Pre-conference courses

A. Full day (9:00-17:00)

A. Full day

1. Causal inference in clinical trials

Presenters: Kelly Van Lancker, Ghent University, Belgium; Alex Ocampo, Novartis Pharma AG, Basel, Switzerland 

Summary: Causal thinking and inference have gained increasing attention in global drug development in light of the recently published ICH E9(R1) guideline on estimands and sensitivity analysis (2019) and the FDA guideline on covariate adjustment (2023). Even so, causal inference remains somewhat of a mystery to many. The goal of this course is to provide insight into the use of causal thinking and methods to better inform decision making. We introduce causal inference methods based on different drug development settings, including estimands, covariate adjustment, and the use of external control data.

Prerequisites: This introductory course is aimed at researchers in the pharmaceutical industry and academia working with clinical trial data; it does not demand prior familiarity with causal inference. We foresee a mix of lectures and hands-on exercises. The participants will strengthen their understanding of the concepts and methods explained during the lectures by analyzing real clinical trial data sets during the practical sessions using R.

Outline: 

Session 1: Pre-specifying the estimand based on counterfactual outcomes 

Important to enable the discussion of causal inference problems is the counterfactual/potential outcomes model. In this session we introduce this framework and explain how it can be helpful to define (causal) estimands.

Session 2: Causal inference methods for treatment-policy estimand: covariate adjustment

Covariate adjustment methods have been claimed to yield more powerful intention-to-treat analyses of randomized trials, at no ‘cost’. The aim of this session is to provide insight into covariate adjustment: how it succeeds to gain power, and when and how it can be used. In this session we will also cover targeted maximum likelihood estimation.  We will also shortly discuss transportability and the use of historical controls.

Session 3: Introduction to DAGs

Another important framework, besides counterfactual outcomes, to enable the discussion of causal inference problems is Directed Acyclic Graphs (DAGs). In this session we explain how to use DAGs to represent the causal relationship that we believe exist between the variables of interest. We also discuss how one can recognize problems of selection or confounding bias. In the hands-on part the participants will discover why standard adjustment for time-varying covariates does not provide a valid adjustment for time-dependent confounding, what instrumental variables are and how to utilize them for detecting causal effects.

Session 4: Causal inference for the hypothetical estimand: time-varying confounding 

Starting from different hypothetical estimands (e.g., due to treatment switching), we review different methods to adjust for time-varying confounding.

Organisers and lecturers:

Kelly Van Lancker is a postdoctoral research fellow in the Department of Applied Mathematics, Computer Science and Statistics of Ghent University (Belgium). She has obtained a PhD in statistics from Ghent University, and was previously a postdoctoral researcher at the Johns Hopkins Bloomberg School of Public Health. Her primary research is at the intersection of causal inference and clinical trials. Dr Van Lancker’s primary research interests are covariate adjustment, estimands, group sequential design and obtaining valid inference when the analysis involves data-adaptive methods, such as variable selection. She has taught several short courses on these topics.
Alex Ocampo is currently a Senior Principal Statistician with Novartis based in Basel, Switzerland. He obtained his bachelor’s degree in Statistics from the University of Michigan and Ph.D. in Biostatistics from Harvard University in 2020 where his doctoral dissertation focused on statistical methods for dealing with missing data when the “Missing at Random” assumption does not hold. His current work at Novartis focuses on promoting causal thinking in the pharmaceutical industry. He works on causal problems in mediation, survival/multistate models, and graphs. He has taught courses in causal inference both internally at Novartis and at the CEN conference in Basel in 2023. He published a Tutorial in Statistics in Medicine in 2023 on using causal graphs in clinical trials.

2. Dynamic Predictions for Longitudinal and Time-to-Event Outcomes, with Applications in R          

Presenters: Dimitrios Rizopoulos, Erasmus University Medical Center, Netherlands
                             Christos Thomadakis, Medical School, National and Kapodistrian University of Athens, Greece

Summary: This course focuses on data collected in follow-up studies. Outcomes from these studies typically include longitudinally measured responses (e.g., biomarkers) and the time until an event of interest occurs (e.g., death, dropout). The aim is often to utilize longitudinal information to predict the risk of the event. An important attribute of these predictions is their time-dynamic nature, i.e., they are updated each time new longitudinal measurements are recorded. In this course, we will introduce the framework of joint models for longitudinal and time-to-event data and explain how it can be used to estimate and evaluate such dynamic risk predictions for the settings of one event and competing risks. We will use the R package JMbayes2 to showcase the capabilities of these models.

Prerequisites: This course assumes knowledge of basic statistical concepts, such as regression models and standard statistical inference using maximum likelihood and Bayesian methods. Also, a basic knowledge of R would be beneficial but is not required. Participants are required to bring their laptops with the battery fully charged. Before the course, instructions will be sent for installing the required software.

