Ozempic May Cause Eye Problems - But I Doubt it

One of the most flawed studies of the year gained major media attention. Yet there are great teaching points for those interested in critical appraisal

The GLP1 agonists (glucagon-like-peptide-1-receptor agonist) may be the biggest medical development of this century.

The success of Western society has resulted in widespread obesity—and obesity-related diseases, such as diabetes, high blood pressure, heart failure and vascular disease. Obesity is clearly not a good thing for human health. The statistics on obesity—especially in children—are staggering.

Yet we have a drug class that helps patients with obesity.

Randomized clinical trials have demonstrated that GLP1a reduce cardiac outcomes in patients with diabetes and most recently in patients with obesity and established cardiac disease. Most of this disease-modifying effect stems from weight loss.

We can debate the absolute benefits, the cost efficacy or the philosophy of using drugs to modify disease largely due to lifestyle. But that is not what this column is about.

This week I write to you about potential harms of these drugs.

The journal JAMA-Ophthalmology published an observational study suggesting that semaglutide (Ozempic) may cause a serious eye condition associated with blindness in adults. The authors danced around the verb cause, writing that they discovered only an association.

The eye disease is non-arteritic anterior ischemic optic neuropathy or NAION. It’s a rare condition; the authors tell us its incidence is in the range of 2-10 in 100,000.

Here is an example of suggestive language (emphasis mine)

The relatively high HRs (4.28 and 7.64 for our T2D and overweight or obese cohorts, respectively) reveal a substantially increased risk of NAION among individuals prescribed semaglutide relative to those prescribed other medications to treat T2D and obesity or overweight. This risk appears not to be due to differences in baseline characteristics between the cohorts.

Important Area of Study

It is possible that billions of people will take these drugs. Knowing whether there are rare adverse effects is important.

RCTs are most effective at determining efficacy. Trials aren’t nearly as good at detecting adverse effects—for two reasons: one is the small numbers of patients and the other is that trials are short.

Uncommon or rare adverse effects are best detected by large observational studies.

So you might expect the JAMA-Ophthalmology paper to have included huge numbers of patients with careful analyses. No. That is not what happened.

The Study

The mostly ophthalmologist-authors from Harvard medical school took patients referred to their clinic. Stop there. Read again. Took patients referred to their special neuro-eye clinic at Harvard.

About 710 patients had diabetes and 980 had obesity. About a third of each group had had exposure to semaglutide. Two-thirds did not. They matched these patients on basic variables.

Then they looked back at how many in each group had NAION and how many did not have NAION.

They found that the incidence of NAION in patients with T2D or obesity was 4-6x for semaglutide vs non-semaglutide patients.

That is it. For good measure, they displayed the Kaplan-Meier plot with truncated y-axis, to accentuate the effect.

Four Reasons This Study Does not Answer the Question

Collider bias: There is a huge collider bias as these were patients referred not just to an eye surgeon, but a specialty clinic at Harvard. If you are referred to such a clinic you are special. Looking only at these patients would be like looking at the role of height and speed in NBA players. Collider bias is also the origin of the false notion of the obesity paradox.

Time-Zero and Immortal Time Bias: This occurs because the authors calculated person-time from the first prescription of semaglutide vs non-GLP-1a diabetic or weight loss medicine. If something else caused the eye issue, then people on the newer drug will have a higher rate on the person-time scale. See this thread from Dr. Venk Murthy.

Unbelievable Effect Sizes: The authors posit that semaglutide increased the risk for a rare eye condition by 4-6x. If that were true, we’d be hearing about a run on blindness.

Poor Controls: Dr. Ed Livingston noted that they don’t control for diabetes severity. Semaglutide is used in more advanced cases of DM. That’s important because T2D and obesity are also risk factors for NAION.

(There were many other problems with this analysis. These four are the major ones.)

Misplaced Incentives - This is Sad

That this paper gets published in a JAMA journal, then JAMA promotes it, then it gets big media attention is problematic.

I would propose that it is 90% due to misplaced incentives and 10% misunderstanding about epidemiologic research.

Readers of Sensible Medicine know about the incentive problem. There are incentives to do such a flawed study, incentive for the journal to publish it because of the attention, and then incentive to cover it in the news.

Perhaps the authors were curious about a potential signal. Yet someone should have said to them that this is not the way to sort this signal out.

Somewhere along the line, perhaps a colleague at Harvard, or a peer-reviewer, or an associate editor would have told the authors to stop.

There are ways to study drug safety. But you need large databases and careful controls.

Looking only at patients from a specialty clinic reminds me of the neurosurgeon who commented after a lecture I gave on oral anticoagulation for stroke prevention in patients with atrial fibrillation that we were killing patients with blood thinners. Because neurosurgeons see brain bleeds. No patient checks in to their clinic or the ER to announce that they have not had a stroke because of the drug.

The best that we can do is learn from these flawed studies. Sensible Medicine hopes to spark change in the appraisal of medical evidence. Thanks for your support. JMM