A clinical diagnostic model based on an eXtreme Gradient Boosting algorithm to distinguish type 1 diabetes

Performance of European prediction models for classification of type 1 and type 2 diabetes in Indians

AIM

We aimed to determine the performance of European prediction models in an Indian population to classify type 1 diabetes(T1D) and type 2 diabetes(T2D).

METHODS

We assessed discrimination and calibration of published models of diabetes classification, using retrospective data from electronic medical records of 83309 participants aged 18-50 years living in India. Diabetes type was defined based on C-peptide measurement and early insulin requirement. Models assessed combinations of clinical measurements: age at diagnosis, body mass index(mean = 26.6 kg/m2), sex(male = 64.9 %), Glutamic acid decarboxylase(GAD) antibody, serum cholesterol, serum triglycerides, and high-density lipoprotein(HDL) cholesterol.

RESULTS

67955 participants met inclusion criteria, of whom 0.8 % had T1D, which was markedly lower than model development cohorts. Model discrimination for clinical features was broadly similar in our Indian cohort compared to the European cohort: area under the receiver operating characteristic curve(AUC ROC) was 0.90 vs. 0.90 respectively, but was lower in the subset of young participants with measured GAD antibodies(n = 2404): and an AUC ROC of 0.87 when clinical features, sex, lipids and GAD antibodies were combined. All models substantially overestimated the likelihood of T1D, reflecting the lower prevalence of T1D in the Indian population. However, good model performance was achieved after recalibration by updating the model intercept and slope.

CONCLUSION

Models for diabetes classification maintain the discrimination of T1D and T2D in this Indian population, where T2D is far more common, but require recalibration to obtain appropriate model probabilities. External validation and recalibration are needed before these tools can be used in non-European populations.

The challenges of identifying and studying type 1 diabetes in adults

Diagnosing type 1 diabetes in adults is difficult since type 2 diabetes is the predominant diabetes type, particularly with an older age of onset (approximately >30 years). Misclassification of type 1 diabetes in adults is therefore common and will impact both individual patient management and the reported features of clinically classified cohorts. In this article, we discuss the challenges associated with correctly identifying adult-onset type 1 diabetes and the implications of these challenges for clinical practice and research. We discuss how many of the reported differences in the characteristics of autoimmune/type 1 diabetes with increasing age of diagnosis are likely explained by the inadvertent study of mixed populations with and without autoimmune aetiology diabetes. We show that when type 1 diabetes is defined by high-specificity methods, clinical presentation, islet-autoantibody positivity, genetic predisposition and progression of C-peptide loss remain broadly similar and severe at all ages and are unaffected by onset age within adults. Recent clinical guidance recommends routine islet-autoantibody testing when type 1 diabetes is clinically suspected or in the context of rapid progression to insulin therapy after a diagnosis of type 2 diabetes. In this moderate or high prior-probability setting, a positive islet-autoantibody test will usually confirm autoimmune aetiology (type 1 diabetes). We argue that islet-autoantibody testing of those with apparent type 2 diabetes should not be routinely undertaken as, in this low prior-prevalence setting, the positive predictive value of a single-positive islet antibody for autoimmune aetiology diabetes will be modest. When studying diabetes, extremely high-specificity approaches are needed to identify autoimmune diabetes in adults, with the optimal approach depending on the research question. We believe that until these recommendations are widely adopted by researchers, the true phenotype of late-onset type 1 diabetes will remain largely misunderstood.

Clinical prediction models combining routine clinical measures identify participants with youth-onset diabetes who maintain insulin secretion in the range associated with type 2 diabetes: The SEARCH for Diabetes in Youth Study

Objective

With the high prevalence of pediatric obesity and overlapping features between diabetes subtypes, accurately classifying youth-onset diabetes can be challenging. We aimed to develop prediction models that, using characteristics available at diabetes diagnosis, can identify youth who will retain endogenous insulin secretion at levels consistent with type 2 diabetes (T2D).

Methods

We studied 2,966 youth with diabetes in the prospective SEARCH study (diagnosis age ≤19 years) to develop prediction models to identify participants with fasting c-peptide ≥250 pmol/L (≥0.75ng/ml) after >3 years (median 74 months) of diabetes duration. Models included clinical measures at baseline visit, at a mean diabetes duration of 11 months (age, BMI, sex, waist circumference, HDL-C), with and without islet autoantibodies (GADA, IA-2A) and a Type 1 Diabetes Genetic Risk Score (T1DGRS).

Results

Models using routine clinical measures with or without autoantibodies and T1DGRS were highly accurate in identifying participants with c-peptide ≥0.75 ng/ml (17% of participants; 2.3% and 53% of those with and without positive autoantibodies) (area under receiver operator curve [AUCROC] 0.95-0.98). In internal validation, optimism was very low, with excellent calibration (slope=0.995-0.999). Models retained high performance for predicting retained c-peptide in older youth with obesity (AUCROC 0.88-0.96), and in subgroups defined by self-reported race/ethnicity (AUCROC 0.88-0.97), autoantibody status (AUCROC 0.87-0.96), and clinically diagnosed diabetes types (AUCROC 0.81-0.92).

Conclusion

Prediction models combining routine clinical measures at diabetes diagnosis, with or without islet autoantibodies or T1DGRS, can accurately identify youth with diabetes who maintain endogenous insulin secretion in the range associated with type 2 diabetes.

