Doctor: 92 million Americans on deadly drug

CONNOR'S HEALTH NOTES

The Number That Belongs on Every Prescription

Connor Hill · InsightfulWord · August 17, 2026

Before anything else: nothing in what follows is a reason to stop taking a prescribed medication. Stopping certain drugs abruptly — blood pressure medication, antidepressants, antiseizure drugs, corticosteroids, anticoagulants — can produce consequences considerably worse than anything the drug was doing. Every question raised here is a question for a prescribing physician, and the useful version of it is asked at an appointment rather than answered alone in a kitchen.

With that established, there is a genuine and underexamined problem in how medication benefit is communicated, and it has nothing to do with any particular drug. It concerns a single arithmetic choice that determines whether a treatment sounds transformative or trivial, and the choice is almost always made in the direction that sounds better.

Consider a hypothetical trial. Over five years, four people in every hundred taking a placebo have a heart attack. Among those taking the drug, three in every hundred do. That result can be stated two entirely accurate ways.

The first: the drug reduces heart attacks by 25 percent. This is the relative risk reduction — the difference expressed as a proportion of the original risk.

The second: the drug prevents one heart attack for every hundred people treated for five years. This is the absolute risk reduction, one percentage point, and its reciprocal is the number needed to treat — a hundred people, five years, one event avoided.

Both describe the same trial. Neither is false. The first appears in nearly every headline, press release and patient conversation; the second appears in almost none of them, and it is the one from which a person can actually reason about whether a medication is worth taking. The asymmetry is not conspiratorial. Relative figures are larger, more memorable, and comparable across trials in a way absolute figures are not. But the effect on a patient's understanding is severe and well documented: the same trial result produces markedly different willingness to take a drug depending on which framing is presented, and the relative framing consistently produces more enthusiasm than the underlying data warrants. The reason this matters more for older readers than for anyone else is that they are prescribed more medications, for longer, on the basis of preventive rather than symptomatic indications — which is precisely the category where the gap between the relative headline and the absolute reality is widest. A drug taken to relieve pain announces whether it is working. A drug taken to prevent an event twenty years out never does.

What follows is how to convert one number into the other, why the same drug helps different people by wildly different amounts, and what to ask at an appointment.

Why the Same Drug Has Many Different Answers

The most important thing absolute risk reduction reveals is that a medication's benefit is not a property of the drug. It is a property of the drug and the patient together.

Preventive medications work by removing a fraction of an existing risk. If the underlying risk is large, removing a quarter of it prevents a great deal. If the underlying risk is small, removing a quarter of it prevents very little, even though the relative figure — the 25 percent — is identical in both cases.

A person with established cardiovascular disease, diabetes and a history of smoking carries a baseline risk many times that of a healthy person of the same age with normal blood pressure and no family history. The same drug, with the same relative effect, produces a number needed to treat of perhaps thirty in the first person and several hundred in the second.

This is why blanket statements about whether a class of medication is worth taking are close to meaningless. The honest answer to whether a preventive drug is worthwhile is a question in return: worthwhile for whom, at what baseline risk, over what horizon.

It is also why the risk calculators that clinicians use exist. Their function is to estimate an individual's baseline risk from age, sex, blood pressure, cholesterol, diabetes status and smoking, so that the relative effect from the trials can be applied to a number that describes the actual person in the room rather than the trial population's average.

The Harms Have the Same Arithmetic

Everything above applies identically to side effects, and this is the half that fear-based content systematically distorts.

A side effect can be described relatively — a drug doubles the rate of some adverse outcome — or absolutely, as the number of additional people harmed per hundred or per thousand treated. A doubling sounds alarming. A doubling of an event that occurs in one person per ten thousand per year produces one additional case per ten thousand per year, which is a different proposition entirely.

The same framing choice is available in both directions, and it is exploited in both directions. Marketing material for a drug tends to present benefits relatively and harms absolutely. Material attacking a drug does precisely the reverse.

The corrective is the same in either case. When any claim about a medication arrives, the question is what the absolute numbers are: how many people out of how many, over how long. If that information is not supplied, the claim has not been made in a form that permits evaluation, regardless of which direction it points.

Where Fear Content Uses the Same Trick

There is a symmetry worth naming plainly, because it is the reason this piece is about arithmetic rather than about any drug.

Content designed to frighten people away from a medication uses the identical statistical latitude that marketing uses to attract them. It reports the relative increase in a harm without the baseline. It cites the aggregate number of people taking a drug — a figure driven by how common the underlying condition is — as though prevalence were itself evidence of over-prescription. It lists adverse outcomes that appear in the literature without distinguishing established causation from reported association, and without indicating frequency.

The tell is consistent: a claim that supplies a percentage without a denominator, or a list of harms without rates, has been constructed to produce an impression rather than to support a decision.

A second tell is the absence of the counterfactual. Every medication decision has two arms, and the risk of not treating is as real as the risk of treating. Material that discusses only one arm is not analysis, whichever arm it discusses.

The Trial Population Is Not the Patient

There is a further gap between trial arithmetic and individual benefit that even the absolute numbers do not close, and it deserves stating because it cuts against overconfidence in both directions.

Randomized trials enrol participants who meet entry criteria and consent to participate. That population is systematically younger, healthier, less burdened with other conditions, and taking fewer other medications than the population to whom the drug is later prescribed. Trials frequently exclude people with significant kidney impairment, advanced age, limited life expectancy, or several simultaneous chronic conditions — which describes a substantial share of the people who end up taking the drug.

The consequence runs in both directions and neither is reassuring. Benefit may be smaller in the real population, because competing causes of death reduce the opportunity for any single prevention to matter, and because adherence outside a trial is lower than inside one. Harm may be larger, because interactions with other medications and reduced clearance in older kidneys are precisely what the exclusion criteria removed from the study.

The horizon compounds it. Preventive medications typically require years to deliver their benefit, because the benefit is the avoidance of an event that would have happened later. A person whose remaining life expectancy is shorter than the trial's follow-up period is being offered the harms on the drug's timescale and the benefits on a timescale they may not reach. That is a legitimate clinical consideration, it is discussed openly in geriatrics, and it almost never comes up in a general appointment.

None of this means the trials do not apply. It means the absolute numbers they produce are a starting estimate for a specific person rather than a finished answer, and that the adjustment is a conversation rather than a calculation.

The Part That Is Genuinely Contested

None of this argues that all criticism of prescribing patterns is manufactured. Some of it is well founded, and it is worth distinguishing the serious version from the sensational one.

The serious critique concerns thresholds. Guidelines define who should be offered preventive treatment, and lowering a threshold expands the treated population into progressively lower-risk people for whom the absolute benefit is progressively smaller while the side effect rate stays roughly constant. At some point those lines cross. Where exactly they cross is a genuine matter of expert disagreement, resolved differently by different guideline bodies in different countries, and it is a legitimate subject of debate.

The serious critique also concerns duration. Many preventive medications are started with an implicit assumption of permanence, and the balance of benefit and harm can shift as a person ages, accumulates other conditions, and adds other prescriptions. Periodic reconsideration is good practice and does not happen as often as it should.

What distinguishes that critique from the sensational version is that it operates on absolute numbers, specifies which patients it concerns, and produces a recommendation to reassess rather than to abandon. It also tends to appear in medical journals and guideline committees rather than in a video, which is where a reader can check it.

The arithmetic at the center of all of it is the same. Four in a hundred became three in a hundred. Whether that is a 25 percent improvement or a one-point improvement depends entirely on who is doing the telling, and the second version is the one from which a decision can be made.