Dissertations / Theses on the topic 'Adaptive trial'

The question we posed at the beginning of this thesis was whether, in the presence of a clinical superiority of one of two possible treatments, it was possible to find an appropriate statistical methodology that would allow us to reach this goal. We were thus led to explore many possibilities to carry out this analysis and randomly assign patients to the two treatments, as required by the particular nature of these experiments. Specifically, we made a close examination of the methods of randomization, especially appreciating the flexibility of the adaptive responses, and could see the strengths of urn models. We started with the study of the urn for excellence, Polya's urn. Next, we analyzed some extensions and generalizations, focusing especially on two kinds of urns with random reinforcement. We exposed the results obtained throughout simulations concerning the convergence of the proportion of the best treatment, which came from the comparison of the models studied. In the end, we showed how the urn model works in a real case, comparing two treatments with continuous response in one ICU trial on Melatonin. We'll see how the properties demonstrated in theory are confirmed in practice. The project ends by giving a hint of a new adaptive model that we have started to idealize in collaboration with the team of Prof. Parmigiani and Prof. Trippa of the "Biostatistics and Computational Biology" Department, Harvard T.H. Chan School of Public Health.

Response-Adaptive (RA) designs are used to adaptively allocate patients in clinical trials. These methods have been generalized to include Covariate-Adjusted Response-Adaptive (CARA) designs, which adjust treatment assignments for a set of covariates while maintaining features of the RA designs. Challenges may arise in multi-center trials if differential treatment responses and/or effects among sites exist. We propose Site-Adjusted Response-Adaptive (SARA) approaches to account for inter-center variability in treatment response and/or effectiveness, including either a fixed site effect or both random site and treatment-by-site interaction effects to calculate conditional probabilities. These success probabilities are used to update assignment probabilities for allocating patients between treatment groups as subjects accrue. Both frequentist and Bayesian models are considered. Treatment differences could also be attributed to differences in social determinants of health (SDH) that often manifest, especially if unmeasured, as spatial heterogeneity amongst the patient population. In these cases, patient residential location can be used as a proxy for these difficult to measure SDH. We propose the Location-Adjusted Response-Adaptive (LARA) approach to account for location-based variability in both treatment response and/or effectiveness. A Bayesian low-rank kriging model will interpolate spatially-varying joint treatment random effects to calculate the conditional probabilities of success, utilizing patient outcomes, treatment assignments and residential information. We compare the proposed methods with several existing allocation strategies that ignore site for a variety of scenarios where treatment success probabilities vary.

In recent years, adaptive designs in clinical trials have been attractive due to their efficiency and flexibility. Response adaptive randomization procedures in phase II or III clinical trials are proposed to appeal ethical concerns by skewing the probability of patient assignments based on the responses obtained thus far, so that more patients will be assigned to a superior treatment group. General response-adaptive randomizations usually assume that the primary endpoint can be obtained quickly after the treatment. However, in real clinical trials, the primary outcome is delayed, making it unusable for adaptation. Therefore, we utilize surrogate and primary endpoints simultaneously to adaptively assign subjects between treatment groups for clinical trials with continuous responses. We explore two types of primary endpoints commonly used in clinical tirials: normally distributed outcome and time-to-event outcome. We establish a connection between the surrogate and primary endpoints through a Bayesian model, and then update the allocation ratio based on the accumulated data. Through simulation studies, we find that our proposed response adaptive randomization is more effective in assigning patients to better treatments as compared with equal allocation randomization and standard response adaptive randomization which is solely based on the primary endpoint.

The most common type of confirmatory trial is a randomised trial comparing the experimental treatment of interest to a control treatment. Confirmatory trials are expensive and take a lot of time in the planning, set up and recruitment of patients. Efficient methodology in clinical trial design is critical to save both time and money and allow treatments to become available to patients quickly. Often there are data available on the control treatment from a previous trial. These historical data are often used to design new trials, forming the basis of sample size calculations, but are not used in the analysis of the new trial. Incorporating historical control data into the design and analysis could potentially lead to more efficient trials. When the historical and current control data agree, incorporating historical control data could reduce the number of control patients required in the current trial and therefore the duration of the trial, or increase the precision of parameter estimates. However, when the historical and current data are inconsistent, there is a potential for biased treatment effect estimates, inflated type I error and reduced power. We propose two novel weights to assess agreement between the current and historical control data: a probability weight based on tail area probabilities; and a weight based on the equivalence of the historical and current control data parameters. For binary outcome data, agreement is assessed using the posterior distributions of the response probability in the historical and current control data. For normally distributed outcome data, agreement is assessed using the marginal posterior distributions of the difference in means and the ratio of the variances of the current and historical control data. We consider an adaptive design with an interim analysis. At the interim, the agreement between the historical and current control data is assessed using the probability or equivalence probability weight approach. The allocation ratio is adapted to randomise fewer patients to control when there is agreement and revert back to a standard trial design when there is disagreement. The final analysis is Bayesian utilising the analysis approach of the power prior with a fixed weight. The operating characteristics of the proposed design are explored and we show how the equivalence bounds can be chosen at the design stage of the current study to control the maximum inflation in type I error. We then consider a design where a treatment arm is added to an ongoing clinical trial. For many disease areas, there are often treatments in different stages of the development process. We consider the design of a two-arm parallel group trial where it is planned to add a new treatment arm during the trial. This could potentially save money, patients, time and resources. The addition of a treatment arm creates a multiple comparison problem. Dunnett (1955) proposed a design that controls the family-wise error rate when comparing multiple experimental treatments to control and determined the optimal allocation ratio. We have calculated the correlation between test statistics for the method proposed by Dunnett when a treatment arm is added during the trial and only concurrent controls are used for each treatment comparison. We propose an adaptive design where the sample size of all treatment arms are increased to control the family-wise error rate. We explore adapting the allocation ratio once the new treatment arm is added to maximise the overall power of the trial.

Adaptive clinical trials are recently gaining more attention. In this thesis, generalizations to the Biomarker-Adaptive Threshold Design (BATD) are studied and applied in the multiple sclerosis (MS) context. The BATD was originally developed for survival outcomes for Phase III clinical trials and allows researchers to both study the efficacy of treatment in the overall group and to investigate the relationship between a hypothesized predictive biomarker and the treatment effect on the primary outcome. We first introduce the original methodology and replicate the authors’ simulation studies to confirm their findings. Then, we generalize the methodology to accommodate count biomarkers and outcomes. Our interest in variables of this form is fuelled by the study of MS, where the number of relapses is a commonly used count outcome for patients with relapsing-remitting MS. Through simulation studies, we find that the BATD has increased power compared with a traditional fixed design under varying scenarios for which there exists a sensitive patient subgroup. As an illustrative example, we consider data from a previously completed trial and apply the methodology for two hypothesized markers: baseline lesion activity and the length of time that a patient has had MS. While we do not find a predictive biomarker relationship between baseline lesion activity and the number of relapses, MS duration does appear to have a predictive biomarker relationship for this dataset. In particular, we consider a randomly chosen subsample of the data for which the overall treatment effect on the outcome was insignificant. When the BATD is applied, a very significant treatment effect is detected and indicates that the effect is strongest for patients that have had MS for less than 7.8 years for this subsample. The methodology holds promise at preserving statistical power when the treatment effect is greatest in a sensitive patient subset.

A study is conducted to test the dynamics of the 33ft Wave-Adaptive Modular Vessel (WAM-V) when outfitted with different suspension systems.  Instrumented with an array of sensors, the vessel is tested with two different suspension arrangements to characterize how they affect WAM-V dynamics, and to ultimately select a suspension that is most suitable for the 33ft WAM-V and other vessels that are planned for the future.<br />Optimizing the suspension can reduce the magnitude of accelerations at the payload tray, benefiting both the operator and the payload.  Reduced accelerations can significantly improve comfort and risk of injury to the operator, while also lessening the likelihood of any damage to any sensitive cargo onboard.  The stock suspension components are characterized through in-house tests conducted at the Center for Vehicle Systems and Safety (CVeSS) at Virginia Tech (VT).  Based on the stock characterizations, new suspension components are chosen to better fit the needs of the 33ft WAM-V.<br />Sea trials are conducted with both suspension systems at the Combatant Craft Division (CCD), a division of the Naval Surface Warfare Center, Carderock Division (NSWCCD), in Norfolk, VA to quantitatively and qualitatively determine the differences between the two suspensions.  The 33ft WAM-V is instrumented with a series of accelerometers and potentiometers for measuring accelerations and displacements.  The data is analyzed for the sea trials conducted at CCD and the results of the analysis indicate that the suspension selection can significantly affect the transmission of shock and vibrations from the pontoons to the operator or payload tray.  Both suspensions are able to mitigate a significant amount of the shocks seen at the pontoons, however, the results do not definitively show which suspension is the better of the two.  This is due to the fact that each suspension is not subjected to the exact same wave conditions, and  <br />therefore the resulting suspension dynamics vary.  For instance, during a 2-foot wave event, the new suspension attenuates more shock than the stock suspension, 76% versus 71%.  However, during a 4-foot wave event, the stock suspension attenuates more shock than the new suspension, 66% versus 60%.<br />Additionally, the suspension selection can significantly influence the ride height.  The stock suspension provides a 70/30 ratio between extension and compression stroke, while the new suspension provides a 50/50 ratio.  The more balanced split between the extension and compression strokes allow for better utilizing the total available stroke for the suspension in both directions.  This significantly reduces the resulting high-g impacts since the suspension does not frequently bottom out when the vessel is subjected to a large wave.<br />It is recommended that the results of this study be extended through laboratory dynamic testing that allows for more repeatable dynamic events than sea trials in order to better establish the influence of each suspension parameter on the vessel dynamics.  Such tests will also allow for a better understanding of the dynamics of the vessel in response to various inputs at the pontoons, both subjectively (visually) and objectively (through measurements).<br>Master of Science

