Week of November 11, 2024
A segment of my patients (who I quietly call my "highly-engaged, high-end healthcare consumers") will pay (out of pocket) for mostly unwarranted screening tests (like total body scans) or medically dubious therapies (such as homeopathic remedies, vitamins sold on Fox news commercials, or intravenous light therapy). Given their kidney disease, these patients often ask me to comment on these tests, the results, or the therapies. The conversations are challenging and time-consuming, explaining pre-test probability, the limited value of screening for patients at low risk of a disease, the trade-offs of following up on unexpected (and usually clinically unimportant) findings, and the absent (or low-quality) data behind many alternative medicines. And though my head's judgy voice screams, "This person has more money than good sense," I try to remain empathetic. I recognize these (typically older) patients' desire for control at a time when their minds are sharp, but their bodies are deteriorating (an "existential crisis of aging"). I wonder what my doctor will say when I buy a 2055 model-year nano crystal-infused quantum healing chamber for my 80th birthday. (I should start a GoFundMe now.)
Here is a thoughtful discussion on direct-to-consumer advertised, cash-based total body scanning - where the benefits do not outweigh the risks:
https://www.webmd.com/a-to-z-guides/features/truth-about-whole-body-scans
Here is an article highlighting a related challenge - accepting known unknowns in a desperate situation - a well-informed scientist who treated herself with experimental therapy. The bet has paid off (for now).
https://www.nature.com/articles/d41586-024-03647-0
Johns Hopkins offers a consumer-oriented discussion about when and why we screen for certain disease states.
https://www.hopkinsmedicine.org/health/treatment-tests-and-therapies/screening-tests-for-common-diseases
A few programming notes:
Several readers commented on the A.I. podcast's value but reflected that the COVID information no longer feels urgent and is "bogging down" the content. So (in the spirit of customer service), I am moving COVID data to the end of the newsletter and will exclude it from the podcast. Feel free to offer feedback (either positive or critical).
---
This week's Google NotebookLM A.I.-generated podcast's "hosts" appear to have achieved some degree of sentience. They now seem to feel they have a mission of "making learning fun." I am happy they are still very enthusiastic about my writing. (When will they turn on me?!?)
(I also note they have comments that imply there should be commercial breaks. Who would be a good sponsor?)
Listen here:
https://drive.google.com/file/d/1GKyp9ab0FnxyY39TvQSzttQAM9d0YM8s/view
About NotebookLM: https://blog.google/technology/ai/notebooklm-audio-overviews/
----
Medical Trends and Technology
The voices of anti-science are about to be the "loudest in the room" (again). The editorial below and a summary of the data on fluoridated water are good forecasts of conversations we may have in 2025. Science (and health care) offer very few trade-off-free solutions. But no one wins when we have fear-based discussions that do not balance the trade-offs or risks of low-probability events compared to beneficial outcomes (i.e., when the benefits DO outweigh the risks). Nuance and details matter. Evaluating and communicating a data set's relative strength and meaning is difficult in a politicized and social-media-driven world.
https://www.science.org/content/blog-post/note-robert-f-kennedy-jr
and
https://sciencebasedmedicine.org/some-data-on-fluoride/
and
https://sciencebasedmedicine.org/fluoride-and-iq/
Fundamentally, artificial intelligence and machine learning help identify associations between variables in large data sets. Though A.I.-generated answers are correlative (a statistical probability of association), the speed and volume of novel variable relationships are remarkable. This article discusses how a machine learning tool can identify existing pharmaceuticals for potentially treating rare diseases. " [The tool] finds shared features across multiple diseases, such as shared genomic aberrations. The A.I. model [might] pinpoint a shared disease mechanism based on common genomic underpinnings, [extrapolating an association] between a well-understood disease [treatment] to a poorly understood [disease] with no treatments."
https://news.harvard.edu/gazette/story/2024/09/using-ai-to-repurpose-existing-drugs-for-treatment-of-rare-diseases/
Infographics
It is an excellent time to review the 2015 compound interest infographic on how to begin understanding the quality of scientific data. Note Bene: This is a bit simplified (for instance, a systemic review may include a mix of randomized trials, cohort studies, and case-control studies), but it is a great starting point for asking questions.
