The end of the year always brings reflection on the last 12 months.
I think it’s pretty clear that the biggest thing in medicine today is the weight loss drug revolution, which isn’t all that interesting or unique of a perspective.
If you’re looking around, you’ll see quite a few takes on some variation of “The Year of Ozempic” as The New Yorker put it.
But 2023 hasn’t just been about the incretin mimetics. It’s the year we have the first CRISPR based gene editing treatment for a human disease and the year when ChatGPT made the New England Journal of Medicine.
Here’s a bit of an end of the year reflection on the biggest things happening in medicine.
The Ozempic data tells us energy toxicity is really bad for us
There are a lot of potential conclusions you can draw from what we’ve learned about the GLP-1 agonists so far.
GLP-1 agonist is shorthand for the drugs like Ozempic, which is the most famous drug in this class. Not every drug in this class just targets only GLP-1, and so the more accurate term for these drugs is “incretin mimetics.”
But I’ll stick with GLP-1 agonists here to keep things simple.
Let’s start with what we know with a fairly high level of confidence - these drugs help a lot of different conditions: they’re effective at treating diabetes, obesity, cardiovascular disease, and kidney disease.
They work across populations - obese/not obese, diabetic/not diabetic, different ages, different sexes, different racial/ethnic groups - and the results have been consistently positive across a lot of clinical trials.
That widespread benefit suggests there’s something about the way that these drugs work that treats numerous different chronic diseases.
My sense is that the broad impact of these drugs across demographics and diseases fits with the hypothesis most chronic disease in the modern world is about energy toxicity - too many calories in and too few calories out.
I suspect that the way that GLP-1 agonists alter this balance by working in our brains to decrease appetite is just a means to an end, rather than being particularly special about the GLP-1 system.
But what about the fact that we see benefits in people who aren’t obese?
You can have energy toxicity without being obese
You can still have energy toxicity without frank obesity - you know this problem as being skinny fat, or, in medical terminology, normal weight obesity.
We all have a certain capacity to store fat before it becomes problematic because not all fat is created equal.
There’s a big difference between the fat we store around our organs (called visceral fat) and the fat we store elsewhere.
Visceral fat accumulation negatively impacts our biology in many ways; the chemical signaling from visceral fat is the way that energy toxicity manifests. When our energy balance is out of whack - we take in more calories than we burn - it’s toxic to almost every organ system.
The data on fat loss suggests that when we start losing fat, the first stuff we start shedding is visceral fat.
And so when you see data showing a benefit of a drug like Ozempic on someone who isn’t obese and doesn’t have diabetes but does have chronic disease, I suspect that it’s because we preferentially lose visceral fat when we start losing fat and this leads to metabolic benefits and thus less chronic disease.
I’ve said before that I think of these drugs as tools for the way we navigate the modern obesogenic world - they work in the brain to modulate appetite (and that’s probably why you’re hearing the anecdotes about their impact on things like addition).
The success of these drugs across so many different populations fits with this hypothesis of energy toxicity as being the root cause of much of the chronic disease that we see.
The other big thing I will take from Ozempic’s success - the end to the diet wars
You may have noticed that none of the studies on Ozempic were testing a specific dietary pattern. The weight loss occurred without forcing people to eat a different diet - no one made the folks on Ozempic go plant based, keto, Mediterranean, or Paleo.
A diet agnostic method of losing weight is fascinating - it suggests to me that most of the people evangelizing a particular dietary pattern as the key to weight loss or health benefits for everyone doesn’t have science on their side.
Anything that lets you live in a different energy balance environment than the majority of people will reduce your risk of chronic disease.
These drugs do that for many, which is why they’re such a gigantic deal.
Every single dietary strategy that has success simple provides rules to avoiding energy toxicity.
Each diet works for some people but not others for a simple reason: different people are different.
Their biology is different, their preferences are different, and it should be pretty clear to all of us that if a single diet was the solution for everyone without exception, we would know it by now.
The fact that Ozempic is diet agnostic, yet still shows benefit across the board suggests that it’s not about the food choices as much as it’s about the food quantity.
