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

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.