Corrections Log
When we get something factually wrong, we fix it and record the fix here. Each entry lists what was wrong and what changed. Our full review and corrections process is described in our editorial policy.
One person writes this site and one person is answerable for what is wrong on it: the founder, Vytautas Jazbutis.
Spotted an error? Report it to support@decodemybio.com.
Log
2026-08-01 · CYP2D6 Gene — 23andMe & AncestryDNA Pharmacogenomics
What was wrong: The page's HowTo structured data, its 'How to Check Your CYP2D6 Status from 23andMe or AncestryDNA' section, several FAQ answers, a 'CYP2D6 and Psychiatric Medications' callout, and a closing 'Get Your CYP2D6 Results' section all told visitors they could obtain a CYP2D6 metabolizer status from consumer raw data — including directly from DecodeMyBio: 'Decode+ includes CYP2D6 phenotype analysis from your existing file,' 'you can upload it to DecodeMyBio to learn your CYP2D6 metabolizer status,' and 'Your Decode+ results will include your CYP2D6 diplotype, activity score, metabolizer phenotype.' None of this is true. The product star-calls CYP2C19 from raw data only; CYP2D6 is not called at all, because its clinical phenotype depends on gene deletions, duplications, and hybrid alleles that a consumer genotyping array cannot detect.
What changed: Rewrote the HowTo schema (renamed and restructured to describe what raw data can and can't show, rather than instructing readers on how to obtain a CYP2D6 status), the 'How to Check Your CYP2D6 Status' section, the affected FAQ answers (JSON-LD and visible copy), the 'How CYP2D6 Is Tested' section, the psychiatric-medications callout, and the closing results section, to state plainly that DecodeMyBio does not call CYP2D6 from raw data, explain why (a consumer array cannot resolve the structural variation that drives CYP2D6 phenotype), and note that Decode+ results still include CYP2C19. Left all CYP2D6 pharmacology (codeine/tamoxifen/antidepressant metabolism, metabolizer phenotypes, population frequencies, CPIC guidelines) unchanged. dateModified was already 2026-08-01 from an earlier correction pass on this page; no further date change needed.
2026-08-01 · Pain & Anesthesia Genetics (product page) and /psychiatric (product page)
What was wrong: Both product pages' FAQPage structured data and visible copy repeatedly claimed 'Decode+ analyzes CYP2D6' — on /pain, alongside OPRM1, COMT, BDNF, and ANKK1 as genes the product 'extracts variants' for and 'assigns genotypes, diplotypes, and phenotype classifications' from; on /psychiatric, alongside CYP2C19. /psychiatric's own Limitations section stated 'Decode+ analyzes CYP2D6 and CYP2C19 only' — itself false, since Decode+ does not call CYP2D6 at all, only CYP2C19.
What changed: Reworded every 'Decode+ analyzes/covers/extracts CYP2D6' instance on both pages (JSON-LD FAQ entries and their visible-copy duplicates, body sections, and callouts — roughly 13 instances on /pain, 15 on /psychiatric) to state that Decode+ analyzes OPRM1, COMT, BDNF, and ANKK1 (pain) or CYP2C19 (psychiatric), and does not call CYP2D6, explaining why and pointing to clinical-grade testing with copy-number analysis for CYP2D6-dependent decisions. Corrected the false Limitations-section line on /psychiatric to state Decode+ analyzes CYP2C19 only. Bumped dateModified on /psychiatric (/pain was already 2026-08-01 from an earlier pass). Left all CYP2D6 pharmacology unchanged.
What was wrong: Twelve medication pages for CYP2D6-affected drugs told visitors 'DecodeMyBio can analyze your CYP2D6 status and report whether [drug] is flagged for your genotype' and that results 'will include your CYP2D6 diplotype, activity score, metabolizer phenotype, and the CPIC recommendation for [drug] specifically' — in each page's 'Understanding Your Results' section, FAQ answers, and 'have your DNA file?' callout boxes. Amitriptyline's version additionally claimed this for CYP2C19 alongside CYP2D6. Separately, /genes/cyp2c19 claimed DecodeMyBio's psychiatric coverage 'maps your CYP2C19 and CYP2D6 results to CPIC guidelines.' None of the CYP2D6 claims are true — DecodeMyBio does not call CYP2D6 from raw data.
What changed: Rewrote the affected callouts, FAQ answers, and 'Understanding Your Results' sections on all 12 medication pages and /genes/cyp2c19 to state that DecodeMyBio does not call CYP2D6 (with the structural-variant reason) and to point CYP2D6-dependent decisions to clinical-grade testing. For amitriptyline specifically, preserved the true claim that DecodeMyBio calls CYP2C19 (and gives the CYP2C19 half of its CPIC recommendation) while correcting the false CYP2D6 half. Bumped dateModified and the matching sitemap.ts entries on all 12 medication pages and /genes/cyp2c19. Left all CYP2D6 pharmacology unchanged.
2026-08-01 · Sample Reports (/sample, /sample/pain, /sample/psychiatric and their underlying sample data files)
What was wrong: The sample reports shown to prospective users to demonstrate what a real report looks like contained fabricated CYP2D6 results presented as genuine output, e.g. 'Your CYP2D6 result is *1/*4 (Intermediate Metabolizer)' and a full CYP2D6 finding card with a diplotype, phenotype, and clinical recommendations. This is a false capability claim in its most direct form — a sample demonstrating a result the product cannot actually produce, since DecodeMyBio does not call CYP2D6 from raw data.
What changed: Replaced every fabricated CYP2D6 finding in the sample data files (sample-pain-data.ts, sample-psych-data.ts, sample-report-data.ts) and the pages that render them with an honest demonstration of the product's actual behavior — a disclosure stating CYP2D6 cannot be called from a genotyping array, matching what the live app actually shows for this gene. Corrected the associated gene-count copy (e.g. '5 genes analyzed' for pain results, which had counted the fabricated CYP2D6 finding, corrected to '4 genes analyzed': OPRM1, COMT, BDNF, ANKK1) and the 'Genes with findings' summary on /sample. These pages carry no dateModified field; none was added.
2026-08-01 · Multiple learn articles (28 articles) referencing CYP2D6 pharmacogenomic testing
What was wrong: Across roughly 28 long-form learn articles, body text, FAQ answers, and sample-report CTA copy repeatedly claimed DecodeMyBio/Decode+ 'analyzes,' 'extracts,' 'covers,' or 'maps' a user's CYP2D6 from raw data — e.g. 'DecodeMyBio's Decode+ extracts CYP2D6 and CYP2C19 variants from your raw data, assigns your diplotype,' 'Decode+ also covers CYP2D6 across 48+ medications,' and 'Upload your 23andMe or AncestryDNA raw data to get your CYP2D6 phenotype.' Two articles ('why-codeine-doesnt-work,' 'tramadol-not-working') had a 'How to Find Out Your CYP2D6 Status' section instructing readers on obtaining a CYP2D6 result the product cannot provide, and one article ('23andme-shutting-down-what-to-do') separately implied all 48+ covered medications, and Pain & Anesthesia results specifically, would include 'your metabolizer status' for CYP2D6-driven drugs like codeine and tramadol.
What changed: Corrected every capability-claim and array-implication instance found across these articles (roughly 60 edits) to state that DecodeMyBio calls CYP2C19 (where true) and does not call CYP2D6, with the reason given (structural variation a consumer array can't resolve). Renamed the two 'How to Find Out Your CYP2D6 Status' sections to describe what raw data can and can't show, mirroring the /genes/cyp2d6 fix. Corrected the 'Medication Safety' and 'Pain & Surgery Prep' sections of '23andme-shutting-down-what-to-do' to distinguish CYP2C19-callable medications from CYP2D6-driven ones. Bumped lastUpdated (and the matching sitemap.ts entries) on the 3 articles that were not already dated August 2026: understanding-your-report, cyp2d6-antidepressants-list, cyp2c19-ssri-metabolism. Left all CYP2D6 science and citations (Gaedigk et al. 2017, PharmVar, CPIC guidelines) unchanged.
