Search

Current treatment guidelines do not provide recommendations to support the selection of treatment for most people with type 2 diabetes. We aimed to develop and validate an algorithm to allow selection of optimal treatment based on glycaemic response, weight change, and tolerability outcomes when choosing between SGLT2 inhibitor or DPP-4 inhibitor therapies.In this retrospective cohort study, we identified patients initiating SGLT2 and DPP-4 inhibitor therapies after Jan 1, 2013, from the UK Clinical Practice Research Datalink (CPRD). We excluded those who received SGLT2 or DPP-4 inhibitors as first-line treatment or insulin at the same time, had estimated glomerular filtration rate (eGFR) of less than 45 mL/min per 1·73 mAmong 10 253 patients initiating SGLT2 inhibitors and 16 624 patients initiating DPP-4 inhibitors in CPRD, baseline HbAA validated treatment selection algorithm for SGLT2 inhibitor and DPP-4 inhibitor therapies can support decisions on optimal treatment for people with type 2 diabetes.BHF-Turing Cardiovascular Data Science Award and the UK Medical Research Council.

ObjectiveTo compare individualized treatment selection strategies based on predicted individual-level treatment effects from a causal forest machine learning algorithm and a penalized regression model.Study Design and SettingCohort 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 trials (SGLT2-inhibitor versus DPP4-inhibitor). 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 UK primary care patients (Clinical Practice Research Datalink).ResultsHeterogeneity in treatment effects was detected in trial participants with both approaches (causal forest: 98.6% & penalized regression: 81.7% predicted to have a benefit on SGLT2-inhibitor therapy over DPP4-inhibitor therapy). 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).ConclusionWhen 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.What is new?QuestionWhat is the comparative utility of machine learning compared to standard regression for identifying variation in patient-level outcomes (treatment effect heterogeneity) due to different treatments?FindingsCausal forest and penalized regression models were developed using trial data to predict glycated hemoglobin [HbA1c]) outcomes with SGLT2-inhibitor and DPP4-inhibitor therapy in 1,428 individuals with type 2 diabetes. In external validation (18,741 patients), penalized regression outperformed causal forest at identifying population strata with a superior glycemic response to SGLT2-inhibitors compared to DPP4-inhibitors.ImplicationsStudies estimating treatment effect heterogeneity should not solely rely on machine learning and should compare results with standard regression.

Precision medicine aims to treat an individual based on their clinical characteristics. A differential drug response, critical to using these features for therapy selection, has never been examined directly in type 2 diabetes. In this study, we tested two hypotheses: (1) individuals with body mass index (BMI) 30 kg/m

Aims/hypothesis Among white European children developing type 1 diabetes, the otherwise common HLA haplotype DR15-DQ6 is rare, and highly protective. Adult-onset type 1 diabetes is now known to represent more overall cases than childhood onset, but it is not known whether DR15-DQ6 is protective in older-adult-onset type 1 diabetes. We sought to quantify DR15-DQ6 protection against type 1 diabetes as age of onset increased. Methods In two independent cohorts we assessed the proportion of type 1 diabetes cases presenting through the first 50 years of life with DR15-DQ6, compared with population controls. In the After Diabetes Diagnosis Research Support System-2 (ADDRESS-2) cohort (n = 1458) clinician-diagnosed type 1 diabetes was confirmed by positivity for one or more islet-specific autoantibodies. In UK Biobank (n = 2502), we estimated type 1 diabetes incidence rates relative to baseline HLA risk for each HLA group using Poisson regression. Analyses were restricted to white Europeans and were performed in three groups according to age at type 1 diabetes onset: 0–18 years, 19–30 years and 31–50 years. Results DR15-DQ6 was protective against type 1 diabetes through to age 50 years (OR p Conclusions/interpretation HLA DR15-DQ6 confers dominant protection from type 1 diabetes across the first five decades of life. Graphical abstract

