Characteristics Associated With Decreased or Increased Mortality Risk From Glycemic Therapy Among Patients With Type 2 Diabetes and High Cardiovascular Risk: Machine Learning Analysis of the ACCORD Trial

The Role of Inflammation and Psycho-Behavioral Factors as Mediators Between Acculturation and Time to Incident Dementia Among Older Mexican Americans

Background

The mechanisms through which acculturation influences the onset of cognitive impairment and dementia are not well understood, especially among older Hispanics.

Objective

To investigate whether inflammation and psycho-behavioral factors mediate the relationship between acculturation and incident dementia among older Mexican Americans.

Methods

We analyzed the Sacramento Area Latino Study on Aging (1998-2007, SALSA), a longitudinal study (N = 1,194) with 10 years of follow-up, and used g-computation for mediation analysis with pooled logistic regression to evaluate whether acculturation (assessed by the Revised Acculturation Rating Scale for Mexican Americans [ARSMA-II]) affected dementia or cognitive impairment but not dementia (CIND) through inflammation (i.e., interleukin 6 [IL-6], tumor necrosis factor-α (TNF-α), high-sensitivity C-reactive protein [hs-CRP]), smoking, alcohol consumption, and depressive symptoms. The potential mediators were assessed at baseline.

Results

The 10-year average adjusted risk ratio (aRR) for the effect of high U.S. acculturation and dementia/CIND was 0.66, 95% CI (0.36, 1.30). The indirect effects were: IL-6 (aRR = 0.98, 95% CI (0.88, 1.05)); TNF-α (aRR:0.99, 95% CI (0.93, 1.05)); hs-CRP: (aRR = 1.21, 95% CI (0.84, 1.95)); current smoking: aRR = 0.97, 95% CI (0.84, 1.16); daily/weekly alcohol consumption (aRR = 1.00, 95% CI (0.96, 1.05)); and depressive symptom score (aRR = 1.03, 95% CI (0.95, 1.26)). hs-CRP yielded a proportion mediated of -26%, suggesting that hs-CRP could suppress the potential effect of high U.S. acculturation. The other factors explored resulted in little to no mediation.

Conclusions

The effect of acculturation on time to incident dementia/CIND varied over time. Our study suggests that inflammation could suppress the effect between high U.S. acculturation and dementia risk.

Heterogeneous Treatment Effects of Intensive Glycemic Control on Kidney Microvascular Outcomes and Mortality in ACCORD

SIGNIFICANCE STATEMENT

Identifying and quantifying treatment effect variation across patients is the fundamental challenge of precision medicine. Here we quantify heterogeneous treatment effects of intensive glycemic control in the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial, considering three outcomes of interest-a composite kidney outcome (driven by macroalbuminuria), all-cause mortality, and first assisted hypoglycemic event. We demonstrate that the effects of intensive glycemic control vary with risk of kidney failure, as predicted by the kidney failure risk equation (KFRE). Participants at highest risk of kidney failure gain the largest absolute kidney benefit of intensive glycemic control but also experience the largest absolute risk of death and hypoglycemic events. Our findings illustrate the value of identifying clinically meaningful treatment heterogeneity, particularly when treatments have different effects on multiple end points.

OBJECTIVE

Clear criteria to individualize glycemic targets in patients with type II diabetes are lacking. In this post hoc analysis of the ACCORD, we evaluate whether the KFRE can identify patients for whom intensive glycemic control confers more benefit in preventing kidney microvascular outcomes.

RESEARCH DESIGN AND METHODS

We divided the ACCORD trial population into quartiles on the basis of 5-year kidney failure risk using the KFRE. We estimated conditional treatment effects within each quartile and compared them with the average treatment effect in the trial. The treatment effects of interest were the 7-year restricted mean survival time (RMST) differences between intensive and standard glycemic control arms on ( 1 ) time-to-first development of severely elevated albuminuria or kidney failure and ( 2 ) all-cause mortality.

RESULTS

We found evidence that the effect of intensive glycemic control on kidney microvascular outcomes and all-cause mortality varies with baseline risk of kidney failure. Patients with elevated baseline risk of kidney failure derived the most from intensive glycemic control in reducing kidney microvascular outcomes (7-year RMST difference of 114.8 [95% confidence interval 58.1 to 176.4] versus 48.4 [25.3 to 69.6] days in the entire trial population) However, this same patient group also experienced a shorter time to death (7-year RMST difference of -56.7 [-100.2 to -17.5] v. -23.6 [-42.2 to -6.6] days).

CONCLUSIONS

We found evidence of heterogenous treatment effects of intensive glycemic control on kidney microvascular outcomes in ACCORD as a function of predicted baseline risk of kidney failure. Patients with higher kidney failure risk experienced the most pronounced reduction in kidney microvascular outcomes but also experienced the highest risk of all-cause mortality.

Heterogeneous treatment effects of metformin on risk of dementia in patients with type 2 diabetes: A longitudinal observational study

INTRODUCTION

Little is known about the heterogeneous treatment effects of metformin on dementia risk in people with type 2 diabetes (T2D).

METHODS

Participants (≥ 50 years) with T2D and normal cognition at baseline were identified from the National Alzheimer's Coordinating Center database (2005-2021). We applied a doubly robust learning approach to estimate risk differences (RD) with a 95% confidence interval (CI) for dementia risk between metformin use and no use in the overall population and subgroups identified through a decision tree model.

RESULTS

Among 1393 participants, 104 developed dementia over a 4-year median follow-up. Metformin was significantly associated with a lower risk of dementia in the overall population (RD, -3.2%; 95% CI, -6.2% to -0.2%). We identified four subgroups with varied risks for dementia, defined by neuropsychiatric disorders, non-steroidal anti-inflammatory drugs, and antidepressant use.

DISCUSSION

Metformin use was significantly associated with a lower risk of dementia in individuals with T2D, with significant variability among subgroups.

Personalized glucose-lowering effect of chiglitazar in type 2 diabetes

Chiglitazar (carfloglitazar) is a peroxisome proliferator-activated receptor pan-agonist presenting non-inferior glucose-lowering efficacy with sitagliptin in patients with type 2 diabetes. To delineate the subgroup of patients with greater benefit from chiglitazar, we conducted a machine learning-based post-hoc analysis in two randomized controlled trials. We established a character phenomap based on 13 variables and estimated HbA1c decline to the effects of chiglitazar in reference to sitagliptin. Out of 1,069 patients, 63.3% were found to have greater reduction in HbA1c levels with chiglitazar, while 36.7% showed greater reduction with sitagliptin. This distinction in treatment response was statistically significant between groups (pinteraction<0.001). To identify patients who would gain the most glycemic control benefit from chiglitazar, we developed a machine learning model, ML-PANPPAR, which demonstrated robust performance using sex, BMI, HbA1c, HDL, and fasting insulin. The phenomapping-derived tool successfully identified chiglitazar responders and enabled personalized drug allocation in patients with drug-naïve diabetes.

Precision subclassification of type 2 diabetes: a systematic review

BACKGROUND

Heterogeneity in type 2 diabetes presentation and progression suggests that precision medicine interventions could improve clinical outcomes. We undertook a systematic review to determine whether strategies to subclassify type 2 diabetes were associated with high quality evidence, reproducible results and improved outcomes for patients.

METHODS

We searched PubMed and Embase for publications that used 'simple subclassification' approaches using simple categorisation of clinical characteristics, or 'complex subclassification' approaches which used machine learning or 'omics approaches in people with established type 2 diabetes. We excluded other diabetes subtypes and those predicting incident type 2 diabetes. We assessed quality, reproducibility and clinical relevance of extracted full-text articles and qualitatively synthesised a summary of subclassification approaches.

RESULTS

Here we show data from 51 studies that demonstrate many simple stratification approaches, but none have been replicated and many are not associated with meaningful clinical outcomes. Complex stratification was reviewed in 62 studies and produced reproducible subtypes of type 2 diabetes that are associated with outcomes. Both approaches require a higher grade of evidence but support the premise that type 2 diabetes can be subclassified into clinically meaningful subtypes.

