Good things continue to happen.
ISCB–GMDS 2026 in Freiburg brought together scientists working in biometry, clinical biostatistics, medical informatics and bioinformatics on a shared scientific platform.
I consider this very important. If the aim is to advance methodology so that we can generate more reliable knowledge and make better decisions in health, no discipline can achieve this alone. Every scientific field, scientific society and methodological community has a role to play.
The presentations we heard in Freiburg made this very clear. Clinical research methodology is no longer limited to study design, sample size and statistical tests. Today, research reliability is considered within a much broader framework, from data quality and missing data to biomarkers and artificial intelligence; from the use of external controls and Bayesian methods to scientific reporting and educational models.
Missing data is no longer only an analysis-stage problem
One of the prominent topics at the congress was missing data. It was once again emphasised that missing data should not be viewed simply as a technical issue to be solved at the analysis stage with a few statistical methods.
Why missing data may arise, which mechanism may underlie it, how study design can reduce this risk, and which analytical approaches can be used under which assumptions should all be considered while the research is still being planned.
This approach once again shows that biostatistics in clinical research is not a service brought in at the end of a study, but a methodological science that should be present from the very beginning.
Companion diagnostics and predictive biomarkers
One presentation that particularly caught my attention focused on predictive biomarker assays and companion diagnostics. In the era of personalised medicine, it is increasingly clear that drug development and biomarker development can no longer be considered independently.
It is no longer enough to ask only, “Is this drug effective?” We also need to ask:
- In which patients is it effective?
- Which biomarker identifies these patients?
- Which cut-off is clinically meaningful?
- What is the level of evidence in biomarker-positive and biomarker-negative patients?
- Does the test used in the clinical trial support the same clinical decision as the test that will be used in real-world practice?
For this reason, a companion diagnostic is not merely a laboratory test. It is directly part of research methodology in defining the target population for a drug, evaluating the benefit–risk balance and supporting clinical decisions.
Analytical validation, clinical validation, clinical utility, cut-off determination, target-population definition and bridging studies therefore emerge as important methodological topics.
Special populations, external controls and Bayesian methods
Another important topic concerned alternative methodological approaches in situations where establishing conventional control groups may be difficult or ethically problematic, particularly in children, pregnant women and certain special or vulnerable populations.
In this context, the integration of external control groups into clinical research and the use of Bayesian methods were particularly noteworthy.
This reminds us once again that methodology is not static. Research design must evolve according to the clinical question, ethical constraints, available evidence and characteristics of the population.
However, using new methods does not mean that methodological standards become less important. On the contrary, when external controls, real-world data or Bayesian methods are used, issues such as bias, exchangeability, data quality, historical differences and uncertainty must be evaluated even more carefully.
Education may have been one of the most important themes
For me, one of the strongest messages of the congress concerned biostatistics education. It was emphasised that teaching statistics cannot be limited to formulas, test names and p-values.
Education should be integrated with:
- real clinical examples,
- how to formulate a research question,
- ethical principles alongside research methodology,
- selection of the appropriate study design,
- use of the appropriate statistical method,
- clinical interpretation of findings,
- scientific writing,
- reporting standards,
- data quality,
- critical appraisal of scientific articles,
- the appropriate use of next-generation AI-supported research and statistical software.
Listening to these discussions inevitably took me back almost 20 years.
This was exactly our starting point when we founded MedicReS
When we founded MedicReS, our core idea was simple: teaching researchers statistics alone is not enough. We need to teach the whole of good research.
We therefore began developing a holistic research approach extending from the research question to study design, from ethical evaluation to sample size, from data management to statistical analysis, and from clinical interpretation of results to scientific publication.
In this process, we developed the Good Medical Research and Good Biostatistical Practice approaches. We then transformed this approach into educational programmes, certification programmes and books.
Later, we took another step forward. We concluded that it was not enough simply to teach these principles; researchers also needed to be able to apply them while conducting their studies. This led us to develop the EPICOS AI platform to support the entire research workflow.
From the research question to design, from ethical planning to power analysis, from data analysis to critical appraisal, and from meta-analysis to synthetic data, we have sought to bring different stages of the research process together within a single methodological framework.
Seeing that many of the topics discussed in Freiburg today sit within this holistic approach that we have advocated for almost 20 years was especially meaningful to me.
Now it is time for KlinBiS CORE 2027
We now want to take this experience to a new stage. As the Clinical Biostatistics Association – KlinBiS, our goal is not to view clinical biostatistics in medical specialty training as a few hours of basic statistics lectures.
We want to develop a specialty-specific core clinical biostatistics education model that takes into account each specialty’s own clinical questions, outcomes, study designs, data structures, sources of bias and evidence-generation challenges.
For this reason, KlinBiS CORE 2027 will not be merely a list of statistical topics. We aim to create a competency framework that brings the research question, ethics, design, data, analysis, clinical interpretation, publication, evidence generation and AI literacy together within one structure.
Our central question will be:
“Which clinical biostatistics and research methodology competencies should a specialist physician have when conducting scientific research in their own field?”
And now we are beginning this work. We decided to start from the beginning of the alphabet. 😊
Our first specialty area: Anaesthesiology
For Anaesthesiology, we will develop our first CORE study by identifying the designs, outcomes, perioperative data, repeated measures, survival outcomes, safety data, prediction models and other specialty-specific methodological requirements most frequently encountered in clinical research.
We will then move on to other specialties. Our aim is not simply to prepare a curriculum, but to build a sustainable Clinical Biostatistics CORE for Türkiye that is integrated into specialty areas and encompasses the clinical research of the future.
In Freiburg, we listened. We compared. We reassessed where we stand. We considered what we need to improve. And we returned with new plans.
Now it is time to act. KlinBiS CORE 2027 begins. First specialty area: Anaesthesiology.
For those who would like to explore this approach in greater depth, my five books on clinical research methodology and clinical biostatistics are available through my Amazon Author Page:
I would also like to share the demo code that remains available for educators and researchers who would like to experience the EPICOS AI research workflow:
EPICOS AI Educator Demo Code: TEACHEPICOS2026
The code provides seven days of full access from the date it is activated and may be activated until 31 December 2026.
Our aim remains the same:
Ask the right question.
Design correctly.
Analyse correctly.
Interpret correctly.
Report correctly.
And bring all of this together with the real needs of clinical practice.