What was wrong: A recurring gene-list pattern across these pages claimed DecodeMyBio 'analyzes,' 'extracts and interprets,' or 'maps' CYP2D6 alongside CYP2C19/CYP2C9/VKORC1/etc. as part of the pharmacogenes the product covers — e.g. /ancestrydna: 'DecodeMyBio extracts and interprets these variants' following a gene list that included CYP2D6; /compare/best-23andme-raw-data-tools and /compare/genesight-alternative-ads: 'DecodeMyBio analyzes 13 pharmacogenes including CYP2D6, CYP2C19...'. /methodology's own pipeline-description used CYP2D6 as a worked example of star-calling ('CYP2C19*2, CYP2D6*4') and CPIC activity-score computation, directly misdescribing what the product's methodology actually does. /raw-dna-analysis and /limitations described CYP2D6's structural-variant gap in terms that implied partial coverage, rather than stating DecodeMyBio does not call the gene at all.
What changed: Removed CYP2D6 from every 'genes DecodeMyBio analyzes/covers/extracts' list on these pages, without altering the other genes named or any gene-count figures (the site's '13' and '19' pharmacogene counts are a separate, unreconciled issue, left as-is). Replaced /methodology's CYP2D6 worked examples with CYP2C19 or CYP2C9 examples and added an explicit statement that DecodeMyBio does not star-call CYP2D6. Strengthened /limitations' and /raw-dna-analysis's structural-variant language to state plainly that CYP2D6 is not called at all, even for its SNP-based star alleles. Bumped dateModified and the matching sitemap.ts entries on /ancestrydna, /compare/best-23andme-raw-data-tools, /methodology, /limitations, /data-sources, /genes/mthfr, /genes, and /medications; /compare/genesight-alternative-ads, /nutrition, and _data/faq-items.ts carry no date field.
2026-08-01 · Privacy Policy
What was wrong: The privacy policy stated 'We do not use tracking cookies or third-party analytics cookies.' That was false: Google Analytics (GA4) and the Google Ads conversion tag were both live on the site, setting their own cookies (_ga, _gcl_au, _gcl_aw) to measure traffic and advertising performance, alongside a first-party attribution cookie.
What changed: Replaced the false no-tracking-cookies claim with an accurate disclosure naming Google Analytics and the Google Ads conversion tag, the specific cookies they set, and the first-party attribution cookie and equivalent localStorage entry used for up to 90 days to record how a visitor reached the site. Stated plainly that none of these cookies contain or are linked to genetic data.
2026-08-01 · Editorial Policy
What was wrong: The editorial policy and the /about page claimed a 'DecodeMyBio Editorial Team' reviewed content, and stated that 'every factual claim is cross-referenced against the cited source before publication.' Neither was true: there is no editorial team and no clinical reviewer, and the citation audit this correction batch itself grew out of found factual claims that had not been checked against their cited sources (the FUT2/B12 inversion and the CYP2D6 gene-page figures documented elsewhere in this log are direct evidence of that).
What changed: Removed the 'Editorial Team' framing from /editorial-policy and /about. The policy now states plainly that one person, founder Vytautas Jazbutis, writes the site, that there is no editorial team and no clinical reviewer, and describes what actually happens instead (sourcing against CPIC/FDA/peer-reviewed literature, and a public corrections log when something is found wrong after publication) rather than an unqualified pre-publication cross-referencing claim.
2026-08-01 · CYP2D6 Gene — 23andMe & AncestryDNA Pharmacogenomics
What was wrong: The why-codeine-doesn't-work correction logged below states this page was corrected the same day the flagship codeine article's CYP2D6 frequency figures were fixed, with Gaedigk et al. 2017 added as a source. That was false. This page's FAQPage structured data, a body paragraph, and its visible FAQ answer all still stated 'approximately 5–10% of European populations, 1–3% of East Asian and African populations' for poor metabolizer frequency — the same superseded figure the flagship article had already corrected to 'up to about 5%, depending on population' — and Gaedigk et al. 2017 appeared zero times on this page despite being the source for the corrected figures.
What changed: Corrected all three instances (FAQPage JSON-LD answer, body paragraph, and visible FAQ answer) to state poor metabolizer frequency ranges from about 0.4% up to roughly 5.4%, and intermediate metabolizer frequency from about 0.4% up to roughly 11%, across world populations — and added Gaedigk et al. 2017 as an inline citation in all three places, where it had not been cited at all before. Bumped dateModified.
2026-08-01 · Why Codeine Doesn't Work for Some People
What was wrong: The article's own meta description (which feeds the meta description, Open Graph description, and Article JSON-LD) still read 'About 10% of people cannot convert codeine into its active form' after the body had already been corrected to 'up to about 5%, depending on population' — the flagship correction contradicting its own description field.
What changed: Corrected the description field to match the corrected body figure ('up to about 5% of people').
2026-08-01 · Pain & Anesthesia Genetics (product page)
What was wrong: The page's hero copy — the first thing a visitor reads — stated 'About 10% of people cannot convert codeine into morphine at all... Another 15% convert it too slowly,' directly contradicting the page's own FAQPage structured data lower on the same page, which already correctly stated 'up to about 5%... up to about 11% (Gaedigk 2017).'
What changed: Corrected the hero copy to 'up to about 5%... up to about 11%,' matching the page's own already-corrected FAQPage data.
2026-08-01 · What Does Poor Metabolizer Mean? Metabolizer Types Explained
What was wrong: The article stated 'Approximately 5–10% of European-ancestry individuals are CYP2D6 poor metabolizers' — the same superseded figure corrected on the flagship codeine article and elsewhere, missed here.
What changed: Corrected to 'up to about 5% of people, depending on population,' and added Gaedigk et al. 2017 as a source (not previously cited on this article). Bumped lastUpdated and the sitemap entry.
2026-08-01 · Genetic Testing for ADHD Medications: What Your DNA Can and Cannot Tell You
What was wrong: The article stated 'CYP2D6 poor metabolizer frequency varies by population: approximately 5–10% in European populations, 1–3% in East Asian and African populations' — the same superseded figure corrected elsewhere, missed here.
What changed: Corrected to the sourced world-population range (about 0.4% up to roughly 5.4%) and added Gaedigk et al. 2017 as a source (not previously cited on this article).
2026-08-01 · Am I a CYP2D6 Poor Metabolizer? 6 Signs + What to Do
What was wrong: The article's description field said '~7% of people are' CYP2D6 poor metabolizers — a third, unexplained figure alongside the '5–10%' used elsewhere on the site. The body and FAQPage structured data separately stated 'Approximately 5–10% of European-ancestry populations,' with East Asian and African frequencies given as flat, uncited percentages.
What changed: Corrected the description to 'up to about 5% of people are, depending on population,' and corrected the body and FAQPage answer to the sourced world-population range (about 0.4% up to roughly 5.4%), removing the uncited per-population breakdown. Added Gaedigk et al. 2017 as a source (not previously cited on this article). Bumped lastUpdated and the sitemap entry.
2026-08-01 · Is Pain Sensitivity Genetic? What COMT, OPRM1, and CYP2D6 Mean for Your Pain
What was wrong: Despite the entry below (dated the same day) stating this article's CYP2D6 figures were sourced to Gaedigk et al. 2017, the article's body still stated 'About 6–10% of people of European descent are CYP2D6 poor metabolizers. Another 1–2% are ultrarapid metabolizers,' and its FAQPage answer for 'Why doesn't codeine work for me?' repeated 'About 6–10% of people of European descent.'
What changed: Corrected both instances to 'up to about 5%' poor metabolizers, depending on population, with the ultrarapid figure corrected to reflect the same source's full range (up to about 21% in some populations).
2026-08-01 · Pharmacogenomics Before Surgery: Why Genetic Testing Can Improve Your Pain Management
What was wrong: The article stated 'About 6–10% of people of European descent cannot activate codeine or tramadol at all,' 'Poor metabolizers (~6–10% of European populations),' and the FAQ repeated 'About 6–10% of people cannot activate these prodrugs at all' — the same superseded figure corrected elsewhere, missed in three places on this article despite the OPRM1 correction already logged for this page below.
