Utilities. If you aren’t using good ones explicitly, you are likely using bad ones implicitly.

I can see my kids adding "but we are still collecting data to increase power".....

Remember kids:

I think this issue is one of the most appealing aspects of Bayes. With Bayes the posterior distribution is the basis for inference, and it is invariant to the design.

And you thought reviewer 2 was rough…

📢 Apply now for the JAMA Network Peer Review Academy! This 9-month remote program connects early-career researchers with JAMA Network editors for hands-on peer review training. https://ja.ma/3TlIeIO

The Staphylococcus aureus Network Adaptive Platform (SNAP) trial has been nominated for a Patient healthcare communication award! Public voting is now open 👉 ccrew.accesscr.com.au/cpir-2025/pu... Find out more about SNAP 🔗 www.snaptrial.com.au Give us a vote!! @thedohertyinst.bsky.social

Saw the animal farm argument of statistical inference presented this morning. “When Bayes and frequentist are equal use frequentist” because apparently some are more equal than others.

I continue to find the endowment argument bizarre. Research grants are paying money for a service. We don’t ask a defense contractor to spend down all their bank accounts before we pay for a plane. Why before we pay for research?

For all clinical trialists who don’t believe in bringing in historical information or hierarchical borrowing across subgroups….a question…Aaron Judge is currently batting about 0.400. Do you take 0.400 as the best estimate of his 2025 average, or something lower? 🧵

3968

The most important thing in any job is that your boss and administration has your back. If you're looking for an academic job in the US...start with this list.

Stay tuned for the first @cmicomms.bsky.social & #Breakpoints collaboration episode on Sun, May 4th 6PM US EST/Mon May 5th 12 AM CET. Hosts Angela Huttner & Erin McCreary interview SNAP trial investigators, Steven Tong &Joshua Davis, on PSSA and MSSA domain results shared at 2025 ESCMID Global.

Please apply if your interested in running a new adaptive platform trial for invasive Streptococcal infections. Will work closely with the #SNAP_trial team. @thedohertyinst.bsky.social Applications close May 20 2025. jobs.unimelb.edu.au/en/job/92003...

Response adaptive randomization adjusts allocation probabilities over the course of a trial, based on accumulating data. RAR increases allocation to arms performing well, and decreases allocation to arms performing poorly. In multiple arm trials, this has advantages over fixed randomization.

RA Fisher on factorial trials. An ongoing missed opportunity in medical research. "No aphorism is more frequently repeated in connection with field trials, than that we must ask Nature few questions, or, ideally, one question at a time. The writer is convinced this view is wholly mistaken... (1/2)

One of my interests is the love/hate relationship statisticians have with data. Want to approve a drug or treat a patient today? Look at the available evidence!! Want to design an experiment...many statisticians answer..IGNORE the available evidence!! Why? www.berryconsultants.com/resource/8-e...

My advice to all my PhD students on defense day….”wait for your committee members to start arguing amongst themselves, and then just watch the show”

I’d be inclined to think hospitalizations are more accurate than case counts. People are more likely to take a really sick kid to a doctor and get counted while mild infections go unreported. Of course hospitalizations are fewer and more variable. Also….

Interesting experimental idea. Curious if the result indicates a prior distribution on “coins are fair” as much as a statement on likelihoods. Would have been interesting to also investigate a “novel” source of randomness.

As best I can tell they can grab me or my kids off the street, move us around fast enough we don’t know where to file a claim, intimidate our potential lawyers to not represent us, and then once we are outside the US claim “oops nothing anyone can do”. How is anyone ok with this?

I believe virtually all trial results should be public. Sponsors pay for trials but participants give precious resources. Society benefits from full transparency. Deidentified of course, likely with delays, but if you experimented on a human you owe humanity the results.

I wouldn’t guess exactly how but wouldn’t be surprised if reporting of cases is very different in the two locales.

This JAMA Guide to Statistics and Methods provides an overview of platform trials, describes the characteristics that make them more efficient for evidence generation, and discusses their limitations. https://ja.ma/3XxvqBC

This workplace has gone 0 days without being denied authorship because we are "just a vendor" (regardless of contribution)

My wife’s business just got a $6000 tariff bill. Where do I send this since I’ve been told US citizens don’t pay tariffs?