Lecturers’ Background:

Dimitris Rizopoulos is a professor of Biostatistics at the Erasmus University Medical Center. He received an M.Sc. in statistics (2003) from the Athens University of Economics and Business and a Ph.D. in Biostatistics (2008) from the Katholieke Universiteit Leuven. Dr. Rizopoulos wrote his dissertation and several methodological and applied articles on various aspects of models for survival and longitudinal data analysis. He is the author of a book on joint models for longitudinal and time-to-event data. He has also written three freely available packages to fit such models in R under maximum likelihood (i.e., package JM) and the Bayesian approach (i.e., packages JMbayes and JMbayes2). He currently serves as co-editor for Biostatistics.

Christos Thomadakis is a post-doctoral researcher at the Department of Hygiene, Epidemiology, and Medical Statistics of the National and Kapodistrian University of Athens (NKUA), as well as an adjunct lecturer at the Athens University of Economics and Business. He earned his M.Sc. in Biostatistics from NKUA in 2015 and completed his Ph.D. in 2022 at the Medical School of NKUA. Dr. Thomadakis’ research focuses on methods for analyzing longitudinal and survival data and applied projects mainly from the HIV epidemiology. He has co-authored several articles on joint modelling for longitudinal and time-to-event data using Bayesian approaches and maximum likelihood. Additionally, he has developed inferential procedures to fit these models in R.

3. ROC analysis for classification and prediction in practice

Presenters: Christos Nakas, University of Thessaly, Greece
                             Constantin Gatsonis, Brown University School of Public Health, USA

Summary: This course will be based on our recently released book on modern ROC methodology (Nakas et al, CRC Press, May 2023).  We will examine the conceptual underpinning and sound interpretation of ROC analysis and will discuss the practical implementation ROC methods in diverse scientific fields, with emphasis on the evaluation of biomarkers, imaging modalities, and machine learning tools. Examples will accompany the methodologic discussion using standard statistical software such as R, Matlab and STATA.  The course will draw on developments in the field during the past two decades, which was a period of intensive growth in both the methods and the applications of ROC analysis. The instructors will provide a contemporary, integrated exposition of ROC methodology for both classification and prediction and survey methods for multiple-class ROC.

The course material would be of interest to graduate level students and researchers from a wide range of disciplines and different industries such as diagnostic medicine, bioinformatics, medical physics, and perception psychology.

Prerequisites: Graduate level course intended for researchers and students with background in the fundamentals of statistical methods, including estimation, hypothesis testing, linear and generalized linear modeling. Some familiarity with basic concepts from the evaluation of diagnosis and prediction would be helpful.

Outline:

1st part:

Fundamental concepts and metrics in the evaluation of diagnostic and predictive accuracy. Construction and interpretation of the ROC curve.

Empirical and model-based estimation of an ROC curve. Inferential procedures for a single ROC curve and its summary measures

Selection of optimal cut-off points for decision-making

2nd part:

Methodology for ROC-based comparison of diagnostic markers

Generalized linear modeling for ROC curves, optimal prediction with combinations of biomarkers (1 hr.).

Multiple-class ROC analysis (1 hr.),

ROC analysis under verification or imperfect gold standard bias

ROC analysis for learning: special topics

All methods will be presented through examples of case studies with implementation using R and Stata.

Lecturer’s background: 

Christos Nakas is a Professor of Biometry at the Laboratory of Biometry, University of Thessaly, Volos, Greece & and a primary investigator at the Department of Clinical Chemistry, Inselspital, University Hospital of the University of Bern, Bern, Switzerland. He has studied in Greece and France and held academic appointments in the USA, Switzerland and Greece. His research interests lie in the fields of ROC analysis, Metabolomics, Experimental Design and general life-sciences applications. He recently published, along with Dr Gatsonis, a book focused on the topic of this proposed pre-conference course.

Constantin Gatsonis is a Professor of Biostatistics, Department of Biostatistics, Brown University School of Public Health, RI, USA. He is the founding Director of the Center for Statistical Sciences and the founding Chair of the Department of Biostatistics. Dr. Gatsonis is a leading authority on the evaluation of diagnostic and screening tests, and has made major contributions to the development of methods for medical technology assessment and health services and outcomes research. He is a world leader in methods for applying and synthesizing evidence on diagnostic tests in medicine and is currently developing methods for Comparative Effectiveness Research in diagnosis and prediction and radiomics.