Les robots opèrent dans le monde réel, dans lequel essayer quelque chose prend beaucoup de temps. Pourtant, les methodes d’apprentissage par renforcement actuels (par exemple, deep reinforcement learning) nécessitent de longues périodes d’interaction pour trouver des politiques efficaces. Dans cette thèse, nous avons exploré des algorithmes qui abordent le défi de l’apprentissage par essai-erreur en quelques minutes sur des robots physiques. Nous appelons ce défi “Apprentissage par renforcement micro-data”. Dans la première contribution, nous avons proposé un nouvel algorithme d’apprentissage appelé “Reset-free Trial-and-Error” qui permet aux robots complexes de s’adapter rapidement dans des circonstances inconnues (par exemple, des dommages) tout en accomplissant leurs tâches; en particulier, un robot hexapode endommagé a retrouvé la plupart de ses capacités de marche dans un environnement avec des obstacles, et sans aucune intervention humaine. Dans la deuxième contribution, nous avons proposé un nouvel algorithme de recherche de politique “basé modèle”, appelé Black-DROPS, qui: (1) n’impose aucune contrainte à la fonction de récompense ou à la politique, (2) est aussi efficace que les algorithmes de l’état de l’art, et (3) est aussi rapide que les approches analytiques lorsque plusieurs processeurs sont disponibles. Nous avons aussi proposé Multi-DEX, une extension qui s’inspire de l’algorithme “Novelty Search” et permet de résoudre plusieurs scénarios où les récompenses sont rares. Dans la troisième contribution, nous avons introduit une nouvelle procédure d’apprentissage du modèle dans Black-DROPS qui exploite un simulateur paramétré pour permettre d’apprendre des politiques sur des systèmes avec des espaces d’état de grande taille; par exemple, cette extension a trouvé des politiques performantes pour un robot hexapode (espace d’état 48D et d’action 18D) en moins d’une minute d’interaction. Enfin, nous avons exploré comment intégrer les contraintes de sécurité, améliorer la robustesse et tirer parti des multiple a priori en optimisation bayésienne. L'objectif de la thèse était de concevoir des méthodes qui fonctionnent sur des robots physiques (pas seulement en simulation). Par conséquent, tous nos approches ont été évaluées sur au moins un robot physique. Dans l’ensemble, nous proposons des méthodes qui permettre aux robots d’être plus autonomes et de pouvoir apprendre en poignée d’essais<br>Robots have to face the real world, in which trying something might take seconds, hours, or even days. Unfortunately, the current state-of-the-art reinforcement learning algorithms (e.g., deep reinforcement learning) require big interaction times to find effective policies. In this thesis, we explored approaches that tackle the challenge of learning by trial-and-error in a few minutes on physical robots. We call this challenge “micro-data reinforcement learning”. In our first contribution, we introduced a novel learning algorithm called “Reset-free Trial-and-Error” that allows complex robots to quickly recover from unknown circumstances (e.g., damages or different terrain) while completing their tasks and taking the environment into account; in particular, a physical damaged hexapod robot recovered most of its locomotion abilities in an environment with obstacles, and without any human intervention. In our second contribution, we introduced a novel model-based reinforcement learning algorithm, called Black-DROPS that: (1) does not impose any constraint on the reward function or the policy (they are treated as black-boxes), (2) is as data-efficient as the state-of-the-art algorithm for data-efficient RL in robotics, and (3) is as fast (or faster) than analytical approaches when several cores are available. We additionally proposed Multi-DEX, a model-based policy search approach, that takes inspiration from novelty-based ideas and effectively solved several sparse reward scenarios. In our third contribution, we introduced a new model learning procedure in Black-DROPS (we call it GP-MI) that leverages parameterized black-box priors to scale up to high-dimensional systems; for instance, it found high-performing walking policies for a physical damaged hexapod robot (48D state and 18D action space) in less than 1 minute of interaction time. Finally, in the last part of the thesis, we explored a few ideas on how to incorporate safety constraints, robustness and leverage multiple priors in Bayesian optimization in order to tackle the micro-data reinforcement learning challenge. Throughout this thesis, our goal was to design algorithms that work on physical robots, and not only in simulation. Consequently, all the proposed approaches have been evaluated on at least one physical robot. Overall, this thesis aimed at providing methods and algorithms that will allow physical robots to be more autonomous and be able to learn in a handful of trials

The clinical trial is the most import study for the development of successful novel drugs. The aim of this dissertation is to develop innovative statistical methods to overcome the three main obstacles in clinical trials: (1) lengthy trial duration and inaccurate maximum tolerated dose (MTD) in phase I trials; (2) heterogeneity in drug effect when patients are given the same prescription and same dose; and (3) high failure rates of expensive phase III confirmatory trials due to the discrepancy in the endpoints adopted in phase II and III trials. Towards overcoming the first obstacle, we originally develop a hybrid design for the time-to-event dose escalation method with overdose control using a normalized equivalent toxicity score (NETS) system. This hybrid design can substantially reduce sample size, shorten study length, and estimate accurate MTD by employing a parametric model and adaptive Bayesian approach. Toward overcoming the second obstacle, we propose a new approach to incorporate patients’ characteristic using our proposed design in phase I clinical trials which considers the personalized information for patients who participant in the trials. To conquer the third obstacle, we propose a novel two-stage screening design for phase II trials whereby the endpoint of percent change in of tumor size is used in an initial screening to select potentially effective agents within a short time interval followed by a second screening stage where progression free survival is estimated to confirm the efficacy of agents. These research projects will substantially benefit both cancer patients and researchers by improving clinical trial efficiency and reducing cost and trial duration. Moreover, they are of great practical meaning since cancer medicine development is of paramount importance to human health care.

Bayesian adaptive designs are emerging as popular approach to develop adaptive clinical trials. In this dissertation, I describe the mathematical steps for computing the theoretical optimal adaptive designs in biomarker-integrated trials and in trials with survival outcomes. Section 1 discusses the optimal design in personalized medicine. The optimal design maximizes the expected trial utility given any pre-specified utility function, though the discussion here focuses on maximizing responses within a given patient horizon. This work provides absolute benchmark for the evaluation of trial designs in targeted therapy with binary treatment outcomes. While treatment efficacy can be measured by a short-term binary outcome in many phase II and phase III trials, patients' progression-free survival time is with significant importance in cancer clinical trials. However, it is often difficult to make a design adaptive to survival outcomes because of the long observation time. In Section 2, an optimal adaptive design is developed so that treatment assignment decision for later patients can be made with complete or partial survival outcomes of early patients. The design also maximizes the expected trial utility given any pre-specified utility function that is of clinical importance. In this section, the focus is on maximizing the expected progression-free survival time. Both Sections1 and 2 include examples of comparing adaptive designs, such as the bayesian adaptive randomization and the play-the-winner rule, in terms of the expected trial utility with respect to the best achievable result. In Section 3, a simulation-based p-value is proposed and can be used to conduct frequentist analysis of Bayesian adaptive clinical trials. The optimal Bayesian design is compared to the equal randomization design in terms of the Type I error and the statistical power. With a fixed trial size and Type I error, the power of the equal randomization design depends on the difference in treatment efficacy, meanwhile the power of the optimal Bayesian design also depends on the size of the patient horizon.

The pharmaceutical industry is currently facing an industry wide problem of high attrition rates for new compounds and rising development costs. As a result of this, there is an emphasis on making the development process more ecient. By learning more about new compounds in the early stages of development, the aim is to stop ineective compounds earlier and improve dose selection for compounds that progress to phase III. One approach to this is to use adaptive designs. The focus of this thesis is on response adaptive designs within phase IIb dose-finding studies. We explore adapting the subject allocations based on accrued data, with the intention of focusing the allocation on the interesting parts of the curve and/or the best dose for phase III. In this thesis we have used simulation studies to assess the operational characteristics of a number of response adaptive designs. We found that there were consistent gains to be made by adapting when we were relatively cautious in our method of adaptation. That is, the adaptive method has the opportunity to alter the subject allocation when there is a clear signal in the data, but maintains roughly equal allocation when there is a lot of variability in the data. When we used adaptive designs that were geared to randomising subjects to a few doses, the results were more varied. In some cases the adaptation led to gains in efficacy whilst in others it was detrimental. One of the key aims of a phase IIb dose-finding study is to identify a dose to take forward into phase III. In the final chapter, we show that the way in which we choose the dose for phase III affects the expected gain, and so begin to consider how we can optimise the decision making process.

