In this QCast episode, co-hosts Jullia and Tom discuss clinical trial simulation and its role in supporting study design before patients, sites, and budgets are committed. Clinical trial simulation uses mathematical and statistical models to test plausible study scenarios, helping teams understand how a design may perform under different assumptions around dose, variability, recruitment, dropout, endpoints, and decision rules. Rather than treating a single forecast as fixed, simulation gives study teams a structured way to explore uncertainty.
The discussion also looks at common pressure points, including the misconception that simulation is mainly about making trials smaller or faster. Used well, simulation can inform early dose escalation, Phase 2 proof-of-concept planning, late-phase sensitivity analyses, interim decision-making, and operational feasibility. Its value depends on choosing the right model for the decision, using credible scenarios, documenting assumptions clearly, and communicating probabilistic outputs in a way that supports practical decisions.
What clinical trial simulation is designed to test
Clinical trial simulation helps teams examine how a proposed study design may behave before the trial is run. It can be used to explore power, bias, precision, probability of success, overdose risk, dropout, recruitment patterns, and other sources of uncertainty. The aim is not to predict the exact trial outcome, but to understand which assumptions matter most and where the design may be vulnerable.
Why model choice should follow the study decision
Different modelling approaches answer different questions. PBPK models may support questions around drug-drug interactions, organ impairment, or paediatric extrapolation, while population PK, PK/PD, exposure-response, and disease progression models may support dose selection, regimen evaluation, endpoint planning, or follow-up decisions. A more complex model is not automatically better if it does not serve the decision the team needs to make.
Where simulation can go wrong in practice
Simulation can become less useful when early data are overfitted, uncertainty is underplayed, or scenarios do not reflect clinical and operational reality. A model may look persuasive while depending on assumptions that would not hold under real trial conditions, such as site workload, patient behaviour, delayed lab uploads, or query turnaround. Clear documentation, verification, and proportionate conclusions help keep simulation outputs grounded and decision-ready.
Jullia
Welcome to QCast, the show where biometric expertise meets data-driven dialogue. I’m Jullia.
Tom
I’m Tom, and in each episode, we dive into the methodologies, case studies, regulatory shifts, and industry trends shaping modern drug development.
Jullia
Whether you’re in biotech, pharma or life sciences, we’re here to bring you practical insights straight from a leading biometrics CRO. Let’s get started.
Tom
Today we’re talking about clinical trial simulation. Now let’s start with the basics. When people say clinical trial simulation, what are they actually referring to?
Jullia
Well to put it simply, clinical trial simulation is the use of mathematical and statistical models to test possible study scenarios before the trial begins. Rather than designing around one fixed assumption, you run many plausible versions of the trial and look at how the design behaves.
That might mean looking at power, bias, precision, probability of success, overdose risk, dropout patterns, or how quickly enough information might accumulate.
Tom
Now that distinction between one assumption and many plausible scenarios feels important. Is that where people sometimes confuse simulation with prediction?
Jullia
Yes, definitely. A simulation is not a guarantee that the trial will unfold in a particular way. It’s more so a way of asking, “If the world looks like this, how does our design perform?”
So if a team simulates a Phase 2 study, the output might show that the proposed design performs well if the true effect is within a certain range, but becomes fragile if variability is higher than anticipated. It helps the team see where the design is robust and where it depends heavily on assumptions being right.
Tom
And clinical trial simulation isn’t the same as clinical simulation, is it?
Jullia
No, and that’s worth saying early. Clinical trial simulation supports research decisions around study design, dosing, endpoints, analysis, and operations.
Clinical simulation, on the other hand, usually refers to healthcare training, such as simulated environments or digital tools used to train clinicians..
Tom
Okay, so for trial teams, where does simulation tend to show up across development? Is it mainly an early-phase activity?
Jullia
It’s very common in early development, but it can be useful across the full programme. In first-in-human and early-phase studies, simulation often supports starting dose justification, escalation planning, expected exposure ranges, and sampling schedules.