CONCLUSION

Critical next steps toward clinical implementation are to test whether subtypes exist in more diverse ancestries and whether tailoring interventions to subtypes will improve outcomes.

Comparison of causal forest and regression-based approaches to evaluate treatment effect heterogeneity: an application for type 2 diabetes precision medicine

OBJECTIVE

Precision medicine requires reliable identification of variation in patient-level outcomes with different available treatments, often termed treatment effect heterogeneity. We aimed to evaluate the comparative utility of individualized treatment selection strategies based on predicted individual-level treatment effects from a causal forest machine learning algorithm and a penalized regression model.

METHODS

Cohort study characterizing individual-level glucose-lowering response (6 month reduction in HbA1c) in people with type 2 diabetes initiating SGLT2-inhibitor or DPP4-inhibitor therapy. Model development set comprised 1,428 participants in the CANTATA-D and CANTATA-D2 randomised clinical trials of SGLT2-inhibitors versus DPP4-inhibitors. For external validation, calibration of observed versus predicted differences in HbA1c in patient strata defined by size of predicted HbA1c benefit was evaluated in 18,741 patients in UK primary care (Clinical Practice Research Datalink).

RESULTS

Heterogeneity in treatment effects was detected in clinical trial participants with both approaches (proportion predicted to have a benefit on SGLT2-inhibitor therapy over DPP4-inhibitor therapy: causal forest: 98.6%; penalized regression: 81.7%). In validation, calibration was good with penalized regression but sub-optimal with causal forest. A strata with an HbA1c benefit > 10 mmol/mol with SGLT2-inhibitors (3.7% of patients, observed benefit 11.0 mmol/mol [95%CI 8.0-14.0]) was identified using penalized regression but not causal forest, and a much larger strata with an HbA1c benefit 5-10 mmol with SGLT2-inhibitors was identified with penalized regression (regression: 20.9% of patients, observed benefit 7.8 mmol/mol (95%CI 6.7-8.9); causal forest 11.6%, observed benefit 8.7 mmol/mol (95%CI 7.4-10.1).

CONCLUSIONS

Consistent with recent results for outcome prediction with clinical data, when evaluating treatment effect heterogeneity researchers should not rely on causal forest or other similar machine learning algorithms alone, and must compare outputs with standard regression, which in this evaluation was superior.

Heterogeneous treatment effects of intensive glycemic control on kidney microvascular outcomes in ACCORD

Objective

Clear criteria to individualize glycemic targets are lacking. In this post-hoc analysis of the Action to Control Cardiovascular Risk in Diabetes trial (ACCORD), we evaluate whether the kidney failure risk equation (KFRE) can identify patients who disproportionately benefit from intensive glycemic control on kidney microvascular outcomes.

Research design and methods

We divided the ACCORD trial population in quartiles based on 5-year kidney failure risk using the KFRE. We estimated conditional treatment effects within each quartile and compared them to the average treatment effect in the trial. The treatment effects of interest were the 7-year restricted-mean-survival-time (RMST) differences between intensive and standard glycemic control arms on (1) time-to-first development of severely elevated albuminuria or kidney failure and (2) all-cause mortality.

Results

We found evidence that the effect of intensive glycemic control on kidney microvascular outcomes and all-cause mortality varies with baseline risk of kidney failure. Patients with elevated baseline risk of kidney failure benefitted the most from intensive glycemic control on kidney microvascular outcomes (7-year RMST difference of 115 v. 48 days in the entire trial population) However, this same patient group also experienced shorter times to death (7-year RMST difference of -57 v. -24 days).

Conclusions

We found evidence of heterogenous treatment effects of intensive glycemic control on kidney microvascular outcomes in ACCORD as a function of predicted baseline risk of kidney failure. Patients with higher kidney failure risk experienced the most pronounced benefits of treatment on kidney microvascular outcomes but also experienced the highest risk of all-cause mortality.

Machine learning improves mortality prediction in three-vessel disease

BACKGROUND AND AIMS

Risk stratification for three-vessel coronary artery disease (3VD) remains an important clinical challenge. In this study, we utilized machine learning (ML), which can address the limitations of traditional regression-based models, to develop a novel model to assess mortality risk in patients with 3VD.

METHODS

This study was based on a prospective cohort of 8943 participants with 3VD consecutively enrolled between 2004 and 2011. An ML-derived random forest model was trained and tested to predict 4-year mortality. The predictability of the model was compared with that of an established model, the Synergy Between Percutaneous Coronary Intervention With Taxus and Cardiac Surgery score II (SSII), among 3VD patients undergoing percutaneous coronary intervention (PCI), coronary artery bypass grafting (CABG), and medical therapy (MT) alone.

RESULTS

The all-cause mortality was 7.5% (667 patients) over the 4-year follow-up period. The correlation-based feature selection algorithm selected 18 of the 94 features to develop the ML model. In the testing dataset, the ML-derived model achieved an area under the curve of 0.81 for 4-year mortality prediction. Its predictability was significantly better than that of the SSII among patients undergoing PCI (0.80 vs. 0.70, p < 0.001) or CABG (0.80 vs. 0.67, p < 0.001). The model also outperformed the SSII in patients receiving MT alone (ML: 0.75 vs. SSII for PCI: 0.70 or SSII for CABG: 0.66, p < 0.001).

CONCLUSIONS

This ML-based approach exhibited better performance in risk stratification for 3VD compared with the conventional method. Further validation studies are needed to confirm these findings.

Review: Machine learning in precision pharmacotherapy of type 2 diabetes-A promising future or a glimpse of hope?

Precision pharmacotherapy of diabetes requires judicious selection of the optimal therapeutic agent for individual patients. Artificial intelligence (AI), a swiftly expanding discipline, holds substantial potential to transform current practices in diabetes diagnosis and management. This manuscript provides a comprehensive review of contemporary research investigating drug responses in patient subgroups, stratified via either supervised or unsupervised machine learning approaches. The prevalent algorithmic workflow for investigating drug responses using machine learning involves cohort selection, data processing, predictor selection, development and validation of machine learning methods, subgroup allocation, and subsequent analysis of drug response. Despite the promising feature, current research does not yet provide sufficient evidence to implement machine learning algorithms into routine clinical practice, due to a lack of simplicity, validation, or demonstrated efficacy. Nevertheless, we anticipate that the evolving evidence base will increasingly substantiate the role of machine learning in molding precision pharmacotherapy for diabetes.

Cardiovascular/Stroke Risk Stratification in Diabetic Foot Infection Patients Using Deep Learning-Based Artificial Intelligence: An Investigative Study

A diabetic foot infection (DFI) is among the most serious, incurable, and costly to treat conditions. The presence of a DFI renders machine learning (ML) systems extremely nonlinear, posing difficulties in CVD/stroke risk stratification. In addition, there is a limited number of well-explained ML paradigms due to comorbidity, sample size limits, and weak scientific and clinical validation methodologies. Deep neural networks (DNN) are potent machines for learning that generalize nonlinear situations. The objective of this article is to propose a novel investigation of deep learning (DL) solutions for predicting CVD/stroke risk in DFI patients. The Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) search strategy was used for the selection of 207 studies. We hypothesize that a DFI is responsible for increased morbidity and mortality due to the worsening of atherosclerotic disease and affecting coronary artery disease (CAD). Since surrogate biomarkers for CAD, such as carotid artery disease, can be used for monitoring CVD, we can thus use a DL-based model, namely, Long Short-Term Memory (LSTM) and Recurrent Neural Networks (RNN) for CVD/stroke risk prediction in DFI patients, which combines covariates such as office and laboratory-based biomarkers, carotid ultrasound image phenotype (CUSIP) lesions, along with the DFI severity. We confirmed the viability of CVD/stroke risk stratification in the DFI patients. Strong designs were found in the research of the DL architectures for CVD/stroke risk stratification. Finally, we analyzed the AI bias and proposed strategies for the early diagnosis of CVD/stroke in DFI patients. Since DFI patients have an aggressive atherosclerotic disease, leading to prominent CVD/stroke risk, we, therefore, conclude that the DL paradigm is very effective for predicting the risk of CVD/stroke in DFI patients.

