Machine learning-based reproducible prediction of type 2 diabetes subtypes.

AIMS/HYPOTHESIS: Clustering-based subclassification of type 2 diabetes, which reflects pathophysiology and genetic predisposition, is a promising approach for providing personalised and effective therapeutic strategies. Ahlqvist's classification is currently the most vigorously validated method because of its superior ability to predict diabetes complications but it does not have strong consistency over time and requires HOMA2 indices, which are not routinely available in clinical practice and standard cohort studies. We developed a machine learning (ML) model to classify individuals with type 2 diabetes into Ahlqvist's subtypes consistently over time. METHODS: Cohort 1 dataset comprised 619 Japanese individuals with type 2 diabetes who were divided into training and test sets for ML models in a 7:3 ratio. Cohort 2 dataset, comprising 597 individuals with type 2 diabetes, was used for external validation. Participants were pre-labelled (T2Dkmeans) by unsupervised k-means clustering based on Ahlqvist's variables (age at diagnosis, BMI, HbA1c, HOMA2-B and HOMA2-IR) to four subtypes: severe insulin-deficient diabetes (SIDD), severe insulin-resistant diabetes (SIRD), mild obesity-related diabetes (MOD) and mild age-related diabetes (MARD). We adopted 15 variables for a multiclass classification random forest (RF) algorithm to predict type 2 diabetes subtypes (T2DRF15). The proximity matrix computed by RF was visualised using a uniform manifold approximation and projection. Finally, we used a putative subset with missing insulin-related variables to test the predictive performance of the validation cohort, consistency of subtypes over time and prediction ability of diabetes complications. RESULTS: T2DRF15 demonstrated a 94% accuracy for predicting T2Dkmeans type 2 diabetes subtypes (AUCs ≥0.99 and F1 score [an indicator calculated by harmonic mean from precision and recall] ≥0.9) and retained the predictive performance in the external validation cohort (86.3%). T2DRF15 showed an accuracy of 82.9% for detecting T2Dkmeans, also in a putative subset with missing insulin-related variables, when used with an imputation algorithm. In Kaplan-Meier analysis, the diabetes clusters of T2DRF15 demonstrated distinct accumulation risks of diabetic retinopathy in SIDD and that of chronic kidney disease in SIRD during a median observation period of 11.6 (4.5-18.3) years, similarly to the subtypes using T2Dkmeans. The predictive accuracy was improved after excluding individuals with low predictive probability, who were categorised as an 'undecidable' cluster. T2DRF15, after excluding undecidable individuals, showed higher consistency (100% for SIDD, 68.6% for SIRD, 94.4% for MOD and 97.9% for MARD) than T2Dkmeans. CONCLUSIONS/INTERPRETATION: The new ML model for predicting Ahlqvist's subtypes of type 2 diabetes has great potential for application in clinical practice and cohort studies because it can classify individuals with missing HOMA2 indices and predict glycaemic control, diabetic complications and treatment outcomes with long-term consistency by using readily available variables. Future studies are needed to assess whether our approach is applicable to research and/or clinical practice in multiethnic populations.

Identifying blood biomarkers for type 2 diabetes subtyping: a report from the ORIGIN trial.

AIMS/HYPOTHESIS: Individuals with diabetes can be clustered into five subtypes using up to six routinely measured clinical variables. We hypothesised that circulating protein levels might be used to distinguish between these subtypes. We recently used five of these six variables to categorise 7017 participants from the Outcome Reduction with an Initial Glargine Intervention (ORIGIN) trial into these subtypes: severe autoimmune diabetes (SAID, n=241), severe insulin-deficient diabetes (SIDD, n=1594), severe insulin-resistant diabetes (SIRD, n=914), mild obesity-related diabetes (MOD, n=1595) and mild age-related diabetes (MARD, n=2673). METHODS: Forward-selection logistic regression models were used to identify a subset of 233 cardiometabolic protein biomarkers that were independent determinants of one subtype vs the others. We then assessed the performance of adding identified biomarkers (one after one, from the most discriminant to the least) to predict each subtype vs the others using area under the receiver operating characteristic curve (AUC ROC). Models were adjusted for age, sex, ethnicity, C-peptide level, diabetes duration and glucose-lowering medication usage at blood collection. RESULTS: A total of 25 biomarkers were independent determinants of subtypes, including 13 for SIDD, 2 for SIRD, 7 for MOD and 11 for MARD (all p<4.3 × 10-5). The performance of the biomarker sets (comprising 1 to 25 biomarkers), assessed through the AUC ROC, ranged from 0.611 to 0.734, 0.723 to 0.861, 0.672 to 0.742, and 0.651 to 0.751, for SIDD, SIRD, MOD and MARD, respectively. No biomarkers other than GAD antibodies were determinants of SAID. CONCLUSIONS/INTERPRETATION: We identified 25 serum biomarkers, as independent determinants of type 2 diabetes subtypes, that could be combined into a diagnostic test for subtyping. TRIAL REGISTRATION: ORIGIN trial, ClinicalTrials.gov NCT00069784.

