Scientific deep-dive

When a Hazard Ratio Is Too Good

Three observational studies this month reported mortality reductions of 78%, 91% and 47%. Nothing in medicine does that. One rule explains all three, and it also shows which rows in the same papers to keep.

By Nora Bissett · Pricing Editor
Editorially reviewed. Not medical advice · How we verify contentLast reviewed
6 min read·1 citation

A study of people with psoriasis found those taking a GLP-1 had 78% lower all-cause mortality than those on other diabetes and obesity drugs.[1] Nothing in medicine reliably cuts all-cause mortality by three quarters, not even statins, blood pressure treatment or smoking cessation. When a number looks like that, it is usually telling you about the patients rather than the drug.

That figure is the third of its kind we have read this month, and a fourth has since turned up in a study of seizure recurrence. The confounding this argument rests on has also now been measured rather than inferred: people with a mental health diagnosis have roughly a quarter the odds of being prescribed one of these drugs and half again the odds of being put on insulin (see a quarter as likely to be offered it). For the constructive counterpart (an observational study reporting effects just as large that does deserve trust, and the specific techniques that earn it), see how to trust an observational study.

The pattern, three times over

All-cause mortality hazard ratios reported in observational GLP-1 studies covered on this site this month.
PopulationHazard ratioComparator
Breast cancer with diabetes0.09Insulin or metformin — and no difference against SGLT2 inhibitors in the same paper
Psoriasis0.219Other antidiabetic and antiobesity drugs
Kidney transplant recipients0.53Non-users, pooled across four studies

Each appeared in a real journal, with matching, adjustment and sensitivity analyses. None is fraudulent. If a drug did produce any of those effects, it would be the most important medicine of the century, which is a useful sanity check on its own.

Ask what a true effect of this size would mean. If the answer is “the greatest advance in the history of medicine”, look at the comparison group.

What produces numbers like this

One mechanism explains all three: the people who receive a newer, more expensive, more demanding drug are not like the people who do not.

  1. They are well enough to be prescribed it. Clinicians do not start a weight-loss drug in someone acutely unwell, frail, or approaching end of life. Those patients stay in the comparison group and die at the rate very sick people die.
  2. They are engaged with care. Getting and staying on one of these drugs requires appointments, prescriptions, prior authorizations and money. That selects for people already doing better on every axis.
  3. The comparator is often a marker of severity. Insulin marks advanced diabetes. “Other antiobesity drugs” can mark people who failed or could not tolerate the newer option.

None of that is measurable in a claims database, so no amount of statistical adjustment removes it. Matching equalizes what was recorded, and the strongest predictors of death (frailty, functional status, whether someone is quietly dying) are largely not recorded.

The tell inside this particular study

The psoriasis paper reports something its authors treat as a finding and which reads more naturally as a warning: the risk reductions were markedly higher in people with psoriasis than in cohorts with obesity or diabetes but without it.[1]

A biological story could be told about that: psoriasis is inflammatory, and these drugs reduce inflammatory markers. But there is a duller explanation. In a population with more comorbidity, prescribing is more selective, so the gap between who gets the newer drug and who does not grows wider. A differential that tracks how sick the population is, rather than tracking anything about the disease mechanism, is the signature of confounding rather than of biology.

Which rows to believe in the same study

This is the part that matters, because the answer is not “ignore the paper”. The same study reports figures that are entirely plausible.

  • Major adverse cardiac events, HR 0.561. Consistent in size and direction with randomized cardiovascular outcome trials. Believable.
  • Alcohol abuse 0.346 and substance abuse 0.510. Larger, but consistent with a randomized trial in alcohol use disorder and with national cohort data on substance use. Worth taking seriously.
  • No increase in typical adverse drug events. A null in a large sample, which is the kind of finding observational data handles well.
  • All-cause mortality 0.219. Not believable as a drug effect, for the reasons above.
All-cause mortality is the outcome most vulnerable to this problem, precisely because everything affects it. A specific endpoint, such as a heart attack, a documented flare or a lab value, is harder for unmeasured frailty to drive. When a study’s specific outcomes look reasonable and its all-cause mortality looks miraculous, the mortality figure is the one to discard.

