This is basically why, at least in medicine, I think it's a fool's errand to pretend like we are estimating effects. I think it's much more realistic to frame it as evaluating claims, and accepting that at some point residual uncertainty will have to be left to clinical judgement.
Worth a read. But.. 😜
- All statistical methods rely on assumptions and thought exercises.
- The good advice here about the importance of context and assumptions, and moving away from resting everything on "a calculation", is widely applicable to sample size evaluation.
We finally have randomized, controlled clinical trial results from ambient AI scribes. But what do these RCTs actually tell us?
Look at the below chart. Time savings are often low, but doctors give the technology rave reviews, esp in anecdotes that people tend to believe more than data 👀 🧵
🖥️🩺
The long-awaited SURPASS-CVOT is here, just in time for the holidays #T2D #CVD #Tirzepatide www.nejm.org/doi/full/10....
In Clinical Scholars- literally one of the first 5 things Rod talked about.
cc: @jeremysussman.bsky.social @iwashyna.bsky.social
The creation of a medical reversal.
Listen in as @richard-lehman.bsky.social and @raj-mehta.bsky.social explore why it took so long for fair tests of treatments to develop, and the transition from anecdotal and authority-based decisions to empirical methodologies in the 18th and 19th centuries.
www.jameslindlibrary.org/blogs/the-fi...
It's finally here! SURPASS CVOT trial results! This is an interesting blend of benefits. The quick summary is tirzepatide is better for renal outcomes, A1c, weight and all cause mortality. For cardiovascular outcomes it's essentially equivalent/slightly better to a monoagonist GLP1. Let's dive in 🧵
Hypertension Notifications Building on its suite of advanced health features, Apple Watch Series 11 introduces groundbreaking hypertension notifications, which can alert users if signs of chronic high blood pressure — or hypertension — are detected.5 Hypertension is the leading modifiable risk factor for heart attack, stroke, and kidney disease, and impacts approximately 1.3 billion adults globally. It is frequently undiagnosed because it often has no symptoms, many people do not see a doctor regularly, and even during a clinical visit, it can be easily missed with a single measurement. Hypertension notifications on Apple Watch use data from the optical heart sensor to analyze how a user’s blood vessels respond to the beats of the heart. The algorithm works passively in the background reviewing data over 30-day periods, and will notify users if it detects consistent signs of hypertension. These notifications provide users with valuable insights into their health as it relates to this widespread condition simply by wearing their Apple Watch, so they can begin making potentially lifesaving behavioral changes, or start treatment to reduce their risk of serious, long-term health events. Replay video: Apple Watch Series 11 Hypertension Notifications Hypertension notifications use data from the optical heart sensor on Apple Watch to analyze how a user’s blood vessels respond to the beats of the heart. Like all of Apple’s health features, hypertension notifications are grounded in rigorous scientific validation. The feature was developed with advanced machine learning and training data from multiple studies totaling over 100,000 participants. Its performance was then validated in a clinical study of over 2,000 participants. While hypertension notifications will not detect all instances of hypertension, with the reach of Apple Watch, the feature is expected to notify over 1 million people with undiagnosed hypertension within the first year.