L’objectif des essais de phase I en oncologie est d’identifier la dose maximale tolérée (DMT). Le schéma « 3+3 » nécessite d’interrompre les inclusions en attendant l’évaluation d’une cohorte de trois patients pour définir la dose à attribuer aux patients suivants. Les investigateurs d’oncologie pédiatrique ont proposé l’adaptation Rolling 6 pour éviter cette suspension temporaire des inclusions. Dans une étude de simulation, nous avons montré qu’un schéma adaptatif avec attribution des doses basées sur un modèle statistique permettait de pallier ce problème, et identifiait plus fréquemment la DMT. Néanmoins ces trois schémas restent limités pour identifier la DMT, notamment du fait que le critère de jugement est un critère binaire, la survenue de toxicité dose-limitante sur un cycle de traitement. Nous avons proposé un nouveau schéma adaptatif utilisant les données ordinales répétées de toxicité sur l’ensemble des cycles de traitement. La dose à identifier est celle associée au taux de toxicité grave maximal par cycle que l’on juge tolérable. Le grade maximal de toxicité par cycle de traitement, en 3 catégories (grave / modéré / nul), a été modélisé par le modèle mixte à cotes proportionnelles. Le modèle est performant à la fois pour détecter un effet cumulé dans le temps et améliore l’identification de la dose cible, sans risque majoré de toxicité, et sans rallonger la durée des essais. Nous avons aussi étudié l’intérêt de ce modèle ordinal par rapport à un modèle logistique mixte plus parcimonieux. Ces modèles pour données longitudinales devraient être plus souvent utilisés pour l’analyse des essais de phase I étant donné leur pertinence et la faisabilité de leur implémentation<br>Phase I dose-finding trials aim at identifying the maximum tolerated dose (MTD). The “3+3” design requires an interruption of enrolment while the evaluation of the previous three patients is pending. In pediatric oncology, investigators proposed the Rolling 6 design to allow for a more continuous enrollment. In a simulation study, we showed that an adaptive dose-finding design, with dose allocation guided by a statistical model not only minimizes accrual suspension as with the rolling 6, and but also led to identify more frequently the MTD. However, the performance of these designs in terms of correct identification of the MTD is limited by the binomial variability of the main outcome: the occurrence of dose-limiting toxicity over the first cycle of treatment. We have then proposed a new adaptive design using repeated ordinal data of toxicities experienced during all the cycles of treatment. We aim at identifying the dose associated with a specified tolerable probability of severe toxicity per cycle. The outcome was expressed as the worst toxicity experienced, in three categories (severe / moderate / no toxicity), repeated at each treatment cycle. It was modeled through a proportional odds mixed model. This model enables to seek for cumulated toxicity with time, and to increase the ability to identify the targeted dose, with no increased risk of toxicity, and without delaying study completion. We also compared this ordinal model to a more parsimonious logistic mixed model.Because of their applicability and efficiency, those models for longitudinal data should be more often used in phase I dose-finding trials

The key original contribution of this work is the use of a Bayes optimisation framework for the decision made at the interim analysis of Adaptive Enrichment trials. Adaptive Enrichment designs make efficient use of pre-identified patient sub-populations. They begin by recruiting from all eligible patients, then at a pre-planned interim analysis select which sub-populations will be recruited from for the remainder of the sample. We ensure strong control of the Familywise Error Rate whichever sub-populations are selected by constructing an overall hypothesis testing structure using both closed testing procedures and combination tests. This allows us to make interim decision by any method we choose. We find the Bayes optimal decision, recruiting the remainder of the trial to optimise the Bayes expected gain of the trial. We compare the Bayes optimal Adaptive Enrichment trials with fixed sampling designs to understand the overall advantage of using adaptive trials. This optimisation framework is very flexible, we evaluate the performance of Bayes optimal Adaptive Enrichment designs for different forms of data: delayed responses, longitudinal analysis and discuss the extension of these methods to survival data. Through this we see that although the information at the interim analysis is reduced the adaptive trials still offer some benefit. Additionally we investigate what may happen when we alter the pattern of recruitment of the Adaptive Enrichment trials, showing that adaptation may be useful in a broad range of scenarios.

A recently developed group-sequential response-adaptive design to compare two treatments with immediate normally distributed responses and known variances is considered. The power function of the test is the same as that under non-adaptive sampling, and significant decreases in the inferior treatment number can be achieved with only minor increases in the average sample number. Reasonably accurate corrected confidence intervals for both the treatment mean difference and the individual means are obtained by constructing approximately pivotal quantities. An approximation to the bias of the maximum likelihood estimator of the treatment mean difference is also studied. When the variances of the response variables are unknown, inaccurate estimates of these can affect the Type II error rate considerably. A new modified version of an existing sample size re-estimation method is developed for group-sequential response-adaptive designs for normal data with unknown variances. The principal modifications involve updating the required sample size at each interim analysis and calculating the test statistic based on current estimates of the variances. Simulation is used to compare the performance of this test with modified versions of two other tests from the recent literature. The power is shown to be more accurately maintained in the new test. An analogous group-sequential response-adaptive design to compare two treatments with immediate dichotomous responses is then developed. Since the variances of the response variables are unknown in binary response trials, due to their dependence on the unknown success probabilities, the new sample size re-estimation method is incorporated into the design. Two parameters of interest are considered, the log odds ratio and the simple difference between the probabilities of success. Three adaptive urn models are studied and their properties are compared to a sequential maximum likelihood estimation rule that minimises the expected number of treatment failures. Simulation results favour the drop-the-loser rule

The primary goal of clinical trials is to collect enough scientific evidence for a new intervention. Despite the widespread use of equal randomization in clinical trials, response-adaptive randomization has attracted considerable interest in terms of ethical concerns. In this thesis, delayed response problems and innovative designs for cytostatic agents in oncology clinical trials are studied. There is typically a prerun of equal randomization before the implementation of response-adaptive randomization, while it is often not clear how many subjects are needed in this prephase, and in practice an arbitrary number of patients are allocated in this equal randomization stage. In addition, real-time response-adaptive randomization often requires patient response to be immediately available after the treatment, while clinical response, such as tumor shrinkage, may take a relatively long period of time to exhibit. In the first part of the thesis, a nonparametric fractional model and a parametric optimal allocation scheme are developed to tackle the common problem caused by delayed response. In addition, a two-stage procedure to achieve a balance between power and the number of responders is investigated, which is equipped with a likelihood ratio test before skewing the allocation probability toward a better treatment. The operating characteristics of the two-stage designs are evaluated through extensive simulation studies and an HIV clinical trial is used for illustration. Numerical results show that the proposed method satisfactorily resolves the issues involved in response-adaptive randomization and delayed response. In phase I clinical trials with cytostatic agents, toxicity endpoints, as well as efficacy effects, should be taken into consideration for identifying the optimal biological dose (OBD). In the second part of the thesis, a two-stage Bayesian mixture modeling approach is developed, which first locates the maximum tolerated dose (MTD) through a mixture of parametric and nonparametric models, and then determines the most efficacious dose using Bayesian adaptive randomization among multiple candidate models. In the first stage searching for the MTD, a beta-binomial model in conjunction with a probit model as a mixture modeling approach is studied, and decisions are made based on the model that better fits the toxicity data. The model fitting adequacy is measured by the deviance information criterion and the posterior model probability. In the second stage searching for the OBD, the assumption that efficacy monotonically increases with the dose is abandoned and, instead, all the possibilities that each dose could have the highest efficacy effect are enumerated so that the dose-efficacy curve can be increasing, decreasing, or umbrella-shape. Simulation studies show the advantages of the proposed mixture modeling approach for pinpointing the MTD and OBD, and demonstrate its satisfactory performance with cytostatic agents.<br>published_or_final_version<br>Statistics and Actuarial Science<br>Master<br>Master of Philosophy

Adaptive clinical trials for new drugs or treatment options promise substantial benefits to both the pharmaceutical industry and the patients, but complicate resource allocation decisions. In this dissertation, we focus on sequential adaptive clinical trials with binary response, which allow for early termination of drug testing for benefit or futility at interim analysis points. The option to stop the trial early enables the trial sponsor to mitigate investment risks on ineffective drugs, and to shorten the development time line of effective drugs, hence reducing expenditures and expediting patient access to these new therapies. In this setting, decision makers need to determine a testing schedule, or the number of patients to recruit at each interim analysis point, and stopping criteria that inform their decision to continue or stop the trial, considering performance measures that include drug misclassification risk, time-to-market, and expected profit. In the first manuscript, we model current practices of sequential adaptive trials, so as to quantify the magnitude of drug misclassification risk. Towards this end, we build a simulation model to realistically represent the current decision-making process, including the utilization of the triangular test, a widely implemented sequential methodology. We find that current practices lead to a high risk of incorrectly terminating the development of an effective drug, thus, to unrecoverable expenses for the sponsor, and unfulfilled patient needs. In the second manuscript, we study the sequential resource allocation decision, in terms of a testing schedule and stopping criteria, so as to quantify the impact of interim analyses on the aforementioned performance measures. Towards this end, we build a stochastic dynamic programming model, integrated with a Bayesian learning framework for updating the drug’s estimated efficacy. The resource allocation decision is characterized by endogenous uncertainty, and a trade-off between the incentive to establish that the drug is effective early on (exploitation), due to a time-decreasing market revenue, and the benefit from collecting some information on the drug’s efficacy prior to committing a large budget (exploration). We derive important structural properties of an optimal resource allocation strategy and perform a numerical study based on realistic data, and show that sequential adaptive trials with interim analyses substantially outperform traditional trials. Finally, the third manuscript integrates the first two models, and studies the benefits of an optimal resource allocation decision over current practices. Our findings indicate that our optimal testing schedules outperform different types of fixed testing schedules under both perfect and imperfect information.<br>Ph. D.