The rule, portable

  1. Find the comparison group before the result. Lower than what?
  2. Distrust hazard ratios below about 0.5 in non-randomized data. They usually describe who was prescribed the drug. One semaglutide-versus-tirzepatide study reached 0.43 only after dropping the people who stopped.
  3. Check whether the effect survives a fair comparator. In the breast cancer study it vanished entirely against a modern one.
  4. Weight specific outcomes above all-cause mortality in observational work.
  5. Ask what a real effect that size would mean. Usually the answer settles it.

Applied across this month’s reading, that rule keeps the cardiovascular, infection and psychiatric findings and discards three mortality figures. It separates evidence about the drugs from arithmetic about who gets prescribed them.

Frequently Asked Questions

No. A psoriasis cohort reported an all-cause mortality hazard ratio of 0.219, but nothing in medicine reliably reduces all-cause mortality by three quarters. Figures like that describe who receives the drug: people well enough to be prescribed it and engaged enough to stay on it.
Because matching only equalizes what was recorded, and the strongest predictors of death (frailty, functional status, whether someone is quietly dying) are largely absent from claims data. You cannot adjust for a variable nobody wrote down.
No, and that is the useful skill. The same study's cardiac outcome of 0.561 is consistent with randomized trials, and its alcohol and substance-use findings match other work. The all-cause mortality figure is the one to discard, not the paper.
Because everything affects it. A specific endpoint, such as a heart attack, a lab value or a documented flare, is harder for unmeasured frailty to drive. When specific outcomes look reasonable and mortality looks miraculous, mortality is the suspect number.
Ask what a real effect of that size would mean. If the answer is that it would be the most important medicine of the century, go and look at the comparison group.

References

  1. 1.Olbrich H, Kridin K, Zirpel H, et al. Glucagon-like peptide-1 receptor agonists and reduced mortality, cardiovascular and psychiatric risks in patients with psoriasis: a large-scale cohort study British Journal of Dermatology. 2025. PMID: 40897378.

Who Loses More Weight on a GLP-1?

Across 64 randomized trials, the only subgroup difference was sex: women lost 10.9% against 6.8% for men. Age, race, ethnicity, starting BMI and starting HbA1c made no significant difference.

5 min read

Stopping Ozempic or Wegovy: Side Effects, Weight Regain and Keeping It Off

Stopping Ozempic or Wegovy causes no withdrawal syndrome on either label. Appetite returns as semaglutide clears over about five weeks, and a 2026 BMJ meta-analysis found regain of about 0.8 kg a month. What helps keep weight off, and what women should know.

10 min read

Exercise Plus a GLP-1: What Each One Actually Changes

In a four-arm randomized trial, the drug alone produced no improvement in physical fitness. Exercise did, with or without the drug, and the strength result needs reading carefully.

6 min read

GLP-1s in People Being Treated for Cancer

Lower mortality and fewer hospitalizations than metformin, at a hazard ratio of 0.875. A second comparison against insulin produced a bigger number, and the bigger number is the less trustworthy one.

5 min read

GLP-1 Drugs and Hidradenitis Suppurativa: Early Studies, No Trial

Early studies of liraglutide and semaglutide report improvement in HS severity and quality of life. There is no randomized trial, and weight loss by other means has had mixed results in this condition, which complicates the obvious explanation.

5 min read

What Instagram Leaves Out

Researchers compared semaglutide posts across five platforms against what the STEP trials recorded. The most useful finding is an omission: the drug's most common side effects were absent from Instagram.

6 min read

Where to get GLP-1 online, safely: sellers our editors have checked

These are telehealth sellers our editors have checked. For each one we hold a price, the form the drug comes in, and the states it reaches.

No insurance needed · vetted by our editors

Some of the links on this page earn us money. If you sign up with a provider after following one, that provider may pay GLP Watchdog a commission. Learn more

9.0

YourEra

Semaglutide at $99/month, well below the typical price

8.7

Eden

Coverage in all 50 states, which most sellers will not confirm

7.3

Found

People who want clinician visits billed to commercial insurance, with compounded medication in the plan price

From $149/mo + membership

Get started →
Read full review →

About this analysis

GLP Watchdog publishes independently. Each seller in our dataset is checked against that seller's own website, and the dataset is updated continuously. Read our full methodology →