Personalised medicine describes the practice of tailoring medical care to the individual characteristics of each patient. Fundamental to this practice is the identification of markers associated with differential treatment response. Such markers can be identified through the assessment of treatment effect modification using statistical methods. Randomised controlled trials provide the optimal setting for evaluating differential response to treatment. Due to restrictions regarding sample size, study length and ethics, observational studies are more appropriate in many circumstances, particularly for the identification of markers associated with adverse side-effects and long term response to treatments. However, the analysis of observational data raises some additional challenges. The overall aim of this thesis was to develop statistical methodology for the analysis of observational data, specifically primary care databases, to identify and evaluate markers associated with differential treatment response. Three aspects of the assessment of treatment effect modification in an observational setting were addressed. The first aspect related to the assessment of treatment effect modification on the additive measurement scale which corresponds to a comparison of absolute treatment effects across patient subgroups. Various ways in which this can be assessed in an observational setting were reviewed and a novel measure, the ratio of absolute effects, which can be calculated from certain multiplicative regression models, was proposed. The second aspect regarded the confounding adjustment and it was investigated how the presence of interactions between the moderator and confounders on both treatment receipt and outcome can bias estimates of treatment effect modification if unaccounted for using Monte Carlo simulations. It was determined that the presence of bias differed across different confounding adjustment methods and, in the majority of settings, the bias was reduced when the interactions between the moderator and confounders were accounted for in the confounding adjustment model. Thirdly, it has been proposed that patient data in observational studies be organised into and analysed as series of nested nonrandomised trials. This thesis extended this study design to evaluate predictive markers of differential treatment response and explored the benefits of this methodology for this purpose. It was suggested how absolute treatment effect estimates can be estimated and compared across patient subgroups in this setting. A dataset comprising primary care medical records of adults with rheumatoid arthritis was used throughout this thesis. Interest lay in the identification of characteristics predictive of the onset of type II diabetes associated with steroid (glucocorticoid) therapy. The analysis in this thesis suggested older age may be associated with a higher risk of steroid-associated type II diabetes, but this warrants further investigation. Overall, this thesis demonstrates how observational studies can be analysed such that accurate and meaningful conclusions are made within personalised medicine research.

Avec la révolution récente de la génomique et la médecine stratifiée, le développement de signatures moléculaires devient de plus en plus important pour prédire le pronostic (biomarqueurs pronostiques) ou l’effet d’un traitement (biomarqueurs prédictifs) de chaque patient. Cependant, la grande quantité d’information disponible rend la découverte de faux positifs de plus en plus fréquente dans la recherche biomédicale. La présence de données de grande dimension (nombre de biomarqueurs ≫ taille d’échantillon) soulève de nombreux défis statistiques tels que la non-identifiabilité des modèles, l’instabilité des biomarqueurs sélectionnés ou encore la multiplicité des tests.L’objectif de cette thèse a été de proposer et d’évaluer des méthodes statistiques pour l’identification de ces biomarqueurs et l’élaboration d’une prédiction individuelle des probabilités de survie pour des nouveaux patients à partir d’un modèle de régression de Cox. Pour l’identification de biomarqueurs en présence de données de grande dimension, la régression pénalisée lasso est très largement utilisée. Dans le cas de biomarqueurs pronostiques, une extension empirique de cette pénalisation a été proposée permettant d’être plus restrictif sur le choix du paramètre λ dans le but de sélectionner moins de faux positifs. Pour les biomarqueurs prédictifs, l’intérêt s’est porté sur les interactions entre le traitement et les biomarqueurs dans le contexte d’un essai clinique randomisé. Douze approches permettant de les identifier ont été évaluées telles que le lasso (standard, adaptatif, groupé ou encore ridge+lasso), le boosting, la réduction de dimension des effets propres et un modèle implémentant les effets pronostiques par bras. Enfin, à partir d’un modèle de prédiction pénalisé, différentes stratégies ont été évaluées pour obtenir une prédiction individuelle pour un nouveau patient accompagnée d’un intervalle de confiance, tout en évitant un éventuel surapprentissage du modèle. La performance des approches ont été évaluées au travers d’études de simulation proposant des scénarios nuls et alternatifs. Ces méthodes ont également été illustrées sur différents jeux de données, contenant des données d’expression de gènes dans le cancer du sein. / With the recent revolution in genomics and in stratified medicine, the development of molecular signatures is becoming more and more important for predicting the prognosis (prognostic biomarkers) and the treatment effect (predictive biomarkers) of each patient. However, the large quantity of information has rendered false positives more and more frequent in biomedical research. The high-dimensional space (i.e. number of biomarkers ≫ sample size) leads to several statistical challenges such as the identifiability of the models, the instability of the selected coefficients or the multiple testing issue.The aim of this thesis was to propose and evaluate statistical methods for the identification of these biomarkers and the individual predicted survival probability for new patients, in the context of the Cox regression model. For variable selection in a high-dimensional setting, the lasso penalty is commonly used. In the prognostic setting, an empirical extension of the lasso penalty has been proposed to be more stringent on the estimation of the tuning parameter λ in order to select less false positives. In the predictive setting, focus has been given to the biomarker-by-treatment interactions in the setting of a randomized clinical trial. Twelve approaches have been proposed for selecting these interactions such as lasso (standard, adaptive, grouped or ridge+lasso), boosting, dimension reduction of the main effects and a model incorporating arm-specific biomarker effects. Finally, several strategies were studied to obtain an individual survival prediction with a corresponding confidence interval for a future patient from a penalized regression model, while limiting the potential overfit.The performance of the approaches was evaluated through simulation studies combining null and alternative scenarios. The methods were also illustrated in several data sets containing gene expression data in breast cancer.

The value proposition of Personalized Medicine is to deliver the “right drug, to the right patient, at the right time”. Companion diagnostics is the required tool for Personalized Medicine used to aid clinical decision making with the aim to identify patients who are most suitable for a given treatment approach and to avoid adverse effects. However, even 16 years after the first co-approval of a therapeutic drug and an associated diagnostic test (trastuzumab (Herceptin1) from Genentech and the HercepTest1 from Dako), the co-development and co-approval of drug-diagnostic pairs is a challenging task.This study has the aim to identify major challenges for diagnostics companies when developing and commercializing companion diagnostics. This is achieved by (1) a literature research and (2) an empirical case study in form of interviews with diagnostics companies. The collected data is analyzed and discussed with focus on current regulatory and reimbursement frameworks in the USA and European Union. The co-development strategies and business models of companion diagnostics developers are identified.The conclusion of this study is that the major hurdles for companion diagnostics development and commercialization are gaps in scientific evidence and lacking regulatory guidelines for co-development and clinical biomarker studies. Companion diagnostics commercialization is further challenged by poor reimbursement levels. The main strategy of diagnostics companies to address these challenges is the demonstration of a beneficial outcome for patients in form of clinical studies. Small companies with limited resources for clinical research receive funding from academic research grants, patient support groups, pharmaceutical industry, and governmental Innovation agencies.Finally the formation of a new “pharma-diagnostics” sectoral innovation system as a result of the emerging paradigm of stratified medicine has been proposed.