The effectiveness of daily humanistic care in pharmaceutical care of patients with type 2 diabetes

Although the current methods of pharmaceutical care have demonstrably improved pathogenic condition and humanistic health outcomes of patients with type 2 diabetes, lacking humanistic care limits its reach and effectiveness. The purpose of this study was to evaluate the effectiveness of daily humanistic care reminders in pharmaceutical care of patients with type 2 diabetes. A total of 1200 patients with type 2 diabetes were recruited. The control group received only pharmaceutical care, whereas the intervention group received extra humanistic care from clinical pharmacists. Biochemical data, levels of quality of life, satisfaction, and self-care behavior and relationships between humanistic care and glycemic control were collected before and after 12-month intervention. Humanistic care increased levels of quality of life, satisfaction, self-efficacy and self-care behavior score, decreased blood pressure, glycosylated hemoglobin A1c (HbA1c), blood lipid, fasting blood glucose, postprandial 2-h blood glucose (PPG), inflammation, and insulin resistance after 12-month follow up. Humanistic care showed a positive correlation with glycemic control and negative relationship with HbA1c in patients with type 2 diabetes. There was a positive correlation between humanistic care and glycemic control (R = 0.641, P < .01) and a negative relationship between humanistic care and HbA1c in patients with type 2 diabetes (r = -0.482, P < .01). In conclusion, humanistic care could improve the control of overall clinical outcomes and enhanced medication adherences, which play an important role in the management of patients with type 2 diabetes.

A machine learning approach identifies modulators of heart failure hospitalization prevention among patients with type 2 diabetes: A revisit to the ACCORD trial

BACKGROUND

To examine patient characteristics that may modulate the heterogeneous treatment effect of intensive systolic blood pressure control (SBP) and intensive glycemic control on incident heart failure (HF) risk in people with type 2 diabetes.

METHODS

We analyzed 10,251 participants from the ACCORD glucose trial, and 4733 from the SBP sub-trial separately. We applied a robust machine-learning (ML) algorithm, namely the causal forest/causal tree analysis, to each trial to identify participants' characteristics that modulate the effectiveness of each trial intervention.

RESULTS

Diastolic blood pressure (DBP) was found to interact with intensive glycemic control and impact outcomes. An increased HF risk associated with intensive glycemic control (absolute risk change (ARC): 2.28 %, 95 % confidence interval (CI): 0.69 % to 3.90 %; relative risk (RR):1.57, 95 % CI: 1.15 to 2.20; P < 0.05) was observed in individuals with baseline DBP at the lowest tertile (45-69 mmHg), while no changes in HF risk associated with intensive glycemic control were observed in individuals with baseline DBP at the middle (70-79 mmHg) and the highest tertiles (80-100 mmHg). Liver function was identified as a modulator of intensive BP control, and baseline Alanine transaminase (ALT) level was a sensitive marker for the modulating effect. Only individuals with baseline ALT at the lowest tertile (8-19 mg/dl) benefited from the intensive BP control for HF prevention (ARC: -1.95 %, 95 % CI: -4.06 % to 0.11 %; RR:0.62. 95 % CI: 0.27 to 0.94; P < 0.05).

CONCLUSIONS

Our study is the first to observe and quantify the potential synergistic harmful effect when low DBP was combined with an intensive blood glucose intervention. Recognizing these may help clinicians develop a more precise approach to such treatments, thus increasing the efficiency and outcomes of diabetes treatments.

Potential misclassification of diabetes and prediabetes in the U.S.: Mismatched HbA1c and glucose in NHANES 2005-2016

AIMS

To assess the prevalence and clinical implications of "mismatches" between HbA1c and glucose levels in the United States across the life course.

METHODS

Participants ages 12-79 years from U.S. National Health and Nutrition Examination Survey (NHANES) 2005-2016 without known diagnosis of diabetes and who had a 75 g oral glucose tolerance test were included. Previously undiagnosed diabetes (DM), prediabetes, and normal glucose metabolism (NGM) were defined using American Diabetes Association cut-points. Mismatches were defined by the hemoglobin glycation index (HGI).

RESULTS

In 10,361 participants, 5% and 41% had diabetes and prediabetes, respectively, by fasting or 2-hour glucose criteria. By HbA1c criteria, the high HGI tertile consisted of mostly abnormal classification (3% DM, 52% prediabetes) and the low HGI tertile contained mostly normal classification (78% NGM). Across all ages, 15% (weighted: 30 million individuals) had clinically significant mismatches of HGI magnitude ≥+0.5% (i.e., high mismatch) or ≤-0.5% (low mismatch). Mismatch was most common in older adults and non-Hispanic Black participants.

CONCLUSIONS

Mismatches of clinically significant magnitude could lead to HbA1c-related misdiagnosis or inappropriate management in up to 30 million Americans. Older adults, non-Hispanic Black individuals, and others with high mismatches may benefit from complementing HbA1c with additional diagnostic and management strategies.

Using machine learning to aid treatment decision and risk assessment for severe three-vessel coronary artery disease

BACKGROUND

Three-vessel disease (TVD) with a SYNergy between PCI with TAXus and cardiac surgery (SYNTAX) score of ≥ 23 is one of the most severe types of coronary artery disease. We aimed to take advantage of machine learning to help in decision-making and prognostic evaluation in such patients.

METHODS

We analyzed 3786 patients who had TVD with a SYNTAX score of ≥ 23, had no history of previous revascularization, and underwent either coronary artery bypass grafting (CABG) or percutaneous coronary intervention (PCI) after enrollment. The patients were randomly assigned to a training group and testing group. The C4.5 decision tree algorithm was applied in the training group, and all-cause death after a median follow-up of 6.6 years was regarded as the class label.

RESULTS

The decision tree algorithm selected age and left ventricular end-diastolic diameter (LVEDD) as splitting features and divided the patients into three subgroups: subgroup 1 (age of ≤ 67 years and LVEDD of ≤ 53 mm), subgroup 2 (age of ≤ 67 years and LVEDD of > 53 mm), and subgroup 3 (age of > 67 years). PCI conferred a patient survival benefit over CABG in subgroup 2. There was no significant difference in the risk of all-cause death between PCI and CABG in subgroup 1 and subgroup 3 in both the training data and testing data. Among the total study population, the multivariable analysis revealed significant differences in the risk of all-cause death among patients in three subgroups.

CONCLUSIONS

The combination of age and LVEDD identified by machine learning can contribute to decision-making and risk assessment of death in patients with severe TVD. The present results suggest that PCI is a better choice for young patients with severe TVD characterized by left ventricular dilation.

Heterogeneous treatment effects of intensive glycemic control on major adverse cardiovascular events in the ACCORD and VADT trials: a machine-learning analysis

BACKGROUND

Evidence to guide type 2 diabetes treatment individualization is limited. We evaluated heterogeneous treatment effects (HTE) of intensive glycemic control in type 2 diabetes patients on major adverse cardiovascular events (MACE) in the Action to Control Cardiovascular Risk in Diabetes Study (ACCORD) and the Veterans Affairs Diabetes Trial (VADT).

METHODS

Causal forests machine learning analysis was performed using pooled individual data from two randomized trials (n = 12,042) to identify HTE of intensive versus standard glycemic control on MACE in patients with type 2 diabetes. We used variable prioritization from causal forests to build a summary decision tree and examined the risk difference of MACE between treatment arms in the resulting subgroups.

RESULTS

A summary decision tree used five variables (hemoglobin glycation index, estimated glomerular filtration rate, fasting glucose, age, and body mass index) to define eight subgroups in which risk differences of MACE ranged from - 5.1% (95% CI - 8.7, - 1.5) to 3.1% (95% CI 0.2, 6.0) (negative values represent lower MACE associated with intensive glycemic control). Intensive glycemic control was associated with lower MACE in pooled study data in subgroups with low (- 4.2% [95% CI - 8.1, - 1.0]), intermediate (- 5.1% [95% CI - 8.7, - 1.5]), and high (- 4.3% [95% CI - 7.7, - 1.0]) MACE rates with consistent directions of effect in ACCORD and VADT alone.