BackgroundMissing data is a common problem in regression modelling. Much of the literature focuses on handling missing outcome variables, but there are also challenges when dealing with missing predictor information, particularly when trying to build prediction models for use in practice.MethodsWe develop a flexible Bayesian approach for handling missing predictor information in regression models. For prediction this provides practitioners with full posterior predictive distributions for both the missing predictor information and the outcome variable, conditional on the observed predictors. We apply our approach to a previously proposed treatment selection model for type 2 diabetes second-line therapies. Our approach combines a regression model and a Dirichlet process mixture model (DPMM), where the former defines the treatment selection model and the latter provides a flexible way to model the joint distribution of the predictors.ResultsWe show that under missing-completely-at-random (MCAR) and missing-at-random (MAR) assumptions (with respect to the missing predictors), the DPMM can model complex relationships between predictor variables, and predict missing values conditionally on existing information. We also demonstrate that in the presence of multiple missing predictors, the DPMM model can be used to explore which variable(s), if collected, could provide the most additional information about the likely outcome.ConclusionsOur approach can provide practitioners with supplementary information to aid treatment selection decisions in the presence of missing data, and can be readily extended to other types of response model.Key MessagesMissing predictor variables present a significant challenge when building and implementing prediction models in clinical practice.Removing individuals with missing information and performing a complete case analysis can lead to imprecision and bias. Multiple imputation approaches typically translate uncertainty through prediction model parameter standard errors, as opposed to a consistent joint probability model.Alternatively, a Bayesian approach using Dirichlet process mixture models (DPMMs) offers a flexible way to model complex joint distributions of predictor variables, which can be used to estimate posterior (predictive) distributions for the missing predictors, conditional on the observed predictors.Using a DPMM, in this way allows uncertainties around missing predictor data to be propagated through to a prediction model of interest using a Bayesian hierarchical framework. This allows prediction models to be developed using datasets with incomplete predictor information (assuming missing-completely-at-random/missing-at-random). Furthermore, predictions can be made on new individuals even if they have incomplete predictor information (under the same assumptions).This approach provides full posterior predictive probability distributions for both missing predictor variables and the outcome variable, allowing a wide range of probabilistic models outputs to be derived to support clinical decision making.

The complexity of each Coupled Model Intercomparison Project grows with every new generation. The Phase 5 effort saw a dramatic increase in the number of experiments that were performed and the number of variables that were requested compared to its previous generation, Phase 3. The large increase in data volume stressed the resources of several centers including at the National Center for Atmospheric Research. During Phase 5, we missed several deadlines and we struggled to get the data out to the community for analysis. In preparation for the current generation, Phase 6, we examined the weaknesses in our workflow and addressed the performance issues with new software tools. Through this investment, we were able to publish approximately 565 TB of compressed data to the community, with another 30 TB yet to be published. When compared to the volumes we produced in the previous generation, 165 TB of uncompressed data, we were able to provide 6 times the amount of data and we accomplish this within one-third of the time. This provided us with an approximate 18 times faster speedup. While this paper discusses the improvements we have made to our own workflow for the Coupled Model Intercomparison Project Phase 6 (CMIP6), we hope to encourage other centers to evaluate and invest in their own workflows in order to be successful in these types of modeling campaigns.

Background Whilst eczema is a common inflammatory skin condition, we lack contemporary estimates of disease incidence and prevalence across the lifespan. Objective To estimate the incidence and prevalence of eczema in children and adults in England and variation by sociodemographic factors (sex, socio-economic status, ethnicity, and geography). Methods We used the Royal College of General Practitioners Research and Surveillance Centre primary care research database of 3.85 million children and adults registered with participating general practitioner practices between 2009 and 2018 inclusive. Eczema incidence was defined as the first-ever diagnosis of eczema recorded in the primary care record, and eczema prevalence was defined as fulfilment of criteria for active eczema (two eczema records appearing in the primary care record within any one-year period). Results Eczema incidence was highest in infants younger than 1 year (15.0 per 100 person-years), lowest in adults aged 40–49 (0.35 p/100 person-years), and increased from middle age to a second smaller peak in people 80 years or older (0.79 p/100 person-years). Eczema prevalence was highest in children aged 2 (16.5%) and lowest in adults aged 30–39 (2.8%). Eczema incidence was higher in male infants ( Conclusions and Clinical Relevance Eczema has a bimodal distribution across the lifespan. We observed differences in incidence and prevalence of eczema by ethnicity, geography, sex, and socio-economic status, which varied in magnitude throughout life.