Skeletal muscle and intermuscular adipose tissue gene expression profiling identifies new biomarkers with prognostic significance for insulin resistance progression and intervention response.

AIMS/HYPOTHESIS: Although insulin resistance often leads to type 2 diabetes mellitus, its early stages are often unrecognised, thus reducing the probability of successful prevention and intervention. Moreover, treatment efficacy is affected by the genetics of the individual. We used gene expression profiles from a cross-sectional study to identify potential candidate genes for the prediction of diabetes risk and intervention response. METHODS: Using a multivariate regression model, we linked gene expression profiles of human skeletal muscle and intermuscular adipose tissue (IMAT) to fasting glucose levels and glucose infusion rate. Based on the expression patterns of the top predictive genes, we characterised and compared individual gene expression with clinical classifications using k-nearest neighbour clustering. The predictive potential of the candidate genes identified was validated using muscle gene expression data from a longitudinal intervention study. RESULTS: We found that genes with a strong association with clinical measures clustered into three distinct expression patterns. Their predictive values for insulin resistance varied substantially between skeletal muscle and IMAT. Moreover, we discovered that individual gene expression-based classifications may differ from classifications based predominantly on clinical variables, indicating that participant stratification may be imprecise if only clinical variables are used for classification. Of the 15 top candidate genes, ST3GAL2, AASS, ARF1 and the transcription factor SIN3A are novel candidates for predicting a refined diabetes risk and intervention response. CONCLUSION/INTERPRETATION: Our results confirm that disease progression and successful intervention depend on individual gene expression states. We anticipate that our findings may lead to a better understanding and prediction of individual diabetes risk and may help to develop individualised intervention strategies.

Correlates of Skeletal Muscle Mass and Differences Between Novel Subtypes in Recent-Onset Diabetes.

CONTEXT: Low skeletal muscle mass (SMM) is associated with long-standing diabetes but little is known about SMM in newly diagnosed diabetes. OBJECTIVE: We aimed to identify correlates of SMM in recent-onset diabetes and to compare SMM between novel diabetes subtypes. METHODS: SMM was normalized to body mass index (SMM/BMI) in 842 participants with known diabetes duration of less than 1 year from the German Diabetes Study (GDS). Cross-sectional associations between clinical variables, 79 biomarkers of inflammation, and SMM/BMI were assessed, and differences in SMM/BMI between novel diabetes subtypes were analyzed with different degrees of adjustment for confounders. RESULTS: Male sex and physical activity were positively associated with SMM/BMI, whereas associations of age, BMI, glycated hemoglobin A1c, homeostatic model assessment for ß-cell function, and estimated glomerular filtration rate with SMM/BMI were inverse (all P < .05; model r2 = 0.82). Twenty-three biomarkers of inflammation showed correlations with SMM/BMI after adjustment for sex and multiple testing (all P < .0006), but BMI largely explained these correlations. In a sex-adjusted analysis, individuals with severe autoimmune diabetes had a higher SMM/BMI whereas individuals with severe insulin-resistant diabetes and mild obesity-related diabetes had a lower SMM/BMI than all other subtypes combined. However, differences were attenuated after adjustment for the clustering variables. CONCLUSION: SMM/BMI differs between diabetes subtypes and may contribute to subtype differences in disease progression. Of note, clinical variables rather than biomarkers of inflammation explain most of the variation in SMM/BMI.

Machine learning based study for the classification of Type 2 diabetes mellitus subtypes.

