Some blood pressure drugs were linked to less dementia than others, in records of 2 million Americans

High blood pressure in midlife is one of the known risk factors for dementia later on, and millions of people take pills to lower it. A new study asks whether the choice of pill might matter for the brain, not just for blood pressure. Using health records of about 2 million Americans, the researchers found that people who started an angiotensin receptor blocker (ARB) were diagnosed with dementia somewhat less often than people who started lisinopril, a widely used drug from a related family, and that the size of that gap varied a lot from one ARB to another.

The study is a preprint, not yet peer-reviewed, and it is observational: nobody was assigned a drug at random. The authors themselves say the differences need to be confirmed in randomized trials.

Two families of drugs

Both drug families act on the same hormone system. Angiotensin II is a hormone that narrows blood vessels and raises blood pressure. ACE inhibitors such as lisinopril stop the body from making it; ARBs such as losartan, candesartan or irbesartan block the receptor it acts on. Doctors usually treat the ARBs as interchangeable, but they differ in how long they last in the body, how tightly they bind to the receptor and how easily they reach the brain. Earlier studies had linked ARBs as a group to lower dementia risk, but had too few people to compare the individual drugs.

What they did

The team used the TriNetX network, a database of anonymised health records and insurance claims from 44 US health care organizations. Out of about 22 million people, 2,051,275 adults aged 55 to 80 had started either an ARB or lisinopril and had never taken a drug from either family before.

They used a design called target trial emulation: they wrote down the rules of the randomized trial they would have liked to run (who is eligible, when follow-up starts, what counts as an outcome) and then mimicked it with existing records. To make the groups comparable, each person starting a given drug was matched with a similar person starting the comparison drug, based on age, sex, other illnesses, other medicines and use of health care. They compared each of six ARBs with lisinopril, and the other five ARBs with losartan, the most prescribed one. The matched groups ranged from 10,113 to 857,773 people, with an average age of about 64, and were followed for a median of 4.7 to 7.2 years.

What they found

• ARBs as a group. People starting an ARB had about a 14% lower rate of a recorded dementia diagnosis than people starting lisinopril (hazard ratio 0.86; a hazard ratio compares how often something happens in two groups over time, and 1 would mean no difference).
• Big differences between ARBs. The gap was largest for candesartan (hazard ratio 0.66) and irbesartan (0.69) and smallest for losartan (0.93). Over ten years, the estimated difference in absolute risk ranged from about 0.7 to 3.1 percentage points, depending on the drug.
• Same order within the family. Compared directly with losartan, all five other ARBs were linked to less dementia, from candesartan (0.68) to valsartan (0.94), and the drugs fell in the same order as in the comparison with lisinopril.
• Stronger for vascular dementia. The links were larger for dementia caused by damage to blood vessels in the brain than for Alzheimer's disease.
• Losartan last throughout. Losartan had the least favourable figures for every outcome, including death from any cause.

What to keep in mind

• Observational. The authors stress that causal effects cannot be established and that hidden differences between the groups may remain. People who were prescribed one drug rather than another may differ in ways the records don't show.
• A warning sign in the checks. As a test for bias, the team looked at conditions no blood pressure drug should affect, carpal tunnel syndrome and hernia. These came out slightly more common in some ARB groups than in the lisinopril groups, a sign of some bias, though they were close to even when ARBs were compared with losartan. A known effect, fewer cases of a type of swelling (angioedema) with ARBs, did show up as expected.
• Record data. Dementia was taken from diagnosis codes without expert review. Blood pressure and whether people actually took their pills were not observed.
• Many comparisons. The analyses were not corrected for testing many drugs and outcomes, and the authors call the secondary results "hypothesis-generating".
• Prescription drugs. All of these are prescription medicines with possible side effects, chosen by a doctor for each patient. This study is not a reason to change or stop a medicine on one's own.

Why it matters

Finding cheap, well-known drugs that could lower the risk of dementia is attractive, because new dementia treatments are expensive and modest in effect. If the differences seen here turned out to be real, the choice between blood pressure drugs that work similarly well could also matter for the brain. For now, this is a large and careful analysis of records, and the authors' conclusion is that randomized trials are needed before anyone can say that one ARB is better for the brain than another.