There are themes. Muscle loss is one of the bigger ones — sometimes subtle, like Ozempic face or Ozempic butt, losing muscle where it is noticeable. Others truly develop sarcopenia; that is rare, typically people who are not exercising when they take it and may have had a predisposition. If I am concerned about someone, it is good to get a DEXA scan beforehand; if muscle mass is inadequate, think twice about the GLP-1 or try another avenue, and definitely start hitting the gym. The other theme is nausea: many get nauseous on higher doses but will not lose weight on the low dose. Some say it stops working — that may be the same dose-intolerance issue. Those are the primary reasons I hear.

Surgery really started back in the '50s — the University of Minnesota, I think, was the first place. The first procedures focused on malabsorption: bypass a portion of the small bowel so you do not absorb your calories. That was the jejunoileal bypass, and it was awful. People did lose weight, but it left a long blind limb with no food going through it — bacterial overgrowth and all sorts of problems. The malabsorbed fat bound calcium in the bowel, so the oxalate that normally binds calcium got absorbed and then bound calcium in the kidneys, causing renal failure. It was a disaster that went on for years. Gastric bypass replaced it in the mid-to-late '60s — Mason, I think, was the surgeon — aiming to avoid those problems while keeping a treatment effect. He thought of it as restriction plus bypass: stapling makes the stomach smaller, some calories are not absorbed — turns out that is not really how it works. From there: lap bands, purely restrictive; true stomach stapling, the VBG; and now the sleeve gastrectomy. These are the real surgeries, and they were created conceptually around restriction or malabsorption — but they work entirely differently than what people thought.

Leptin never panned out in large part because of leptin resistance. The hypothalamus and the brain itself just become resistant to it, because leptin is so high in people with obesity for so long. There is low-grade inflammation we know in those tissues, and eventually you just do not respond to it anymore.

With biopsies you have to be very targeted to get the tissue of interest — even in the pancreas, an area that looks like a lesion may not yield cancer cells. Once you do get the tissue, the stains can tell you a great deal: genetic predisposition, whether it will respond to one drug over another. That is phenomenal, but diagnostic studies need to become less invasive so they can scale. Procedure-based treatment I love — going through the mouth rather than incising the abdomen — but procedure-based diagnosis has a scaling problem: a doctor spending an hour or two with one patient cannot screen everyone. So I would love to see innovation toward minimally invasive diagnosis — smartphone and AI, minimally invasive scans — while blood tests remain great because they are quick and easy.

Most devices we see are incremental improvements — a better wire, a more ergonomic tool. What I see happening more recently is AI starting to have an impact by coaching you through procedures. On a heads-up display it does more than show the images: it can highlight the structures you want to work on, point to where the stitch should go, count the stitches and tell you if they are close enough together, and change the shape of the stomach as you work on it to show whether you are getting a good treatment effect — never possible in real time before. It is not widely available yet; this is in research centers. The other hope is robotics. In our lab we randomized trainees learning a complicated colon-tumor resection: they struggle horribly the traditional way — that is why it takes two or three years — but sit down with the robot and they are almost as good as an expert, shortening the learning curve dramatically. Layering AI and automation onto the robots may be the big win. Intuitive Surgical's robots showed this when they first came out: they democratized the field, took mediocre surgeons and made them excellent, and really helped the struggling.

I think they are supportive of robots, but I think that a truly exceptional surgeon is probably just going to be better without the robot — the robot is just going to make you worse.

That is common across all medicine — even inside a massive medical center, knowing who truly is the best is complicated. We rely partly on volume: how many procedures they do, and more importantly how many over their career. Volume is not the whole story, but it is important. Medicine needs to move toward objective metrics — one thing AI can do for us. I am involved in a platform called Everself where doctors doing these procedures are held to metrics: weight-loss outcomes, stitches placed per procedure, procedure time, complications. The next layer is AI on top: it can coach you through the procedure and grade you at the end — how many full-thickness sutures (you want 100%; maybe the doctor puts in 70%), whether sutures are close enough together, the volume of stomach you reduce. The idea next is to share that grade — with the credentialing bodies, so true underperformers get a refresher, and with patients, so they can select doctors on objective metrics. Adenoma detection rates in colonoscopy show this: publishing them was stopped after patients flocked to the highest-rated doctors, wait times became enormous, and they could not get access. But there should be a reasonable cutoff of required expertise, and AI hopefully will help us get there.

The AI is trained on thousands and thousands of procedures — more than I have done — so it recognizes patterns. You still use your clinical judgment; you are not using it blindly, you might ignore it sometimes, and you do not have to follow it. Automating operative robots with AI would be different. But where it is just suggesting where you put a stitch or showing where a blood vessel is, it is a huge advantage. We do very technical procedures for achalasia — tunneling under the esophageal mucosa, down to the bottom of the esophagus, and cutting through the muscle. The vessels in there are hard to see, and AI can see them with pattern recognition and color them for you, reducing your chances of hitting a blood vessel.

