On the complications people actually notice, modestly and in two directions at once. Diabetic foot ulcers occurred in 2.2% of GLP-1 patients at one year against 2.7% of matched patients on a DPP-4 inhibitor, a hazard ratio of 0.813 [1]. Charcot neuroarthropathy moved the opposite way in the same cohort. Nerve damage itself was not measured, and the endpoints counted are what the damaged nerve leads to. What else these drugs are recorded as moving is catalogued in the outcomes atlas.
The comparison that found both results
Adults with type 2 diabetes carrying a neuropathy diagnosis were identified in a US research network as they started either a GLP-1 or a DPP-4 inhibitor. Matching on demographics, other conditions and medications left 19,770 in each arm, followed for one and two years, with a Bonferroni correction applied to the significance threshold.
The ulcer result was a hazard ratio of 0.813, 95% CI 0.716 to 0.922. In absolute terms that is half a percentage point, or five fewer people per thousand in a year, which is real and smaller than an 18.7% relative reduction sounds. All-cause mortality was lower too, and the authors label that one hypothesis-generating rather than a finding.
Charcot neuroarthropathy is the collapse of bones and joints in a foot that has lost protective sensation. It occurred in 0.3% of GLP-1 patients against 0.2%, a hazard ratio of 1.993, 95% CI 1.297 to 3.064. That is roughly double, and roughly one extra case per thousand people per year. Both descriptions are accurate, and a limb-threatening problem does not stop mattering because its absolute number is small.
The hardest endpoints did not move at all. Amputation ran at 0.5% in both arms, hazard ratio 1.073, 95% CI 0.810 to 1.421, and foot and ankle osteomyelitis ran at 0.4% in both, hazard ratio 0.978, 95% CI 0.708 to 1.349. A benefit on ulcers, a signal on Charcot, and no change in amputation is a picture that resists a headline. That is usually what a real safety signal looks like before anyone has sorted it out.
Walking distance, measured in a randomized trial
The neuropathy evidence is observational, so the randomized result next to it is worth setting out. Between October 2020 and July 2024, 792 patients with symptomatic peripheral artery disease and type 2 diabetes were randomized to semaglutide or placebo. The primary outcome was maximum walking distance at week 52 [2].
The semaglutide arm reached 1.21 times its own baseline distance and the placebo arm 1.08 times, giving an estimated treatment ratio of 1.13, 95% CI 1.06 to 1.21, p = 0.0004. The drug did not make people walk 21% further. It made them walk about 13% further than they would have anyway, and the placebo arm improving is exactly what a trial exists to reveal.
Serious treatment-related adverse events were rare on both sides, at six events in five participants on semaglutide against nine events in six on placebo, with no treatment-related deaths. Everyone in that trial had type 2 diabetes, and the authors call for studies in peripheral artery disease without it.
The hand, and a number a buyer can use
Trigger finger is a thickened tendon catching as it slides through a pulley in the palm, and diabetes makes it commoner. Adults with type 2 diabetes who had already had a steroid injection for it were matched 5,863 per arm [3]. At six months nothing separated, at 6.1% going on to surgery against 6.8%.
At twelve months the gap opened to 10.6% against 11.9%, a relative risk of 0.89, 95% CI 0.81 to 0.99. The authors then did what almost nobody does and converted it: 71 people take the drug so that one of them avoids the operation at a year, and 59 at two years. The twelve-month interval runs to 0.99, so one more surgery in the treated arm would cross it.
Who these results are actually about
Every population here had diagnosed type 2 diabetes, and most people buying a GLP-1 from a telehealth seller do not. If you have neuropathy, this is a conversation for whoever manages your feet, and it is not one a subscription intake form will start. What a prescriber is supposed to check while you are on the drug is set out in the monitoring nobody quotes you a price for.
Both arms of the ulcer study are drug classes rather than products, so nothing in it separates semaglutide from tirzepatide. That limitation runs through most of the non-weight evidence, including the inflammation findings and the kidney evidence. The same exposure problem applies as in every claims cohort, including the surgical outcome studies: a filled prescription is a filter, not a randomization.