The Prescribing Cascade: When a Drug's Side Effect Gets Its Own Prescription

NU ranks records over spin. This page summarizes the primary medical literature on a named, documented error called the "prescribing cascade" — it is not medical advice, not a diagnosis, and not a treatment plan. Drugs discussed here (thiazides, beta-blockers, diuretics) help millions of people and are first-line for good reasons. Nothing below tells you to take or stop anything. Talk to your doctor.

The term is real, not a fringe idea

"Prescribing cascade" is an established term in the geriatrics and pharmacology literature — not a wellness buzzword. A EuropePMC search for the exact phrase `"prescribing cascade"` returns 520 records (checked June 2026), including peer-reviewed titles like "Prescribing Cascade as a Therapeutic Error: A Danger for Geriatric Patients with Multimorbidity" and "Decision-Making and Downstream Outcomes of the Gabapentinoid-Diuretic Prescribing Cascade." (If you drop the quotes and search the two words separately, you get several thousand hits, but most of those just happen to contain both words and are not about the cascade — so we report the exact-phrase number.)

The mechanism is simple and uncontroversial: a drug causes a side effect, the side effect is read as a new disease, and a second drug is prescribed for it — instead of revisiting the first drug. Now the patient is on two drugs, each with its own side-effect profile, and the loop can repeat. Evidence level: well-documented in human observational studies and case series. This is descriptive epidemiology, not a lab curiosity.

A note on what these counts mean: a EuropePMC hit count is a literature-search proxy, not a curated tally of evidence quality. It tells you a topic is discussed, not how strong the studies are. We use it to show a field exists, nothing more.

A concrete, sourced example: blood-pressure drugs and blood sugar

The classic textbook cascade involves older blood-pressure drugs nudging blood sugar upward, which then gets read as new diabetes and treated with a glucose-lowering drug.

The signal that some antihypertensives are associated with new-onset diabetes is genuinely in the human literature. (A keyword search pairing terms like thiazide / beta-blocker / insulin resistance / new-onset diabetes returns several thousand co-occurring records — but treat that as "this is a discussed topic," not as a count of cascade studies, because keyword searches sweep in any paper mentioning the words.) Real titles in this body include work on antihypertensive therapy and metabolic impact in diabetes, and case reports of multi-drug interactions such as thiazide-induced hypokalemia interacting with statins and SGLT2 inhibitors.

Honest evidence level here matters a lot:

• That thiazide diuretics and older (non-vasodilating) beta-blockers are associated with a modest increase in new-onset diabetes risk is supported by human randomized and observational data — this is one of the better-established metabolic side effects in cardiology.
• That this association is clinically large enough to change first-line therapy for any given patient is NOT settled — the cardiovascular benefit of these drugs is also proven and often outweighs the metabolic cost. This is a trade-off, not a scandal.
• The full "cascade" — drug A raises sugar → clinician adds drug B without reconsidering drug A — is documented in case reports and observational / nursing-home cohort studies, not in large randomized trials designed to measure the cascade itself.

Do not read "associated with new-onset diabetes" as "these drugs give you diabetes." Association in a population is not causation in your body.

How big is the cascade problem? Be careful with the numbers

The cascades that have been quantified tend to be specific drug pairs in specific populations. The gabapentinoid–diuretic cascade (a nerve-pain drug causing fluid retention, then a water pill added) has a small but real evidence base — an exact-phrase search for that specific pair returns only on the order of 15–20 records, including nursing-home studies of the gabapentinoid–loop-diuretic cascade. The exact count shifts with how you phrase the query, which is itself the point: this is a niche, early literature.

What you will hear quoted in talks and social posts — sweeping figures like "X% of elderly hospital admissions are caused by prescribing cascades" — we could not source to a specific primary record and are flagging as unverified. The honest statement is narrower: cascades are real, repeatedly documented, and concentrated in older adults on many medications, but a single clean population-wide percentage is not something we can stand behind here. Treat any precise viral stat as unverified until you see the actual study.

The incentive structure — the actual thesis

Here is the part NU cares about, and it is about money and study design, not secrecy.

The fix for a cascade is often deprescribing — carefully stopping or swapping the offending drug under supervision. Deprescribing is unpatentable. Nobody can sell it. There is no revenue stream that funds a billion-dollar trial proving "stopping drug A is better than adding drug B."

You can see the asymmetry in the records themselves:

• Deprescribing in older adults is a substantial and growing field in the published literature — but mostly observational, guideline, and pilot work rather than large randomized trials.
• On ClinicalTrials.gov, the exact phrase `prescribing cascade` returns only 50 registered studies total (checked June 2026), and `thiazide new-onset diabetes` returns just 5 — and several of the top hits aren't even about cascades. The interventional-trial machinery is overwhelmingly pointed at new molecules, not at un-prescribing old ones.

That gap is the story. Not "they're hiding a cure" — there is no cure here to hide. The honest claim is structural: a profitable intervention (add a drug) gets industrial-scale randomized evidence; an unprofitable one (review and maybe remove a drug) leans on observational data and clinician judgment. Under-patentable tends to mean under-funded tends to mean under-studied. That is a real bias in the evidence base, and it is admissible without any conspiracy.

Safety — read this before doing anything

• Never stop a blood-pressure drug, beta-blocker, or any prescription on your own. Abruptly stopping some beta-blockers can cause dangerous rebound effects. The cascade is a clinician's decision to unwind, slowly, with monitoring.
• The metabolic side effects discussed are trade-offs, not proof your medication is wrong for you. For many people the cardiovascular benefit is the bigger number.
• "Deprescribing" is itself a medical intervention with risks if done wrong. The literature on it is real but still maturing — mostly observational and pilot-stage, not a settled randomized playbook.

Bottom line

• Proven / well-documented: The prescribing cascade is a real, named therapeutic error in the peer-reviewed literature (520 exact-phrase records), concentrated in older adults on multiple drugs.
• Human-trial / observational evidence: Older antihypertensives (thiazides, non-vasodilating beta-blockers) are genuinely associated with a modest rise in new-onset diabetes risk — but they also have proven cardiovascular benefit. It's a trade-off.
• Limited / early evidence: That a given patient's "new diabetes" was caused by their BP drug, and that stopping it is the right move, is case-by-case and supported more by observational data than randomized trials. Specific cascades (e.g. gabapentinoid–diuretic) have only a handful of focused studies.
• Unverified — flagged: Any precise population-wide percentage for cascade-caused harm. We did not source it; don't repeat it as fact.
• The real thesis (structural, not conspiratorial): Deprescribing is unpatentable, so the unprofitable fix gets far fewer registered trials (50 for "prescribing cascade" on ClinicalTrials.gov) than the profitable habit of adding another drug. Under-funded ≠ wrong; it means we have weaker evidence than we should.

Bring this to your doctor as questions, not conclusions. Don't start or stop any treatment based on this page.