OBJECTIVE To describe the relationship between type 2 diabetes and all-cause mortality among adults with coronavirus disease 2019 (COVID-19) in the critical care setting. RESEARCH DESIGN AND METHODS This was a nationwide retrospective cohort study in people admitted to hospital in England with COVID-19 requiring admission to a high dependency unit (HDU) or intensive care unit (ICU) between 1 March 2020 and 27 July 2020. Cox proportional hazards models were used to estimate 30-day in-hospital all-cause mortality associated with type 2 diabetes, with adjustment for age, sex, ethnicity, obesity, and other major comorbidities (chronic respiratory disease, asthma, chronic heart disease, hypertension, immunosuppression, chronic neurological disease, chronic renal disease, and chronic liver disease). RESULTS A total of 19,256 COVID-19–related HDU and ICU admissions were included in the primary analysis, including 13,809 HDU (mean age 70 years) and 5,447 ICU (mean age 58 years) admissions. Of those admitted, 3,524 (18.3%) had type 2 diabetes and 5,077 (26.4%) died during the study period. Patients with type 2 diabetes were at increased risk of death (adjusted hazard ratio [aHR] 1.23 [95% CI 1.14, 1.32]), and this result was consistent in HDU and ICU subsets. The relative mortality risk associated with type 2 diabetes decreased with higher age (age 18–49 years aHR 1.50 [95% CI 1.05, 2.15], age 50–64 years 1.29 [1.10, 1.51], and age ≥65 years 1.18 [1.09, 1.29]; P value for age–type 2 diabetes interaction = 0.002). CONCLUSIONS Type 2 diabetes may be an independent prognostic factor for survival in people with severe COVID-19 requiring critical care treatment, and in this setting the risk increase associated with type 2 diabetes is greatest in younger people.

Despite the known heterogeneity of type 2 diabetes and variable response to glucose lowering medications, current evidence on optimal treatment is predominantly based on average effects in clinical trials rather than individual-level characteristics. A precision medicine approach based on treatment response would aim to improve on this by identifying predictors of differential drug response for people based on their characteristics and then using this information to select optimal treatment. Recent research has demonstrated robust and clinically relevant differential drug response with all noninsulin treatments after metformin (sulfonylureas, thiazolidinediones, dipeptidyl peptidase 4 [DPP-4] inhibitors, glucagon-like peptide 1 [GLP-1] receptor agonists, and sodium–glucose cotransporter 2 [SGLT2] inhibitors) using routinely available clinical features. This Perspective reviews this current evidence and discusses how differences in drug response could inform selection of optimal type 2 diabetes treatment in the near future. It presents a novel framework for developing and testing precision medicine–based strategies to optimize treatment, harnessing existing routine clinical and trial data sources. This framework was recently applied to demonstrate that “subtype” approaches, in which people are classified into subgroups based on features reflecting underlying pathophysiology, are likely to have less clinical utility compared with approaches that combine the same features as continuous measures in probabilistic “individualized prediction” models.

Objectives Electronic health records (EHR) provide a valuable resource for assessing drug side-effects, but treatments are not randomly allocated in routine care creating the potential for bias. We conduct a case study using the Prior Event Rate Ratio (PERR) Pairwise method to reduce unmeasured confounding bias in side-effect estimates for two second-line therapies for type 2 diabetes, thiazolidinediones, and sulfonylureas. Study Design and Settings Primary care data were extracted from the Clinical Practice Research Datalink (n = 41,871). We utilized outcomes from the period when patients took first-line metformin to adjust for unmeasured confounding. Estimates for known side-effects and a negative control outcome were compared with the A Diabetes Outcome Progression Trial (ADOPT) trial (n = 2,545). Results When on metformin, patients later prescribed thiazolidinediones had greater risks of edema, HR 95% CI 1.38 (1.13, 1.68) and gastrointestinal side-effects (GI) 1.47 (1.28, 1.68), suggesting the presence of unmeasured confounding. Conventional Cox regression overestimated the risk of edema on thiazolidinediones and identified a false association with GI. The PERR Pairwise estimates were consistent with ADOPT: 1.43 (1.10, 1.83) vs. 1.39 (1.04, 1.86), respectively, for edema, and 0.91 (0.79, 1.05) vs. 0.94 (0.80, 1.10) for GI. Conclusion The PERR Pairwise approach offers potential for enhancing postmarketing surveillance of side-effects from EHRs but requires careful consideration of assumptions.