Increased gut permeability is 100% real. The phrase "leaky gut" is irritating because you do not know what the user means — someone might blame whole constellations of symptoms on it, and lay literature's "leaky gut is associated with XYZ" may not mean what I mean. A study last year grew organoids — three-dimensional cultures — from small-bowel biopsy stem cells. In obesity and MASH, tight junctions were less developed and disorganized; transcriptomics showed they were not even making RNA for tight-junction proteins — clear transcriptional downregulation. Another group used tracers: 51 chromium EDTA is small but should not cross tight junctions. In MASH patients it zipped into the bloodstream at far above-normal levels; without MASH, and in treated celiac disease, it stayed out; in fatty liver disease it was getting in, probably playing a role. What crosses matters: the gut's first stop is the liver via the portal circulation, and bacterial products like LPS — lipopolysaccharide from gram-negative membranes — are inflammatory. They interact with toll-like receptor 4 and start inflammatory cascades through NF-kB. A Duke group injected LPS into healthy people: inflammatory markers went through the roof, and clamp studies showed insulin resistance. So yes, increased gut permeability is associated with real illness, with hardcore good science — I just do not want to blame it for everything.

I think they are better than high fructose corn syrup, for sure — high fructose corn syrup is the thing we should be treating like alcohol. Fructose in fruit is fine because it comes with a matrix around it — not a rush of fructose into your liver. Fructose can only be processed by the liver, and in a beverage it is absorbed rapidly, goes directly to the liver, gets trapped, and has to be dealt with. Even juices are kind of processed fruit; minimally processed is better. The problem with artificial sweeteners is that they come in foods that are highly processed as it is, and you cannot separate the two. That is also why people were down on polyunsaturated fats for a while — they come in a bar full of other stuff that is not good for you. The food is bad for you; the polyunsaturated fat itself has been shown to reduce LDL and has health benefits.

I do the same. Olive oil is the best, obviously, then small amounts of butter — large is bad; the problem is the overall amount of saturated fat. Polyunsaturated fats have a lot of proof they are very safe. Problems come from a big container that oxidizes in the sun before you finish it — you do not want oxidized oil in your body — or repeated deep frying, which generates trans fats. Even the old omega-3-to-omega-6 ratio worry is unnecessary: a minimal amount of omega-3s, and fish once in a while covers it.

One thing that is very exciting is gene therapy. We talked about GLP-1s being mega dosing — super physiologic and not nutrient responsive. A company is working on a new gene-therapy approach, and I was involved in the very early work. They built a viral vector carrying the GLP-1 gene under the beta-cell insulin promoter. So when a patient secretes insulin in a nutrient-responsive way, they simultaneously secrete GLP-1 into the same vesicles — and at mealtime the vesicles release GLP-1 and insulin together. Viral vectors are a beautiful tool of biology: genetic cargo in a virus that causes no problems, driving stable expression of certain proteins in a cell. Delivery is with endoscopic ultrasound — the same kind of device we developed to biopsy the pancreas — a fine-needle injection of the viruses into the tail of the pancreas. You would not take it intravenously because the vectors end up in other tissue; animal studies with green fluorescent protein confirmed they stay in the pancreas tail. The only place it becomes active is in the beta cells, not the alpha cells. And the drug is not made in the L cells where it must travel up through the liver — it is made where it is needed, at the pancreas: paracrine, and much faster.

That is what I am hoping. It is very exciting: they just entered clinical trials in the Netherlands, and a one-time GLP-1 injection looks very promising. It could also augment the gastric or small-bowel procedures — another tool in the armamentarium. It might be more use for diabetes than weight loss. It is early, but certainly very encouraging.

My mother would attest to that, unfortunately. I took my motorcycle apart in high school and could not get it back together — it had to be flatbedded away and fixed. I fixed parts in my car that ended up bursting into flames, so I am much better at dealing with patients. I always tinkered with things and needed to do things with my hands. That is why in medical school I could not be a general internist: I needed to solve problems with my hands, and that is fulfilling. I do not like managing a slow demise. In internal medicine I felt we were giving people a reason to continue with their current life instead of addressing the problem — blood pressure high, take a medicine, rather than really helping them address it. We treat high LDL and ApoB very effectively, but we took our eye off the ball: fatty liver is up, diabetes is up, people are still dying in greater numbers. We need to address the underlying problem. And I just like doing things with my hands — a large part of why I was headed into cardiology or interventional gastroenterology.