PURPOSE: Data-driven diabetes research has increased its interest in exploring the heterogeneity of the disease, aiming to support in the development of more specific prognoses and treatments within the so-called precision medicine. Recently, one of these studies found five diabetes subgroups with varying risks of complications and treatment responses. Here, we tackle the development and assessment of different models for classifying Type 2 Diabetes (T2DM) subtypes through machine learning approaches, with the aim of providing a performance comparison and new insights on the matter. METHODS: We developed a three-stage methodology starting with the preprocessing of public databases NHANES (USA) and ENSANUT (Mexico) to construct a dataset with N = 10,077 adult diabetes patient records. We used N = 2,768 records for training/validation of models and left the remaining (N = 7,309) for testing. In the second stage, groups of observations -each one representing a T2DM subtype- were identified. We tested different clustering techniques and strategies and validated them by using internal and external clustering indices; obtaining two annotated datasets Dset A and Dset B. In the third stage, we developed different classification models assaying four algorithms, seven input-data schemes, and two validation settings on each annotated dataset. We also tested the obtained models using a majority-vote approach for classifying unseen patient records in the hold-out dataset. RESULTS: From the independently obtained bootstrap validation for Dset A and Dset B, mean accuracies across all seven data schemes were [Formula: see text] ([Formula: see text]) and [Formula: see text] ([Formula: see text]), respectively. Best accuracies were [Formula: see text] and [Formula: see text]. Both validation setting results were consistent. For the hold-out dataset, results were consonant with most of those obtained in the literature in terms of class proportions. CONCLUSION: The development of machine learning systems for the classification of diabetes subtypes constitutes an important task to support physicians for fast and timely decision-making. We expect to deploy this methodology in a data analysis platform to conduct studies for identifying T2DM subtypes in patient records from hospitals.

Mid-pregnancy sleep disturbances are not associated with mid-pregnancy maternal glycemia.

Background: In pregnancy, epidemiological data have consistently shown strong associations between sleep quality and duration and maternal glycemia. However, other sleep disturbances such as difficulty falling asleep and staying asleep are common in pregnancy. They may contribute to impaired maternal glycemia through sympathetic nervous system activity, systemic inflammation, and hormonal pathways. However, there is little research examining associations between these specific sleep disturbances and maternal glycemia. Objective: This study aimed to investigate the associations of sleep disturbances during mid-pregnancy and mid-pregnancy maternal glycemia and gestational diabetes subtypes. Study Design: This is a secondary data analysis of the Comparison of Two Screening Strategies for Gestational Diabetes trial. Participants (n = 828) self-reported the frequency of sleep disturbances (i.e., trouble falling asleep, trouble staying asleep, waking several times per night, and waking feeling tired or worn out) in mid-pregnancy. Gestational diabetes was diagnosed using either the International Associations of Diabetes and Pregnancy Study Groups or Carpenter-Coustan approach. We defined gestational diabetes subtypes based on the degree of insulin resistance and beta-cell dysfunction. We used multinomial logistic regression to examine associations of sleep disturbances with gestational diabetes status (i.e., normal, mild glycemic dysfunction, and gestational diabetes) and gestational diabetes subtypes (i.e., neither insulin resistance or beta-cell dysfunction, insulin resistance only, beta-cell dysfunction only, and insulin resistance and beta-cell dysfunction). Results: A total of 665 participants (80%) had normal glycemia, 81 (10%) mild hyperglycemia, and 80 (10%) had gestational diabetes. Among participants with gestational diabetes, 62 (78%) had both insulin resistance and beta-cell dysfunction, 15 (19 %) had insulin resistance only, and 3 had beta-cell dysfunction only or neither insulin resistance nor beta-cell dysfunction. Sleep disturbance frequency was not associated with maternal glycemia or gestational diabetes subtypes. Conclusions: Sleep disturbances in mid-pregnancy were not associated with maternal glycemia during mid-pregnancy. Future research should collect data on sleep disturbances at multiple time points in pregnancy and in combination with other sleep disturbances to determine whether sleep plays any role in maternal glycemic control.

Newer Glucose-Lowering Therapies in Older Adults with Type 2 Diabetes.

Diabetes is prevalent in older adults and older adults with diabetes are more likely to have multiple comorbidities. It is, therefore, important to personalize diabetes management in this group. Newer glucose-lowering drugs, including dipeptidyl peptidase-4 inhibitors, sodium-glucose cotransporter 2 inhibitors, and glucagon-like peptide-1 receptor agonists can be safely used in older patients and are preferred choices in many cases due to their safety, efficacy, and low risk of hypoglycemia.

Diabetes reversal: Update on current knowledge and proposal of prediction score parameters for diabetes remission.

BACKGROUND AND AIM: The field of diabetes reversal is continuously evolving. Strategies for implementing diabetes care towards diabetes reversal are still being worked out. We aim to analyse data from available literature to ascertain factors allowing patient centric dietary approach to achieve diabetes reversal in clinical practice. METHODS: In this exploratory review, an update on current knowledge is presented to delineate factors driving diabetes remission in an individual based on major studies in the field. This knowledge is then applied to subtypes of type 2 diabetes to optimise dietary approach for reversal of diabetes. RESULTS AND CONCLUSION: Shorter duration of diabetes, lesser number of medicines needed to achieve euglycemia and 15 kg weight loss are common factors favouring diabetes remission in all major studies. A patient centric approach to diabetes reversal taking into account the recently described diabetes subtypes is being proposed to improve the proportion of patients achieving remission. We also propose the parameters of a novel diabetes remission prediction score, based on patient motivation, interaction with the care-team, level of diabetes self-care and the intent of the care-team.