https://i0.wp.com/www.compoundchem.com/wp-content/uploads/2015/04/A-Rough-Guide-to-Types-of-Scientific-Evidence.png
from
https://www.compoundchem.com/2015/04/09/scientific-evidence/
FYI - I've been playing with Anthropic’s Claude LLM this week. Claude does an excellent job summarizing medical studies from PubMed - see
https://claude.site/artifacts/12e1c0b1-5ea3-40a1-bcaa-a851f7c599a0
I asked Claude to help me understand the available clinical trials on Berberine and Citrus Bergamot, two OTC supplements used to lower cholesterol.
Things I learned this week
Chimpanzees are somewhat self-conscious - "[When testing Chimps performance on a series of computer tasks] chimpanzee performance was influenced by the number and types of audience present. Performance increased for the most difficult task as the experimenter count increased, while for the easiest task, performance decreased [with an increased number of familiar audience members]".
https://cosmosmagazine.com/nature/animals/under-pressure-chimps-perform-tasks-differently-with-an-audience/
and
https://www.cell.com/iscience/fulltext/S2589-0042(24)02416-7
I missed the buzz in 2019 when TikTok unleashed the 1982 video of a Tyrolean Tongue Choir to the internet.
https://www.classicfm.com/music-news/videos/austrian-tyrolean-tongue-choir-moscow-nights/
In exploring this, I learned about Umngqokolo (African throat singing) and am again impressed by the wide range of human expression.
https://www.themusicstudio.ca/a-world-of-music-voice-part-ii/
Unlike merging Mongolian throat singing with Western musical styles (for instance, The Hu, a Mongolian folk-metal band https://en.wikipedia.org/wiki/The_Hu), I only found traditional (non-electronic) African overtone singing.
https://spotify.link/DWwxjwptpOb
Living with A.I.
Geometry-informed neural networks (GINNs) allow the development of shapes by informing the engine of the end function and geometric limits. GINNs allow for rapid iteration of numerous solutions that demonstrate a wide range of solutions (with varying trade-offs and uses of space).
https://x.com/artuursberzins/status/1854853972055494856
and
https://arxiv.org/abs/2402.14009
related - a computationally-engineered rocket engine:
https://www.tctmagazine.com/additive-manufacturing-3d-printing-industry-insights/aerospace-insights/leap-71-world-first-hot-fire-test-computationally-engineered-rocket-engine/
Microsoft's research team highlighted their evolving A.I. agent tools. Magentic-One is a central, generalist agent that can oversee other agents, "where a lead agent, the Orchestrator, directs other agents to solve tasks. The Orchestrator plans, tracks progress, and re-plans to recover from errors, while directing specialized agents to perform tasks like operating a web browser, navigating local files, or writing and executing Python code." Think about this as the evolution of CoPilot spanning multiple steps and applications, "please gather all the emails from John in Outlook, summarize them into a word document that includes the Q3 sales data and the YakYak company's 2024 white paper on the Yak wool weaving market and email the file as an attachement to Susan."
https://www.microsoft.com/en-us/research/articles/magentic-one-a-generalist-multi-agent-system-for-solving-complex-tasks/
A.I. art of the week (A visual mashup of topics from the newsletter, now using ChatGPT to summarize the newsletter, suggest prompts, and make the images).
"A mountain scene with a rocket launching in the background, with Tyrolean monks in traditional robes performing throat singing in the foreground. A chimpanzee dressed as a choral director stands in front of the monks, holding a baton and directing them enthusiastically."
https://drive.google.com/file/d/1Px5h6iOPH1NBe1A15f73SxoLVK2tR9rd/view
---
COVID rates will likely trend upward as we get closer to Thanksgiving and Christmas.
The Pandemic Mitigation Collaborative (PMC) website uses wastewater levels to forecast 4-week predictions of COVID rates.
https://pmc19.com/data/
based upon https://biobot.io/data/
Wastewater Scan offers a multi-organism wastewater dashboard with an excellent visual display of individual treatment plant-level data.
https://data.wastewaterscan.org/
------
Clean hands and sharp minds, team,
Adam
Comments
Post a Comment