Recent findings regarding the synergistic effects of emodin and its analogs with other bioactive compounds: Insights into new mechanisms

CONTEXT

Emodin is a natural bioactive ingredient mainly extracted from traditional Chinese herbs. Increasing lines of evidence suggest that emodin and its analogs exert notable synergistic pharmacological effects with other bioactive compounds.

OBJECTIVE

This review provides an overview of the pharmacological activity of emodin and its analogs in combination with other physiologically active substances, describes the related molecular mechanisms, and discusses future prospects in this field.

METHODS

Information from multiple scientific databases, such as PubMed, the China Knowledge Resource Integrated Database from the China National Knowledge Infrastructure (CNKI), the Web of Science, Google Scholar, and Baidu Scholar, was collected between January 2006 and August 2022. The subject terms used in the literature search were emodin, pharmaceutical activities, analogs, aloe emodin, rhein, and synergistic effects.

RESULTS

The comprehensive literature analysis suggested that combinations of emodin or its analogs with other bioactive compounds exert notable synergistic anticancer, anti-inflammatory, and antimicrobial effects and that such combinations improve glucose and lipid metabolism and central nervous system diseases.

DISCUSSION AND CONCLUSIONS

Further assessments of the dose-effect relationship and the differences in the efficacy of emodin or its analogs with other bioactive compounds among various modes of administration are needed, and a drug safety evaluation of these combinations needs to be carefully performed. Future studies should also focus on determining the optimal drug combinations for specific diseases.

A novel 6-metabolite signature for prediction of clinical outcomes in type 2 diabetic patients undergoing percutaneous coronary intervention

Background

Outcome prediction tools for patients with type 2 diabetes mellitus (T2DM) undergoing percutaneous coronary intervention (PCI) are lacking. Here, we developed a machine learning-based metabolite classifier for predicting 1-year major adverse cardiovascular events (MACEs) after PCI among patients with T2DM.

Methods

Serum metabolomic profiling was performed in a nested case–control study of 108 matched pairs of patients with T2DM occurring and not occurring MACEs at 1 year after PCI, then the matched pairs were 1:1 assigned into the discovery and internal validation sets. External validation was conducted using targeted metabolite analyses in an independent prospective cohort of 301 patients with T2DM receiving PCI. The function of candidate metabolites was explored in high glucose-cultured human aortic smooth muscle cells (HASMCs).

Results

Overall, serum metabolome profiles differed between diabetic patients with and without 1-year MACEs after PCI. Through VSURF, a machine learning approach for feature selection, we identified the 6 most important metabolic predictors, which mainly targeted the nicotinamide adenine dinucleotide (NAD⁺) metabolism. The 6-metabolite model based on random forest and XGBoost algorithms yielded an area under the curve (AUC) of ≥ 0.90 for predicting MACEs in both discovery and internal validation sets. External validation of the 6-metabolite classifier also showed good accuracy in predicting MACEs (AUC 0.94, 95% CI 0.91–0.97) and target lesion failure (AUC 0.89, 95% CI 0.83–0.95). In vitro, there were significant impacts of altering NAD⁺ biosynthesis on bioenergetic profiles, inflammation and proliferation of HASMCs.

Conclusion

The 6-metabolite model may help for noninvasive prediction of 1-year MACEs following PCI among patients with T2DM.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12933-022-01561-1.

Diabetes mellitus risk prediction in the presence of class imbalance using flexible machine learning methods

Background

Early detection and prediction of type two diabetes mellitus incidence by baseline measurements could reduce associated complications in the future. The low incidence rate of diabetes in comparison with non-diabetes makes accurate prediction of minority diabetes class more challenging.

Methods

Deep neural network (DNN), extremely gradient boosting (XGBoost), and random forest (RF) performance is compared in predicting minority diabetes class in Tehran Lipid and Glucose Study (TLGS) cohort data. The impact of changing threshold, cost-sensitive learning, over and under-sampling strategies as solutions to class imbalance have been compared in improving algorithms performance.

Results

DNN with the highest accuracy in predicting diabetes, 54.8%, outperformed XGBoost and RF in terms of AUROC, g-mean, and f1-measure in original imbalanced data. Changing threshold based on the maximum of f1-measure improved performance in g-mean, and f1-measure in three algorithms. Repeated edited nearest neighbors (RENN) under-sampling in DNN and cost-sensitive learning in tree-based algorithms were the best solutions to tackle the imbalance issue. RENN increased ROC and Precision-Recall AUCs, g-mean and f1-measure from 0.857, 0.603, 0.713, 0.575 to 0.862, 0.608, 0.773, 0.583, respectively in DNN. Weighing improved g-mean and f1-measure from 0.667, 0.554 to 0.776, 0.588 in XGBoost, and from 0.659, 0.543 to 0.775, 0.566 in RF, respectively. Also, ROC and Precision-Recall AUCs in RF increased from 0.840, 0.578 to 0.846, 0.591, respectively.

Conclusion

G-mean experienced the most increase by all imbalance solutions. Weighing and changing threshold as efficient strategies, in comparison with resampling methods are faster solutions to handle class imbalance. Among sampling strategies, under-sampling methods had better performance than others.