They could have equally shown exposure to the Macarena, NSYNC, or Beverly Hills Cop III was linked to autism. Thanks to @jsm2334.bsky.social for doing these. It's a lot of work to go through the details and read critically. Much appreciated!

Interested in adaptive platform trials? INCEPT (www.incept.dk) recently got approved along with the first domain, and we expect to start inclusion soon. A protocol/design paper is available at doi.org/10.1111/aas.... along with the full protocol docs at incept.dk/documents @andersperner.bsky.social

I see no route to randomize. Who would enroll on either side of the vaccine “debate”? So it’s either observational and prospective (which provides cover to say “we don’t have an answer yet”) or retrospective in which case hard to see value added (can see value subtracted…) Happy to be proven wrong.

If you rearrange the letters in my name you get Dr. Eevil. Sadly, I lack the funds….

The clear malicious delight in inciting a mob against people doing their job. The lies about firings based on made up “poor performance”. No decent person treats other people this way. And then this cowardly bs. There are FOUR lights!

Viewpoint: Proposals to lower or eliminate the @fda.gov's preapproval effectiveness requirements for drugs are unnecessary and dangerous, as they undermine patient safety and public health. https://ja.ma/4brnk39

Reading about academics lamenting their students might have to take industry jobs. First, I sympathize with anyone who fits well in academia and can’t find a job now. I also sympathize with anyone who fits well in industry but has been told/implied they are failure if they work there. (1/5)

Tim makes a great general point here, that you should always write keeping in mind how something will be read. We communicate complicated stuff, we should make the readers job as easy as possible. My own pet peeves....

I’m speaking in a panel discussion on simulating clinical trials on Friday, moderated by Mike Krams and featuring Andy Grieve and Tobias Mielke. We will be discussing simulating individual trials and development programs. us02web.zoom.us/webinar/regi...

Great advice. Closely related if you are going to try to balance, then severely limit the number of variables involved. Same principle…trying to balance too much actually results in imbalance.

Working on 6th grade math with my son. I’d hate math if I had it today. Questions are REALLY about non mathematical terminology they have constructed around the math, NOT the math itself.

Berry Consultants now has a podcast... berryconsultants.com/podcast First episode live... Challenges faced in pioneering adaptive trials Influence of FDA approvals in adaptive trials Impact of platform trials during COVID Aspirations for the next 25 years of clinical trial design

Multiple publications on HEALEY-ALS in JAMA. 2 result papers (2 more in JAMA Network), 2 editorials, and a podcast “So how do we judge these first results of the HEALEY ALS platform? The platform set out to test 4 interventions rapidly and efficiently, and did so.” jamanetwork.com/journals/jam...

With 331 estimates (really any modest number) the highest estimates tend to be biased high and the lowest are biased low. Whatever the truth is, the estimates tend to be more spread out than the truth. A Bayesian hierarchical model would be tailor made for this.

Seeing proposals that FDA should approve drugs based only on a proof of safety, not efficacy. I’m not sure this is even a thing. There is no short, small trial that shows a drug is safe, and even if there were, a “safety only” policy comes with other problems….a thread. @elizabethjacobs.bsky.social

My first bluesky thread…Some have called for FDA requiring only one trial instead of two. Some quick….naive…calculations involving trial performance and Bayes theorem. If we require X trials, what is the probability an approved therapy actually works?

I'm filling out reports and noting we have clinical trials "in every continent but Antarctica"....anyone in Antarctica want a clinical trial? Please call....

Clinical trials that use adaptive designs are challenging to report, and this Adaptive designs CONSORT Extension is there to help researchers when reporting their randomised adaptive trial. https://www.bmj.com/content/369/bmj.m115

On December 6th, I'll be participating in an ISBS roundtable to discuss adaptive platform trials. www.linkedin.com/posts/isbios...

First post here...let's see if I can do this :) FDA Draft master protocol guidance is out. I expect the most impact from the control group, informed consent, and blinding sections.... www.berryconsultants.com/2024/01/11/b...