For example, a team might compare different dose-escalation approaches under a range of toxicity assumptions. One approach might move quickly but carry more overdose risk. While another might be more cautious but slower to reach informative dose levels. Simulation helps make that trade-off visible.
Tom
Could you give an example of how that might affect the protocol?
Jullia
Sure. So imagine an early dose-escalation study where cohorts are staggered and each escalation decision depends on safety review after a set observation window. A simulation could test how different review timelines, cohort sizes, or stopping rules affect the speed and safety profile of the study.
That might lead to a change in escalation rules, the timing of safety review meetings, or the amount of PK sampling needed at each dose level.
Tom
Now that’s quite useful because it pulls the modelling into very ordinary trial activity, not just the statistical section.
Jullia
Exactly that. Good simulation work often sits between statistics, pharmacometrics, clinical operations, and clinical strategy. It might start with a modelling question, but the answer can affect visit schedules, monitoring requirements, interim review timing, or how much flexibility the protocol can realistically support.
Tom
Now how does the role change once you get into Phase 2?
Jullia
Well Phase 2 is often about proof of concept and learning. The question becomes, “How likely are we to detect a clinically meaningful effect, given what we currently know and do not know?”
Simulation can explore a range of treatment effects rather than using a single point estimate. It can compare continuous and categorical endpoints, test biomarker-enriched populations, or look at how different dose arms behave under several possible dose-response shapes.
Tom
There’s a common misconception there, I think. People may assume simulation is mainly used to make a study smaller or faster. Is that too narrow?
Jullia
Yes, I’d say that’s too narrow. Sometimes simulation supports a more efficient design, but efficiency should not be treated as the only goal.
The real purpose is better decision-making. In some cases, that means simplifying. In others, it means adding safeguards, choosing a different endpoint, revising assumptions, or accepting that the question needs more data than originally hoped.
Tom
And in late-phase trials, the stakes are different again. What does simulation add there?
Jullia
So late-phase simulation often focuses on robustness and interpretability. Teams may test time-to-event assumptions in survival studies, explore non-proportional hazards, assess sensitivity to missing data, or look at the consequences of protocol deviations.
That could mean simulating different dropout mechanisms to understand whether the planned sensitivity analyses are adequate.
Tom
You mentioned different model types earlier. Can we briefly walk through the main families?
Jullia
Yes, so one group is PBPK models, which stands for physiologically based pharmacokinetic models. These use mechanistic descriptions of the body and the drug to explore things like drug to drug interactions, renal or hepatic impairment, or paediatric extrapolation scenarios.
Then you have population PK or PK/PD models, which look at variability in exposure and response across individuals.
Tom
And exposure-response modelling sits slightly differently?
Jullia
Yes, and this is because exposure-response modelling focuses directly on the relationship between drug concentration and effect. That might help define the dose-response curve, explore the therapeutic window, or compare benefit-risk across regimens.
Then disease progression models are another important group. They describe how outcomes change over time, with and without treatment.
Tom
So the model choice depends more so on the decision being made, not on whichever modelling approach is most sophisticated?
Jullia
Pretty much. Really, the decision should drive the model. If the question is about dose escalation, a PK/PD or toxicity-based simulation may be appropriate. If the question is whether a visit schedule captures disease change properly, a disease progression model may be more relevant.
So a complex model can be impressive, but if it doesn’t answer the decision question, it can distract rather than help.
Tom
What would a sensible workflow look like? Say a sponsor wants simulation to support a study design decision.
Jullia
Well first, I’d define the decision context very clearly. Is it dose selection, sample size, endpoint choice, interim rules, or operational feasibility?
Then assemble and check the data, which may include clinical data, non-clinical evidence, literature information, and prior programme data.
Tom
And how about the scenarios?
Jullia
Yes, so the team defines which uncertainties to vary. That might include treatment effect, variability, dropout, adherence, recruitment rate, assessment timing, or protocol deviations.
