Why psychiatric medications make stopping hard, why hyperbolic tapering works, and what the research actually shows. Every claim below links to its source.
The science
Long-term benzodiazepines physically rewire the brain's calming system
Benzodiazepines amplify GABA-A receptors, the brain's main 'brakes.' With chronic use the brain compensates by down-regulating and physically removing those receptors while glutamate (the 'accelerator') rises. The result is a nervous system wired to be hyper-excitable the moment the drug is reduced. This is a physical neuroadaptation, not weakness and not 'the original anxiety returning', which is why it feels so hard, and why a taper has to give the brain time to rebuild.
The evidence ›
Diazepam selectively down-regulates the GABA-A receptor α1-subunit gene and drives receptor endocytosis via a PLC-δ/Ca²⁺/calcineurin cascade, reducing inhibitory tone while glutamatergic signaling rises.
The last few milligrams are the hardest, because of pharmacology, not willpower
Dose and receptor occupancy relate hyperbolically, not linearly. For citalopram, dropping from 20mg to 15mg changes serotonin-transporter occupancy by only ~3%, but dropping the final 5mg to 0 changes it by ~58%. A 'small' end dose is doing a huge amount of work in your brain. Cutting by a fixed amount feels easy at first, then slams you at the end. Hyperbolic tapering makes each step smaller as the dose lowers, so the brain feels roughly equal ~10% changes the whole way down.
Even mg cuts → a cliff at the end Hyperbolic cuts → even ~10% steps
The evidence ›
Horowitz & Taylor (Lancet Psychiatry 2019): SSRI dose–occupancy is hyperbolic. Citalopram 20→15mg ≈ 3% SERT drop; 5→0mg ≈ 58% drop. A hyperbolic regimen (20, 9.1, 5.4, 3.4, 2.3, 1.5, 0.8, 0.4, 0 mg) yields ~10% occupancy steps.
Tiny daily steps produced 'very little' withdrawal; big weekly steps produced far more
This isn't just theory. In 608 patients, those tapering in tiny daily steps (~4.5%/day) rated withdrawal around 2–3 on a 1–7 scale, while those using large weekly steps had significantly worse withdrawal, especially in the first three months. The size and shape of the steps materially change how it feels, and a platform that recalculates micro-steps and tracks symptoms daily is operationalizing exactly the approach that performed best.
Big weekly cuts Daily micro-steps (Subside’s approach)
The evidence ›
Van Os & Groot (2023), n=608, 32,368 observations: daily tiny-step tapering (mean 4.5%/day) → withdrawal ~2–3/7; weekly large steps (mean 33.4%/step) → significantly higher withdrawal, especially months 1–3 (p<0.001).
With the hyperbolic method, ~72% came off successfully, even though most had failed before
Hyperbolic tapering works for people who previously couldn't stop. In 824 patients using hyperbolic tapering strips, 72% discontinued short-term and 66% were still off 1–5 years later, and 71% had already tried and failed without the method. First-time taperers succeeded 76% of the time; even those with 3+ prior failed attempts succeeded 57% of the time. The method, not 'willpower,' was the difference, and it's been replicated across independent cohorts.
The evidence ›
Groot & van Os (2021), n=824: 72% short-term and 66% long-term (1–5 yr) discontinuation; 71% had previously failed; first-timers 76%, 3+ prior attempts 57%. Third independent replication.
Stopping benzodiazepines cold turkey can cause seizures and delirium that are life-threatening
When the brain is in a GABA-underactive, hyper-excitable state, abruptly removing the drug can tip it into seizures, up to status epilepticus, plus withdrawal delirium and psychosis. Benzodiazepines and alcohol are among the few drugs where the withdrawal itself, not overdose, can kill. That's exactly why a gradual, controlled taper exists, and why our reductions are engineered to keep you inside a safe zone.
The evidence ›
Abrupt cessation after long-term/high-dose use can produce withdrawal seizures and delirium that may be fatal; near-uniform consensus that abrupt discontinuation is unacceptable for long-term users.