This thesis is motivated by an adaptive design which was developed to inoculate healthy volunteers with nontypeable Haemophilus influenzae. The goal was to estimate the doses at which 50% (HCD50) and 90% (HCD90) of subjects became colonized. A fifteen-subject study was designed in two stages, with the first six subjects allocated sequentially. The design was chosen based on scientific and statistical arguments, however, due to limited time, heuristic decisions were made for expedience. This design and a number of alternative designs are evaluated in depth by simulation, under both Bayesian and frequentist criteria. In this thesis, Bayesian myopic strategies with one-step- , two-step- and three-step-look-ahead procedures are investigated. The optimal design is defined as the one with minimum expected loss where the loss is the sum of the posterior variance of the HCD50 and HCD90. The higher the expected loss, the worse the design. Designs using different prior distribution are examined. In addition, the toxicity-response relationship can also be incorporated in selecting the optimal design. A new model considering both colonization (efficacy) and adverse event (toxicity) is proposed, and design procedures developed. Furthermore, restrictions on the probability of toxicity are implemented. The results from simulations show that it is beneficial to look more steps ahead in determining the optimal dose although the benefit may not be large. The is true for both univariate (colonization) and bivariate (colonization and toxicity) models. For the bivariate model, as the restriction becomes more conservative (the probability of toxicity is constrained to be smaller), the expected loss becomes larger and early stopping may occur. Non-sequential designs are also found and examined using D and A criteria for optimal design. The expected loss is computed to evaluate the designs and to compare with sequential strategies. From the simulation results, it shows that using sequential design strategies does improve the performance of the design compared to using non-sequential strategies, and the improvement may be large.

Introduction: Adaptive designs (ADs) are underused, particularly in publicly funded confirmatory trials, despite their promising benefits and methodological prominence given in the statistical literature. Research Question: This thesis investigates why ADs are underused in the publicly funded setting, explores facilitators, and proposes recommendations to improve their appropriate use. Methods: Confirmatory ADs are reviewed from a statistical and practical perspective. Cross-disciplinary key stakeholders are then interviewed to explore roadblocks to the use of ADs. Based on the interview findings, follow-up quantitative surveys are undertaken to explore wider perceptions on barriers, concerns, and facilitators aimed to generalise the findings. The surveys targeted CTUs (Clinical Trials Units), private sector organisations, and Public Funders in the UK. In view of some of the findings, case studies of applied confirmatory ADs are reviewed to highlight their scope and characteristic, and to investigate the state of reporting of the most common AD. The design and implementation of selected ADs is demonstrated using retrospective and prospective planned case studies. Lessons learned are highlighted to enhance the design of future trials of similar characteristics. Results: The main barriers to the use of ADs include the lack of funding support accessible to UK CTUs to aid their design; limited practical knowledge; preference for traditional mainstream designs; difficulties in marketing ADs to key stakeholders; limited time to support ADs relative to other competing priorities; lack of applied training; and insufficient access to case studies of undertaken ADs, which would facilitate practical learning and successful implementation. Researchers’ inadequate description of AD-related aspects (such as rationale, scope, and decision-making criteria to guide the planned AD) in grant proposals was viewed among the major obstacles by Public Funders. Suboptimal reporting of the design and conduct of undertaken ADs appears to influence concerns about their robustness in decision-making and credibility to change practice. Conclusions: Most obstacles appear connected to a lack of practical implementation knowledge and applied training, and limited access to adequately reported case studies to facilitate practical learning. Assurance of scientific rigour through transparent adequate reporting is paramount to the credibility of findings from adaptive trials. There is a need for a consensus guidance document on ADs and an AD-tailored CONSORT statement to enhance their reporting and conduct. This thesis provides detailed recommendations to improve the appropriate use of ADs and areas for future related research.

Thesis: M.B.A., Massachusetts Institute of Technology, Sloan School of Management, 2014.<br>Cataloged from PDF version of thesis.<br>Includes bibliographical references (pages 69-72).<br>There is great competition for clinical research funding. This is in part due to the National Institute of Health's reduced budget to support such initiatives. It has resulted in a growing trend for clinical research to use adaptive design models to accelerate clinical trials and at the same time reduce overall cost. Although such models have existed for several years, the pace of adoption remains slow, especially for early-stage clinical research. Through a review of relevant literature and interviews with industry experts, this thesis explores the barriers that inhibit the adoption of adaptive design of clinical trials. Reasons uncovered include: a lack of novel funding mechanisms, regulatory uncertainty, logistical difficulties, overly technical communications, a lack of collaboration among stakeholders, and an inability to recruit and retain patients. Then follows a series of possible solutions - some already functioning, others possible - for each of the barriers. This research found that unless efforts are devoted to addressing these underlying barriers, the widespread adoption of adaptive designs for clinical trials will not occur. The thesis concludes with recommendations and suggestions for future research.<br>by Ayesha N. Khalid.<br>M.B.A.

A method of designing early clinical trials is developed for finding an optimum dose level of a new drug to be recommended for use in later phases. During the trial, the efficacious doses are allocated to the patients more often and those with a high probability of toxicity are less likely to be chosen. The method proposed is adaptive in the sense that the statistical models are updated after the data from each cohort of patients are collected and the dose level is adjusted at each stage based on the current data. Two classes of designs are presented. Although both are for efficacy and toxicity responses, one of them also considers pharmacokinetic information. The dose optimisation criteria are based on the probability of success and on the determinant of the Fisher information matrix for estimation of the dose-response parameters. They can be constrained by both acceptable levels of the probability of toxicity and desirable levels of the area under the concentration curve or the maximum concentration. The method presented is general and can be applied to various dose-response and pharmacokinetic models. To illustrate the methodology, it is applied to two different classes of models. In both cases, the pharmacokinetic model incorporates the population variability by making appropriate assumptions about the model parameters, while the dose responses are assumed to be either trinomial or bivariate binomial. Various design properties of the method are examined by simulation studies. Efficiency measures and the sensitivity of the designs to the assumed prior parameter values are presented. All of the computations are conducted in R, where the D- v optimal sampling time points are obtained by using the package PFIM. The results show that the proposed adaptive method works well and could be appropriate as a seamless phase IB/IIA trial design.

This thesis deals with sequential and adaptive methods for clinical trials, and how such methods can be used to achieve efficient clinical trial designs. The efficiency gains that can be achieved through non-adaptive group sequential methods are well established, while the newer adaptive methods seek to combine the best of the classical group sequential framework with an approach that gives increased flexibility. Our results show that the adaptive methods can provide some additional efficiency, as well as increased possibilities to respond to new internal and external information. Care is however needed when applying adaptive methods. While sub-optimal rules for adaptation can lead to inefficiencies, the logistical challenges can also be considerable. Efficient non-adaptive group sequential designs are often easier to implement in practice, and have for the cases we have considered been quite competitive in terms of efficiency. The four problems that are presented in this thesis are very relevant to how clinical trials are run in practice. The solutions that we present are either new approaches to problems that have not previously been solved, or methods that are more efficient than the ones currently available in the literature. Several challenging optimisation problems are solved through numerical computations. The optimal designs that are achieved can be used to benchmark new methods proposed in this thesis as well as methods available in the statistical literature. The problem that is solved in Chapter 5 can be viewed as a natural extension to the other problems. It brings together methods that we have used to the design of individual trials, to solve the more complex problem of designing a sequence of trials that are the core part of a clinical development program. The expected utility that is maximised is motivated by how the development of new medicines works in practice.

This master thesis deals with the topic of new employee orientation and onboarding process. The emphasis is placed on explanation of how the onboarding process and other human resource management activities affect the overall economic performance of a company and achieving its goals. Recent surveys and aspects of the current labor market are taken into account in order to highlight the increasing importance of effectively onboarding new employees. The practical part analyzes the onboarding process in a chosen international company. Based on the outcomes of the research a new proposal to optimize the existing process is introduced.

Response adaptive randomization represents a major advance in clinical trial methodology that helps balance the benefits of the collective and the benefits of the individual and improves efficiency without undermining the validity and integrity of the clinical research. Response adaptive designs use information so far accumulated from the trial to modify the randomization procedure and deliberately bias treatment allocation in order to assign more patients to the potentially better treatment. No attention has been paid to incorporating the problem of errors-in-variables in adaptive clinical trials. In this work, some important issues and methods of response adaptive design of clinical trials in the presence of mismeasurement are examined. We formulate response adaptive designs when the dichotomous response may be misclassified. We consider the optimal allocations under various objectives, investigate the asymptotically best response adaptive randomization procedure, and discuss effects of misclassification on the optimal allocation. We derive explicit expressions for the variance-penalized criterion with misclassified binary responses and propose a new target proportion of treatment allocation under the criterion. A real-life clinical trial and some related simulation results are also presented.

Statistics<br>Ph.D.<br>Clinical trials are major and costly undertakings for researchers. Planning a clinical trial involves careful selection of the primary and secondary efficacy endpoints. The 2010 draft FDA guidance on adaptive designs acknowledges possible study design modifications, such as selection and/or order of secondary endpoints, in addition to sample size re-estimation. It is essential for the integrity of a double-blind clinical trial that individual treatment allocation of patients remains unknown. Methods have been proposed for re-estimating the sample size of clinical trials, without unblinding treatment arms, for both categorical and continuous outcomes. Procedures that allow a blinded estimation of the treatment effect, using knowledge of trial operational characteristics, have been suggested in the literature. Clinical trials are designed to evaluate effects of one or more treatments on multiple primary and secondary endpoints. The multiplicity issues when there is more than one endpoint require careful consideration for controlling the Type I error rate. A wide variety of multiplicity approaches are available to ensure that the probability of making a Type I error is controlled within acceptable pre-specified bounds. The widely used fixed sequence gate-keeping procedures require prospective ordering of null hypotheses for secondary endpoints. This prospective ordering is often based on a number of untested assumptions about expected treatment differences, the assumed population variance, and estimated dropout rates. We wish to update the ordering of the null hypotheses based on estimating standardized treatment effects. We show how to do so while the study is ongoing, without unblinding the treatments, without losing the validity of the testing procedure, and with maintaining the integrity of the trial. Our simulations show that we can reliably order the standardized treatment effect also known as signal-to-noise ratio, even though we are unable to estimate the unstandardized treatment effect. In order to estimate treatment difference in a blinded setting, we must define a latent variable substituting for the unknown treatment assignment. Approaches that employ the EM algorithm to estimate treatment differences in blinded settings do not provide reliable conclusions about ordering the null hypotheses. We developed Bayesian approaches that enable us to order secondary null hypotheses. These approaches are based on posterior estimation of signal-to-noise ratios. We demonstrate with simulation studies that our Bayesian algorithms perform better than existing EM algorithm counterparts for ordering effect sizes. Introducing informative priors for the latent variables, in settings where the EM algorithm has been used, typically improves the accuracy of parameter estimation in effect size ordering. We illustrate our method with a secondary analysis of a longitudinal study of depression.<br>Temple University--Theses