In other words, it’s the calories, stupid.
CRISPR is real, but I’m not sure if it’s spectacular
CRISPR becoming reality is a pretty amazing and gigantic story.
The speed of advancement here is astounding - discovering the first CRISPR systems in single celled creatures and turning them into cures for genetic disease in humans in under two decades is almost unbelievable.
But here we are.
I suspect it’s going to be a long time before we’re regularly editing genes for anything other than single gene genetic disorders with pretty serious outcomes and poor/nonexistent current therapies.
The technical challenges are massive - delivering this stuff safely and reliably is going to be a huge scientific lift.
There’s also the gigantic unknown of what editing the genome means in the long term - How does this impact long term cancer risk? How confident can we feel that we’re not going to have off target edits or edits in a different organ than the anticipated target?
These unknowns can’t be answered quickly, which is why I suspect broad implementation of this technology is a long way away. And that doesn’t even touch on issues of cost or ethics, which are going to be incredibly thorny to tackle.
I’m certainly pretty far away from being convinced gene editing is going to do much to treat heart disease, even if there are single gene mutations that greatly impact heart disease risk.
After all, most of the diseases that humans get (not to mention almost all of the genetically governed traits we have) are the result of many different genes interacting with many different environmental signals.
Our understanding of any of that is nowhere close to where it needs to be before we turn into Gattaca.
Long term, CRISPR may certainly prove civilization altering, but in the short term I think the use cases are going to be few and far between.
Artificial intelligence is here
A lot of folks in healthcare are excited and bullish about the impact of AI on our future healthcare.
Will AI cure disease? Create new drugs? Develop new ways of detecting cancer early?
I’m in the skeptical group, at least from the perspective of what a large language model can do.
Large language models like ChatGPT are great at synthesizing large amount of information, summarizing complicated topics, and doing all kinds of tasks that humans currently do.
I can certainly picture an AI tool doing about 75% of my current job - I spend a lot of time in the medical chart checking boxes to confirm diagnoses, ordering tests, answering message, and writing prescriptions.
A large language model could probably do all of that easily. ChatGPT would probably make fewer mistakes with it than I do.
But the important things doctors do aren’t anywhere close to being touched by these models and will require some pretty gigantic advances to actually be impacted, even if you totally ignore the human side of medicine.
First, a lot of our decisions are made without a clinical trial to guide us - they’re judgment calls in a data free zone.
Second, separating signal from noise is really hard. I don’t want my doctor to have read 50,000 papers on the disease that I’ve been diagnosed with. Most of that information is useless - the bulk of the information is from low quality, confounded observational studies that teach us nothing about treatments, causation, or anything useful.
I would need a huge amount of confidence that an AI doctor hasn’t had their brain poisoned by useless data that might be wrong, and this doesn’t even touch on the hallucination problem.
Third, and probably most importantly, there’s a huge amount that we don’t know that we can’t know without high quality clinical trials being run.
One of the biggest issues that I learned about during the pandemic is that too many important people in medicine and public health are willing to make treatment decisions and even policy decisions in the absence of high quality data. A large part of this is because of overconfidence in the reliability of low quality data.
I worry that models that can synthesize all the available information are going to make this same mistake on a much deeper level.
Reading every paper doesn’t give you judgment and a thousand observational studies are almost never as good as a single well conducted randomized trial.
If an AI doctor can find and integrate all the crappy data in the world to make a treatment recommendation, I have zero confidence that it will lead to better decisions and a fair amount of confidence that it will sometimes lead to worse decisions.
Gaining new knowledge doesn’t happen from knowing everything that’s been written about a topic - it comes from running experiments and seeing the results, which are sometimes the opposite of what the world experts would have predicted they would be.
So even though these AI models may be able to incorporate gigantic amounts of information, they aren’t going to know things that no one knows about biology.
As the saying goes, all models are wrong, but some models are useful.
Here’s to an exciting year ahead
I want to conclude by thanking you all for reading (and sharing) this newsletter!
2024 promises to have a lot more exciting medicine in the pipeline - I’m looking forward to seeing what’s next.