~56% of people get withdrawal coming off antidepressants, you're not imagining it
For years, antidepressant withdrawal was dismissed as mild and brief. A major systematic review found roughly 56% of people experience withdrawal when stopping, with a large share rating it severe. (Estimates vary by method and the debate is ongoing, we show that honestly.) If stopping has been hard, that matches the evidence, which is exactly why a slow, supported, tracked taper is worth it.
The evidence ›
Davies & Read (2019): withdrawal incidence across 14 studies 27–86%, weighted ~56%; ~46% of those affected rated it most-severe. Later meta-analyses (e.g. Henssler 2024) estimate lower; disclosed honestly.
About 1 in 4 older adults given a benzodiazepine ends up on it long-term
This is not a fringe problem. Guidelines say benzodiazepines should rarely be used beyond 2–4 weeks, yet long-term use is widespread, and most people were never given a structured way to come off. If you've been on these longer than planned and feel stuck, you're part of a very large group, and the gap is in the support system, not in you.
The evidence ›
~25–30% of older adults prescribed a benzodiazepine take it longer than recommended; guidelines advise against use beyond 4–12 weeks; dependence among long-term users ranged 40–63% in a Dutch study.
SSRIs also drive neuroadaptation, they need a gentle exit, not a switch-flip
Like benzodiazepines, SSRIs/SNRIs change the brain: with continuous high serotonin-transporter occupancy, downstream signaling adapts. Remove the drug abruptly and that adapted system is suddenly unmasked, producing discontinuation symptoms (flu-like feelings, insomnia, nausea, imbalance, 'brain zaps,' hyperarousal). Because occupancy is hyperbolic, the biggest change is at the very end of the dose range, so the exit has to be engineered around pharmacology, not a calendar.
The evidence ›
Clinical SSRI doses produce ~80% serotonin-transporter occupancy following an Emax/hyperbolic relationship; this underpins why abrupt removal unmasks adapted signaling and why low-dose reductions have outsized effects.
For a minority, withdrawal can last months, which is why we go gently from the very start
Most withdrawal is time-limited, but a real subset experience protracted (post-acute) withdrawal lasting far longer. For antidepressants, documented protracted cases ranged 5–166 months; for benzodiazepines, symptoms usually resolve within 6–18 months, with rarer multi-year cases. The risk of a drawn-out course is part of why going slowly and gently, rather than crashing the dose, is the safer bet.
The evidence ›
Systematic review of PAWS after antidepressants: protracted duration 5–166 months (mean ~37); evidence sparse/low-certainty. Benzodiazepine symptoms typically resolve within 6–18 months, rarer protracted cases for years.
Why a tracked, personalized taper beats usual care
Daily micro-step tapering with symptom tracking matches the exact method that produced the lowest withdrawal in real cohorts, generic care almost never does this.
Best real-world outcomes came from tiny daily steps (~4.5%/day) tracked over time (withdrawal ~2–3/7) vs large weekly steps (n=608). Prescription pads can’t recalculate sub-milligram daily doses; software can.
A personalized hyperbolic schedule beats willpower and beats linear tapering, because the difficulty is pharmacological, not motivational.
72% came off using hyperbolic tapering strips, including 57% of those who failed 3+ times before. Even-mg approaches systematically hit a wall at the end (5→0mg = 58% occupancy drop).
Structured tapering support roughly triples success versus usual care.
Supervised tapering: 62% discontinuation vs 21% usual care; adding structured psychological support pushed it to 76% vs 25% for slow taper alone. Most long-term users were never offered any structured program.
Daily symptom data lets the schedule adapt and pause, a feedback loop a static, written-once prescription cannot offer.
Withdrawal severity concentrates in early months and at the low-dose end; individuals vary enormously (protracted mean ~37 months). A platform ingesting daily ratings can slow or hold steps when symptoms rise.
Educational information, not medical advice. Taper only with a qualified prescriber. In crisis, call or text 988 (US) or your local emergency number. Safety & crisis resources →