What changed: Corrected all three instances to 'up to about 5% of people, depending on population,' citing Gaedigk et al. 2017 (already cited elsewhere in this article's sources).
2026-08-01 · Tramadol Not Working? CYP2D6 Genetics May Be Why
What was wrong: Despite the entry below correcting this article's intermediate/ultrarapid metabolizer figures, the article still stated 'for about 5 to 10% of the population, tramadol genuinely does not work,' 'Approximately 5 to 10% of people of European descent are CYP2D6 poor metabolizers,' 'approximately 5 to 10% in Caucasians, 1 to 2% in East Asians, and 3 to 8% in African Americans,' and an FAQ answer repeated 'About 5–10% of Caucasians lack functional CYP2D6' — four instances of the same superseded poor-metabolizer figure, all missed by the earlier pass on this article.
What changed: Corrected all four instances to 'up to about 5% of people, depending on population,' citing Gaedigk et al. 2017 (already cited in this article's sources).
2026-08-01 · Anesthesia and Genetics: Why It Affects People Differently
What was wrong: Despite the RYR1/postoperative-pain correction already logged for this article below, it separately stated 'about 1 to 2% of Caucasians... are CYP2D6 ultrarapid metabolizers' (not itself wrong, but uncited) and, further down, 'If you are a CYP2D6 poor metabolizer (5 to 10% of Caucasians)' — the same superseded poor-metabolizer figure missed here.
What changed: Corrected the poor-metabolizer figure to 'up to about 5% of people, depending on population,' and added Gaedigk et al. 2017 as a source (not previously cited on this article).
2026-08-01 · Antidepressant Not Working? Your Genetics May Be Why
What was wrong: Despite the STAR*D and institutional-program corrections already logged for this article below, it separately stated in a body bullet '5 to 10% are CYP2D6 poor metabolizers' and, in an FAQ answer, 'About 5 to 10% of Caucasians are CYP2D6 poor metabolizers' — the same superseded figure missed in two places here.
What changed: Corrected both instances to 'up to about 5% of people, depending on population,' and added Gaedigk et al. 2017 as a source (not previously cited on this article).
2026-08-01 · Best MTHFR Supplements
What was wrong: Separately from the PMID corrections logged below, the article stated that L-methylfolate is 'more effective than folic acid at raising plasma folate and lowering homocysteine in individuals with the TT genotype,' citing Prinz-Langenohl et al. 2009. The paper's own title states its finding applies to women with 'the homozygous or wild-type' genotype — i.e. both groups, not TT specifically — and its measured endpoints were plasma folate pharmacokinetics (AUC, Cmax, Tmax); it did not measure homocysteine at all.
What changed: Reworded the claim to state that L-methylfolate raises plasma folate more effectively than folic acid in both the TT and wild-type genotypes, and noted that this specific study measured plasma folate pharmacokinetics, not homocysteine.
2026-08-01 · Nutrition & Methylation (product page)
What was wrong: The FUT2 bullet on the product page, and the corresponding FUT2 finding in the /sample/nutrition sample report, both stated that non-secretor variants 'are associated with lower circulating B12 levels' (product page) or that non-secretors 'may have reduced vitamin B12 absorption... potentially leading to lower B12 levels over time' (sample report). Both are backwards: Hazra et al. 2008, Chery et al. 2013, and Velkova et al. 2017 all report that the non-secretor-linked allele is associated with higher, not lower, measured plasma B12. The sample report also cited the wrong PMID for this claim (19706858, an unrelated CYP2C19/clopidogrel paper) and its result label read 'Non-Secretor (Reduced B12 Absorption).'
What changed: Reworded the /nutrition FUT2 bullet to state the association is with higher circulating B12 levels, likely reflecting altered absorption and transport rather than better B12 status. Reworded the /sample/nutrition finding to the same effect, retitled the result 'Non-Secretor (Higher Measured B12),' corrected the citation to PMID 18776911 (Hazra et al. 2008), and replaced the now-inconsistent 'eat more B12' considerations (in both the finding and the action-plan tier) with guidance that this result does not indicate a deficiency risk and that a blood test is the reliable way to check B12 status.
2026-08-01 · Raw DNA Analysis 2026 (pillar guide)
What was wrong: This page was a whole-page survivor of two claim classes already corrected elsewhere on the site: 'consumer array accuracy is typically 99%+ per SNP' appeared in its FAQPage structured data and twice more in visible copy (an FAQ answer and a 'Common Misconceptions' section), plus a third accuracy claim in body prose ('Call accuracy is typically 99%+ per SNP'); a fixed '600,000 to 700,000' SNP-count figure appeared in its FAQPage structured data and in body prose; and its FAQPage structured data separately quoted the superseded '$250–$2,000+' clinical-testing price range.
What changed: Reworded all four accuracy instances (FAQPage JSON-LD, two visible-copy instances, and the 'Common Misconceptions' section) to state that genotyping calls are generally reliable for common, well-characterized variants, with coverage — not accuracy — as the real limitation. Reworded both fixed-count instances to describe the SNP count as varying by chip version. Corrected the price range to '$100–$2,000+' to match the sourced range used elsewhere on the site. Bumped dateModified.
What was wrong: Six pages carried an unqualified 'well-covered' / 'can reliably determine your genotype' claim about consumer-array coverage of specific pharmacogenomic variants (VKORC1 rs9923231; SLCO1B1 rs4149056; CYP2C9 *2/*3 on the celecoxib page; CYP2C19 *2/*3/*17 on the pantoprazole page; SLCO1B1 rs4149056 and ABCG2 rs2231142 on the rosuvastatin page; SLCO1B1 rs4149056 on the simvastatin page) across both FAQPage structured data and visible copy — twelve instances total — with no chip-version caveat, the same overstatement already corrected elsewhere (e.g. the MTHFR C677T article).
What changed: Reworded all twelve instances (and a matching instance in a learn article, articles.tsx:2006, about CYP2C9 *2/*3) to state coverage is typical on 'most standard' consumer arrays but can vary by chip version, and that determination is 'typically' rather than unconditionally reliable. Bumped dateModified on all six pages.
2026-08-01 · Pharmacogenomic Testing Cost 2026: Real Prices from Free to $2,000 (and related pages)
What was wrong: Three different clinical-PGx price ranges coexisted across the site against this article's own sourced range of '$100–$2,000+.' The 'what-is-pharmacogenomics' article quoted '$250-$2,000+' in a body list item and two FAQ answers (3 instances). The 'is-pharmacogenomic-testing-worth-it' article quoted a separate, narrower '$250 to $1,000+' figure for 'lab-ordered clinical testing' in a comparison table and two body/FAQ passages (3 instances). 'pharmacogenomics-before-surgery' quoted '$250–$2,000+' once, and 'antidepressant-not-working' quoted a fourth variant, '$300 to $2,000+,' in its FAQ. /compare/pharmacogenomic-testing-vs-23andme quoted '$250–2,000+' in four places (a body paragraph, a comparison table cell used twice, and a second body paragraph), the homepage quoted '$250–$2,000+' once, and _data/faq-items.ts quoted it once more.
What changed: Standardised all 14 instances across these seven files on '$100–$2,000+,' matching this article's own cited range, with no other wording changes beyond the figure itself. Bumped dateModified on /compare/pharmacogenomic-testing-vs-23andme.
