This is every correction we have made to a published article, with the date and what changed. We do not quietly edit. If an article said something wrong, the correction stays here after the article is fixed, because a corrections page that only lists things nobody noticed is not a corrections page.
If you find something we have missed, email [email protected]. We will fix it and log it here, and we will credit you if you want to be credited.
23 September 2026
Six citation errors in “How Much Can I Actually Change My Scoliosis?”
Article: How Much Can I Actually Change My Scoliosis? What the Research Shows. An internal citation audit checked every study cited in this article against PubMed, Crossref, Europe PMC and OpenAlex. Six things were wrong. All six are now fixed in the article. This is what they were.
1. We cited a meta-analysis that does not exist
The article credited a pooled 3.2-degree Cobb change to “Fan Y, et al., Frontiers in Surgery, 2023;10:1029012″. No such paper exists. That DOI returns nothing at Crossref, and the Frontiers article ID belongs to a cardiology paper. The numbers were real but belonged to a different study: Ceballos-Laita L, et al., European Journal of Physical and Rehabilitation Medicine 2023;59(2):228-236 (PMID 36692412), which pooled six randomized trials covering 144 adolescents and found a mean change of 3.18 degrees. Fixed: the citation now points to the real paper, and we have added the authors’ own caveat that the change did not exceed the minimum clinically important difference.
2. We cited a second paper that does not exist
The article said “Schreiber’s 2019 follow-up study found that Schroth gains attenuate when supervised practice stops (Physiotherapy Canada 2019;71(4):345-353)”. There is no such paper. Those page numbers in that issue belong to three unrelated articles, and none of Schreiber’s thirteen indexed scoliosis papers is a long-term follow-up. Fixed: the claim and the reference are removed. The honest position, now in the article, is that nobody has tested what happens to Schroth gains after supervised practice stops. The one trial that did follow patients after a program ended (Monticone 2014) found the change held for at least a year, which is the opposite of what we had written.
3. We cited a real paper for the opposite of what it found
The article said “Collagen has a half-life of approximately 100 days (Heinemeier et al. 2013, FASEB Journal)”. That paper is real, and it found close to the opposite: using radiocarbon dating, the core of adult Achilles tendon is essentially not renewed after height growth finishes. No source for a 100-day figure exists that we could find. Fixed: the sentence is rewritten to say what the evidence actually shows, including that intervertebral disc collagen has a half-life measured in decades (Sivan et al. 2008), and the 12-to-18-month figure is now labelled as our working estimate rather than a measured one.
4. We presented adolescent data as adult data
The table listed Fishman’s 9.2-degree lumbar change with the population “Adults”. That study (Fishman 2021) was 56 adolescents, mean age 14. Fixed: the row now reads “AIS, mean age 14”. This matters because the article is titled in the first person and is read by adults, so an adolescent number in an adult row reads as a promise we cannot make.
5. We described the wrong study design, twice
The article twice called the Fishman yoga study a “serial case report”. It is a non-randomized controlled trial, and the “controls” were the patients who reported not doing the poses. The serial case report is a different Fishman paper from 2014 that contains no 9.2-degree figure. Fixed: both descriptions now state the real design and its weaknesses, including that fewer than half of those enrolled returned a follow-up X-ray.
6. Two reference list errors
The Fishman 2014 paper was listed with the year changed to 2021, which merged two different studies into one entry. The Katzman 2017 page range was given as 2831-2843; it is 2831-2841. Fixed: both corrected, and the Fishman 2021 paper now has its own entry.
What we changed about how we work
These six errors shared one cause: a knowledge base note that carried the bad citations, which then seeded every article written from it. Checking the article was not enough, so the note has been corrected too. Every citation in the corrected article was re-verified against PubMed directly, by record, not from memory.
We are publishing this in full rather than quietly fixing it because this site asks you to trust its citations. That claim is worth nothing if the first real test of it is handled privately.