Utility of oral glucose tolerance test in predicting type 2 diabetes following gestational diabetes: Towards personalized care.

BACKGROUND: Women with gestational diabetes mellitus (GDM) are at a seven-fold higher risk of developing type 2 diabetes (T2D) within 7-10 years after childbirth, compared with those with normoglycemic pregnancy. Although raised fasting blood glucose (FBG) levels has been said to be the main significant predictor of postpartum progression to T2D, it is difficult to predict who among the women with GDM would develop T2D. Therefore, we conducted a cross-sectional retrospective study to examine the glycemic indices that can predict postnatal T2D in Emirati Arab women with a history of GDM. AIM: To assess how oral glucose tolerance test (OGTT) can identify the distinct GDM pathophysiology and predict possible distinct postnatal T2D subtypes. METHODS: The glycemic status of a cohort of 4603 pregnant Emirati Arab women, who delivered in 2007 at both Latifa Women and Children Hospital and at Dubai Hospital, United Arab Emirates, was assessed retrospectively, using the International Association of Diabetes and Pregnancy Study Groups (IADPSG) criteria. Of the total, 1231 women were followed up and assessed in 2016. The FBG and/or the 2-h blood glucose (2hrBG) levels after a 75-g glucose load were measured to assess the prevalence of GDM and T2D, according to the IADPSG and American Diabetes Association (ADA) criteria, respectively. The receiver operating characteristic curve for the OGTT was plotted and sensitivity, specificity, and predictive values of FBG and 2hrBG for T2D were determined. RESULTS: Considering both FBG and 2hrBG levels, according to the IADPSG criteria, the prevalence of GDM in pregnant Emirati women in 2007 was 1057/4603 (23%), while the prevalence of pre-pregnancy T2D among them, based on ADA criteria, was 230/4603 (5%). In the subset of women (n = 1231) followed up in 2016, the prevalence of GDM in 2007 was 362/1231 (29.6%), while the prevalence of pre-pregnancy T2D was 36/1231 (2.9%). Of the 362 pregnant women with GDM in 2007, 96/362 (26.5%) developed T2D; 142/362 (39.2%) developed impaired fasting glucose; 29/362 (8.0%) developed impaired glucose tolerance, and the remaining 95/362 (26.2%) had normal glycemia in 2016. The prevalence of T2D, based on ADA criteria, stemmed from the prevalence of 36/1231 (2.9%) in 2007 to 141/1231 (11.5%), in 2016. The positive predictive value (PPV) for FBG suggests that if a woman tested positive for GDM in 2007, the probability of developing T2D in 2016 was approximately 24%. The opposite was observed when 2hrBG was used for diagnosis. The PPV value for 2hrBG suggests that if a woman was positive for GDM in 2007 then the probability of developing T2D in 2016 was only 3%. CONCLUSION: FBG and 2hrBG could predict postpartum T2D, following antenatal GDM. However, each test reflects different pathophysiology and possible T2D subtype and could be matched with a relevant T2D prevention program.

Linking glycemic dysregulation in diabetes to symptoms, comorbidities, and genetics through EHR data mining.

Diabetes is a diverse and complex disease, with considerable variation in phenotypic manifestation and severity. This variation hampers the study of etiological differences and reduces the statistical power of analyses of associations to genetics, treatment outcomes, and complications. We address these issues through deep, fine-grained phenotypic stratification of a diabetes cohort. Text mining the electronic health records of 14,017 patients, we matched two controlled vocabularies (ICD-10 and a custom vocabulary developed at the clinical center Steno Diabetes Center Copenhagen) to clinical narratives spanning a 19 year period. The two matched vocabularies comprise over 20,000 medical terms describing symptoms, other diagnoses, and lifestyle factors. The cohort is genetically homogeneous (Caucasian diabetes patients from Denmark) so the resulting stratification is not driven by ethnic differences, but rather by inherently dissimilar progression patterns and lifestyle related risk factors. Using unsupervised Markov clustering, we defined 71 clusters of at least 50 individuals within the diabetes spectrum. The clusters display both distinct and shared longitudinal glycemic dysregulation patterns, temporal co-occurrences of comorbidities, and associations to single nucleotide polymorphisms in or near genes relevant for diabetes comorbidities.