Skip to Main content Skip to Navigation

Challenges in personalized evidence-based medicine, applications in type 2 diabetes

Abstract : Evidence based medicine requires randomized clinical trials for estimating a mean treatment effect. The personalization of this treatment effect needs prognostic biomarker for assessing the spontaneous risk of the disease and the absolute benefit of the treatment; and the search for potential theranostic biomarker, associated with a different relative treatment effect. Surrogate endpoints are also proposed, as their measure would reflect the treatment effect on the clinical outcome of interest. Taking care of patients with type 2 diabetes is based on hypoglycemic drugs. Several of them have been retrospectively associated with serious adverse events. They need to be assessed with cardiovascular outcome trials. Taking care of those patients also include handling other cardiovascular risk factor, as high blood pressure. Antihypertensive treatment is based on a “target to treat” strategy, which raise several questions. Finally, many theranostic biomarkers of the hypoglycemic drugs effect have been studied, with conflicting results. Statistical power is a high challenge in randomized trial looking for such interaction. We aimed to provide a mean treatment effect estimation of hypoglycemic drugs on cardiovascular outcomes and to explore potential tools for personalizing the treatment effect estimation. The first part of this thesis reports a network meta-analysis assessing the contemporary hypoglycemic drugs in type 2 diabetes patients on overall mortality, cardiovascular mortality and major adverse cardiovascular events. We confirmed the superiority of SGLT2 inhibitors and of GLP1 receptor agonists compared to control and to DPP4 inhibitors. We also showed the need for direct comparison, especially for clarifying the position of metformin in the pharmacological strategy. The second part of this thesis reports a meta-regression analysis, assessing the association between the decrease in blood pressure through antihypertensive drugs and the risk of cardiovascular events. We confirmed the association between the blood pressure control and the risk of stroke, but did not find any association regarding overall mortality, cardiovascular mortality and myocardial infarction. The third part reports a statistical comparison of the parallel group design and the cross-over design, regarding their capacity to assess a potential theranostic biomarker. We showed that the advantage of the cross-over for reducing the sample size lead on the intra-subject correlation, as already known for estimating the treatment effect. Finally, we highlighted the need for comparisons of hypoglycemic drugs for preventing macrovascular events. We emphasized pitfalls in estimating benefit—risk balance. Individual patient data meta-analyses would help better assessing the effect of glucose control on macrovascular events. High-throughput genome sequencing technologies would help to identify both prognostic and theranostic biomarkers. Lastly, we proposed an extended version of the effect model, which allow to grasp the benefit—risk balance of a treatment, according to different biomarkers. To conclude, assessing a mean and a stratified treatment effect should be conducted taking into account the global benefit—risk balance estimation
Document type :
Complete list of metadatas

Cited literature [274 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Tuesday, March 3, 2020 - 3:32:08 PM
Last modification on : Thursday, March 5, 2020 - 1:13:41 AM
Long-term archiving on: : Thursday, June 4, 2020 - 4:10:20 PM


Version validated by the jury (STAR)


  • HAL Id : tel-02497292, version 1



Guillaume Grenet. Challenges in personalized evidence-based medicine, applications in type 2 diabetes. Pharmacology. Université de Lyon, 2019. English. ⟨NNT : 2019LYSE1233⟩. ⟨tel-02497292⟩



Record views


Files downloads