2026-08-01 · Celiac & Gluten Screening (product page) and its sample preview
What was wrong: After the learn articles were corrected today to stop claiming a negative consumer tag-SNP screening result can rule out celiac disease, the commercial and sample pages that sell this same screening were not updated and continued to make the false claim. The /celiac product page stated in its metadata, Open Graph tags, and FAQPage structured data that the product can 'rule out celiac with >99% confidence'; its hero section said a negative result means celiac is 'essentially ruled out — with greater than 99% confidence'; a 'Key Clinical Evidence' bullet claimed 'negative predictive value >99%: absence of both HLA-DQ2 and HLA-DQ8 essentially excludes celiac disease'; and a comparison table listed the screening as able to 'rule out (>99% NPV)' celiac disease. The /sample/celiac preview page repeated the same claim in a headline banner ('Based on HLA-DQ Research with >99% Negative Predictive Value') and its underlying sample data file described a negative result as 'essentially excluded (>99% negative predictive value).' A repo-wide sweep also found the identical unhedged claim ('Rule out celiac disease from your existing DNA data' / 'Rule out celiac disease with HLA-DQ2 and HLA-DQ8 genetic screening') in cross-sell blurbs linking to /celiac on three unrelated commercial pages: /cannabis, /pain, and /sample. None of these passages distinguished this product's consumer tag-SNP method from clinical HLA typing, the test that actually has a >99% negative predictive value. A reader with symptoms could have read this as license to skip a diagnostic workup.
What changed: Reworded every instance across /celiac and /sample/celiac (visible copy and JSON-LD/FAQPage structured data) so the >99% negative predictive value is attributed specifically to clinical HLA typing, and this product's tag-SNP screening is described as informative but not a rule-out on its own — a negative result 'should not, by itself, be used to rule out celiac disease,' with guidance to talk to a doctor, especially with symptoms. Added the same citation used in the corrected learn articles (Baaqeel et al. 2021, PMID 34042155: tag-SNP genotyping misclassified more than 32% of confirmed celiac patients as non-carriers in a non-European cohort) to the product page's Limitations section and references list, and strengthened the page's disclaimer to state plainly that a negative result should not be used to rule out celiac disease or any gluten-related condition. Reworded the three cross-sell blurbs on /cannabis, /pain, and /sample to describe the screening as checking HLA-DQ2/DQ8 status rather than claiming a rule-out. Separately, /celiac's reference list cited Megiorni & Pizzuti with the right authors but the wrong paper — it gave World J Gastroenterol 18(33):4523-31 / PMID 22969222, when the paper being described is actually J Biomed Sci 19(1):88 / PMID 23050549 — and this was corrected as part of a site-wide citation audit; this fix was omitted when this entry was first logged. Bumped /celiac's dateModified and its sitemap entry.
2026-08-01 · GeneSight Alternative, GeneSight Alternative (Ads), and GeneSight vs 23andMe (comparison pages)
What was wrong: Three comparison pages carried an unsourced '>99% concordance' figure for how accurately consumer genotyping arrays match clinical-grade pharmacogenomic testing — in visible FAQ copy, FAQPage structured data, and body text — the same kind of unsupported precision figure already corrected elsewhere on the site (e.g. the 23andMe raw data articles) but missed on these three pages.
What changed: Replaced '>99% concordance' with 'generally reliable for the variants tested' (or equivalent phrasing) in all instances across /compare/genesight-alternative (FAQ JSON-LD, FAQ visible copy, and a body paragraph — 3 instances), /compare/genesight-alternative-ads (FAQ JSON-LD and FAQ visible copy — 2 instances), and /compare/genesight-vs-23andme (FAQ JSON-LD, a comparison-table cell, and a body paragraph — 3 instances, not the single 'one remaining FAQ answer' originally logged here). Bumped dateModified where the page carries one.
2026-08-01 · Nutrition & Methylation (product page)
What was wrong: The page's FAQ — in both the visible copy and the FAQPage structured data — stated that 'consumer genotyping arrays reliably cover the key SNPs for MTHFR, COMT, FUT2, VDR, BCMO1, TCN2, and other nutrigenomic variants Decode+ analyzes,' and a second FAQ answer repeated 'consumer arrays reliably cover the nutrigenomic variants Decode+ analyzes' — the same chip-version-dependency overstatement already corrected in our MTHFR C677T article. A 'Consumer DNA Data Coverage' section repeated the same overstatement twice more: 'reliably cover the key SNPs for the nutrigenomic genes Decode+ analyzes' and, further down, that these variants 'are well-covered by standard genotyping arrays.' Separately, the COMT bullet in the 'Genes Analyzed' section stated the Val158Met variant 'may influence methylation demand' without noting that this extension is speculative, the same claim already hedged in our nutrigenomics article.
What changed: Reworded all four array-coverage passages (the two FAQ answers — including the FAQPage JSON-LD — and the two 'Consumer DNA Data Coverage' section passages) to state that array coverage 'typically includes' or 'commonly includes' these SNPs but depends on the chip version used for a given test. Reworded the COMT bullet to flag the methylation-demand extension as speculative and not established in controlled studies, matching the hedge already applied in the nutrigenomics article. Confirmed the page's VDR bullets describe receptor response, not circulating vitamin D levels, so no change was needed there.
2026-08-01 · DNA Test for Medication
What was wrong: A comparison table repeated the same stale GeneSight turnaround figure already corrected in our GeneSight cost article — '~36 hours' after lab receipt, alongside '~3–5 days' for Genomind. The same figure also appeared twice on the GeneSight vs 23andMe compare page: once in its own comparison table ('~36 hours after lab receipt') and again in a separate bullet list ('Results typically available within 36 hours after the lab receives the sample').
What changed: Corrected the comparison table in this article, and both instances (the table and the bullet list) on /compare/genesight-vs-23andme, to describe GeneSight and Genomind turnaround in qualitative terms (a few days, lab-set and subject to change) rather than a specific figure, matching the correction already made in the GeneSight cost article. Bumped the compare page's dateModified and last-updated copy to reflect the change.
2026-08-01 · GeneSight Cost, Insurance Coverage & Alternatives
What was wrong: The article stated GeneSight has had 'over 2 million tests administered' — Myriad Genetics' own public statements put the count well above that as of late 2024, and the true figure changes over time. It also stated GeneSight delivers results within '36 hours' of the lab receiving the sample, in both the body and FAQ; GeneSight's own current published turnaround estimate is about three days.
What changed: Replaced the test-count figure with qualitative language ('several million tests administered to date') and a note to check Myriad's current figures directly rather than relying on a number that will go stale. Corrected the turnaround-time claim in both the body and FAQ to describe results as available within a few days, matching GeneSight's current published estimate, and noted that turnaround times are set by the lab and can change. Extended the article's caveat to tell readers to verify current pricing, coverage, and turnaround directly with the vendor and their insurer, and added a dated, citable source for the UnitedHealthcare 2025 coverage-policy claim.
2026-08-01 · What Is Nutrigenomics?
What was wrong: The VDR bullet stated that variants are 'associated with lower circulating vitamin D levels in population studies' — large genome-wide association studies attribute most of the genetic variance in circulating 25-hydroxyvitamin D levels to other genes (GC, CYP2R1, DHCR7, CYP24A1), not VDR. The COMT bullet stated that the Val158Met variant 'may influence methylation demand, with potential implications for B-vitamin utilization' without noting that this extension beyond COMT's established catechol-metabolism role is speculative. The FUT2 bullet gave non-secretor frequency as 'about 20% of the population' without noting the figure is specific to European ancestry, and separately stated non-secretor variants 'are associated with lower B12 levels' — backwards from what the literature (Hazra et al. 2008; Chery et al. 2013; Velkova et al. 2017) reports, which is higher measured plasma B12 in non-secretors.
What changed: Reworded the VDR bullet to describe its established role in receptor function and downstream signaling (bone density, calcium handling) and to state that circulating vitamin D level GWAS implicate GC, CYP2R1, DHCR7, and CYP24A1 rather than VDR. Reworded the COMT bullet to flag the methylation-demand extension as speculative and not established in controlled studies. Qualified the FUT2 frequency as specific to European ancestry, and corrected the B12 direction to state non-secretor status is associated with higher measured plasma B12 levels — likely reflecting altered absorption and transport rather than better B12 status, with the functional significance still unsettled. Added two supporting sources: a 79,366-person 25-hydroxyvitamin D GWAS and a VDR polymorphism review; the FUT2/B12 citation (Hazra et al. 2008) was already present in this article's sources.
