Week 3: Special Modules

Summer Course 2018
31st Residential Summer Course in Epidemiology

2 July – 6 July 2018

Week 3, parallel morning module 1
Advanced statistical topics
Per Kragh Andersen with Corrado Lagazio and Michaela Baccini

  • Competing risks
  • Recurrent events and longitudinal data
  • Cohort sampling
  • Propensity score
  • Causal modelling
Week 3, parallel morning module 2
Genetic epidemiology: genome-wide association studies and whole genome approaches to the study of disease
David Evans and Gibran Hemani 
Genetic epidemiology is concerned with the genetic aetiology of medically relevant traits and diseases in human populations. In this course we will learn how to design, conduct and analyse human genome-wide association studies (GWAS). Over the last decade, GWAS have completely revolutionized the field of genetic epidemiology, providing seminal insights into the biological basis of health and disease, and simultaneously identifying a wealth of drug targets for existing and novel pharmacotherapies. Indeed many of the genetic variants identified by GWAS are now finding useful roles in Mendelian randomization studies which investigate possible causal relationships between risk factors and medically relevant outcomes (see the Aetiological Epidemiology course run by Lawlor and Tilling). Whilst a substantial portion of this module focuses on using GWAS to identify genetic variants involved in the aetiology of complex traits and diseases, we will also show how GWAS data can be leveraged to predict individuals’ trait values, estimate heritability, and discover genetic relationships between different diseases that may be difficult to measure in the same study
 Week 3, parallel morning module 3
Advanced topics in epidemiology
Jan Vandenbroucke with Suzanne Cannegieter and Irene Petersen

The origins and usefulness of several advance study desings and methods of analyses will be studied – each time ending with a current positioning: what is state-of-the art and what are the potential applications and pitfalls (with practical examples)

  • New User Designs and Case-only studies: Case-crossover and self-controlled case-series. Some ideas behind these designs/analyses already existed, but they find increasingly new applications in today’s epidemiology; with STATA exercises.
  • Instrumental Variable and Regression Discontinuity Design. Both methods are extensively used in economics and econometrics but there is increasing interest in applications to epidemiology over the last decade. These methods aim to ‘control for confounding’ without having to specify the confounders.
  • Imputation part I. Bridging ideas about the necessity for imputation based on ideas about assessing causality.
  • Imputation part II. An in-depth session on applied imputation with exercises in STATA


Week 3, parallel morning module 4
From epidemiology to the burden of disease: putting risks in perspective
Nino Künzli

Epidemiology is a core science to investigate and quantify the association between risk factors and health outcomes. However, public health professionals and policy makers need to understand the public health relevance of risks to plan and prioritize prevention and policy making. The epidemiology-based assessment of the risk related burden of disease provides the bridge between public health science and policy. This course will familiarize students with the use of epidemiology in quantitative risk assessment and the comparison across risks. Based on a range of examples, students will i) learn how epidemiology contributes to quantitative risk assessment, ii) understand the tools to assess the health burden and iii) critically interpret the derived outputs. Exercises based on the burden of disease data will train critical thinking for the comparison of different public health risks.

Approach: Lectures on concepts, self-studies with on-line exercises and group discussions will foster the understanding of how epidemiology is used in risk assessment. Students are expected to bring their personal note books to independently work on line, on the web.


Week 3, parallel afternoon module 1
Clinical Epidemiology
For a long time, clinical epidemiology has largely ignored the evaluation of medical tests, which has received far less attention than the evaluation of new drugs and other interventions in clinical trials.

Increasingly, guideline developers, reimbursement agencies and other agencies emphasize clinical effectiveness: the ability of medical tests to improve health outcomes for patients undergoing testing., This has stimulated a series of interesting methodological development, because study designs must consider the purpose of testing and the role of new tests in the clinical pathway. Tests are also seldom used in isolation, but usually form part of a testing strategy that includes multiple tests.

In this one week course we will highlight the differences between analytical performance, clinical performance and clinical effectiveness. We will discuss the study design considerations for evaluations of diagnostic tests, prognostic tests, test for treatment selection, and screening tests, as well as challenges for evaluating tests in randomized trials.

Approach: A combination of short lectures and personal assignments, to be completed in pairs and discussed in the group.

Week 3, parallel afternoon module 2
Aetiological epidemiology:
Mendelian randomization and triangulation of causal methods
Debbie Lawlor and Kate Tilling

Aetiological epidemiology is concerned with causes of disease and health related conditions in populations and provides the evidence base for prevention and treatment targets for which interventions can be developed and their effectiveness tested. Mendelian randomization (MR) is a relatively new aetiological epidemiology method. It refers to the use of genetic variants as instrumental variables to understand causal effects of non-genetic modifiable risk factors (including ‘physiological’ risk factors such as blood pressure, fasting glucose, CRP and lifestyle risk factors such as smoking, caffeine consumption and alcohol intake). It is a method that is increasingly used in epidemiology, particularly since the development of methods that can be used to explore the key sources of bias in MR, and control for these. Other approaches to causal inference, including cross-cohort comparisons, negative control studies, and matching designs such as within sibship analyses are also increasingly used. A substantial part of this module will be concerned with MR, but we will also discuss other approaches and their key sources of bias and the value of triangulating different approaches that each have different sources of bias, in order to improve causal inference.


Week 3, parallel afternoon module 3
Environmental epidemiology
Josep M. Antó and Jordi Sunyer

  • Epidemiological design and applications
  • Exposure measurements : ecological and individual
  • Spatial and temporal clusters
  • Effects measurements. Biological markers

Risk assessment

Week 3, parallel afternoon module 4
Concepts and methods in causal mediation analysis
Bianca De Stavola and Rhian Daniel

In diverse fields of empirical research attempts are made to decompose the effect of an exposure on an outcome into its effects via a number of different pathways.

Path analysis has a long tradition in dealing with this type of enquiry, but more recent contributions in the causal inference literature have led to greater understanding of relevant parameters and the assumptions under which they can be identified. For this reason the module first starts with an introduction to causal language and causal diagrams before embarking on a comparison of the two schools.

There will be four sessions:

  • Motivation, causal language and causal diagrams (with non-computer practical)
  • Traditional mediation analysis (with computer practical in Stata)
  • Counterfactual-based mediation analysis (with computer practical in Stata)

Further issues and discussion (with computer practical in Stata)