.026], p<0.0001) but higher than that in type 2 diabetes (mean [SD] 0.229 [0.034], p<0.0001). Autoantibody-negative adults were more likely to have the more protective HLA DR15-DQ6 genotype (15% vs 3%, p<0.0001), were less likely to have the high-risk HLA DR3-DQ2/DR4-DQ8 genotype (6% vs 19%, p<0.0001) and had a lower non-HLA T1DGRS (p<0.0001) than autoantibody-positive adults. In contrast to children, autoantibody-negative adults were more likely to be male (75% vs 59%), had a higher BMI (27 vs 24 kg/m²) and were less likely to have other autoimmune conditions (2% vs 10%) than autoantibody-positive adults (all p<0.0001). In both adults and children, type 1 diabetes genetic risk was unaffected by the number of autoantibodies (p>0.3). These findings, along with the identification of seven misclassified adults with monogenic diabetes among autoantibody-negative adults and the results of a sensitivity analysis with and without measurement of ZnT8A, suggest that the intermediate type 1 diabetes genetic risk in autoantibody-negative adults is more likely to be explained by the inclusion of misclassified non-autoimmune diabetes (estimated to represent 67% of all antibody-negative adults, 95% CI 61%, 73%) than by the presence of unmeasured autoantibodies or by a discrete form of diabetes. When these estimated individuals with non-autoimmune diabetes were adjusted for, the prevalence of autoantibody positivity in adult-onset type 1 diabetes was similar to that in children (93% vs 91%, p=0.4).

CONCLUSIONS/INTERPRETATION

The inclusion of non-autoimmune diabetes is the most likely explanation for the observed lower rate of autoantibody positivity in clinician-diagnosed adult-onset type 1 diabetes. Our data support the utility of islet autoantibody measurement in clinician-suspected adult-onset type 1 diabetes in routine clinical practice.

Identifying type 1 and 2 diabetes in research datasets where classification biomarkers are unavailable: assessing the accuracy of published approaches

OBJECTIVES

We aimed to compare the performance of approaches for classifying insulin-treated diabetes within research datasets without measured classification biomarkers, evaluated against two independent biological definitions of diabetes type.

STUDY DESIGN AND SETTING

We compared accuracy of ten reported approaches for classifying insulin-treated diabetes into type 1 (T1D) and type 2 (T2D) diabetes in two cohorts: UK Biobank (UKBB) n = 26,399 and Diabetes Alliance for Research in England (DARE) n = 1,296. The overall performance for classifying T1D and T2D was assessed using: a T1D genetic risk score and genetic stratification method (UKBB); C-peptide measured at >3 years diabetes duration (DARE).

RESULTS

Approaches' accuracy ranged from 71% to 88% (UKBB) and 68% to 88% (DARE). When classifying all participants, combining early insulin requirement with a T1D probability model (incorporating diagnosis age and body image issue [BMI]), and interview-reported diabetes type (UKBB available in only 15%) consistently achieved high accuracy (UKBB 87% and 87% and DARE 85% and 88%, respectively). For identifying T1D with minimal misclassification, models with high thresholds or young diagnosis age (<20 years) had highest performance. Findings were incorporated into an online tool identifying optimum approaches based on variable availability.

CONCLUSION

Models combining continuous features with early insulin requirement are the most accurate methods for classifying insulin-treated diabetes in research datasets without measured classification biomarkers.

Explainable diabetes classification using hybrid Bayesian-optimized TabNet architecture

Diabetes is a deadly chronic disease that occurs when the pancreas is not able to produce ample insulin or when the body cannot use insulin effectively. If undetected, it may lead to a host of health complications. Hence, accurate and explainable early-stage detection of diabetes is essential for the proper administration of treatment options in leading a healthy and productive life. For this, we developed an interpretable TabNet model tuned via Bayesian optimization (BO). To achieve model-specific interpretability, the attention mechanism of TabNet architecture was used, which offered the local and global model explanations on the influence of the attributes on the outcomes. The model was further explained locally and globally using more robust model-agnostic LIME and SHAP eXplainable Artificial Intelligence (XAI) tools. The proposed model outperformed all benchmarked models by obtaining high accuracy of 92.2% and 99.4% using the Pima Indians diabetes dataset (PIDD) and the early-stage diabetes risk prediction dataset (ESDRPD), respectively. Based on the XAI results, it was clear that the most influential attribute for diabetes classification using PIDD and ESDRPD were Insulin and Polyuria, respectively. The feature importance values registered for insulin was 0.301 (PIDD) and for polyuria 0.206 was registered (ESDRPD). The high accuracy and ancillary interpretability of our objective model is expected to increase end-users trust and confidence in early-stage detection of diabetes.

Customized and Automated Machine Learning-Based Models for Diabetes Type 2 Classification

This study aims to develop models to accurately classify patients with type 2 diabetes using the Practice Fusion dataset. We use Random Forest (RF), Support Vector Classifier (SVC), AdaBoost classifier, an ensemble model, and automated machine learning (AutoML) model. We compare the performance of all models in a five-fold cross-validation scheme using four evaluation measures. Experimental results demonstrate that the AutoML model outperformed individual and ensemble models in all evaluation measures.

Type 2 diabetes subgroups and potential medication strategies in relation to effects on insulin resistance and beta-cell function: A step toward personalised diabetes treatment?

Background

Type 2 diabetes is a syndrome defined by hyperglycaemia that is the result of various degrees of pancreatic β-cell failure and reduced insulin sensitivity. Although diabetes can be caused by multiple metabolic dysfunctions, most patients are defined as having either type 1 or type 2 diabetes. Recently, Ahlqvist and colleagues proposed a new method of classifying patients with adult-onset diabetes, considering the heterogenous metabolic phenotype of the disease. This new classification system could be useful for more personalised treatment based on the underlying metabolic disruption of the disease, although to date no prospective intervention studies have generated data to support such a claim.

Scope of Review

In this review, we first provide a short overview of the phenotype and pathogenesis of type 2 diabetes and discuss the current and new classification systems. We then review the effects of different anti-diabetic medication classes on insulin sensitivity and β-cell function and discuss future treatment strategies based on the subgroups proposed by Ahlqvist et al.

Major Conclusions

The proposed novel type 2 diabetes subgroups provide an interesting concept that could lead to a better understanding of the pathophysiology of the broad group of type 2 diabetes, paving the way for personalised treatment choices based on understanding the root cause of the disease. We conclude that the novel subgroups of adult-onset diabetes would benefit from anti-diabetic medications that take into account the main pathophysiology of the disease and thereby prevent end-organ damage. However, we are only beginning to address the personalised treatment of type 2 diabetes, and studies investigating the effects of current and novel drugs in subgroups with different metabolic phenotypes are needed to develop personalised treatment of the syndrome

•

Novel subgroups of type 2 diabetes provide a concept that could lead to a better understanding of its pathophysiology.

•

Treatment strategies would benefit from anti-diabetic medications that influence the main pathophysiology of diabetes.

•

Here, we review different anti-diabetic medications classes affecting insulin sensitivity and β-cell function.

•

We suggest that future treatment strategies could benefit by taking into account subgroups provided by Ahlqvist et al.