2026-08-01 · MTHFR C677T: What It Means and What to Do
What was wrong: The article stated that 'natural food folate is already in a form that does not require the MTHFR enzyme step that the variant affects' — an overstatement, since dietary folate is a mix of reduced folate forms, and a portion of it still passes through the same MTHFR-dependent conversion step as synthetic folic acid. It also stated that consumer genotyping arrays 'reliably cover' the MTHFR C677T variant (rs1801133), in both the body and FAQ, without noting that coverage depends on the chip version used for a given test.
What changed: Reworded the food-folate claim to describe it accurately as a mix of forms — some already active as 5-MTHF, some still requiring MTHFR-dependent conversion — while noting food folate avoids the unmetabolized-folic-acid buildup associated with high-dose synthetic supplementation. Reworded the array-coverage claims in the body and FAQ to note that coverage depends on the chip version used. Added a supporting source (Scaglione & Panzavolta 2014) distinguishing folate, folic acid, and 5-methyltetrahydrofolate.
2026-08-01 · Celiac Disease vs. Gluten Sensitivity
What was wrong: Separately from the tag-SNP correction logged above, the article stated 'an estimated 80% of cases are undiagnosed' for celiac disease, in both the body and an FAQ answer — a specific figure that does not match the ACG guideline cited alongside it, which describes only 'the majority' of cases as undiagnosed. It also gave non-celiac gluten sensitivity (NCGS) prevalence as '0.5% to 13%' without noting that the upper end of that range comes from self-reported symptom surveys rather than confirmed NCGS diagnoses. The FUT2 passage stated flatly that non-secretor variants 'reduce B12 bioavailability' and that 'about 20% of people' carry them, without noting the 20% figure is specific to European ancestry. A same-day pass reworded that FUT2 claim to associative language and added a supporting source (Hazra et al. 2008), but kept the direction backwards — it stated the association was with 'lower circulating B12 levels,' which is the opposite of what Hazra et al. 2008 actually reports: carriers of the non-secretor-linked allele have higher, not lower, measured plasma B12.
What changed: Replaced the 80% figure with 'the majority' in both the body and FAQ, matching the cited ACG guideline's own language rather than an unsourced specific percentage. Reworded the NCGS prevalence passage and its summary table to distinguish self-reported prevalence (about 10% globally, per a newly added 2025 meta-analysis, ranging under 1% to over 30% by country) from clinically confirmed NCGS, which is lower — a newly added source shows only 16% of suspected NCGS patients confirm gluten-specific symptoms on a blinded gluten challenge. Reworded the FUT2 claim to state it is associated with higher measured plasma B12 levels — likely reflecting altered absorption and transport rather than better B12 status, with the functional significance still unsettled — correcting the inverted direction from the earlier same-day pass, and qualified the 20% figure as specific to European ancestry.
2026-08-01 · Why Codeine Doesn't Work for Some People
What was wrong: Several pages stated that about 10% of people of European descent are CYP2D6 poor metabolizers, that 10-15% are intermediate metabolizers, and that together up to 25% of people get a suboptimal response to codeine for genetic reasons. Gaedigk et al. 2017, which estimates CYP2D6 phenotype frequencies from genotype across world populations, reports poor metabolizer status at 0.4-5.4% and intermediate at 0.4-11%. The figures we published exceeded the highest frequency reported in any population.
What changed: Corrected to the published ranges — up to about 5% poor metabolizers and up to about 11% intermediate, both depending on population — and removed the combined 25% figure rather than substituting an arithmetic estimate of our own. The same correction was applied to the Pain & Anesthesia product page's FAQPage structured data, with Gaedigk et al. 2017 added as a source there. This entry previously stated the CYP2D6 gene page and the 'Is Pharmacogenomic Testing Worth It?' article had also been corrected with Gaedigk added — that was false. The gene page's FAQ, allele-frequency prose, and visible FAQ answer still carried the stale 'approximately 5–10% of European populations' figure with zero Gaedigk citations, and the flagship article's own meta description still read 'About 10%' while its body already said 'up to about 5%.' Both are now fixed — see the entries below. The 'Is Pharmacogenomic Testing Worth It?' article does not state a specific CYP2D6 frequency figure and needed no change.
2026-08-01 · Pain & Anesthesia Genetics (product page)
What was wrong: The page told prospective buyers that patients with the OPRM1 G allele 'may need 30-50% more morphine equivalents,' citing 'Crist & Berrettini, 2014.' The CPIC guideline for OPRM1 and opioids (already cited elsewhere on the page) describes the actual effect as modest — on the order of a 10% increase in post-operative morphine dose in some studies — and states it is too small on its own to support a dosing recommendation. Separately, three references in the page's reference list resolved to the wrong papers: the PMID given for the Crews et al. 2021 CPIC guideline pointed to an unrelated paper on tooth growth rings, the PMID given for Crist & Berrettini's 'Pharmacogenetics of OPRM1' pointed to an unrelated paper on alcoholism genetics, and the PMID given for Trescot & Faynboym's pain-genetics review (25247901) belongs to a different paper — the correct PMID is 25247900. A fourth reference, Hwang et al.'s OPRM1 meta-analysis, was cited under the wrong title ('OPRM1 A118G polymorphism and postoperative opioid consumption') — the paper's actual title is 'OPRM1 A118G gene variant and postoperative opioid requirement,' and its own findings are about opioid requirement, not consumption.
What changed: Reworded the OPRM1 claim to state that CPIC characterizes the effect as modest (about 10% in some studies) and insufficient on its own for a specific dosing recommendation, citing Crews et al. 2021 alongside Crist & Berrettini. Corrected all three misattributed PMIDs in the reference list to the papers they were meant to cite (Crews et al., Crist & Berrettini, and Trescot & Faynboym), and corrected the Hwang et al. reference title to match the actual published paper.
2026-08-01 · CYP2D6 Gene — 23andMe & AncestryDNA Pharmacogenomics
What was wrong: In four separate places on the page (an FAQ answer, a body paragraph, an allele-frequency list item, and a comparison table), CYP2D6 ultrarapid metabolizer prevalence in North African and Middle Eastern populations was given as 'up to 29%.' The CPIC-compiled world-population frequency study already used elsewhere on the site for this figure caps it at 21%.
What changed: Corrected all four instances from 29% to 21% (or the equivalent range, e.g. '1–21%' in the table), matching the cited world-population frequency analysis.
2026-08-01 · What Is Pharmacogenomics? Complete Guide to DNA-Based Medication Safety
What was wrong: The article stated that 'up to 29% of certain East African populations are ultrarapid metabolizers.' The CPIC-compiled world-population frequency study already cited in this article's sources caps ultrarapid metabolizer prevalence at 21%.
What changed: Corrected the figure to 21%, matching the cited source, and clarified that the higher end of the range spans North African, Middle Eastern, and East African populations.
2026-08-01 · Pharmacogenomic Testing Cost 2026: Real Prices from Free to $2,000
What was wrong: The article stated 'the average patient tries 2 to 3 antidepressants before finding one that works,' with no supporting source.
What changed: Replaced with the actual STAR*D remission-cascade data already cited in this article's sources: 36.8% of patients remitted on their first antidepressant trial, with remission rates declining on each subsequent trial (30.6%, 13.7%, 13.0%).
2026-08-01 · Tramadol Not Working? CYP2D6 Genetics May Be Why
What was wrong: The article stated that '10 to 17%' of people are CYP2D6 intermediate metabolizers and that ultrarapid metabolizer status reaches 'up to 10 to 29%' in some North African and Middle Eastern populations. The cited source, a CPIC-compiled analysis of allele frequencies across world populations, reports a different range: genotypic intermediate metabolizer prevalence of 0.4% to 11%, and ultrarapid metabolizer prevalence up to 21%.
What changed: Corrected both figures to match the cited source: intermediate metabolizer prevalence is now given as roughly 0.4% to 11% depending on ancestry, and the North African/Middle Eastern ultrarapid metabolizer ceiling was corrected from 29% to 21%.