CONCLUSIONS

This data-driven analysis provides evidence supporting the diabetes treatment guideline recommendation of intensive glucose lowering in diabetes patients with low cardiovascular risk and additionally suggests potential benefits of intensive glycemic control in some individuals at higher cardiovascular risk.

High Hemoglobin Glycation Index Is Associated With Telomere Attrition Independent of HbA1c, Mediated by TNFα

CONTEXT

The hemoglobin glycation index (HGI) is correlated with metabolic diseases and inflammation. Whether the HGI is associated with the aging process and how inflammation and oxidative stress affect the relationship remain unclear.

OBJECTIVE

We aimed to analyze links between the HGI and aging biomarkers, and to explore a potential role of inflammation and oxidative stress in the correlations.

METHODS

A cross-sectional study of 434 subjects with different glucose intolerances in a rural community was enrolled. The HGI was calculated as the difference between the measured and predicted hemoglobin A1c (HbA1c). The population was categorized into tertiles of the HGI. Telomere length (LTL) and mitochondrial DNA copy number (mtDNAcn) determined by polymerase chain reaction assay. Tumor necrosis factor (TNF) α and interleukin (IL) 6, 8-oxo-2'-deoxyguanosine (8-oxo-dG), superoxide dismutase (SOD) activities, and glutathione reductase (GR) were measured.

RESULTS

Participants in the high HGI group were older and reported a shorter LTL, higher levels of TNFα, SOD activities, and HbA1c. Correlation analyses demonstrated that HGI was correlated with LTL (r = -0.25, P < .001) and TNFα (r = 0.19, P < .001) regardless of HbA1c levels. No relationship was found between HGI and mtDNAcn. HGI (β = -0.238, 95% CI -0.430, -0.046, P = .015) and TNFα (β = -0.02, 95% CI -0.030, -0.014, P < .001) were proved to be correlated with LTL independently, using multiple linear regression analysis. Ordinal logistic regression models showed that compared with subjects the high HGI group, the possibilities of a higher-level LTL was 5.29-fold in the low HGI group (OR 5.29, 95% CI (2.45, 11.41), P < .001), 2.41-fold in the moderate HGI group (OR 2.41, 95% CI 1.35, 4.30, P = .003) after controlling for confounding variables. Mediation analyses indicated that TNFα accounted for 30.39% of the effects of the HGI on LTL.

CONCLUSION

HGI was negatively related to telomere attrition, independent of HbA1c. TNFα acted as a mediator of the relationship between HGI and LTL.

Artificial intelligence perspective in the future of endocrine diseases

In recent years, artificial intelligence (AI) shows promising results in the diagnosis, prediction, and management of diseases. The move from handwritten medical notes to electronic health records and a huge number of digital data commenced in the era of big data in medicine. AI can improve physician performance and help better clinical decision making which is called augmented intelligence. The methods applied in the research of AI and endocrinology include machine learning, artificial neural networks, and natural language processing. Current research in AI technology is making major efforts to improve decision support systems for patient use. One of the best-known applications of AI in endocrinology was seen in diabetes management, which includes prediction, diagnosis of diabetes complications (measuring microalbuminuria, retinopathy), and glycemic control. AI-related technologies are being found to assist in the diagnosis of other endocrine diseases such as thyroid cancer and osteoporosis. This review attempts to provide insight for the development of prospective for AI with a focus on endocrinology.

Generalizability of heterogeneous treatment effects based on causal forests applied to two randomized clinical trials of intensive glycemic control

Purpose Machine learning is an attractive tool for identifying heterogeneous treatment effects (HTE) of interventions but generalizability of machine learning derived HTE remains unclear. We examined generalizability of HTE detected using causal forests in two similarly designed randomized trials in type II diabetes patients. Methods We evaluated published HTE of intensive versus standard glycemic control on all-cause mortality from the Action to Control Cardiovascular Risk in Diabetes study (ACCORD) in a second trial, the Veterans Affairs Diabetes Trial (VADT). We then applied causal forests to VADT, ACCORD, and pooled data from both studies and compared variable importance and subgroup effects across samples. Results HTE in ACCORD did not replicate in similar subgroups in VADT, but variable importance was correlated between VADT and ACCORD (Kendall's tau-b 0.75). Applying causal forests to pooled individual-level data yielded seven subgroups with similar HTE across both studies, ranging from risk difference of all-cause mortality of -3.9% (95% CI -7.0, -0.8) to 4.7% (95% CI 1.8, 7.5). Conclusions Machine learning detection of HTE subgroups from randomized trials may not generalize across study samples even when variable importance is correlated. Pooling individual-level data may overcome differences in study populations and/or differences in interventions that limit HTE generalizability.

Association Between Long-Term Visit-to-Visit Hemoglobin A1c and Cardiovascular Risk in Type 2 Diabetes: The ACCORD Trial

Background: To explore the association between visit-to-visit variability of glycated hemoglobin (HbA1c) and cardiovascular outcomes in the patients with type 2 diabetes mellitus (T2DM) of the Action to Control Cardiovascular Risk in Diabetes (ACCORD) study.

Methods: We conducted a post-hoc analysis on the ACCORD population including 9,544 participants with T2DM. Visit-to-visit variability of HbA1c was defined as the individual SD, coefficient of variation (CV), and variability independent of the mean (VIM) across HbA1c measurements. The clinical measurements included primary outcome [the first occurrence of non-fatal myocardial infarction (MI), non-fatal stroke or cardiovascular death], total mortality, cardiovascular death, non-fatal MI event, non-fatal stroke, total stroke, heart failure, macrovascular events, and major coronary events (CHD).

Results: Over a median follow-up of 4.85 years, 594 and 268 participants experienced all-cause mortality and cardiovascular mortality, respectively. After adjusting for baseline HbA1c levels and confounding factors, the adjusted hazard ratio (HR) comparing patients in the highest vs. the lowest quartile CV of HbA1c variability was 1.61 (95% CI 1.29–2.00) for the primary outcome. Similar trends for secondary outcome were also observed. There was no association between HbA1c fluctuation and non-fatal stroke. Noticeably, there was 66% greater risk for the all-cause mortality among patients in the highest vs. the lowest quartile (HR 1.66, 95% CI 1.27–2.17).

Conclusions: Greater variability of HbA1c is associated with higher risk for cardiovascular complications and all-cause death in T2DM. Our study stresses the significance of well-controlled glycemic levels for improving cardiovascular outcomes. Further randomized clinical trials are required to confirm these findings.

Hemoglobin glycation index, calculated from a single fasting glucose value, as a prediction tool for severe hypoglycemia and major adverse cardiovascular events in DEVOTE

INTRODUCTION

Hemoglobin glycation index (HGI) is the difference between observed and predicted glycated hemoglobin A1c (HbA1c), derived from mean or fasting plasma glucose (FPG). In this secondary, exploratory analysis of data from DEVOTE, we examined: whether insulin initiation/titration affected the HGI; the relationship between baseline HGI tertile and cardiovascular and hypoglycemia risk; and the relative strengths of HGI and HbA1c in predicting these risks.

RESEARCH DESIGN AND METHODS

In DEVOTE, a randomized, double-blind, cardiovascular outcomes trial, people with type 2 diabetes received once per day insulin degludec or insulin glargine 100 units/mL. The primary outcome was time to first occurrence of a major adverse cardiovascular event (MACE), comprising cardiovascular death, myocardial infarction or stroke; severe hypoglycemia was a secondary outcome. In these analyses, predicted HbA1c was calculated using a linear regression equation based on DEVOTE data (HbA1c=0.01313 FPG (mg/dL) (single value)+6.17514), and the population data were grouped into HGI tertiles based on the calculated HGI values. The distributions of time to first event were compared using Kaplan-Meier curves; HRs and 95% CIs were determined by Cox regression models comparing risk of MACE and severe hypoglycemia between tertiles.