Hypertension notifications use data from the optical heart sensor on Apple Watch to analyze how a user’s blood vessels respond to the beats of the heart. Like all of Apple’s health features, hypertension notifications are grounded in rigorous scientific validation. The feature was developed with advanced machine learning and training data from multiple studies totaling over 100,000 participants. Its performance was then validated in a clinical study of over 2,000 participants. While hypertension notifications will not detect all instances of hypertension, with the reach of Apple Watch, the feature is expected to notify over 1 million people with undiagnosed hypertension within the first year. “Hypertension is the leading preventable cause of heart attack and stroke, yet millions remain undiagnosed,” said Harlan Krumholz, MD, SM, cardiologist and scientist at Yale University and Yale New Haven Hospital. “Making accurate detection easy and part of daily life can help people get care earlier and prevent avoidable harm.” A user’s iPhone and Apple Watch Series 11 show “Possible Hypertension” and offer more details. Hypertension notifications were developed with advanced machine learning and training data from multiple studies totaling over 100,000 participants. If users receive a hypertension notification, it is recommended that they log their blood pressure for seven days using a third-party blood pressure cuff and share the results with their provider at their next visit, which is consistent with the latest American Heart Association guidelines for the diagnosis and management of hypertension. Clearance for hypertension notifications from the FDA and other regulators is expected soon, and the feature will be available in more than 150 countries and regions — including the U.S. and the EU — this month. Hypertension notifications will be available on Apple Watch Series 9 and later, and Apple Watch Ultra 2 and later, with watchOS 26.
Hypertension Notifications Hypertension notifications use data from the optical heart sensor on Apple Watch to analyze how a user’s blood vessels respond to the beats of the heart. Like all of Apple’s health features, hypertension notifications are grounded in rigorous scientific validation. The feature was developed with advanced machine learning and training data from multiple studies totaling over 100,000 participants. Its performance was then validated in a clinical study of over 2,000 participants. While hypertension notifications will not detect all instances of hypertension, with the reach of Apple Watch, the feature is expected to notify over 1 million people with undiagnosed hypertension within the first year. “Hypertension is the leading preventable cause of heart attack and stroke, yet millions remain undiagnosed,” said Harlan Krumholz, MD, SM, cardiologist and scientist at Yale University and Yale New Haven Hospital. “Making accurate detection easy and part of daily life can help people get care earlier and prevent avoidable harm.” A user’s iPhone and Apple Watch Series 11 show “Possible Hypertension” and offer more details. Hypertension notifications were developed with advanced machine learning and training data from multiple studies totaling over 100,000 participants. If users receive a hypertension notification, it is recommended that they log their blood pressure for seven days using a third-party blood pressure cuff and share the results with their provider at their next visit, which is consistent with the latest American Heart Association guidelines for the diagnosis and management of hypertension. Clearance for hypertension notifications from the FDA and other regulators is expected soon, and the feature will be available in more than 150 countries and regions — including the U.S. and the EU — this month. Hypertension notifications will be available on Apple Watch Series 9 and later, and Apple Watch Ultra 2 and later, with watchOS 26.
The new ⌚️ will have hypertensive notifications
This doesn’t seem like a cuffless BP - but more like a notification ‘you may have hypertension, go check your BP’ ?
From www.apple.com/newsroom/202...
@jordybc.bsky.social @drwanpen.bsky.social @spjuraschek.bsky.social @daichishimbo.bsky.social?
Some key changes that stood out to me (got the opportunity to help w writing these!):
1) Stronger recs for single pill combos
2) Broader recs for screening for primary aldo (now consider in anyone with stage 2 HTN — i.e., BP >=140/90)
3) Just. Say. No… to cuffless BP devices… and to EtOH
2/
should we screen for CKD in the general population?
Finally, flozins, finerenone and GLP1s allow a good cost-effectiveness case
www.nephjc.com/news/2025/7/... for #NephJC shorts
Time to stop just being dialysis-ologists and flozinate, glipinate!
#NephSky
Heard from #AHRQ that the entire grants staff has been let go! They are unable to issue funds for grants already funded, let alone fund new grants. This is outrageous and will make Americans poorer and sicker. Why? A brief thread about just some recent AHRQ-funded research: #MedSky #HealthPolicy
This is what we were waiting for. A direct comparison between apixaban and rivaroxaban for the treatment of acute VTE. Apixaban reduces bleeding risk in the first 3 months by >50%!
Practice changing investigator-initiated RCT.
#ISTH2025
Is generative #AI capable of clinical reasoning?
Our new piece @thelancet.bsky.social open-access
@adamrodmanmd.bsky.social
www.thelancet.com/journals/lan...