2026-08-01 · Anesthesia and Genetics: Why It Affects People Differently
What was wrong: The article stated that 'RYR1 has over 400 known pathogenic variants' and that 'approximately 20 to 25% of patients experience inadequate pain control after surgery,' neither of which had a supporting source. On the RYR1 figure, malignant hyperthermia expert curation panels have formally classified a much smaller number of variants as pathogenic or likely pathogenic — the 400+ figure conflates all reported variants (most of uncertain significance) with the pathogenic subset. On postoperative pain, a widely cited review reports a considerably higher figure than the one used. Separately, after the >80% figure was sourced to Gan 2017, the article added an unsupported clause attributing the shortfall in part to 'pharmacogenomic variation' — the cited review actually attributes it to surgery type, the intervention used, and time elapsed since the procedure, not genetics.
What changed: The RYR1 claim was reworded to state that RYR1 has hundreds of reported variants associated with malignant hyperthermia susceptibility, only a subset of which have been formally reviewed and classified as pathogenic or likely pathogenic, without asserting an unverified count. The postoperative pain statistic was corrected and sourced to a 2017 review reporting that postoperative pain is not adequately managed in more than 80% of patients in the U.S. Removed the unsupported pharmacogenomic-variation clause and replaced it with the factors the cited review actually names.
2026-08-01 · Pharmacogenomics Before Surgery: Why Genetic Testing Can Improve Your Pain Management
What was wrong: The article stated that 'clinical studies consistently show' OPRM1 G-allele carriers require '30–50% more morphine' for equivalent pain relief, and repeated a similar framing in its introduction ('meaning standard morphine doses provide inadequate relief'). The CPIC guideline for OPRM1 and opioids, already cited in the article, describes the effect very differently: a small increase in post-operative morphine requirement (on the order of 10% in some studies) that CPIC calls too modest to be clinically actionable, with no dosing recommendation based on OPRM1 genotype alone. Separately, the article called ondansetron 'the most commonly prescribed anti-nausea medication during and after surgery' without a source, and described the FDA's codeine boxed warning as applying broadly to 'ultrarapid metabolizers' without noting that the warning itself is scoped to children and breastfeeding.
What changed: Both OPRM1 passages were rewritten to reflect what CPIC's own guideline says: a modest, non-actionable increase in morphine requirement rather than a 30–50% figure. The ondansetron claim was reworded to describe it as a commonly used first-line antiemetic rather than asserting it is the single most prescribed one. The codeine boxed-warning passage was rewritten to state that the FDA warning specifically addresses pediatric patients (with a contraindication under age 12) and that CPIC separately extends the same avoidance recommendation to ultrarapid metabolizers of any age.
2026-08-01 · Is Pain Sensitivity Genetic? What COMT, OPRM1, and CYP2D6 Mean for Your Pain
What was wrong: The article's COMT 'up to four-fold' activity-difference claim had no direct citation, though it is a well-established figure. Separately, it stated OPRM1 G-allele carrier frequency as 'About 10–15%' in people of European descent and 'up to 40%' in East Asian descent, without a source, and cited a real, correctly-linked meta-analysis (Hwang 2014) as the basis for a claim that G-allele carriers require '30–50% more morphine' — but that meta-analysis reports its findings as a standardized mean difference, not a percentage, and CPIC's own guideline describes the effect as a modest, non-actionable increase (on the order of 10%).
What changed: Added the original enzyme-kinetics source (Lotta et al. 1995) for the COMT four-fold claim. Replaced the unsourced OPRM1 carrier-frequency figures with the sourced allele frequencies from a population-genetics study: about 15% in European-descent populations and 40–60% in East Asian populations. Reworded the '30–50% more morphine' claim to describe a modest, non-actionable increase in morphine requirement, consistent with what the cited meta-analysis and the CPIC OPRM1 guideline actually report. Added the CPIC opioid-genotype guideline and the CYP2D6 world-population frequency study as sources for the adjacent CYP2D6 figures.
2026-08-01 · Ultrarapid Metabolizer: What It Means and Why It Matters
What was wrong: The article stated CYP2D6 ultrarapid metabolizer prevalence reaches 'up to 10 to 29%' in some North African and Middle Eastern populations — the cited world-population frequency study caps this figure at 21%. It also attributed codeine's CYP2D6-specific boxed warning language to a single 'FDA Safety Communication, 2017' without noting the FDA's separate 2013 pediatric restriction that preceded it.
What changed: Corrected the ultrarapid metabolizer ceiling from 29% to 21%, matching the cited source, in both the article body and FAQ. Reworded the codeine boxed-warning passage to note the FDA's 2013 pediatric tonsillectomy/adenoidectomy restriction and the 2017 update that added the CYP2D6-specific boxed warning language and breastfeeding warning. Added the PharmVar CYP2D6 GeneFocus review (for the consumer-array copy-number-detection limitation) and the CPIC atomoxetine guideline as sources.
2026-08-01 · Intermediate Metabolizer: What It Means for Your Medications
What was wrong: The article stated CYP2D6 intermediate metabolizer prevalence as 'Approximately 10–15%' in European-ancestry individuals and 'up to 40–50%' in East Asian populations, in both the body and FAQ. The CPIC-compiled world-population frequency study already cited elsewhere in the article reports a different range: 0.4% to 11% globally. The CYP2C19 intermediate metabolizer frequency figures ('18–25%' / 'up to 50%') had no dedicated frequency source.
What changed: Corrected the CYP2D6 intermediate metabolizer figure to the sourced range (approximately 0.4% to 11%, varying by ancestry) in both the body and FAQ. Added a dedicated worldwide frequency meta-analysis (Koopmans et al. 2021) as the source for the CYP2C19 figures and qualified them as estimates that vary by study and cohort — which also changed the European-ancestry CYP2C19 intermediate metabolizer figure from the previous '18–25%' to '20–25%,' matching Koopmans et al.'s reported range; this range change was not mentioned when this entry was first logged.
2026-08-01 · SSRI Side Effects and Genetics: Why Antidepressants Affect People Differently
What was wrong: The article stated that 'the average patient tries 2 to 3 antidepressants before finding one that works adequately' and that 'up to 50% of patients discontinue their first antidepressant within 6 months, most commonly due to side effects' — neither figure had a supporting source.
What changed: Replaced the antidepressant-count claim with sourced data from the STAR*D trial: 36.8% of patients remitted on their first antidepressant, with remission rates declining on each subsequent trial (30.6%, 13.7%, 13.0%). Reworded the discontinuation claim to state that research suggests roughly half of patients discontinue within six months, with side effects among the most commonly cited reasons (rather than asserting it is the single most common reason), and added the supporting sources.
2026-08-01 · Antidepressant Not Working? Your Genetics May Be Why
What was wrong: The article repeated the same unsourced '2 to 3 antidepressants' and '50% discontinue... most commonly due to side effects' claims found in our SSRI side effects article. It also stated that Mayo Clinic, St. Jude Children's Research Hospital, and Vanderbilt University Medical Center 'have integrated pharmacogenomic testing into routine care for psychiatric medications' — those institutions do run real preemptive pharmacogenomic programs, but the programs are broader than psychiatric care specifically, and the claim was uncited. The bipolar-misdiagnosis and combination-therapy claims in the article were accurate but had no citations. Separately, the combination-therapy claim (citing Cuijpers et al. 2014) stated that medication plus psychotherapy 'produces better outcomes than either alone' — Cuijpers 2014 compares combined therapy against antidepressant medication alone; it does not establish that combination therapy beats psychotherapy alone.
