No randomized trial has tested one for it. The evidence is a matched cohort of 4,153 patients per arm, and it found what a trial would not [1]. Blood pressure fell 5.7 mmHg on a GLP-1 against 6.3 mmHg on a mineralocorticoid receptor antagonist. On the thing both drugs were prescribed to do, the two groups landed in the same place. Everything afterward differed: MACE at HR 0.63, kidney events at 0.64, all-cause mortality at 0.34. A two-thirds reduction in death with the treated quantity equalized is a signature of confounding.
Resistant hypertension means pressure still uncontrolled on three drugs at once. It is a specific diagnosis, not a description of a bad reading. The question here is whether a GLP-1 belongs as the fourth drug, and the general blood-pressure and vascular evidence is a different matter, set out in the vessel-function measurements.
What was compared
Adults with overweight or obesity and resistant hypertension, across 67 US healthcare organizations, who started a fourth drug between 2017 and 2025 [1]. Those starting semaglutide or tirzepatide were matched against those starting spironolactone or eplerenone, the guideline-recommended option. Of 213,309 eligible patients, 22,694 started a GLP-1 and 5,673 an MRA. Propensity matching left 4,153 in each group.
Over a median 1.4 years the GLP-1 group had lower risks across the board. MACE HR 0.63, 95% CI 0.52 to 0.78. Cardiovascular events HR 0.74, 95% CI 0.59 to 0.92. Major adverse kidney events HR 0.64, 95% CI 0.46 to 0.88. Acute kidney injury HR 0.62, 95% CI 0.46 to 0.83. All-cause mortality HR 0.34, 95% CI 0.21 to 0.55.
And the blood pressure
Systolic pressure fell 5.7 mmHg on the GLP-1 and 6.3 mmHg on the MRA at twelve weeks, with overlapping intervals. On the thing both drugs were prescribed to do, the two groups were the same. The MRA arm did marginally better.
Consider what it takes to be in each group. Spironolactone is a decades-old generic costing a few dollars a month. A clinician can add it this afternoon. Getting semaglutide or tirzepatide approved as a fourth-line blood pressure drug takes an insurer’s cooperation, a prescriber willing to pursue it, documentation, and often the patient’s own money.
The people who clear that are better insured, more engaged with their care, and well enough to be considered for an elective addition. Those characteristics predict survival on their own. Propensity matching adjusts for what was recorded, and none of that is recorded. It is the same shape of confounding visible from the access side in the uptake registry.
What the trials reported, which is nothing
A Cochrane review now in its fourth update pooled eight randomized trials of long-term weight drugs in adults with hypertension, about 13,000 participants [2]. For both semaglutide and tirzepatide it records that no results were reported for all-cause mortality, cardiovascular morbidity, serious adverse events or adverse events. Liraglutide could not be entered at all, because no trial has published separate figures for hypertensive participants.
That is not a safety finding. It says the trials did not report the outcomes a reviewer needs, in a population where the question is unusually live. The older drugs did report. Orlistat probably increases serious adverse events at RR 1.45, 95% CI 1.10 to 1.91, across three trials and 1,476 participants. The full reading is in the Cochrane update.
What the randomized evidence does support
In broader populations the hard outcomes are real and much smaller than the cohort suggests. A meta-analysis of 11 trials and 85,373 people found MACE at HR 0.87, 95% CI 0.81 to 0.93, alongside kidney benefits [3]. That is a 13% relative reduction, not a 37% one, and it comes from randomization rather than from matching. Stroke measured on its own is thinner still, as the stroke evidence sets out.
So the practical position is narrow. A GLP-1 achieved about the same pressure reduction as the guideline drug in a population where pressure is hard to move. The outcome gap beside it is not credible at the size reported. The authors say prospective studies are needed to determine whether these drugs should be part of treatment for resistant hypertension, and that is the correct conclusion from their own data.
Nobody on this roster sells anything for resistant hypertension. What sellers publish about who they will and will not prescribe to is thin. The ordinary-care results show what happens to trial figures once access decides who is in the sample.