Kert Viele
@kertviele.bsky.social
Director of Research, Berry Consultants. Clinical trial designer specializing in Bayesian adaptive trials. Fellow of the American Statistical Association.
In some situations you may actually want two trials (different populations, etc.). Then requiring one trial loses meaningful information.
December 5, 2025 at 3:35 AM
In some situations you may actually want two trials (different populations, etc.). Then requiring one trial loses meaningful information.
For example...p1=0.024 and p2=0.024 is two trial successes, but p1=0.026 and p2=0.001 is not, despite the latter being stronger evidence of efficacy. If the bar in such a situation were equivalent evidence to two p=0.025, I'm ok with this. That said.....
December 5, 2025 at 3:35 AM
For example...p1=0.024 and p2=0.024 is two trial successes, but p1=0.026 and p2=0.001 is not, despite the latter being stronger evidence of efficacy. If the bar in such a situation were equivalent evidence to two p=0.025, I'm ok with this. That said.....
Depending on the actual details of the final proposal...this may not be unreasonable. Lots of sponsors run two identical trials (sometimes even just administratively different at the same sites). Needing two p<0.025 in this scenarios leads to odd conclusions.
December 5, 2025 at 3:35 AM
Depending on the actual details of the final proposal...this may not be unreasonable. Lots of sponsors run two identical trials (sometimes even just administratively different at the same sites). Needing two p<0.025 in this scenarios leads to odd conclusions.
We've been hurt in different ways :). I do feel everyone underestimates the task they do every day, to their own detriment.
December 1, 2025 at 3:32 PM
We've been hurt in different ways :). I do feel everyone underestimates the task they do every day, to their own detriment.
I've had lots of people say "I finally understand power (or fill in another operating characteristics)" after seeing this built up in pieces. The average over a distribution is a pretty complicated concept, especially if your life has been built around reading individual datasets.
December 1, 2025 at 3:16 PM
I've had lots of people say "I finally understand power (or fill in another operating characteristics)" after seeing this built up in pieces. The average over a distribution is a pretty complicated concept, especially if your life has been built around reading individual datasets.
Often seeing the individual trials in more complicated setting is the way clinicians see how the decision rules all interact, where risks exist (for example trials with insufficient complete data at interims), and also how operating characteristics arise from the collection of trials.
December 1, 2025 at 3:16 PM
Often seeing the individual trials in more complicated setting is the way clinicians see how the decision rules all interact, where risks exist (for example trials with insufficient complete data at interims), and also how operating characteristics arise from the collection of trials.
You can see Bayesian properties as well, if your inputs are simulated from the prior. I also would not underestimate the communication that goes on in presenting simulations, particularly individual simulated trials within a collection. Often only then are the consequences of decision rules clear.
December 1, 2025 at 3:16 PM
You can see Bayesian properties as well, if your inputs are simulated from the prior. I also would not underestimate the communication that goes on in presenting simulations, particularly individual simulated trials within a collection. Often only then are the consequences of decision rules clear.
I once saw an argument for eliminating a statistics department as "I have psychologists who know psychology and statistics, I have economists who know economics and statistics...why have people who just know statistics?". These fields overlap of course, but they aren't the same.
December 1, 2025 at 3:01 PM
I once saw an argument for eliminating a statistics department as "I have psychologists who know psychology and statistics, I have economists who know economics and statistics...why have people who just know statistics?". These fields overlap of course, but they aren't the same.
I don't necessarily consider designing studies to be intrinsic to being a scientist. It's a distinct skill set, especially for anything beyond the simplest studies. You can have incredibly insights into biology, etc., without necessarily recognizing immortal time bias and so on.
December 1, 2025 at 2:59 PM
I don't necessarily consider designing studies to be intrinsic to being a scientist. It's a distinct skill set, especially for anything beyond the simplest studies. You can have incredibly insights into biology, etc., without necessarily recognizing immortal time bias and so on.
Not sure of the scientific background here but the most promising methods are platform and basket trials. Platforms allow you to test multiple therapies at once and focus on the best, basket trials allow you do to that in patient subgroups (ISPY 2 did both)
October 29, 2025 at 2:40 AM
Not sure of the scientific background here but the most promising methods are platform and basket trials. Platforms allow you to test multiple therapies at once and focus on the best, basket trials allow you do to that in patient subgroups (ISPY 2 did both)
“A few subgroup analyses”
October 12, 2025 at 4:53 PM
“A few subgroup analyses”
It’s a really bizarre business model to me. It seems aimed at SEEMING like our standard data analysis paradigms, but it’s much like me deciding to fit a linear regression and instead of telling you the coefficients, I instead give you a bunch of (X,Y) pairs generated from my fitted model.