RESULTS

Changes in HGI were observed at 12 months after insulin initiation and stabilized by 24 months for the whole cohort and insulin-naive patients. There were significant differences in MACE risk between baseline HGI tertiles; participants with high HGI were at highest risk (low vs high, HR: 0.73 (0.61 to 0.87)95% CI; moderate vs high, HR: 0.67 (0.56 to 0.81)95% CI; p<0.0001). No significant differences between HGI tertiles were observed in the risk of severe hypoglycemia (p=0.0911). With HbA1c included within the model, HGI no longer significantly predicted MACE.

CONCLUSIONS

High HGI was associated with a higher risk of MACE; this finding is of uncertain significance given the association of HGI with insulin initiation and HbA1c.

TRIAL REGISTRATION NUMBER

NCT01959529.

Effects of the hemoglobin glycation index on hyperglycemia diagnosis: Results from the REACTION study

AIMS

This study aimed to assess the effects of the hemoglobin glycation index (HGI) on hyperglycemia diagnosis and summarize the general characteristics of patients with a high-HGI phenotype.

METHODS

The fasting plasma glucose and glycated hemoglobin (HbA1c) levels of participants (n = 47,648) were used to estimate a linear regression equation and determine the baseline HGI. Overall, 42,317 participants without a history of diabetes were included in the final analysis. The participants were divided into three groups according to the tertiles (low, moderate, and high) of baseline HGI. Proportions and variables were compared among the three HGI groups. A multivariate ordered logistic regression model was used to explore associations between related variables and the high-HGI phenotype.

RESULTS

Regression analysis indicated that the high-HGI phenotype was positively associated with female sex, advanced age, obesity, increased low-density lipoprotein and triglyceride levels, decreased high-density lipoprotein cholesterol, and postprandial glycemic excursion levels (all P < 0.05). The prevalence of hyperglycemia increased from the low- to the high-HGI groups when using HbA1c for diagnosis.

CONCLUSIONS

Individuals with high HGI have similar clinical characteristics. Measuring HbA1c alone for diagnosis could lead to inappropriate diabetes management decisions in people with low or high HGI.

Applications of Machine Learning in Bone and Mineral Research

In this unprecedented era of the overwhelming volume of medical data, machine learning can be a promising tool that may shed light on an individualized approach and a better understanding of the disease in the field of osteoporosis research, similar to that in other research fields. This review aimed to provide an overview of the latest studies using machine learning to address issues, mainly focusing on osteoporosis and fractures. Machine learning models for diagnosing and classifying osteoporosis and detecting fractures from images have shown promising performance. Fracture risk prediction is another promising field of research, and studies are being conducted using various data sources. However, these approaches may be biased due to the nature of the techniques or the quality of the data. Therefore, more studies based on the proposed guidelines are needed to improve the technical feasibility and generalizability of artificial intelligence algorithms.

Development and validation of optimal phenomapping methods to estimate long-term atherosclerotic cardiovascular disease risk in patients with type 2 diabetes

AIMS/HYPOTHESIS

Type 2 diabetes is a heterogeneous disease process with variable trajectories of CVD risk. We aimed to evaluate four phenomapping strategies and their ability to stratify CVD risk in individuals with type 2 diabetes and to identify subgroups who may benefit from specific therapies.

METHODS

Participants with type 2 diabetes and free of baseline CVD in the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial were included in this study (N = 6466). Clustering using Gaussian mixture models, latent class analysis, finite mixture models (FMMs) and principal component analysis was compared. Clustering variables included demographics, medical and social history, laboratory values and diabetes complications. The interaction between the phenogroup and intensive glycaemic, combination lipid and intensive BP therapy for the risk of the primary outcome (composite of fatal myocardial infarction, non-fatal myocardial infarction or unstable angina) was evaluated using adjusted Cox models. The phenomapping strategies were independently assessed in an external validation cohort (Look Action for Health in Diabetes [Look AHEAD] trial: n = 4211; and Bypass Angioplasty Revascularisation Investigation 2 Diabetes [BARI 2D] trial: n = 1495).

RESULTS

Over 9.1 years of follow-up, 789 (12.2%) participants had a primary outcome event. FMM phenomapping with three phenogroups was the best-performing clustering strategy in both the derivation and validation cohorts as determined by Bayesian information criterion, Dunn index and improvement in model discrimination. Phenogroup 1 (n = 663, 10.3%) had the highest burden of comorbidities and diabetes complications, phenogroup 2 (n = 2388, 36.9%) had an intermediate comorbidity burden and lowest diabetes complications, and phenogroup 3 (n = 3415, 52.8%) had the fewest comorbidities and intermediate burden of diabetes complications. Significant interactions were observed between phenogroups and treatment interventions including intensive glycaemic control (p-interaction = 0.042) and combination lipid therapy (p-interaction < 0.001) in the ACCORD, intensive lifestyle intervention (p-interaction = 0.002) in the Look AHEAD and early coronary revascularisation (p-interaction = 0.003) in the BARI 2D trial cohorts for the risk of the primary composite outcome. Favourable reduction in the risk of the primary composite outcome with these interventions was noted in low-risk participants of phenogroup 3 but not in other phenogroups. Compared with phenogroup 3, phenogroup 1 participants were more likely to have severe/symptomatic hypoglycaemic events and medication non-adherence on follow-up in the ACCORD and Look AHEAD trial cohorts.

CONCLUSIONS/INTERPRETATION

Clustering using FMMs was the optimal phenomapping strategy to identify replicable subgroups of patients with type 2 diabetes with distinct clinical characteristics, CVD risk and response to therapies.

A Type 2 Diabetes Subtype Responsive to ACCORD Intensive Glycemia Treatment

OBJECTIVE

Current type 2 diabetes (T2D) management contraindicates intensive glycemia treatment in patients with high cardiovascular disease (CVD) risk and is partially motivated by evidence of harms in the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial. Heterogeneity in response to intensive glycemia treatment has been observed, suggesting potential benefit for some individuals.

RESEARCH DESIGN AND METHODS

ACCORD was a randomized controlled trial that investigated whether intensively treating glycemia in individuals with T2D would reduce CVD outcomes. Using a novel approach to cluster HbA1c trajectories, we identified groups in the intensive glycemia arm with modified CVD risk. Genome-wide analysis and polygenic score (PS) were developed to predict group membership. Mendelian randomization was performed to infer causality.

RESULTS

We identified four clinical groupings in the intensive glycemia arm, and clinical group 4 (C4) displayed fewer CVD (hazard ratio [HR] 0.34; P = 2.01 × 10-3) and microvascular outcomes (HR 0.86; P = 0.015) than those receiving standard treatment. A single-nucleotide polymorphism, rs220721, in MAS1 reached suggestive significance in C4 (P = 4.34 × 10-7). PS predicted C4 with high accuracy (area under the receiver operating characteristic curve 0.98), and this predicted C4 displayed reduced CVD risk with intensive versus standard glycemia treatment (HR 0.53; P = 4.02 × 10-6), but not reduced risk of microvascular outcomes (P < 0.05). Mendelian randomization indicated causality between PS, on-trial HbA1c, and reduction in CVD outcomes (P < 0.05).

CONCLUSIONS

We found evidence of a T2D clinical group in ACCORD that benefited from intensive glycemia treatment, and membership in this group could be predicted using genetic variants. This study generates new hypotheses with implications for precision medicine in T2D and represents an important development in this landmark clinical trial warranting further investigation.

Application of Machine Learning in Translational Medicine: Current Status and Future Opportunities

The exponential increase in our ability to harness multi-dimensional biological and clinical data from experimental to real-world settings has transformed pharmaceutical research and development in recent years, with increasing applications of artificial intelligence (AI) and machine learning (ML). Patient-centered iterative forward and reverse translation is at the heart of precision medicine discovery and development across the continuum from target validation to optimization of pharmacotherapy. Integration of advanced analytics into the practice of Translational Medicine is now a fundamental enabler to fully exploit information contained in diverse sources of big data sets such as “omics” data, as illustrated by deep characterizations of the genome, transcriptome, proteome, metabolome, microbiome, and exposome. In this commentary, we provide an overview of ML applications in drug discovery and development, aligned with the three strategic pillars of Translational Medicine (target, patient, dose) and offer perspectives on their potential to transform the science and practice of the discipline. Opportunities for integrating ML approaches into the discipline of Pharmacometrics are discussed and will revolutionize the practice of model-informed drug discovery and development. Finally, we posit that joint efforts of Clinical Pharmacology, Bioinformatics, and Biomarker Technology experts are vital in cross-functional team settings to realize the promise of AI/ML-enabled Translational and Precision Medicine.