Phase I clinical trials in oncology aim to evaluate the toxicity risk of new therapies and identify a safe but also effective dose for future studies. Traditional Phase I trials of chemotherapies focus on estimating the maximum tolerated dose (MTD). The rationale for finding the MTD is that better therapeutic effects are expected at higher dose levels as long as the risk of severe toxicity is acceptable. With the advent of a new generation of cancer treatments such as the molecularly targeted agents (MTAs) and immunotherapies, higher dose levels no longer guarantee increased therapeutic effects, and the focus has shifted to estimating the optimal biological dose (OBD). The OBD is a dose level with the highest biologic activity with acceptable toxicity. The search for OBD requires joint evaluation of toxicity and efficacy. Although several seamleass phase I/II designs have been published in recent years, there is not a consensus regarding an optimal design and further improvement is needed for some designs to be widely used in practice. In this dissertation, we propose a modification to an existing seamless phase I/II design by Wages and Tait (2015) for locating the OBD based on binary outcomes, and extend it to time to event (TITE) endpoints. While the original design showed promising results, we hypothesized that performance could be improved by replacing the original adaptive randomization stage with a different randomization strategy. We proposed to calculate dose assigning probabilities by averaging all candidate models that fit the observed data reasonably well, as opposed to the original design that based all calculations on one best-fit model. We proposed three different strategies to select and average among candidate models, and simulations are used to compare the proposed strategies to the original design. Under most scenarios, one of the proposed strategies allocates more patients to the optimal dose while improving accuracy in selecting the final optimal dose without increasing the overall risk of toxicity. We further extend this design to TITE endpoints to address a potential issue of delayed outcomes. The original design is most appropriate when both toxicity and efficacy outcomes can be observed shortly after the treatment, but delayed outcomes are common, especially for efficacy endpoints. The motivating example for this TITE extension is a Phase I/II study evaluating optimal dosing of all-trans retinoic acid (ATRA) in combination with a fixed dose of daratumumab in the treatment of relapsed or refractory multiple myeloma. The toxicity endpoint is observed in one cycle of therapy (i.e., 4 weeks) while the efficacy endpoint is assessed after 8 weeks of treatment. The difference in endpoint observation windows causes logistical challenges in conducting the trial, since it is not acceptable in practice to wait until both outcomes for each participant have been observed before sequentially assigning the dose of a newly eligible participant. The result would be a delay in treatment for patients and undesirably long trial duration. To address this issue, we generalize the time-to-event continual reassessment method (TITE-CRM) to bivariate outcomes with potentially non-monotonic dose-efficacy relationship. Simulation studies show that the proposed TITE design maintains similar probability in selecting the correct OBD comparing to the binary original design, but the number of patients treated at the OBD decreases as the rate of enrollment increases. We also develop an R package for the proposed methods and document the R functions used in this research. The functions in this R package assist implementation of the proposed randomization strategy and design. The input and output format of these functions follow similar formatting of existing R packages such as "dfcrm" or "pocrm" to allow direct comparison of results. Input parameters include efficacy skeletons, prior distribution of any model parameters, escalation restrictions, design method, and observed data. Output includes recommended dose level for the next patient, MTD, estimated model parameters, and estimated probabilities of each set of skeletons. Simulation functions are included in this R package so that the proposed methods can be used to design a trial based on certain parameters and assess performance. Parameters of these scenarios include total sample size, true dose-toxicity relationship, true dose-efficacy relationship, patient recruit rate, delay in toxicity and efficacy responses.

Thesis (PhD)--Stellenbosch University, 2015<br>ENGLISH ABSTRACT : In 2010, young people aged 15–24 years accounted for 42% of new HIV infections globally. In 2009, about five million (10%) of the total South African population was estimated to be aged 15–19 years. Current South African national sero-prevalence data estimate the prevalence of HIV to be 5.6% and 0.7% among adolescent girls and boys aged 15–19 years, respectively. HIV infections are mainly transmitted via sexual transmission. Adolescent sexuality is multi-faceted and influenced at multiple levels. In preparing to enroll adolescents in future biomedical HIV prevention trials, particularly prophylactic HIV vaccine trials, it is critical to provide counseling services appropriate to their needs. At the time of writing, there was no developed psychosocial intervention in South Africa for use among adolescent vaccine trial participants. Thus, the aim of the present study is to adapt and pilot-test a psychosocial intervention, namely, the Centers for Disease Control and Prevention (CDC) risk reduction counseling intervention of Project Respect, an intervention tasked at being developmentally and contextually appropriate among potential adolescent participants in HIV biomedical trials in the future. To achieve this overall aim, I qualitatively explored adolescent sexuality and risk factors for HIV among a diverse sample of participants aged 16–18 from Soweto. Thereafter, I developed a composite HIV risk scale in order to measure the variance in HIV risk among the sample of adolescents studied. The study followed a two-phased, mixed method research design and was informed by ecological systems theory and integrative model of behavioral prediction. The aim of Phase 1, split into phases 1a and b, was to conduct focus group discussions (FGDs) and to undertake a cross-sectional survey, respectively, to determine psychological (for example, self-esteem and depression), behavioral (specifically, sexual behavior) and social (specifically, social support, parent-adolescent communication) contexts that placed adolescents at risk for HIV infection. Phase 1a was qualitative, with data collected via nine FGDs: three involved parents of adolescents, four involved adolescents aged 16–18 years and two counselors. Nine key themes related to adolescent sexuality and risks for HIV acquisition were identified, namely: (1) dating during adolescence; (2) adolescent girls dating older men; (3) condom use amongst adolescents; (4) teenage pregnancies; (5) views about homosexuality; (6) parent-adolescent communication about sexual health; (7) the role of the media; (8) discipline and perceived government influence; and (9) group sex events. Phase 1b was quantitative and the data were collected via a cross-sectional survey to investigate the variance of risk for HIV. For Phase 1b, the sample consisted of 506 adolescents with a mean age of 17 years (interquartile range [IQR]: 16–18). More than half the participants were female (59%, n = 298). I used a three-step hierarchical multiple regression model to investigate the variance in risk for HIV. In step 3, the only significant predictors were “ever threatened to have sex” and “ever forced to have sex”, the combination of which explained 14% (R2 = 0.14; F (12, 236) = 3.14, p = 0.00). Depression and parentadolescent communication were added to steps 2 and 3, respectively, with both variables insignificant in these models. In Phase 2, I adapted and pilot tested the CDC risk reduction counseling intervention. The intervention was intended to be developmentally and contextually appropriate among adolescents from Soweto aged 16–18 years, viewed as potential participants in future HIV biomedical trials. Participants in Phase 2 were aged 16–18 years; the sample was mainly female (52%, n = 11) and most (91%, n = 19) were secondary school learners in grades 8 to 12. Participants provided feedback about their experiences of the adapted counseling intervention through in-depth interviews. I identified three main themes in this regard, namely: benefits of HIV testing services, reasons for seeking counseling and HIV testing services, and participants’ evaluation of the study visits and counseling sessions. The adapted CDC risk reduction counseling intervention was found to be acceptable with favorable outcomes for those adolescents who participated in the piloting phase. This study adds to the literature on risks for HIV among adolescents in Soweto, South Africa, by considering multiple levels of influence. Reaching a more complete understanding of ecological factors contributing to sexual risk behaviors among adolescents in the pilot-study enabled the development of a tailored counseling intervention. The findings showed the adapted CDC risk reduction counseling intervention to be feasible and acceptable among adolescents likely to be participants and eligible to participate in future HIV biomedical prevention trials. Thus, this study provides a much needed risk reduction counseling intervention that can be used among adolescents, an age group likely to participate in future HIV vaccine prevention research.<br>AFRIKAANSE OPSOMMING : In 2010 het jongmense tussen die ouderdomme van 15 en 24 jaar 42% van nuwe MIV-infeksies wêreldwyd uitgemaak. In 2009 was omtrent 5 miljoen mense (10%) van die Suid-Afrikaanse bevolking tussen 15 en 19 jaar oud. Volgens data oor die huidige Suid-Afrikaanse nasionale sero-voorkoms, word die voorkoms van MIV onderskeidelik op 5.6% en 0.7% onder tienermeisies en -seuns tussen die ouderdomme van 15 tot 19 jaar beraam. MIV-infeksies word hoofsaaklik deur seks oorgedra. Adolessente seksualiteit het baie fasette en word op verskeie vlakke beïnvloed. Ter voorbereiding van die werwing van adolessente vir toekomstige biomediese proewe, veral proewe oor profilaktiese MIVentstowwe, is dit van kritiese belang dat beradingsdienste verskaf word wat geskik is vir hul behoeftes. Op die tydstip wat hierdie tesis geskryf is, het daar nog geen psigososiale intervensie in Suid-Afrika bestaan vir gebruik onder adolessente deelnemers aan entstofproewe nie. Daarom is die doel van hierdie studie om ʼn psigososiale intervensie ‒ die Centers for Disease Control and Prevention (CDC) se Projek Respek, ʼn beradingsintervensie vir die vermindering van risiko ‒ aan te pas en met ʼn loodsprojek te toets. Hierdie intervensie is geskik vir die ontwikkelings- en kontekstuele vlak van adolessente deelnemers aan toekomstige MIV- biomediese proewe. Ten einde hierdie oorkoepelende doelwit te bereik, het ek adolessente seksualiteit en die risikofaktore vir MIV onder ʼn diverse steekproef deelnemers tussen die ouderdomme van 16 en 18 jaar van Soweto kwalitatief ondersoek. Daarna het ek ʼn saamgestelde MIV-risikoskaal ontwikkel om die variansie van MIV-risiko onder die groep adolessente te meet. Die studie se navorsingsontwerp het uit twee fases en gemengde metodes bestaan, en is gebaseer op ekologiesestelsel-teorie en die integrerende gedragsvoorspellingsmodel. Die doel van fase 1, wat in fases 1a en 1b verdeel is, was om onderskeidelik fokusgroepbesprekings te hou en om ʼn deursnitopname te doen om die sielkundige kontekste (byvoorbeeld elemente van selfbeeld en depressie), gedragskontekste (spesifiek seksuele gedrag) en sosiale kontekste (spesifiek sosiale ondersteuning en ouer-adolessent-kommunikasie) te bepaal waarin adolessente die risiko loop om MIV-infeksie op te doen. Fase 1a was kwalitatief en data is deur middel van nege fokusgroepbesprekings ingesamel: drie met die ouers van adolessente, vier met adolessente tussen 16 en 18 jaar oud en twee met beraders. Nege sleuteltemas is geïdentifiseer wat verband hou met adolessente seksualiteit en risiko’s om MIV op te doen: (1) verhoudings tydens adolessensie, (2) tienermeisies wat verhoudings met ouer mans het, (3) die gebruik van kondome onder adolessente, (4) tienerswangerskappe, (5) sienings oor homoseksualiteit, (6) ouer-adolessent-kommunikasie oor seksuele gesondheid, (7) die rol van die media, (8) dissipline en die ervaarde regeringsinvloed en (9) groepseksgeleenthede. Fase 1b was kwantitatief en data is deur middel van ’n deursnitopname ingesamel om die variansie van risiko vir MIV te ondersoek. Vir Fase 1b het die steekproef bestaan uit 506 adolessente met ’n gemiddelde ouderdom van 17 jaar (interkwartielwydte [IKW]: 16–18). Meer as die helfte van die deelnemers was vroulik (59%, n = 298). Ek het ’n hiërargiese meervoudige regressiemodel met drie stappe gebruik om die variansie van risiko vir MIV te ondersoek. Die enigste beduidende voorspellers in stap 3 was “ooit gedreig om seks te hê” en “ooit geforseer om seks te hê”. Die kombinasie hiervan het 14% (R2 = 0.14; F (12, 236) = 3.14, p = 0.00) verklaar. Depressie en oueradolessent- kommunikasie is onderskeidelik in stappe 2 en 3 bygevoeg, en albei veranderlikes was onbeduidend in hierdie modelle. In Fase 2 het ek die CDC se intervensie vir die verlaging van risiko aangepas en met ’n loodsprojek getoets. Die intervensie was bedoel om geskik te wees vir die ontwikkelings- en kontekstuele vlakke van 16- tot 18-jarige adolessente van Soweto wat beskou is as potensiële deelnemers aan toekomstige MIV- biomediese proewe. Deelnemers in Fase 2 was 16 tot 18 jaar oud, die steekproef was hoofsaaklik vroulik (52%, n = 11) en die meeste van die deelnemers (91%, n = 19) was in grade 8 tot 12 op hoërskool. Deelnemers het tydens indringende onderhoude terugvoering oor hulle ervarings van die aangepaste beradingsintervensie verskaf. Ek het drie hooftemas in hierdie verband geïdentifiseer, wat die volgende insluit: voordele van MIV-toetsingsdienste, redes waarom berading en MIV-toetsingsdienste verlang word, en die deelnemers se evaluering van die studiebesoeke en beradingsessies. Daar is bevind dat die aangepaste beradingsintervensie van die CDC aanvaarbaar was en gunstige uitkomste gelewer het vir die adolessente wat aan die loodsfase deelgeneem het. Hierdie studie dra by tot die literatuur oor MIV-risiko’s vir adolessente in Soweto, Suid-Afrika, deur meervoudige invloedsvlakke te oorweeg. Die feit dat ’n meer volledige begrip tydens die loodsondersoek verkry is van die interaksie van die ekologiese faktore wat tot seksuele risikogedrag onder adolessente bydra, het die ontwikkeling van ʼn doelgemaakte intervensie deur berading moontlik gemaak. Die bevindings het getoon dat die aangepaste beradingsintervensie van die CDC lewensvatbaar en aanvaarbaar is vir gebruik onder adolessente wat waarskynlik geskikte deelnemers aan toekomstige biomediese proewe oor MIV-voorkoming kan wees. Hierdie studie verskaf dus ʼn noodsaaklike beradingsintervensie om die MIV-risiko onder adolessente ‒ ʼn ouderdomsgroep wat waarskynlik aan toekomstige biomediese navorsing oor MIV-voorkoming sal deelneem ‒ te verminder.

