In this episode, I talk with my long-time friend and frequent guest, Kaspar Rufibach, about a skill that quietly determines how much impact we really have: presenting and communicating our work.

We walk through how Kaspar prepares his talks (including why he starts months in advance), how he structures messages so stakeholders actually remember and act on them, and why overcrowded slides are often just a sign that we haven’t done the hard thinking yet.

We also get honest about something many statisticians feel but rarely discuss: the fear of public speaking, the frustration of bad meetings, and the “personal brand” you build every time you present—whether you intend to or not.

If you’ve ever walked out of a meeting thinking “I don’t think they really understood what I meant,” this episode is for you.

In this episode, I’m joined by Deepa Jahagirdar, Associate Research Principal at Cytel, to explore what it really takes to build a good external control arm (ECA). Deepa brings a fascinating background from social epidemiology, where causal questions often need to be answered without running randomized trials. That experience translates directly into today’s growing need for ECAs, especially when we rely on real-world data to support single-arm trials, extension phases, or situations where randomization simply isn’t possible.

Together, we discuss how to choose the right data source, how target trial emulation works in practice, what to do about confounding, and how to judge whether an ECA is truly robust. If you’re working with real-world evidence, complex study designs, or causal inference, this episode will give you clarity and confidence in approaching ECAs the right way.

Why this episode made our all-time Top 9: If you’ve ever thought “non-parametric = Wilcoxon/Mann-Whitney and that’s it,” this conversation will happily destroy that myth. Frank shows how rank-based methods unlock rigorous analyses for skewed data, outliers, ordinal endpoints, small samples, composites/estimands—and how to communicate effects without relying on means.

Adaptive designs let us learn earlier, stop smarter, and protect patients—but they also make communication tricky. In this episode, Kaspar Rufibach and I dig into what “still correct” looks like when you try to explain results from group-sequential and other adaptive trials to regulators, clinicians, and scientific audiences. We unpack conditional vs. unconditional bias, median-unbiased estimation, stage-wise ordering for p-values, confidence intervals in multi-stage settings, and what to do with secondary endpoints and multiplicity. We also touch on ICHE20 (Adaptive Clinical Trials) and why pre-specification isn’t just a box-tick—it’s what builds trust.

In this episode, I’m joined once again by my friend and frequent guest, Kaspar Rufibach, to talk about a topic that’s been around for decades but is gaining fresh attention thanks to the new ICH E20 draft guideline—adaptive designs in confirmatory clinical trials.

Kaspar and I discuss why and when we should consider adapting a clinical trial, what kinds of adaptations are statistically valid and meaningful in a regulatory context, and why these designs—despite their efficiency—are still not used as often as they could be.

We also dive into the statistical foundations behind adaptive designs, such as p-value combination methods and meta-analytic thinking, and explore how adaptive approaches can help us make faster and smarter decisions in drug development.

This episode is a little different because Alun turns the microphone toward me. After 456 episodes, it feels both strange and exciting to be the “guest” on my own show. Together, we reflect on the journey so far and then dive into a topic close to both our hearts: the human skills that make statisticians and quantitative scientists truly effective.

We talk about leadership as helping others accomplish something, how to influence people across functions (not just departments), why being known inside your organization matters, and how presentation skills can make or break your impact. We wrap up with three actions you can start applying right away.

I’m excited to reshare one of our most-played conversations—the one where Norwegian regulator/HTA leader Anja Schiel and I get very practical about when single-arm trials fail decision-makers and what comparative, smarter alternatives look like for regulators, HTA bodies, payers, clinicians, and—most importantly—patients.

As statisticians in pharma, one of the most important professional relationships we can build is with our physician colleagues. When this partnership works well, studies run smoother, decisions are better, and our impact for patients grows.

In this all-time Top 7 replay, Benjamin Piske and I talk about what makes this collaboration effective, the challenges you may face, and how to establish yourself as a true partner rather than “just the statistician.”

This is one of our most downloaded episodes ever, and I’m excited to bring it back in this replay. In this conversation, I spoke with Lara Wolfson (MSD) and Anders Gorst-Rasmussen (Novo Nordisk) about EU HTA (European Union Health Technology Assessment): what it is, why it’s coming, and why statisticians like us must pay attention.

If you’ve ever wondered whether your approach to safety analysis is leading to misleading conclusions, this episode is a must-listen.

We’re bringing back one of our most downloaded episodes ever – a deep dive into how adverse events should be analyzed properly. This conversation with Jan Beyersmann and Kaspar Rufibach is packed with methodological insights and practical implications for statisticians working in clinical trials.

Adverse event (AE) analysis has long been approached differently from efficacy analysis, often using overly simplistic methods that can bias results. In this episode, we discuss why that’s a problem – and how the SAVVY collaboration (Survival analysis for AdVerse events with Varying follow-up times) is pushing the field forward.

Together with academia and multiple pharma companies, this collaboration tackled the issue of AE analysis using real randomized trial data, not just simulations. The findings show how common methods can underestimate or overestimate event probabilities and how established statistical methods can be applied more consistently to ensure fair benefit–risk assessments.

If you’ve ever wondered whether your approach to safety analysis is leading to misleading conclusions, this episode is a must-listen.