What changed: Replaced the antidepressant-count claim with the STAR*D remission-cascade data (36.8%, 30.6%, 13.7%, 13.0% across four steps) and reworded the discontinuation claim the same way as the SSRI side effects article. Reworded the institutional claim to state precisely what is documented: these centers have run preemptive pharmacogenomic programs genotyping patients for CYP2D6 and/or CYP2C19, the genes relevant to the antidepressants discussed, among other medications. Added citations for the St. Jude PG4KDS, Vanderbilt PREDICT, and Mayo RIGHT programs, plus sources for the bipolar-misdiagnosis statistic and the combination-therapy (medication plus psychotherapy) claim. Reworded the combination-therapy claim to state medication plus psychotherapy produces better outcomes than medication alone, matching what Cuijpers 2014 actually compares.
2026-08-01 · Genetic Testing for Antidepressants: What Your DNA Can and Cannot Tell You
What was wrong: The article claimed blood levels of the same SSRI at the same dose can differ 'in some cases, a 10-fold difference' between two people, with no supporting source. It also said pharmacogenomic testing 'costs less than a single therapy session' (an unsourced pricing claim), stated 'the average person with depression tries two to three antidepressants before finding one that works adequately' (unsourced, in both body and FAQ), and attributed declining response to repeated antidepressant failures to 'adherence drops and hopelessness' — an editorial causal claim not supported by the cited STAR*D source.
What changed: Removed the unsourced 10-fold figure, replacing it with qualitative language about meaningfully different blood levels. Removed the therapy-session pricing comparison. Replaced the antidepressant-count claim (body and FAQ) with the actual STAR*D remission-cascade data already cited in the article's sources: 36.8% remission on the first trial, declining to 30.6%, 13.7%, and 13.0% on subsequent trials, for a 67% cumulative remission rate. Removed the 'adherence drops and hopelessness' causal claim, replacing it with the STAR*D authors' own finding that participants needing more treatment steps had higher relapse rates during follow-up.
2026-08-01 · Genetic Testing for ADHD Medications: What Your DNA Can and Cannot Tell You
What was wrong: The article stated that 'amphetamine and lisdexamfetamine are metabolized through multiple non-CYP pathways,' which is imprecise: the FDA label for amphetamine states it is metabolized, to some degree, by CYP2D6.
What changed: Reworded to state that amphetamine and lisdexamfetamine are metabolized mainly through non-CYP pathways, while noting the FDA label's statement that amphetamines are metabolized to some degree by CYP2D6, in both the body text and the corresponding FAQ answer. Added the Adderall XR prescribing information as a source.
2026-08-01 · Genetic Testing for Anxiety Medication: What Your DNA Reveals About Treatment
What was wrong: The article stated that buspirone is metabolized by CYP3A4 and that paroxetine is FDA-approved for GAD, social anxiety disorder, panic disorder, and OCD. Both statements are accurate per the drugs' FDA labels, but neither was cited.
What changed: Added the BuSpar and Paxil FDA prescribing information as sources for these two claims. No wording changes were needed.
2026-08-01 · Sample Cannabis & CBD Results (Decode+ preview)
What was wrong: The sample preview repeated the same overstatement as our Cannabis & CBD product page. It presented a finding titled 'Elevated Risk of CBD Side Effects', flagged at the highest severity, stating that a CYP3A5 and CYP2C19 genotype combination 'is associated with a significantly higher rate of gastrointestinal side effects from CBD' and explaining it through a 'shunting effect' producing excess 7-OH-CBD. It told the reader to 'be aware of an elevated risk of GI side effects' and that 'this genetic interaction may be a contributing factor' if they had GI issues with CBD. The summary panel displayed 'Side Effect Risk: Yes' in red as a headline statistic. The underlying study is real and was correctly cited, but our interpretation of it was wrong — the study could not analyze the two genes separately, its authors explicitly declined to claim a cause-and-effect relationship, and they concluded that pharmacogenomic testing may hold questionable utility for predicting these events. The page also stated that CYP2C9 'handles approximately 70% of THC clearance', a figure with no supporting source; described a single reduced-function CYP2C9 copy as producing effects that the cited study demonstrated only in people with two copies, without noting that distinction or that the *2 variant showed no effect; claimed cannabis smokers have 'up to 2.5x lower clozapine concentrations', which the cited case report does not state; cited a PMC identifier belonging to a different paper as the source for its CBD metabolism finding; and showed a summary statistic of 'Poor Metabolizer' for THC that contradicted the Intermediate Metabolizer result displayed directly beneath it.
What changed: The side-effect finding is now titled 'CYP3A5 + CYP2C19 Genotype Context', is labelled 'Context' rather than a severity grade, and states in its own description that it is not a prediction that CBD will cause side effects. Its clinical context now gives the raw counts (7 of 18 versus 1 of 15), the single event in the comparison group, the fact that CYP2C19 accounted for 7 of the 8 diarrhea cases and that CYP3A5 could not be analyzed separately, the authors' refusal to claim causation, and their conclusion about questionable utility. The 'Side Effect Risk: Yes' statistic was replaced with the neutral CYP3A5 genotype result, and the 'actionable findings' badge with a count of genes reported. The '70% of THC clearance' figure was removed in favour of the sourced statement that CYP2C9 is the primary enzyme forming 11-OH-THC, and the CYP2C9 finding now states that the threefold exposure difference was observed in people with two reduced-function copies, that the *2 variant made no difference, and that the effect of a single copy is less certain. The unsupported 2.5x clozapine figure was replaced with what the cited paper reports. The misattributed PMC identifier was replaced with the correct sources, and a Limitations entry now states plainly that Decode+ does not predict CBD side effects. The contradictory THC summary statistic was corrected to match the finding.
2026-08-01 · Cannabis & CBD Genetics (product page)
What was wrong: This page sold the Cannabis & CBD analysis partly on a claim we overstated. It said poor CYP3A5 metabolizers 'may experience 5.6x higher rate of CBD side effects,' described a 'CYP3A5 + CYP2C19 gene-gene interaction,' and listed 'CBD Side Effect Risk' as a product feature. The underlying study is real and was correctly cited, but our interpretation of it was wrong: the study could not analyze the two genes separately because the groups overlapped, its authors explicitly declined to claim a cause-and-effect relationship, they described the pharmacokinetic differences as mild, and they concluded that pharmacogenomic testing 'may hold questionable utility' for predicting these adverse events. The 39% versus 7% comparison also rested on a single case of diarrhea in a group of 15, in 33 volunteers taking prescription-strength CBD at 5 mg/kg twice daily — not the doses sold over the counter. Three of these statements sat inside FAQPage structured data, meaning they were eligible to appear in search results with no surrounding context. Separately, the page stated that CYP2C9 handles '~70% of THC clearance' — a figure with no supporting source — and three of its five references were inaccurate: the Etkins citation gave the wrong title and author initial; the Nasrin citation gave the wrong year (2024 rather than 2021) and a PMC identifier belonging to a different paper; and the Zullino citation gave the wrong title, journal, volume and pages, attached to a claim that smoking lowers clozapine levels 'by up to 2.5x' that the cited case report does not make.
What changed: Rewritten so the page no longer claims genotype predicts CBD side effects, in both the visible copy and the JSON-LD. The 5.6x figure and the 'gene-gene interaction' framing are gone throughout. The FAQ entry now asks whether genetics can predict CBD side effects and answers no, giving the study's raw counts, the overlap problem, the absence of a causal claim, and the authors' own conclusion. The 'CBD Side Effect Risk' feature is now 'CYP3A5 + CYP2C19 Genotype Context' and is described as context, not a risk prediction. A new Limitations bullet states plainly that Decode+ does not predict CBD side effects. The '~70% of clearance' figure was removed and replaced with the sourced statement that CYP2C9 is the primary enzyme forming THC's main active metabolite, 11-OH-THC. The three inaccurate references were corrected against the source records, the published correction to the Etkins paper was added, the unsupported 2.5x clozapine figure was replaced with what the cited paper actually reports, and all nine references are now hyperlinked to their source rather than listed as unverifiable plain text.