Then virtual trials are run repeatedly, often using Monte Carlo simulation, and the outputs are summarised using decision metrics. Those metrics might include probability of success, expected power, bias, or precision.
Tom
Can you give an operational example? Something outside the usual endpoint and sample size discussion.
Jullia
Well, I suppose a practical one is recruitment and data flow around an interim analysis. Suppose a study has an interim look once a certain number of events or assessments are available. The team might simulate recruitment variability across regions, delayed lab uploads, dropout, and query turnaround time.
The output could show whether the interim analysis is likely to happen when expected, whether the available data are mature enough, or whether delays could create pressure to make decisions on incomplete information.
Tom
Now where do teams get into trouble with simulation?
Jullia
So one common issue is overfitting sparse early data. Early-phase datasets can support several explanations, especially when there are few dose levels and limited outcomes. If the model fits noise too closely, the simulation can appear more certain than the evidence really allows.
Another problem is using scenarios that are statistically neat but clinically unrealistic. If an assumption doesn’t reflect plausible biology, patient behaviour, site workload, or protocol feasibility, the output becomes less useful for decision-making.
Parameter uncertainty is another one to look out for. People can anchor on a point estimate, which can be risky.
So if the probability of success changes dramatically when variability increases slightly, decision-makers need to see that. Otherwise, the study may look stronger than it really is.
Tom
What about documentation?
Jullia
Well as you'd imagine, documentation is central if the output is going to influence important decisions. Teams need a clear trail from data inputs and model versions through to scenario definitions, code, outputs, and conclusions.
Verification also matters. The team needs to know that the code behaves as intended and that the model is adequate for the question being asked.
Tom
Now if we had to cover some key takeaways, what should listeners keep in mind?
Jullia
First, clinical trial simulation is decision support under uncertainty, not a way to remove uncertainty. Second, the most useful simulations begin with a clear decision question and scenarios that reflect credible clinical and operational realities.
And third, the output is only as strong as the assumptions, documentation, verification, and communication around it.
Tom
That last point on communication is interesting. How should technical teams present probabilistic outputs to non-technical stakeholders?
Jullia
They need to translate the output into decisions and trade-offs. Rather than only saying, “This scenario gives a certain probability of success,” explain what changes when assumptions shift. For example, what happens if dropout rises, if an endpoint is noisier than expected, or if recruitment is slower in one region?
Tom
Are there times when simulation is simply not needed?
Jullia
Yes, for example if a straightforward analytical calculation answers the question adequately, a complex simulation may not add much. Simulation also has limited value when the scenario set is poorly defined, when data are too weak to support the intended conclusion, or when teams are not prepared to act on the findings.
Tom
Now as we come towards the end, what would you want a sponsor team to do before commissioning or starting this kind of work?
Jullia
I’d start by agreeing on the decision the simulation needs to support. Then identify the data available, the assumptions that matter most, and the stakeholders who need to accept the output.
It also helps to decide early how the results will be used. Will they affect dose selection, the protocol, the statistical analysis plan, operational planning, or a regulatory discussion?
Jullia
With that, we’ve come to the end of today’s episode on clinical trial simulation. If you found this discussion useful, don’t forget to subscribe to QCast so you never miss an episode and share it with a colleague. And if you’d like to learn more about how Quanticate supports data-driven solutions in clinical trials, head to our website or get in touch.
Tom
Thanks for tuning in, and we’ll see you in the next episode.
QCast by Quanticate is the podcast for biotech, pharma, and life science leaders looking to deepen their understanding of biometrics and modern drug development. Join co-hosts Tom and Jullia as they explore methodologies, case studies, regulatory shifts, and industry trends shaping the future of clinical research. Where biometric expertise meets data-driven dialogue, QCast delivers practical insights and thought leadership to inform your next breakthrough.
Subscribe to QCast on Apple Podcasts or Spotify to never miss an episode.