courses.qmd

---
title: "Pre-conference Courses"
---

```{r, echo=F, results=F, message=F, warning=F}
library(tibble) 
library(dplyr)
library(knitr)
library(kableExtra)

courses <- tribble(
  ~Time, ~Course, ~Instructors,
  "Full-day", "Targeted Learning", "Oliver Dukes,<br>Stijn Vansteelandt,<br>Shaun Seaman",
  "Full-day", "The Design of Simulation Studies", "Tim P. Morris,<br>Ian R. White",
  "Full-day", "Network Meta-Analysis:<br>From Key Concepts to Advanced Methods", "Virginia Chiocchia,<br>Konstantina Chalkou,<br>Orestis Efthimiou,<br>Tasnim Hamza,<br>Georgia Salanti", 
  "Full-day", "Good Software Engineering Practice for R Packages", "Audrey Te-ying Yeo,<br>Alessandro Gasparini,<br>Daniel Sabanés Bové",
  "Morning", "Bayesian Methods for Precision Medicine", "Peter F. Thall",  
  "Afternoon", "Multi-State Models:<br>Theory, Applications and New Developments", "Liesbeth de Wreede,<br>Hein Putter",
) |> 
  mutate(
    Course = paste0(
      "<a href='#", 
      Course |>
        gsub(pattern = "<br>", replacement = "-", fixed = TRUE) |> 
        gsub(pattern = " ", replacement = "-", fixed = TRUE) |> 
        gsub(pattern = ":", replacement = "", fixed = TRUE) |> 
        gsub(pattern = ",", replacement = "", fixed = TRUE) |> 
        tolower()
      , 
      "'>", Course, "</a>"
    )
  )

```
## Overview

The following full-day and half-day (morning or afternoon) pre-conference courses will be held **in-person-only** on **Sunday, 24<sup>th</sup> of August 2025** (the day before the start of the main conference). Note that main conference registration is a prerequisite to be able to book pre-conference courses. See the [registration](registration.qmd#courses) page for details on the registration costs for the pre-conference courses.

```{r, echo=F, include=T, results = "asis", message=F, warning=F}
kbl(courses, escape = F, format = "html", table.attr = "quarto-disable-processing=true") %>%
  column_spec(2, width = "55%") %>%
  column_spec(3, width = "35%") %>%
  kable_styling(bootstrap_options = c("striped", "hover"))
cat("\n") 
```

## Targeted Learning

**Instructors:**

* Oliver Dukes, Ghent University, Belgium 
* Stijn Vansteelandt, Ghent University, Belgium
* Shaun Seaman, University of Cambridge, United Kingdom

Evaluating treatment effects using observational data or
trials with complex intercurrent events may require accounting for
high-dimensional confounding. This course will describe how, in these
challenging situations, machine learning and variable selection
procedures can be used to infer causal effects. The first part is a high
level overview of how and why this methodology works, touching on recent
developments in 'double machine learning' and 'targeted maximum
likelihood estimation'. In the second part the participants will
exercise the concepts and methods explained during the first part via
the analysis of real data sets.

This introductory course is aimed at researchers in the (pharmaceutical)
industry and academia working with observational as well as trial data;
a basic understanding of causal inference can be helpful but is not
necessary. We foresee a mix of lectures and hands-on exercises using R.

## The Design of Simulation Studies

**Instructors:**

* Tim P. Morris, UCL, United Kingdom
* Ian R. White, UCL, United Kingdom

Simulation studies are an invaluable tool for 
biostatistical research, as you will see at the ISCB46 conference this
week, where they will play a prominent role in many presentations. You
may even be presenting one of your own. Do you have the confidence to
thoughtfully critique someone's simulation study, or to defend yours?
Although it is tempting to think of a simulation studies as a mere
coding exercise, we have a rather different view. This short course will
focus on the design of simulation studies, following the ADEMP framework
(Aims, Data-generating mechanisms, Estimands, Methods of Analysis,
Performance measures). Practical sessions will be interactive rather
than computer-based, involving group discussions and debates on issues
around the design of simulation studies.

## Network Meta-Analysis: From Key Concepts to Advanced Methods

**Instructors:**

* Virginia Chiocchia, Institute of Social and Preventive Medicine, University of Bern, Switzerland
* Konstantina Chalkou, Institute of Social and Preventive Medicine, University of Bern, Switzerland
* Orestis Efthimiou, Institute of Primary Health Care (BIHAM), University of Bern, Switzerland 
* Tasnim Hamza, Institute of Social and Preventive Medicine, University of Bern, Switzerland
* Georgia Salanti, Institute of Social and Preventive Medicine, University of Bern, Switzerland

Network meta-analysis is an extension of pairwise
meta-analysis that allows us to compare three or more interventions
simultaneously, by combining direct and indirect evidence from a network
of studies. Network meta-analysis can be used to estimate the relative
treatmenteffects between any pair of interventions in the network, it
increases precision compared to using only direct evidence, and it can
produce a hierarchy of the interventions.

In the morning session of this full-day course we will demonstrate the
assumptions and methods of network meta-analysis and network
meta-regression with interactive lectures and practical exercise. In the
afternoon we will introduce advanced topics in network meta-analysis,
such as the use of individual participant data, component network
meta-analysis for composite interventions, and dose-response analysis.