Rubio terminated 5800 USAID contracts – more than 90% of its foreign aid programs – in defiance of the courts.
Here’s a list of just some of the lifesaving awards that were terminated. Nearly all were Congressionally mandated. They’ve saved millions of lives. 🧵
A randomized trial of tailored A.I.-guided ablation of atrial fibrillation vs standard-of-care, anatomical ablation demonstrated superiority of AI guidance
www.nature.com/articles/s41...
I wrote a book The Creative Destruction of Medicine (2013, title adopted after Schumpeter).
This title of the book for what we are now seeing in the U.S. is just The Destruction of Medicine.
Indirects support infrastructure for translational research and clinical trials. Look for clinical trials to be paused or halted because universities will no longer have funds to keep them open. This will be devastating for cancer patients at NCI-designated comprehensive cancer centers, for example.
I have a piece half-done on the entire world of indirect costs. It's too big an endeavor to finish now, in the middle of the night, when this problem *that was always a live grenade with a half-pulled pin* is suddenly a topic du jour. Shit of a thing.
Here are some highlights.
A randomized trial of GPT-4 vs 92 physicians with or without this #AI LLM for performance on patient care tasks.
AI improved physician performance, on par with AI alone (based on 5 clinical vignettes) nature.com/articles/s41...
Figure 2. Sodium Content on Packaging in a Basket of Food Products in the Agincourt-Bushbuckridge by Wave and Legislative Target
From JAMA Cardiology:
South African regulations limiting sodium in processed foods were associated with reductions in sodium consumption and blood pressure among rural adults aged 40 years or older.
ja.ma/4hMWyUW
#CardioSky
Editorial commentary title in AJKD by Bill Mitch and colleagues ‘Revisiting Protein Restrictions in Early CKD: Did we get it wrong’?
In @ajkd.bsky.social
www.ajkd.org/article/S027...
Revisiting protein restriction: did they get it wrong?
Yes they did it wrong.
But they don’t admit that in this commentary, sadly 😕
At 12-month follow-up, primary outcome data were available for 846 (90%) patients. Treatment failure occurred in 153 (34%) of 452 patients in the antibiotic group, compared with 28 (7%) of 394 in the appendicectomy group (difference 26·7%, 90% CI 22·4–30·9).
Finally!
Antibiotics are not non-inferior to appendectomy despite a generous 20% non-inferiority margin
They are actually inferior: www.thelancet.com/journals/lan... in the @thelancet.bsky.social
#MedSky
The potential importance of reactivation of herpes simplex virus type 1 (HSV-1) as a contributing factor to developing Alzheimer's disease
www.science.org/doi/10.1126/...
Published an opinion piece in @bmj.com with @raj-mehta.bsky.social. I bet no one has opinions about the topic.
www.bmj.com/content/388/...
A protein intake of 1.5 g/kg/day enhances net protein balance in older adults on a two-meal eating pattern.
by Church DD, Hirsch KR (...) Ferrando AA et 5 al. in J Nutr #MedSky
📖 read the article:
We need some of these phrases for nephrology!
Kidney failure = Filter futility
Peritoneal dialysis = Abdominal cleaning
Hyponatremia = Salt shortage
Adam Curtis Title Card: We Too Are Trapped In A System. Text over a background of a large office building
THREAD
This Is The Story Of The The Pernicious Rise of AI-Generated Papers and their Online Impact
An Incomplete History Told In The Voice of Documentarian Adam Curtis
1/31
IMO the key finding is that the best LLMs (i.e. chatgpt) are able to perform with minimum bias even if those biases may be present in the training data.
Lowering rate of error and bias is critical to safe deployment and efficiency gains.
It's an EBM level advancement that we can't ignore.
This is a really neat article, but also akin to a massive meta-analysis that lumps together a lot of very different papers.
Study confirms general views that:
(1) instruction prompt matters
(2) Large model >> Small models