L'hétérogénéité phénotypique observée dans des populations de cellules isogéniques est une cause immédiate de résistance non-génétique aux anticancéreux qui reste difficile à définir en raison de sa nature transitoire. Dans une première partie, nous avons étudié l’origine moléculaire de l'hétérogénéité phénotypique de réponse au TNF-related apoptosis-inducing ligand (TRAIL). Pour cela nous avons développé une méthode permettant d’isoler les cellules résistantes avant qu’elles n’engagent une induction génétique par le traitement, et les cellules sensibles avant qu’elles ne détériorent leur contenu lors de la mort cellulaire. La méthode emploie trois technologies en cellule unique : une mesure prédictive de réponse cellulaire par vidéo-microscopie, une isolation de chaque cellule in situ par capture laser, suivi d’un profilage transcriptomique. La prédiction de réponse thérapeutique permet alors d’isoler et profiler toutes les cellules, quelle que soit l’issue de leur réponse pharmacologique tardive (mort cellulaire pour les sensibles), et ainsi nous a permis d’identifier les gènes à l’origine de la différence d’efficacité de TRAIL dans une population de cellules clonales. Puis dans un second projet, nous avons étudié la résistance transitoire induite lors de traitement par TRAIL. Nous avons pu mettre en évidence le rôle de TAK-1 dans la mise en place de cette résistance transitoire. Il a été décrit que la voie apoptotique déclenchée par TRAIL pouvait aussi bifurquer sur la voie de la nécroptose, une mort cellulaire fortement immunogène. Nous avons pu démontrer une communication antagoniste de ces voies : la résistance induite lors de l’apoptose des cellules traitées par TRAIL permettait de sensibiliser les cellules survivantes au traitement nécroptotique. Ces travaux pourraient permettre de suggérer de nouvelles stratégies thérapeutiques afin de pallier les résistances inhérentes à ces voies de signalisations<br>The phenotypic heterogeneity observed in populations of isogenic cells is an immediate cause of non-genetic resistance to anticancer drugs which remains difficult to profile due to its transient nature. In a first part of this work, we have studied the molecular origins of the phenotypic heterogeneity observed in response to TNF-related apoptosis-inducing ligand (TRAIL). For that, we have developed a method to isolate resistant cells before they initiate genetic induction by the treatment, and sensitive cells before they deteriorate their content during cell death. The method is composed of three single-cell technologies: predictive measurement of cell response by video-microscopy, isolation of each cell in situ by laser capture, followed by transcriptomic profiling. The prediction of therapeutic response makes it possible to isolate and profile all the cells, regardless of the outcome of their late pharmacological response (cell death for sensitive cells), and thus enabled us to identify the genes at the origin of the difference in efficacy of TRAIL. Then in a second part of this work, we studied the transient resistance induced during treatment with TRAIL. We were able to demonstrate the role of TAK-1 in the establishment of this transient resistance. It has been described that the apoptotic pathway triggered by TRAIL could also branch off onto the pathway of necroptosis, a highly immunogenic cell death. We were able to demonstrate an antagonistic communication of these pathways: the resistance induced during apoptosis of cells treated with TRAIL made it possible to sensitize the surviving cells to the necroptotic treatment. This work could suggest new therapeutic strategies in order to overcome the resistance inherent in these signaling pathways

