What We Know About Long-Term Safety

“We don’t know the long-term effects” is said about these drugs constantly, usually as though it were a decisive argument. It is worth unpacking, because the statement is partly true, partly false, and almost never quantified.

How much data actually exists

The class is considerably older than the current attention implies.

So: roughly two decades of clinical use for the class, with large randomised trials following participants for 2 to 4 years under close monitoring, plus accumulating registry and post-marketing data across millions of person-years.

That is materially different from a new drug approved last year. It is also not a lifetime.

The outcome trials are safety data, not just efficacy data

This point gets missed. A cardiovascular outcome trial is not only asking “does it help?” — it is watching thousands of people for years and counting everything that happens to them.

SELECT randomised over 17,000 participants with overweight or obesity and established cardiovascular disease, without diabetes, and followed them for a mean of about 3.3 years. It found a reduction in major adverse cardiovascular events. LEADER and SUSTAIN-6 did the same in type 2 diabetes. FLOW did it in chronic kidney disease, reporting a 24% reduction in major kidney events.

When a drug reduces hard endpoints across several years in tens of thousands of closely watched people, that is simultaneously evidence of benefit and evidence that no large, common, serious harm is being missed. It does not exclude a rare or a late one. The detail is in the heart data and the trial table.

What has actually emerged over years

Mostly the familiar things, persisting or fading:

• Gastrointestinal effects — usually worst during titration, improving substantially at a stable dose
• Gallbladder disease — a genuine increase, driven largely by the rate of weight loss rather than by the drug directly, as covered in GLP-1s and your gallbladder
• A small rise in heart rate — a few beats per minute, not associated with worse cardiovascular outcomes in the trials
• Pancreatitis — remains in the warnings; large studies have not established the clear signal once feared
• Thyroid C-cell tumours — the rodent finding has not translated into a demonstrated human risk, and the human data remains mixed, as set out in the thyroid cancer warning explained
• Lean mass and bone density loss — real, expected with any substantial weight loss, and largely modifiable

That last one is the long-term issue I think about most, and it is the one most within your control. It is the entire argument of muscle is the whole game and the bone health rankings.

What is genuinely unknown

I would rather name these than pretend they do not exist.

Decades of continuous exposure. Nobody has taken semaglutide for 20 years, because it has not existed for 20 years. What continuous use from 35 to 60 does is unknown and will remain so until people have done it.

Younger, non-diabetic, long-duration users. The trial populations skewed older and sicker than many current users. Someone starting at 32 with a BMI of 32 and no other conditions is not well represented in the evidence base.

Cumulative musculoskeletal effects. Repeated cycles of loss and regain, over decades, with variable attention to protein and training — that is the real-world pattern and it is not what trials measured.

Pregnancy exposure. These drugs are not for use in pregnancy, data is limited by design, and unplanned pregnancies happen — particularly because fertility often returns as weight falls, which is why the contraception point in GLP-1s and your other medications matters.

What happens across repeated stopping and starting, which is now extremely common for reasons of cost and supply rather than clinical ones.

The comparison that is usually missing

“We don’t know the long-term effects” is only half an argument, because it is always made against an implied alternative of no risk.

The alternative is not no risk. The long-term consequences of obesity, type 2 diabetes, obstructive sleep apnea and their complications are extremely well characterised — decades of data, large effect sizes, and outcomes nobody disputes. Those risks are running right now, with certainty, while the uncertainty about 20 years of drug exposure remains hypothetical.

That is not an argument that everyone should take these drugs. It is an argument that the uncertainty has to be weighed against something rather than against zero.

It is also worth noting that we do this constantly and without anxiety elsewhere. Statins, antihypertensives, inhaled steroids and oral contraceptives are all taken for decades on the basis of trials lasting a few years. The reasoning applied to this class is not usually applied to those.

What I would actually do

Assume it is long-term treatment. The regain data after discontinuation is clear, and framing this as a course you complete sets you up for the outcome in what happens when you stop.

Get the monitoring right, because that is what converts an unknown into something you would notice: baseline and periodic kidney and liver function, HbA1c, lipids, vitamin D, B12 and ferritin. Reading your own labs covers it.

Protect muscle and bone deliberately. This is the long-term risk you can most change, and almost nobody does it seriously.

Stay findable. Post-marketing surveillance depends on people reporting things. If something unusual happens, report it rather than assuming somebody else will.

Revisit the decision annually rather than treating it as settled. What is known will change, and it should be allowed to change your mind in either direction.