October 1, 2025 at 2:09 PM
It’s a really bizarre business model to me. It seems aimed at SEEMING like our standard data analysis paradigms, but it’s much like me deciding to fit a linear regression and instead of telling you the coefficients, I instead give you a bunch of (X,Y) pairs generated from my fitted model.
None of this is properly “data”. At best it is a snapshot of a predictive distribution from some black box AI model fit. What you are estimating from such things is, at best, the AI model fit from the actual data (which the AI could just provide directly instead of “data” but often doesn’t).
October 1, 2025 at 2:05 PM
None of this is properly “data”. At best it is a snapshot of a predictive distribution from some black box AI model fit. What you are estimating from such things is, at best, the AI model fit from the actual data (which the AI could just provide directly instead of “data” but often doesn’t).
Feels a bit like dominantly inherited Alzheimer’s. Genetic basis, somewhat predictable age of onset, etc.
September 25, 2025 at 2:11 PM
Feels a bit like dominantly inherited Alzheimer’s. Genetic basis, somewhat predictable age of onset, etc.
From a modeling perspective we can have a debate. But in reality sites are far from random. Sponsors have preexisting arrangements with productive sites etc.
September 3, 2025 at 3:23 PM
From a modeling perspective we can have a debate. But in reality sites are far from random. Sponsors have preexisting arrangements with productive sites etc.
Other applications of hierarchical models here for borrowing across different regions of the world.
cdn.who.int/media/docs/d...
cdn.who.int/media/docs/d...
cdn.who.int
August 29, 2025 at 5:49 PM
Other applications of hierarchical models here for borrowing across different regions of the world.
cdn.who.int/media/docs/d...
cdn.who.int/media/docs/d...
After a month I can’t remember what my own code does by reading the source (well at least it takes a while, and I don’t think I’m the worst coder). My help files are for me. If they help someone else too, great.
August 29, 2025 at 5:39 PM
After a month I can’t remember what my own code does by reading the source (well at least it takes a while, and I don’t think I’m the worst coder). My help files are for me. If they help someone else too, great.
It's unclear the conclusions in that paper...that essentially the extra complexity isn't worthwhile...apply outside of the context of that paper, particularly the small sample sizes.
August 28, 2025 at 4:50 PM
It's unclear the conclusions in that paper...that essentially the extra complexity isn't worthwhile...apply outside of the context of that paper, particularly the small sample sizes.
I don't know that I have seen a writeup specific to historical borrowing (too many things like that...). However the models are essentially identical to those in the basket trial literature like this paper.
www.tandfonline.com/doi/abs/10.1...
www.tandfonline.com/doi/abs/10.1...
A Comparison of Different Approaches to Bayesian Hierarchical Models in a Basket Trial to Evaluate the Benefits of Increasing Complexity
A basket trial is a trial that investigates a single experimental therapy simultaneously across multiple different patient population groups. When groups are heterogeneous, the optimal analysis is ...
www.tandfonline.com
August 28, 2025 at 4:50 PM
I don't know that I have seen a writeup specific to historical borrowing (too many things like that...). However the models are essentially identical to those in the basket trial literature like this paper.
www.tandfonline.com/doi/abs/10.1...
www.tandfonline.com/doi/abs/10.1...
There are multiple versions. Some are simply a hierarchical model over studies. Others fit a two component mixture with one component having the current study independent of others. Others fit more complex models like dirichlet processes.
August 26, 2025 at 6:03 AM
There are multiple versions. Some are simply a hierarchical model over studies. Others fit a two component mixture with one component having the current study independent of others. Others fit more complex models like dirichlet processes.
Of course, seeing people submit false AI references can be very informative about their quality control….we have already caught false references in potential FDA submissions.
August 17, 2025 at 3:31 AM
Of course, seeing people submit false AI references can be very informative about their quality control….we have already caught false references in potential FDA submissions.
But...but....I really like my
Bayesian Adaptive Design And Subject Simulator
software....
Bayesian Adaptive Design And Subject Simulator
software....
August 15, 2025 at 2:48 PM
But...but....I really like my
Bayesian Adaptive Design And Subject Simulator
software....
Bayesian Adaptive Design And Subject Simulator
software....