Le design standard des essais randomisés de phase III suppose le recrutement d'un grand nombre de patients pour assurer un risque α de 0.025 unilatéral et une puissance d'au moins 80%. Ceci s'avérer difficile dans les maladies rares, ou encore si le traitement cible une population spécifique définie par un sous-type moléculaire rare. Nous avons évalué par simulation la performance d'une série d'essais randomisés. Au terme de chaque essai, s'il est associé à une amélioration significative, le traitement expérimental devient le contrôle de l'essai suivant. Les designs ont été évalués pour différents taux de recrutement, différentes sévérités de la maladie, et différentes distributions hypothétiques des effets d'un futur traitement. Nous avons montré, que sous des hypothèses raisonnables, une série d'essais de plus petite taille et avec un risque α relâché est associée à un plus grand bénéfice à long terme que deux essais de design standard. Nous avons enrichi cette approche avec des designs plus flexibles incluant des analyses intermédiaires d'efficacité et/ou futilité, et des designs adaptatifs à trois bras avec sélection de traitement. Nous avons montré qu'une analyse intermédiaire avec une règle d'arrêt pour futilité était associé à un gain supplémentaire et à une meilleure maitrise du risque, contrairement aux règles d'arrêt pour efficacité qui ne permettent pas d'améliorer la performance. Les séries d'essais à trois bras sont systématiquement plus performants que les séries d'essais à deux bras. Dans la troisième de la thèse, nous avons étudié les essais randomisés évaluant un algorithme de traitement plutôt que l'efficacité d'un seul traitement. Le traitement expérimental est déterminé selon la mutation. Nous avons comparé deux méthodes basées sur le modèles de Cox à effets aléatoires pour l'estimation de l'effet traitement dans chaque mutation : Maximum Integrated Partial Likellihood (MIPL) en utilisant le package coxme et Maximum H-Likelihood (MHL) en utilisant le package frailtyHL. La performance de la méthode MIPL est légèrement meilleure. En présence d'un effet traitement hétérogène, les deux méthodes sousestime l'effet dans les mutations avec un large effet, et le surestime dans les mutations avec un modeste effet<br>Large sample sizes are required in randomized trials designed to meet typical one-sided α-level of 0.025 and at least 80% power. This may be unachievable in a reasonable time frame even with international collaborations. It is either because the medical condition is rare, or because the trial focuses on an uncommon subset of patients with a rare molecular subtype where the treatment tested is deemed relevant. We simulated a series of two-arm superiority trials over a long research horizon (15 years). Within the series of trials, the treatment selected after each trial becomes the control treatment of the next one. Different disease severities, accrual rates, and hypotheses of how treatments improve over time were considered. We showed that compared with two larger trials with the typical one-sided α-level of 0.025, performing a series of small trials with relaxed α-levels leads on average to larger survival benefits over a long research horizon, but also to higher risk of selecting a worse treatment at the end of the research period. We then extended this framework with more 'flexible' designs including interim analyses for futility and/or efficacy, and three-arm adaptive designs with treatment selection at interim. We showed that including an interim analysis with a futility rule is associated with an additional survival gain and a better risk control as compared to series with no interim analysis. Including an interim analysis for efficacy yields almost no additional gain. Series based on three-arm trials are associated with a systematic improvement of the survival gain and the risk control as compared to series of two-arm trials. In the third part of the thesis, we examined the issue of randomized trials evaluating a treatment algorithm instead of a single drugs' efficacy. The treatment in the experimental group depends on the mutation, unlike the control group. We evaluated two methods based on the Cox frailty model to estimate the treatment effect in each mutation: Maximum Integrated Partial Likellihood (MIPL) using package coxme and Maximum H-Likelihood (MHL) using package frailtyHL. MIPL method performs slightly better. In presence of a heterogeneous treatment effect, the two methods underestimate the treatment effect in mutations where the treatment effect is large, and overestimates the treatment effect in mutations where the treatment effect is small

In applications for generic medicinal products the concept of bioequivalence is fundamental. Two medicinal products, i.e. a test and a reference drugs containing the same active substance are considered bioequivalent if their bioavailability (rate and extent of absorption of an active substance that is absorbed from a drug product and becomes available at the site of action) after the administration of both products produce a similar therapeutic effect. The assessment of bioequivalence is based upon 90% confidence intervals for the ratio of the population geometric means (test/reference) for the parameters under consideration which should be contained within the limits 80%-125%. It is recommended using randomized, two-period, two-sequence, single dose crossover designs (2x2 crossover designs) The number of subjects to be included should be based on an appropriate sample size calculation, though the number of evaluable subjects should not be less than 12. Sometimes, there are drugs whose rate and extent of absorption is highly variable dose to dose within the same subject. The main problem with highly variable drugs is that to declare bioequivalence it requires a study with an unacceptably larger sample size. In this case, the usual approach to determine bioequivalence is ‘Reference Scaled Average Bioequivalence’ (RSABE), which is based on expanding the limits as a function of the within-subject variability in the reference formulation. But, using 2x2 crossover designs, it is not possible to estimate separately the test and reference variabilities, and thus it requires using more complex designs like replicated or semi-replicated crossover designs. On the other hand, regulations also allow using common 2×2 crossover designs based on two-stage adaptive designs (TSD) with sample size re-estimation at an interim analysis. At an interim look (stage 1), if average bioequivalence is not declared with an initial sample size, they allow to increase it based on the intra-subject estimated variability and to enroll additional subjects at a stage 2, or to stop for futility in case of poor likelihood of bioequivalence. This is crucial because both parameters must clearly be pre-specified in protocols, and the strategy agreed with regulatory agencies in advance with emphasis on controlling the overall type I error. Using Monte Carlo simulations, we show that RSABE and TSD methodologies achieve comparable statistical power, though the scaled method usually requires less sample size, but at the expense of each subject being exposed more times to the treatments. With an adequate initial sample size (not too low, e.g., 24 subjects), TSDs are a flexible and efficient option to consider: They have enough power (e.g., 80%) at the stage 1 for non-highly variable drugs and, if otherwise, they provide the opportunity to step up to a stage 2 that includes additional subjects. Based on TSDs, we also present an iterative method to adjust the significance levels at each stage which preserves the overall type I error for a wide set of scenarios which should include the true unknown variability value, and which provides a power of at least 80%. TSDs work particularly well for coefficients of variation below 0.3 which are especially useful due to the balance between the power and the percentage of studies proceeding to stage 2. We present an R package to adjust the significance levels at each stage in order to control the overall type I error.<br>En aplicacions per a medicaments genèrics el concepte de bioequivalència és fonamental. Dos productes, un test i un de referència, amb el mateix principi actiu, es consideren bioequivalents si la seva biodisponibilitat (quantitat Cmax i velocitat Tmax d’una substància activa que s’absorbeix d’un fàrmac i està disponible en el seu lloc d¿acció) després de l'administració d’ambdós productes produeix un efecte terapèutic similar. Per això, l’interval de confiança del 90% per a la ràtio de les mitjanes (mitjanes geomètriques poblacionals) dels productes test i referència de les mesures farmacocinètiques han d’estar dins dels límits de bioequivalència 80%-125%. Es recomana utilitzar dissenys aleatoritzats encreuats 2x2, és a dir, de dos períodes i dues seqüències (en anglès 2x2 crossover designs) El nombre de subjectes que s’inclouen es basa en un càlcul adequat de la grandària mostral, tot i que aquest nombre sol ser petit però mai inferior a 12 subjectes. Però en cas de productes/fàrmacs d’alta variabilitat cal incloure molts més subjectes per aconseguir una potència estadística adequada, de manera que la bioequivalència es determina amb pocs subjectes però a través de l’escalat dels límits de bioequivalència (RSABE, Reference Scaled Average Bioequivalence), expandits en funció de la variabilitat intra-subjecte en el grup de referència. En aquest cas, amb dissenys 2x2 no és possible estimar per separat la variabilitat dels productes test i referència i cal fer servir dissenys més complexos com ara dissenys encreuats replicats o semi-replicats. Les agències reguladores també permeten utilitzar dissenys encreuats 2x2 adaptatius de dues etapes amb re-estimació de la grandària mostral en la primera (anàlisi provisional, en anglès interim analysis). Llavors, si no podem declarar bioequivalència a la primera etapa amb una grandària mostral inicial petita, podem incrementar la mostra en funció de la variabilitat intra-subjecte estimada i afegir nous subjectes en la segona, o parar l’estudi per futilitat si la probabilitat de declarar bioequivalència és finalment petita. Aquesta estratègia ha d’estar definida en el protocol, i prèviament acordada amb les agències reguladores amb especial èmfasi en el control de l’error de tipus I. Mitjançant simulacions de Monte Carlo, mostrem que les metodologies basades en RSABE i dissenys adaptatius bietàpics proporcionen una potència estadística similar, tot i que els mètodes escalats normalment requereixen menys grandària mostral tot i que cal exposar més vegades els subjectes als tractaments. Amb una grandària mostral inicial adequada (no molt petita, per exemple 24 subjectes), els dissenys bietàpics són una opció molt flexible i eficient a considerar: proporcionen una potència raonable (per exemple del 80%) a la primera etapa per fàrmacs que no són altament variables, i en cas contrari, proporcionen l’oportunitat de saltar a una segona etapa que inclou subjectes addicionals. Basant-nos en aquests dissenys adaptatius bietàpics, presentem un mètode iteratiu per ajustar el nivell de significació a cada etapa que preserva l’error de tipus I global per a un conjunt d’escenaris que molt probablement inclouen el vertader valor desconegut de la variabilitat intra-subjecte, i que proporciona una potència estadística d’almenys el 80%. Aquests dissenys funcionen particularment bé per coeficients de variació per sota de 0.3 pel balanç que proporcionen entre la potència estadística i el percentatge d’estudis que salten a la segona etapa. Presentem un paquet d’R que ens permet ajustar els nivells de significació a cada etapa i que controla l’error de tipus I global.<br>En aplicaciones para medicamentos genéricos el concepto de bioequivalencia es fundamental. Dos productos, uno ‘test’ y uno de ‘referencia’, con el mismo principio activo, se consideran bioequivalentes si su biodisponibilidad (cantidad ‘Cmax’ y velocidad ‘Tmax’ de una sustancia activa que se absorbe de un medicamento y está disponible en su lugar de acción) después de la administración de ambos productos produce un efecto terapéutico similar. Para ello, el intervalo de confianza del 90% para la ratio de las medias (medias geométricas poblacionales) de los productos test y referencia de las medidas farmacocinéticas tienen que estar dentro de los límites de bioequivalencia 80%-125%. Se recomienda utilizar diseños aleatorizados cruzados 2x2, de dos períodos y dos secuencias (en inglés 2x2 crossover designs). El número de sujetos que se incluyen se basa en un cálculo adecuado del tamaño de muestra, aunque este número suele ser pequeño, pero nunca inferior a 12 sujetos. Pero en el caso de productos/medicamentos de alta variabilidad es necesario incluir muchos más sujetos para conseguir una potencia estadística adecuada, de forma que la bioequivalencia se determina con pocos sujetos, pero a través del escalado de los límites de bioequivalencia (RSABE, ‘Reference Scaled Average Bioequivalence’), expandidos en función de la variabilidad intra-sujeto en el grupo de referencia. En este caso, con diseños 2x2 no es posible estimar por separado la variabilidad de los productos test y referencia y se requieren diseños más complejos como diseños cruzados replicados o semi-replicados. Las agencias reguladoras también permiten usar diseños cruzados 2x2 adaptativos de dos etapas con re-estimación del tamaño del a muestra en la primera (análisis provisional, en inglés interim analysis). Entonces, si no podemos declarar bioequivalencia en la primera etapa con un tamaño de muestra inicial pequeño, podemos incrementar la muestra en función de la variabilidad intra-sujeto estimada y añadir nuevos sujetos en la segunda, o parar el estudio por futilidad si la probabilidad de declarar bioequivalencia es finalmente pequeña. Esta estrategia se define en el protocolo, previo acuerdo con las agencias reguladoras con especial énfasis en el control del error de tipo I. Mediante simulaciones de Monte Carlo, mostramos que las metodologías basadas en RSABE y diseños adaptativos bietápicos proporcionan una potencia estadística similar, aunque los métodos escalados habitualmente requieren menos tamaño de muestra aun siendo necesario exponer más veces al sujeto a los tratamientos. Con un tamaño de muestra inicial adecuado (no muy pequeño, por ejemplo 24 sujetos), los diseños bietápicos son una opción muy flexible y eficiente a considerar: proporcionan una potencia razonable (por ejemplo, del 80%) en la primera etapa para medicamentos que no son altamente variables, y en caso contrario, proporcionan la oportunidad de saltar a una segunda etapa e incluir sujetos adicionales. Basándonos en éstos diseños adaptativos bietápicos, presentamos un método iterativo para ajustar el nivel de significación en cada etapa que preserva el error de tipo I global para un conjunto de escenarios que muy probablemente incluyen el verdadero valor desconocido de la variabilidad intra-sujeto, y que proporciona una potencia estadística de al menos el 80%. Estos diseños funcionan particularmente bien para coeficientes de variación por debajo de 0.3 dado el balance que proporcionan entre la potencia estadística y el porcentaje de estudios que saltan a la segunda etapa. Presentamos un paquete de R que nos permite ajustar los niveles de significación en cada etapa y que controla el error de tipo I global.