The course is designed for participants who are familiar with
meta-analysis and Bayesian statistics. By the end of the course,
participants will be able to:

-   Understand and assess the assumptions underlying the validity of
    indirect comparisonsand network meta-analysis

-   Estimate the relative treatment effects between any pair of
    interventions within a network of studies and present them in a
    transparent way

-   Assess and test for inconsistency within a network of interventions

-   Formulate a network meta-regression model and interpret the results
    and output

-   Obtain an overview of advanced methods and extensions in network
    meta-analysis

The practical exercises will be performed in the statistical software R.


## Good Software Engineering Practice for R Packages

**Instructors:**

* Audrey Te-ying Yeo, Finc Research
* Alessandro Gasparini, Red Door Analytics AB
* Daniel Sabanés Bové, RCONIS

The vast majority of statisticians in academia and
industry alike write statistical software daily. Nonetheless, software
engineering principles are often neglected in biostatistics: most
biostatisticians know a programming language (such as R) but lack formal
training in writing reusable and reliable code.

This course aims to equip participants with the essential software
engineering practices required to develop and maintain robust R
packages. With the growing demand for reproducible research and the
increasing complexity of statistical methods developed for
multidimensional data, writing high-quality R packages has become a
critical skill for statisticians to prototype, develop, and disseminate
novel methods and push their adoption in practice. The course will focus
on the key principles of software engineering, such as workflows,
modular design, version control, testing, documentation, and quality
indicators. Focussing on these aspects ensures the reliability and
sustainability of R packages.

Participants will learn how to structure their R packages following best
practices and making use of tools that streamline the development
process. The course will also cover version control using Git, allowing
participants to manage code changes effectively and collaborate with
others. A significant emphasis will be placed on writing and running
unit tests, ensuring that packages are error-free and behave as expected
across different environments and over time.

Furthermore, the course will cover quality indicators for R packages and
explore techniques for writing effective documentation, enabling users
to pick, understand, and use statistical software packages effectively.

By the end of the course, participants will have a solid understanding
of good software engineering principles tailored to R package
development, enabling them to build packages that are not only
functional but also reliable, reusable, and easy to maintain.

## Bayesian Methods for Precision Medicine

**Instructor:**

* Peter F. Thall, M.D. Anderson Cancer Center, United States of America

This half day short course will
present statistical concepts and methods related to precision, or
'personalized' medicine, which uses individual patient covariates to
choose treatment or doses. The topics are drawn from the book, 'Bayesian
Precision Medicine' published by Chapman and Hall in 2024. To start, the
problem of comparing immunotherapy to prayer for treating a severe
disease will be discussed. Basic concepts of causal inference will be
reviewed, including bias correction methods for analyzing observational
data, causal diagrams, with both toy and real-world illustrative
examples. Two clinical trial designs will be reviewed that aim to
identify optimal subgroup-specific doses or treatments in particular
medical settings, each using a utility of a multivariate outcome. The
first is a phase 1-2 design that uses the joint utility of five
time-to-event outcomes to optimize patient subgroup-specific natural
killer cell doses for treating advanced leukemia or lymphoma. The second
design does phase 2 treatment screening and selection, illustrated by a
randomized three-arm trial to compare targeted agents. Two data analyses
that apply Bayesian nonparametric regression models to identify optimal
covariate-specific treatments then will be presented. The first analysis
uses observational data to identify optimal covariate-specific doses of
intravenous busulfan as part of the preparative regimen for allogeneic
stem cell transplantation. The second is a re-analysis of a published
dataset from a randomized trial, with a joint utility of progression
free survival time and total toxicity burden used to choose optimal
personalized targeted therapies for advanced breast cancer.

## Multi-State Models: Theory, Applications and New Developments

**Instructors:**

* Liesbeth de Wreede, Leiden University Medical Center, Netherlands
* Hein Putter, Leiden University Medical Center, Netherlands

The Multi-state models play an increasingly important role in the
analysis of time-to-event data. They provide a comprehensive framework
to analyze and understand complex medical phenomena, making them
invaluable in research aimed at improving patient care, guiding public
health policies, and advancing medical science. Extensions of the
Nelson-Aalen estimator of the cumulative hazard and of the Kaplan-Meier
estimator of the survival function allow for a detailed assessment of
the dynamics of complex disease processes and patient trajectories, and
the effect of covariates on these patterns.

In the first half we offer a brief coverage of basic concepts and
techniques in multi-state models, focusing on non- and semi-parametric
(Cox-model based) Markov models. We start with an introduction to
important concepts, in particular transition intensities (rates) and
transition probabilities (risks) and the relation between them, viewing
multi-state models as an extension of competing risks models. We
continue with methods for assessment of the effect of covariates on the
transition intensities through proportional hazards models. Throughout
the course, all steps needed for a multi-state analysis will be
illustrated with examples and syntax based on the mstate package in R.

In the second half we focus on two selected topics related to more
recent advancements in multi-state modelling. The first concerns the
Markov assumption. We discuss formal tests for the Markov assumption. We
explore estimation of transition probabilities that are consistent also
when the Markov assumption is violated, in particular the landmark
Aalen-Johansen estimator, and extensions like the hybrid landmark
Aalen-Johansen estimator. The second topic concerns incorporation of
relative survival in multi-state models. This allows to split mortality
in excess and background mortality with and without intermediate events.
Two different models for assessing the impact of covariates on the
excess hazard will be introduced: the Cox model and Aalen's additive
hazards model.