Potential value and impact of data mining and machine learning in clinical diagnostics

Data mining involves the use of mathematical sciences, statistics, artificial intelligence, and machine learning to determine the relationships between variables from a large sample of data. It has previously been shown that data mining can improve the prediction and diagnostic precision of type 2 diabetes mellitus. A few studies have applied machine learning to assess hypertension and metabolic syndrome-related biomarkers, as well as refine the assessment of cardiovascular disease risk. Machine learning methods have also been applied to assess new biomarkers and survival outcomes in patients with renal diseases to predict the development of chronic kidney disease, disease progression, and renal graft survival. In the latter, random forest methods were found to be the best for the prediction of chronic kidney disease. Some studies have investigated the prognosis of nonalcoholic fatty liver disease and acute liver failure, as well as therapy response prediction in patients with viral disorders, using decision tree models. Machine learning techniques, such as Sparse High-Order Interaction Model with Rejection Option, have been used for diagnosing Alzheimer's disease. Data mining techniques have also been applied to identify the risk factors for serious mental illness, such as depression and dementia, and help to diagnose and predict the quality of life of such patients. In relation to child health, some studies have determined the best algorithms for predicting obesity and malnutrition. Machine learning has determined the important risk factors for preterm birth and low birth weight. Published studies of patients with cancer and bacterial diseases are limited and should perhaps be addressed more comprehensively in future studies. Herein, we provide an in-depth review of studies in which biochemical biomarker data were analyzed using machine learning methods to assess the risk of several common diseases, in order to summarize the potential applications of data mining methods in clinical diagnosis. Data mining techniques have now been increasingly applied to clinical diagnostics, and they have the potential to support this field.

Postchallenge glucose increment was associated with hemoglobin glycation index in subjects with no history of diabetes

We investigated the association between postchallenge glucose increment and hemoglobin glycation index (HGI), the difference between observed and predicted glycated hemoglobin (HbA1c), in subjects with no history of diabetes. We enrolled 1381 subjects who attended our outpatient clinic for an oral glucose tolerance test (OGTT) to screen for diabetes. HGI was defined as observed HbA1c minus predicted HbA1c. The predicted HbA1c was calculated by entering fasting plasma glucose (FPG) level into an equation [HbA1c(%)=FPG(mg/dL)*0.029+2.9686] determined from an HbA1c versus FPG regression analysis using data from an independent cohort of 2734 subjects with no history of diabetes. The association between 2-hour glucose increment and HGI was analyzed using linear regression analyses with adjustment of relevant parameters. Overall, the proportions of subjects with normal glucose tolerance, pre-diabetes, and newly diagnosed diabetes were 42.3%, 41.3%, and 16.4%, respectively. Compared with subjects who had an HGI≤0, subjects with an HGI>0 had a lower FPG (95.0±13.3 vs 98.5±15.3 mg/dL, p<0.001) but a higher 2-hour plasma glucose (151.1±52.8 vs 144.6±51.4 mg/dL, p=0.027) and 2-hour glucose increment (56.1±46.1 vs 46.1±45.0 mg/dL, p<0.001). The 2-hour glucose increment after an OGTT was independently associated with HGI (β coefficient 0.003, 95% CI 0.002 to 0.003, p<0.001). Our findings suggested that postchallenge glucose increment was independently associated with HGI in subjects with no history of diabetes.

Potential predictors of type-2 diabetes risk: machine learning, synthetic data and wearable health devices

BACKGROUND

The aim of a recent research project was the investigation of the mechanisms involved in the onset of type 2 diabetes in the absence of familiarity. This has led to the development of a computational model that recapitulates the aetiology of the disease and simulates the immunological and metabolic alterations linked to type-2 diabetes subjected to clinical, physiological, and behavioural features of prototypical human individuals.

RESULTS

We analysed the time course of 46,170 virtual subjects, experiencing different lifestyle conditions. We then set up a statistical model able to recapitulate the simulated outcomes.

CONCLUSIONS

The resulting machine learning model adequately predicts the synthetic dataset and can, therefore, be used as a computationally-cheaper version of the detailed mathematical model, ready to be implemented on mobile devices to allow self-assessment by informed and aware individuals. The computational model used to generate the dataset of this work is available as a web-service at the following address: http://kraken.iac.rm.cnr.it/T2DM .

Use of Machine Learning Approaches in Clinical Epidemiological Research of Diabetes

PURPOSE OF REVIEW

Machine learning approaches-which seek to predict outcomes or classify patient features by recognizing patterns in large datasets-are increasingly applied to clinical epidemiology research on diabetes. Given its novelty and emergence in fields outside of biomedical research, machine learning terminology, techniques, and research findings may be unfamiliar to diabetes researchers. Our aim was to present the use of machine learning approaches in an approachable way, drawing from clinical epidemiological research in diabetes published from 1 Jan 2017 to 1 June 2020.

RECENT FINDINGS

Machine learning approaches using tree-based learners-which produce decision trees to help guide clinical interventions-frequently have higher sensitivity and specificity than traditional regression models for risk prediction. Machine learning approaches using neural networking and "deep learning" can be applied to medical image data, particularly for the identification and staging of diabetic retinopathy and skin ulcers. Among the machine learning approaches reviewed, researchers identified new strategies to develop standard datasets for rigorous comparisons across older and newer approaches, methods to illustrate how a machine learner was treating underlying data, and approaches to improve the transparency of the machine learning process. Machine learning approaches have the potential to improve risk stratification and outcome prediction for clinical epidemiology applications. Achieving this potential would be facilitated by use of universal open-source datasets for fair comparisons. More work remains in the application of strategies to communicate how the machine learners are generating their predictions.

The trials and tribulations of determining HbA1c targets for diabetes mellitus

Glycated haemoglobin (HbA_1c) is considered the gold standard for predicting glycaemia-associated risks for the microvascular and macrovascular complications of diabetes mellitus over 5-10 years. The value of HbA_1c in the care of patients with type 1 diabetes mellitus (T1DM) and type 2 diabetes mellitus (T2DM) is unassailable, yet HbA_1c targets remain contentious. Guidelines from diabetes care organizations recommend conflicting HbA_1c targets - generally between 6.5% and 8%. However, all such organizations advocate for individualization of HbA_1c targets, leaving both health-care providers and their patients confused about what HbA_1c target is appropriate in an individual patient. In this Review, we outline the landmark T1DM and T2DM trials that informed the current guidelines, we discuss the evidence that drives individualized HbA_1c targets, we examine the limitations of HbA_1c, and we consider alternatives for monitoring glycaemic control. Ultimately, in synthesizing this literature, we argue for an HbA_1c target of <7% for most individuals, but emphasize the importance of helping patients determine their own personal goals and determinants of quality of life that are independent of a particular glycaemic target. We also recognize that as newer technologies and anti-hyperglycaemic therapies emerge, glycaemic targets will continue to evolve.

Non-invasive detection of diabetic complications via pattern analysis of temporal facial colour variations

BACKGROUND AND OBJECTIVE

Diabetes mellitus is a common disorder amounting to 400 million patients worldwide. It is often accompanied by a number of complications, including neuropathy, nephropathy, and cardiovascular diseases. For example, peripheral neuropathy is present among 20-30% of diabetics before the diagnosis is substantiated. For this reason, a reliable detection method for diabetic complications is crucial and attracts a lot of research attention.