Le cancer du sujet âgé est un réel problème de santé publique. L’incidence du cancer augmentant avec l’âge couplée au vieillissement général de la population font que plus de la moitié des tumeurs diagnostiquées aujourd’hui le sont chez des patients de plus de 65 ans. Cependant, cette population hétérogène a longtemps été exclue des essais cliniques et le manque de données prospectives rend difficile la prise en charge de ces patients. Plusieurs publications soulignent l’importance et la complexité de réaliser des essais cliniques dans cette population. Les schémas classiques ne prenant pas en compte l’hétérogénéité, les essais de phase II spécifiques aux sujets âgés sont rares et généralement stratifiés en sous-groupes définis selon un critère gériatrique ce qui augmente le nombre de patients à inclure et donc diminue la faisabilité. L’objectif de cette thèse est de présenter, comparer et développer des schémas de phase II adaptatifs stratifiés permettant de prendre en compte l’hétérogénéité de la population âgée. L’utilisation de ce type d’approche permet de réduire le nombre de patients à inclure tout en maintenant la puissance statistique et en contrôlant le risque d’erreur de type I. Ce qui implique une diminution du coût et de la durée de l’étude et donc une augmentation de la faisabilité. Afin d’améliorer l’efficacité de la recherche clinique en oncogériatrie, il est donc primordial d’utiliser des schémas adaptatifs stratifiés prenant en compte l’hétérogénéité de la population et permettant d’identifier un sous-groupe d’intérêt susceptible de pouvoir bénéficier (ou non) de la nouvelle thérapeutique<br>Elderly cancer is a real public health problem. With the overall aging population and the increased incidence of cancer, more than half of all tumors diagnosed today are in patients aged 65 years or older. However, this heterogeneous population has long been excluded from clinical trials and the lack from prospective data makes it difficult managing these patients. Many publications highlight the importance and the complexity of conducting clinical trials in this population. As classical phase II designs do not take into account the heterogeneity, elderly specific phase II clinical trials are very uncommon and generally conducted in specific subgroups defined by geriatric criteria which increases the number of patients to be included and thus reduces the feasibility. The objective of this thesis is to present, compare and develop stratified adaptive designs that address the heterogeneity of the elderly population. The use of this methodology can minimize the number of patients to be included while maintaining statistical power and controlling the type I error risk. This implies a reduction in the cost and duration of the study and thus increases the feasibility. In order to improve the efficiency of clinical research in geriatric oncology, it is essential to use stratified adaptive designs that take into account the heterogeneity of the population and make it possible to identify a subgroup of interest that might benefit (or not) from the new therapeutic

Le développement majeur de la thèse a été consacré au problème d’optimisation du choix de dose dans les essais de recherche de dose, en phase II. Nous avons considéré ce problème sous l’angle des fonctions d’utilité. Nous avons alloué une valeur d’utilité aux doses, le problème pour le sponsor étant de trouver la meilleure dose, c’est-à-dire celle dont l’utilité est la plus élevée.Dans ce travail, nous nous sommes limités à une seule fonction d’utilité, intégrant deux composantes: une composante liée à l’efficacité (la POS=puissance d’un essai de phase III de 1000 patients de cette dose contre placebo) et une autre liée à la safety. Pour cette dernière, nous avons choisi de la caractériser par la probabilité prédictive d’observer un taux de toxicité inférieur ou égal à un certain seuil (que nous avons fixé à 0.15) en phase III (toujours pour un essai de 1000 patients au total). Cette approche a l’avantage d’être similaire aux concepts utilisés dans les essais de phase I en oncologie qui ont notamment pour objectif la recherche de la dose liée à une toxicité limite (notion de ”Dose limiting Toxicity”).Nous avons retenu une approche bayésienne pour l’analyse des données de la phase II.Mis à part les avantages théoriques connus de l’approche bayésienne par rapport à l’approche fréquentiste (respect du principe de vraisemblance, dépendance moins grande aux résultats asymptotiques, robustesse), nous avons choisi l’approche bayésienne pour plusieurs raisons:• Combinant, par définition même de l’approche bayésienne, une information a priori avec les données disponibles, elle offre un cadre plus flexible la prise de décision du sponsor: lui permettant notamment d’intégrer de manière plus ou moins explicite les informations dont il dispose en dehors de l’essai de la phase II.• L’approche bayésienne autorise une plus grande flexibilité dans la formalisation des règles de décision.Nous avons étudié les propriétés des règles de décisions par simulation d’essais de phase II de différentes tailles: 250, 500 et 1000 patients. Pour ces deux derniers design nous avons aussi évalué l’intérêt de d’effectuer une analyse intermédiaire lorsque la moitié des patients a été enrôlée (c’est-à-dire avec respectivement les premiers 250 et 500 patients inclus). Le but était alors d’évaluer si, pour les essais de phase II de plus grande taille, s’autoriser la possibilité de choisir la dose au milieu de l’étude et de poursuivre l’étude jusqu’au bout si l’analyse intermédiaire n’est pas concluante permettait de réduire la taille de l’essai de phase II tout en préservant la pertinence du choix de dose final<br>The main development of the thesis was devoted to the problem of dose choice optimization in dose-finding trials, in phase II. We have considered this problem from the perspective of utility functions. We have allocated a utility value to the doses itself, knowing that the sponsor’s problem was now to find the best dose, that is to say, the one having the highest utility. We have limited ourselves to a single utility function, integrating two components: an efficacy-related component (the PoS = the power of a phase III trial - with 1000 patients - of this dose versus placebo) and a safety-related component. For the latter, we chose to characterize it by the predictive probability of observing a toxicity rate lower or equal to a given threshold (that we set to 0.15) in phase III (still for a trial of 1000 patients in total). This approach has the advantage of being similar to the concepts used in phase I trials in Oncology, which particularly aim to find the dose related to a limiting toxicity (notion of "Dose limiting Toxicity").We have adopted a Bayesian approach for the analysis of phase II data. Apart from the known theoretical advantages of the Bayesian approach compared with the frequentist approach (respect of the likelihood principle, less dependency on asymptotic results, robustness), we chose this approach for several reasons:• It provides a more flexible framework for the decision-making of the sponsor because it offers the possibility to combine (by definition of the Bayesian approach) a priori information with the available data: in particular, it offers the possibility to integrate, more or less explicitly, the information available outside the phase II trial.• The Bayesian approach allows greater flexibility in the formalization of the decision rules.We studied the properties of decision rules by simulating phase II trials of different sizes: 250, 500 and 1000 patients. For the last two designs (500 and 1000 patients in phase II), we have also evaluated the interest of performing an interim analysis when half of the patients are enrolled (i.e. with the first 250and the first 500 patients included respectively). The purpose was then to evaluate whether or not, for larger phase II trials, allowing the possibility of choosing the dose in the middle of the study and continuing the study to the end if the interim analysis is not conclusive, could reduce the size of the phase II trial while preserving the relevance of the final dose choice