Β. Half-Day (9:00-12:30 or 13:30-17:00)

1. Adverse events with survival time outcome: clinical questions and methods for statistical analysis based on hazard functions

Presenter: Laura Antolini,  Laura Antolini, Università Milano Bicocca School of Medicine, Monza, Italy

Summary: In the study of novel treatments with a survival time outcome the analysis of the occurrence of adverse events is of crucial importance since it may imply the treatment interruption or, if fatal or even the patient death. This justifies the throughout effort in data collection on adverse events that is surprisingly often followed by statistical analyses only based on descriptive methods, such as simple proportions. In this course the analysis of adverse events will be tackled starting from the clinical question, followed by the identification of the estimand of interest and methods for estimation and Inference. Two different approaches will be discussed:
1. the observed occurrence of AE as first event, requiring a competing risk/multistate framework, with a risk benefit interpretation
2. the direct impact of the treatment in the risk of developing the adverse event, requiring a latent time variable framework, with an interpretation on causality

Outline:

The course will start by a critically review the commonly used standard theoretical quantities and estimators with reference to their appropriateness for dealing with approaches 1 or 2 in the presence of non fatal and fatal AE. Then the course will follow with the presentation of the suitable methods to address approaches 1 and 2. The theoretical explanation will be paralleled by motivating applications using the softwares Stata and R with data and code fragments made available to students.
Target audience:
Applied biostatisticians/epidemiologists or graduate students familiar with basic survival analysis.

Lecturer’s Background:

Laura Antolini is a professor of Medical Statistics at the University of Milano-Bicocca, Milan, Italy. She received an M.Sc in Statistical and Economical Sciences (1999) from the University of Bologna and a Ph.D. in Statistics (2003) from the University of Padua. Prof. Antolini wrote her dissertation and several methodological and applied articles on various aspects of models for survival analysis, particularly focusing on illness death models, competing risks and model assessment. She also participates in research projects in the experimental laboratory and clinical observational fields, from designing and writing protocols to the statistical analysis of the data that emerged.

2. Statistical and practical aspects of the design and analysis of Multi-Arm Multi-Stage (MAMS) Platform Trials

Presenters: Dr Babak Choodari-Oskooei, University College London, MRC Clinical Trials and Methodology Unit at UCL, UK
                             Professor Mahesh Mahesh Parmar, MRC Clinical Trials Unit at UCL and the Institute of Clinical Trials and Methodology at UCL, UK

Summary:  This course aims to help participants:

Understand the motivation behind these designs,
Learn how to choose the design parameters and stopping boundaries, both for lack-of-benefit and efficacy,
Learn how to deal with overwhelming efficacy,
Learn about stopping randomisation to research arms,
Learn how to add a new research arm, and how to control Type I and II error rates in both pre-planned and unplanned addition of a new arm,
Learn about MAMS designs in which arms are ranked and selectively chosen to continue.

MAMS platform trials:

Typically, in these protocols, randomisation is stopped to insufficiently active treatment arms at interim stages and new research arms can be added during the course of the trial. The MAMS approach is one of the few adaptive designs being deployed in a number of trials and across a range of disease in the phase III setting, including STAMPEDE (prostate cancer), CompARE (TB), TRUNCATE-TB (TB), RAMPART (renal cancer), and ROSSINI-II (wound surgery).

Outline:

This half-day course consists of two main sessions, and four lectures in total. The first session introduces MAMS platform designs. The second session focuses on the implementation of the statistical aspects of such trials and provides guidelines on the design and analysis of such trials. It will also explore further design issues such as adding new research arms, and designs in which research arms are ranked and selectively chosen to continue.

Lecturers’ background:

Babak Choodari-Oskooei, is a Principal Research Fellow and senior statistician at the UCL’s Institute of Clinical Trials and Methodology in London. He obtained his PhD (2009) in Statistics from the UCL Statistical Sciences department. He then worked as a postdoc on the area of his expertise, on multi-arm multi-stage (MAMS) adaptive trial designs, at UK’s Medical Research Council (MRC) Clinical Trials Unit. He has published many articles in prestigious journals, including a book chapter in the clinical trials compendium “Principles and Practice of Clinical Trials”. At UCL, Babak teaches postgraduate courses and leads short courses in advanced and adaptive trial designs. He is also an associate editor of Clinical Trials journal and the chair on the data monitoring and trial steering committees for several trials.

Mahesh Parmar is Professor of Medical Statistics and Epidemiology and Director of the MRC Clinical Trials Unit at University College London and the Institute of Clinical Trials and Methodology at University College London. He has more than 400 publications in peer-reviewed journals, many of which have had a direct impact on policy, clinical practice, and improving outcomes for patients. The MRC Clinical Trials Unit he directs is at the forefront of resolving internationally important questions, particularly in infectious diseases and cancer, and also aims to deliver swifter and more effective translation of scientific research into patient benefits. Examples of his methodological contributions include the development and implementation of the MAMS platform and DURATIONS designs.