Summary

The basic Statistics course prepares learners to apply core statistical concepts in the context of clinical research. Participants will learn to apply key statistical principles, distinguish between common study designs and justify their selection based on specific research objectives, and construct research questions that are aligned to statistical principles. They will develop the ability to outline and critically review the key components of the statistic aspects of a study protocol and a statistical analysis plan. In addition, learners will assess when advanced trial designs—such as adaptive trials—are warranted and appraise their implications for study implementation and interpretation.

Statistics is fundamental to clinical research. No clinical study can be implemented without minimal requirements on statistical approaches. Otherwise, they are flawed and misinterpreted. And yes, ChatGPT cannot replace the key understanding of statistical elements in clinical research and the multidisciplinary dialogues with clinicians, data managers, monitors, study managers and other stakeholders to align statistical concepts for clinical research. Numbers alone do not speak for themselves; they need a correct interpretation and communication with all their advantages and limitations.

This module has the overarching learning outcome that you can implement research questions and hypotheses which can be statistically evaluated, communicated and interpreted by adequate study designs. We divided the module into four topics:

We start with the basic statistical concepts where we will introduce design and statistical tools. Design tools are key for planning clinical studies and you are likely already familiar with – or at least heard – of some tools. For example, the PICO framework which helps in precisely defining the population, intervention, control and outcome. From a statistical perspective PICO is not enough, because it allows of different interpretations of study results. This led to the introduction of the estimand framework in the ICH E9 guidelines back in the year 2020. Yet, before we can start using such frameworks, we have go one step back and introduce the statistical objectives in clinical research: Description, modelling and causal inference. Without knowing the key statistical objectives in clinical research it is not possible to follow the arguments of trial statisticians.

Clinical research is about creating evidence to guide patient care and support clinical decision-making. But single observations by themselves can be misleading. Patients are different, diseases progress in different ways and random variation is always present. Statistics gives us the tools to separate the "signal" from random noise. Without statistics, we risk drawing the wrong conclusions from limited or biased observations. Such tools, like hypothesis testing, p-values and others, we introduce in the subtopic of statistical tools. They are key principles for a correct interpretation and communication of study results.

Basic statistical concepts in clinical research

Module Summary

The basic Statistics course prepares learners to apply core statistical concepts in the context of clinical research. Participants will learn to apply key statistical principles, distinguish between common study designs and justify their selection based on specific research objectives, and construct research questions that are aligned to statistical principles. They will develop the ability to outline and critically review the key components of the statistic aspects of a study protocol and a statistical analysis plan. In addition, learners will assess when advanced trial designs—such as adaptive trials—are warranted and appraise their implications for study implementation and interpretation.

Observational data

Module Summary

In the previous topic, we introduced the basic statistical concepts in clinical research. You applied the framework of estimands, intercurrent events, hypothesis testing and confidence intervals to formulate and align a research question with core statistical concepts. We also introduced three different statistical objectives, description, modelling and causal inference.

In observational data the exposure or treatment of interest is not controlled by the researcher as in a randomized controlled trial. Thus, correct conclusions from observational data are more challenging. Are we looking at associations? Or predictions? Or causal claims?

Often clinicians aim to make causal conclusions from observational data. For example, Kenchaiah et al. investigated in a New England Journal of Medicine article whether obesity is associated with heart risk failure. For several reasons the authors did not implement a randomized controlled trial to answer the question: “Does obesity cause heart failures?” – A legitimate causal question. However, the authors report associations as their findings. Often randomized controlled trials are not possible or feasible. For example, in rare disease it is often not feasible to conduct a randomized controlled trial and case-control study is more appropriate to provide evidence for future research.

In this topic we give an introduction to observational data. We discuss challenges in interpretation, study designs and the causal framework.

Assessment and certificate