METHODS

In this paper, we introduce a non-invasive detection framework for patients with diabetic complications that only requires short video recordings of faces from a standard commercial camera. We employed multiple image processing and pattern recognition techniques to process video frames, extract relevant information, and predict the health status. To evaluate our framework, we collected a dataset of 114 video files from diabetic patients, who were diagnosed with diabetes for years and 60 video files from the control group. Extracted features from videos were tested using two conceptually different classifiers.

RESULTS

We found that our proposed framework correctly identifies patients with diabetic complications with 92.86% accuracy, 100% sensitivity, and 80% specificity.

CONCLUSIONS

Our study brings a novel perspective on diagnosis procedures in this field. We used multiple techniques from image processing, pattern recognition, and machine learning to robustly process video frames and predict the health status of our subjects with high efficiency.

Heterogeneous Treatment Effects on Cardiovascular Diseases With Dipeptidyl Peptidase-4 Inhibitors Versus Sulfonylureas in Type 2 Diabetes Patients

This study explored heterogeneous treatment effects (HTEs) of the real-world use of dipeptidyl peptidase-4 inhibitors (DPP-4is) vs. sulfonylureas (SUs) on cardiovascular diseases (CVDs) and mortality in patients with type 2 diabetes. Utilizing Taiwan's National Health Insurance Research Database, 19,853 propensity score-matched pairs of DPP-4i and SU stable users were identified. Classification and regression tree analyses and Cox models were applied to explore HTEs, according to various patient characteristics, on the composite CVDs, three-point major adverse cardiovascular event (MACE), and all-cause mortality. The absolute risk difference (ARD), hazard ratio (HR), and 95% confidence interval (CI) were estimated for comparing treatment effects. CVD history, ischemic stroke, or transient ischemic attack (IS/TIA) history, and age at treatment initiation were significant treatment effect modifiers. Patients with prior IS/TIA but without any other prior CVDs benefited most in reduced risks of composite CVDs from using DPP-4i vs. SU (ARD -4.31%, 95% CI -7.48% to -1.14%, HR 0.81, 95% CI 0.69 ~ 0.95), followed by those without prior IS/TIA and CVDs and initiated with DPP-4i at age < 69.3 years (ARD -0.90%, 95% CI -1.47% to -0.32%, HR 0.86, 95% CI 0.77 ~ 0.97). Patients with prior IS/TIA benefited most in reduced risks of three-point MACE from using DPP-4i vs. SU (ARD -4.22%, 95% CV -6.66% to -1.78%, HR 0.80, 95% CI 0.69 ~ 0.93), followed by those without prior IS/TIA and initiated with DPP-4i at age < 69.3 years (ARD -0.68%, 95% CI -1.08% to -0.29%, HR 0.81, 95% CI 0.70 ~ 0.93). Consideration of CVD and IS/TIA histories and age could facilitate individualized diabetes management of using DPP-4i vs. SU. Future studies are warranted given the hypothesis-generating nature in this exploratory research.

Value of a Machine Learning Approach for Predicting Clinical Outcomes in Young Patients With Hypertension

Risk stratification of young patients with hypertension remains challenging. Generally, machine learning (ML) is considered a promising alternative to traditional methods for clinical predictions because it is capable of processing large amounts of complex data. We, therefore, explored the feasibility of an ML approach for predicting outcomes in young patients with hypertension and compared its performance with that of approaches now commonly used in clinical practice. Baseline clinical data and a composite end point—comprising all-cause death, acute myocardial infarction, coronary artery revascularization, new-onset heart failure, new-onset atrial fibrillation/atrial flutter, sustained ventricular tachycardia/ventricular fibrillation, peripheral artery revascularization, new-onset stroke, end-stage renal disease—were evaluated in 508 young patients with hypertension (30.83±6.17 years) who had been treated at a tertiary hospital. Construction of the ML model, which consisted of recursive feature elimination, extreme gradient boosting, and 10-fold cross-validation, was performed at the 33-month follow-up evaluation, and the model’s performance was compared with that of the Cox regression and recalibrated Framingham Risk Score models. An 11-variable combination was considered most valuable for predicting outcomes using the ML approach. The C statistic for identifying patients with composite end points was 0.757 (95% CI, 0.660–0.854) for the ML model, whereas for Cox regression model and the recalibrated Framingham Risk Score model it was 0.723 (95% CI, 0.636–0.810) and 0.529 (95% CI, 0.403–0.655). The ML approach was comparable with Cox regression for determining the clinical prognosis of young patients with hypertension and was better than that of the recalibrated Framingham Risk Score model.

The present and future scope of real-world evidence research in diabetes: What questions can and cannot be answered and what might be possible in the future?

The last decade has witnessed an exponential growth in the opportunities to collect and link health-related data from multiple resources, including primary care, administrative, and device data. The availability of these "real-world," "big data" has fuelled also an intense methodological research into methods to handle them and extract actionable information. In medicine, the evidence generated from "real-world data" (RWD), which are not purposely collected to answer biomedical questions, is commonly termed "real-world evidence" (RWE). In this review, we focus on RWD and RWE in the area of diabetes research, highlighting their contributions in the last decade; and give some suggestions for future RWE diabetes research, by applying well-established and less-known tools to direct RWE diabetes research towards better personalized approaches to diabetes care. We underline the essential aspects to consider when using RWD and the key features limiting the translational potential of RWD in generating high-quality and applicable RWE. Only if viewed in the context of other study designs and statistical methods, with its pros and cons carefully considered, RWE will exploit its full potential as a complementary or even, in some cases, substitutive source of evidence compared to the expensive evidence obtained from randomized controlled trials.

Implications of the Hemoglobin Glycation Index on the Diagnosis of Prediabetes and Diabetes

OBJECTIVE

Fasting plasma glucose (FPG), 2-hour plasma glucose (2hPG) from a 75-g oral glucose tolerance test (OGTT) and glycated hemoglobin (HbA1c) can lead to different results when diagnosing prediabetes and diabetes. The Hemoglobin Glycation Index (HGI) quantifies the interindividual variation in glycation resulting in discrepancies between FPG and HbA1c. We used data from the Vitamin D and Type 2 Diabetes (D2d) study to calculate HGI, to identify HGI-associated variables, and to determine how HGI affects prediabetes and diabetes diagnosis.

MEASUREMENTS

A linear regression equation [HbA1c (%) = 0.0164 × FPG (mg/dL) + 4.2] was derived using the screening cohort (n = 6829) and applied to calculate predicted HbA1c. This was subtracted from the observed HbA1c to determine HGI in the baseline cohort with 2hPG data (n = 3945). Baseline variables plus prediabetes and diabetes diagnosis by FPG, HbA1c, and 2hPG were compared among low, moderate, and high HGI subgroups.

RESULTS

The proportion of women and Black/African American individuals increased from low to high HGI subgroups. Mean FPG decreased and mean HbA1c increased from low to high HGI subgroups, consistent with the HGI calculation; however, mean 2hPG was not significantly different among HGI subgroups.

CONCLUSIONS

High HGI was associated with Black race and female sex as reported previously. The observation that 2hPG was not different across HGI subgroups suggests that variation in postprandial glucose is not a significant source of population variation in HGI. Exclusive use of HbA1c for diagnosis will classify more Black individuals and women as having prediabetes compared with using FPG or 2hPG.

Machine Learning Applications in Endocrinology and Metabolism Research: An Overview

Machine learning (ML) applications have received extensive attention in endocrinology research during the last decade. This review summarizes the basic concepts of ML and certain research topics in endocrinology and metabolism where ML principles have been actively deployed. Relevant studies are discussed to provide an overview of the methodology, main findings, and limitations of ML, with the goal of stimulating insights into future research directions. Clear, testable study hypotheses stem from unmet clinical needs, and the management of data quality (beyond a focus on quantity alone), open collaboration between clinical experts and ML engineers, the development of interpretable high-performance ML models beyond the black-box nature of some algorithms, and a creative environment are the core prerequisites for the foreseeable changes expected to be brought about by ML and artificial intelligence in the field of endocrinology and metabolism, with actual improvements in clinical practice beyond hype. Of note, endocrinologists will continue to play a central role in these developments as domain experts who can properly generate, refine, analyze, and interpret data with a combination of clinical expertise and scientific rigor.

Identifying treatment effects of an informal caregiver education intervention to increase days in the community and decrease caregiver distress: a machine-learning secondary analysis of subgroup effects in the HI-FIVES randomized clinical trial

BACKGROUND

Informal caregivers report substantial burden and depressive symptoms which predict higher rates of patient institutionalization. While caregiver education interventions may reduce caregiver distress and decrease the use of long-term institutional care, evidence is mixed. Inconsistent findings across studies may be the result of reporting average treatment effects which do not account for how effects differ by participant characteristics. We apply a machine-learning approach to randomized clinical trial (RCT) data of the Helping Invested Family Members Improve Veteran's Experiences Study (HI-FIVES) intervention to explore how intervention effects vary by caregiver and patient characteristics.

METHODS

We used model-based recursive partitioning models. Caregivers of community-residing older adult US veterans with functional or cognitive impairment at a single VA Medical Center site were randomized to receive HI-FIVES (n = 118) vs. usual care (n = 123). The outcomes included cumulative days not in the community and caregiver depressive symptoms assessed at 12 months post intervention. Potential moderating characteristics were: veteran age, caregiver age, caregiver ethnicity and race, relationship satisfaction, caregiver burden, perceived financial strain, caregiver depressive symptoms, and patient risk score.

RESULTS

The effect of HI-FIVES on days not at home was moderated by caregiver burden (p < 0.001); treatment effects were higher for caregivers with a Zarit Burden Scale score ≤ 28. Caregivers with lower baseline Center for Epidemiologic Studies Depression Scale (CESD-10) scores (≤ 8) had slightly lower CESD-10 scores at follow-up (p < 0.001).

CONCLUSIONS

Family caregiver education interventions may be less beneficial for highly burdened and distressed caregivers; these caregivers may require a more tailored approach that involves assessing caregiver needs and developing personalized approaches.

TRIAL REGISTRATION

ClinicalTrials.gov, ID:NCT01777490. Registered on 28 January 2013.

Prediction of Nephropathy in Type 2 Diabetes: An Analysis of the ACCORD Trial Applying Machine Learning Techniques

Applying data mining and machine learning (ML) techniques to clinical data might identify predictive biomarkers for diabetic nephropathy (DN), a common complication of type 2 diabetes mellitus (T2DM). A retrospective analysis of the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial was intended to identify such factors using ML. The longitudinal data were stratified by time after patient enrollment to differentiate early and late predictors. Our results showed that Random Forest and Simple Logistic Regression methods exhibited the best performance among the evaluated algorithms. Baseline values for glomerular filtration rate (GFR), urinary creatinine, urinary albumin, potassium, cholesterol, low-density lipoprotein, and urinary albumin to creatinine ratio were identified as DN predictors. Early predictors were the baseline values of GFR, systolic blood pressure, as well as fasting plasma glucose (FPG) and potassium at month 4. Changes per year in GFR, FPG, and triglycerides were recognized as predictors of late development. In conclusion, ML-based methods successfully identified predictive factors for DN among patients with T2DM.

Feature Selection in Predictive Modeling: A Systematic Study on Drug Response Heterogeneity for Type II Diabetic Patients

With the rapid development of computer hardware and software technologies, more and more electronic health data from insurance claims, clinical trials and hospitals are becoming readily available. These data provide a rich resource for developing various healthcare analytics algorithms, among which predictive modeling is of key importance in many real health problems. One important issue for data-driven predictive modeling is high dimensionality, and feature selection is one effective strategy to reduce the number of independent variables and control the confounding factors. However, most of the existing studies just pick one feature selection approach without comprehensive investigations. In this paper, we investigate the issue of drug response heterogeneity for type II diabetes mellitus (T2DM) patients using a large scale clinical trial data. Our goal is to find out the important factors that may lead to the response heterogeneity for three popular T2DM drugs, Metformin, Rosiglitazone and Glimepiride. We implemented 8 different feature selection approaches and compared their performances with various measures including prediction error and the consistency of the identified important factors. Finally, we ensemble all factor lists picked by different algorithms and obtain a final set of factors that contribute to the drug response heterogeneities and verified them through existing literature.

Variations in glucose/C-peptide ratio in patients with type 2 diabetes associated with renal function

INTRODUCTION

Accurate dosing of medications for glycemic control is a challenge for clinicians in diabetic patients with kidney disease. Diminishing glomerular filtration rates are associated with decreased renal clearance of insulin and increased prevalence of hypoglycemic episodes. Measurement of glucose/C peptide ratios may be useful to guide dosing in those patients who receive powerful insulin secretogogues as glomerular function decreases with age and disease.

METHODS

In order to determine the relationship between glucose, C-peptide and renal function, we reviewed the records of patients with type 2 diabetes followed in our kidney hypertension clinic who met the following criteria: age 35-90 years, requirement of medications to control glycemia, at least 4 simultaneous measurements of C peptide, HbA1c, creatinine and blood glucose.

RESULTS

87 patients (67 males, 20 females), ages 67.1 ± 10.6 years, BMI 32.5 ± 5.2, A1c 8.2 ± 1.2%, eGFR 73 ± 27.2 ml/min, had glucose/C-peptide ratios 60.7 ± 46.4. 59% of the total group were taking insulin secretogogues. Patients were divided into groups based upon mean eGFR and use or absence of insulin secretogogues. Glucose C-peptide ratios were lowest in the quartile of patients with the lowest eGFR (<50 ml/min).

CONCLUSION

Diminished renal function and advanced age are associated with the lowest glucose/C-peptide ratios, independent of achieved glycemic control. With similar use of secretogogues, glucose/C-peptide ratio were lower when eGFR was ≤49 ml/min compared to >50-80 ml/min. Use of secretogogues was associated with decreased glucose/C-peptide levels. In patients with reduced renal function (eGFR < 50 ml/min), use of insulin secretogogues may be associated with lower glucose/C-peptide ratios associated with higher risks for hypoglycemic reactions.

Hemoglobin glycation index predicts cardiovascular disease in people with type 2 diabetes mellitus: A 10-year longitudinal cohort study

BACKGROUND AND AIMS

Previous studies have suggested that the hemoglobin glycation index (HGI) can be used as a predictor of diabetes-related complications. We examined the prognostic significance of a high HGI for cardiovascular disease (CVD) in an ongoing hospital-based cohort.

METHODS

From March 2003 to December 2004, 1302 consecutive patients with type 2 diabetes and without a prior history of CVD were enrolled. CVD was defined as the occurrence of coronary artery disease or ischemic stroke. The HGI was calculated as the measured glycated hemoglobin (HbA1c) minus predicted HbA1c. Predicted HbA1c were calculated for 1302 participants by inserting fasting blood glucose (FBG) into the equation, Predicted HbA1c level = 0.02106 × FBG [mg/dL] + 4.973. Cox proportional hazards models were used to identify the associations between the HGI and CVD after adjusting for confounding variables.

RESULTS

During 11.1 years of follow-up, 225 participants (17.2%) were newly diagnosed with CVD. The baseline HGI was significantly higher in subjects with incident CVD than in those without CVD, although the baseline FBG levels did not differ according to the occurrence of CVD. Compared with patients without CVD, those with CVD were older, had a longer duration of diabetes and hypertension, and used more insulin at baseline. A Cox hazard regression analysis revealed that the development of CVD was significantly associated with baseline HGI (hazard ratio [HR], 1.94; 95% confidence interval [CI], 1.31-2.87; p < 0.001, comparing the highest and lowest quartiles of HGI). This relationship was unchanged after additional adjustment for baseline HbA1c level (HR, 1.74; 95% CI, 1.08-2.81). The HRs of HbA1c in relation to outcomes were similar to or lower than those seen for HGI. After adjustment for HGI, the effect of the highest HbA1c on incident CVD disappeared.

CONCLUSIONS

High HGI was independently associated with incident CVD in patients with type 2 diabetes. Patients with high HGI at baseline had a higher inherent risk for CVD.