2026-08-01 · Cannabis & CBD Genetics from 23andMe or AncestryDNA Data
What was wrong: The article carried the same overstatement as our Cannabis & CBD product page: that a CYP3A5 and CYP2C19 genotype combination 'increases CBD side effect risk by 5.6x,' that this was 'driven by a metabolic shunting effect,' and that 'if CBD has given you GI issues, this gene-gene interaction may be the reason.' The study behind this is real and was correctly cited, but our reading of it was wrong — see the /cannabis entry above for the detail. The article also stated that CYP2C9 'handles approximately 70% of THC clearance,' a figure with no supporting source; that people with the *3 variant 'clear THC up to 3 times more slowly,' which misdescribes a finding about total THC exposure and did not mention that the CYP2C9*2 variant showed no effect; and that 'about 15-20% of people of European descent carry at least one reduced-function CYP2C9 allele,' which understates the published allele frequencies.
What changed: The 5.6x figure, the shunting mechanism and the gene-gene framing were removed. The section is now titled 'Can Genetics Explain CBD Side Effects?' and answers no on current evidence, giving the raw counts (7 of 18 versus 1 of 15), the fact that CYP2C19 accounted for 7 of the 8 diarrhea cases and that CYP3A5 could not be analyzed separately, the authors' refusal to claim causation, and their conclusion about questionable utility. The '70% of THC clearance' figure was removed in favour of the sourced statement that CYP2C9 is the primary enzyme forming 11-OH-THC. The CYP2C9 passage now reports what the study found — roughly threefold higher total THC exposure in *3 homozygotes, a trend toward greater sedation, and no effect of *2. The 15–20% figure was replaced with published allele frequencies in Europeans (*1 about 82%, *2 about 12%, *3 about 6%), with a note that the earlier figure understated it. Added the FDA Epidiolex label, the NCBI dronabinol pharmacogenomics summary, the allele-frequency meta-analysis and the published correction to the Etkins paper as sources, and added a caveat noting the small single-study basis.
2026-08-01 · CBD Side Effects and Genetics
What was wrong: The article was built around a 2026 clinical study, presented as showing a 'CYP3A5 + CYP2C19 gene-gene interaction' that 'dramatically increases CBD side effect risk' — summarised in the title, description and FAQ as 'a 5.6-fold difference in GI side effect risk based entirely on genotype.' The study is real and correctly cited, but the article misrepresented it. It claimed 'neither gene alone predicts the side effect risk,' where the paper reports that CYP2C19 intermediate/normal metabolizers accounted for 7 of the 8 diarrhea cases and that CYP3A5 could not be analyzed separately because of overlap between the groups. It presented 7-OH-CBD accumulation as the cause of gastrointestinal side effects, which the authors explicitly decline to claim. It omitted that the 7% comparison figure was a single case out of 15, that the study was a secondary analysis of 33 healthy volunteers, that participants took prescription-strength CBD at 5 mg/kg twice daily, that the authors described the pharmacokinetic effects as mild, and that they concluded pharmacogenomic testing 'may hold questionable utility' for predicting these events. Separately, the article named the CYP3A subfamily including CYP3A5 as 'the primary metabolic pathway for CBD,' handling 'the largest share of CBD clearance'; stated that 80–90% of people of European descent carry CYP3A5*3/*3; advised readers to try 10–15 mg instead of 25–50 mg, to space doses further apart, and to use sublingual CBD because it 'partially bypasses first-pass liver metabolism'; made an unsourced claim about CYP1A2 induction; and carried a hand-written in-body 'Clinical Reference' section duplicating the References list.
What changed: Rewritten around what the sources support. The 5.6-fold framing, and the title, description and FAQ answers built on it, were removed. The study's actual figures (7 of 18 versus 1 of 15, p = 0.0463) are now reported alongside the authors' own limitations: the single event in the comparison group, the inability to separate CYP3A5 from CYP2C19, their explicit refusal to claim cause and effect, the prescription-strength dosing, and their conclusion about questionable utility. The article now states that the paper carries a published correction (transposed figures, not affecting these results). CBD's primary metabolizing enzymes are stated as CYP2C19 and CYP3A4 per the FDA Epidiolex label, with an explicit note that CYP3A5 is not an established primary CBD enzyme. Removed: the CYP3A5*3/*3 frequency figure, all dosing advice, the sublingual first-pass claim, the CYP1A2 claim, and the in-body Clinical Reference section. The article now states plainly that the evidence does not support using a CYP3A5 or CYP2C19 result to predict who will get diarrhea from CBD, and the product call-to-action was reworded to match. Added sourced coverage of the better-documented risk — CBD's inhibition of CYP2C8, CYP2C9 and CYP2C19, and the clobazam interaction.
2026-08-01 · Why Don’t Edibles Work the Same for Everyone?
What was wrong: The article stated that 11-OH-THC 'may be 1.5 to 7 times more potent than delta-9-THC.' No source supports that figure; the available human data indicates the two are roughly equipotent. It described CYP2C9*1/*2 heterozygotes as intermediate metabolizers who clear THC more slowly, when the study it relied on found no difference in THC pharmacokinetics by *2 status. Its title ('It’s Your CYP2C9 Gene, Not Your Dose') and several passages asserted that efficient CYP2C9 metabolism explains why edibles do nothing for some people — no published study supports a genetic explanation for a null response to oral THC, and the documented CYP2C9 effect runs in the opposite direction, toward higher exposure. It stated that CBD is primarily metabolized by CYP2C19 and CYP3A5, and described our CBD article as covering a 'CYP3A5/CYP2C19 gene-gene interaction that predicts gastrointestinal side effects.' It also carried a section of practical dosing advice — start at 2.5–5 mg, wait 2 or more hours before re-dosing, account for fat content — that no cited source supported, plus similar guidance in one FAQ answer.
What changed: The 1.5-to-7-times potency figure was removed and replaced with what the evidence shows: intravenous administration in humans found THC and 11-OH-THC equally potent, and a 2024 controlled animal study found equal or somewhat greater activity while noting how little research-based evidence exists on the metabolite. The CYP2C9 section now reports what the 43-volunteer study actually found — roughly threefold higher total THC exposure in CYP2C9*3 homozygotes, a trend toward greater sedation, and no difference by *2 — and states that this rests on a single study. The title was changed, and a new section states plainly that there is no established genetic explanation for edibles doing nothing, and that low and variable oral bioavailability, absorption differences and degradation in the stomach are sufficient explanations. CBD's primary enzymes were corrected to CYP2C19 and CYP3A4, and the description of the linked CBD article was corrected. The dosing-advice section and the dosing guidance in the FAQ were removed entirely.
2026-08-01 · Can You Test for Celiac Disease from 23andMe Raw Data?
What was wrong: The article claimed a negative consumer DNA screening result provides 'the same high negative predictive value as a clinical HLA test,' described the genetic information as 'equivalent' to clinical HLA typing, cited an uncited $100–300 figure for clinical HLA typing cost, and — in a separate passage ('What the Test Actually Checks') — stated outright that lacking the haplotypes means 'celiac disease is essentially ruled out — with greater than 99% negative predictive value,' without distinguishing this product's tag-SNP method from clinical HLA typing. A residual passage elsewhere in the article was missed by this pass: 'if your genetic screening rules out celiac, and you still have gluten-related symptoms, NCGS may be the explanation' — restating the exact rule-out claim the rest of the correction retracted.
What changed: Reworded throughout to state that the tag-SNP approach used by consumer DNA is informative but not equivalent to clinical HLA typing. The 'What the Test Actually Checks' section now attributes the >99% negative predictive value to clinical HLA typing specifically, before introducing this product's tag-SNP method as a separate, less-certain proxy. Added a citation showing the tag-SNP method is substantially less reliable outside people of European descent (over 32% of confirmed celiac patients in one non-European population were misclassified as non-carriers), removed the unsourced cost figure, and clarified throughout that a negative result should not, by itself, be used to rule out celiac disease. The residual passage was reworded in a later pass to state that a genetic screening not showing the celiac-associated HLA genes does not, by itself, rule out celiac disease.
2026-08-01 · Should I Go Gluten-Free? What Your DNA Can Tell You
What was wrong: The article stated a DNA-based screening could 'rule out celiac disease with >99% confidence,' described its genetic information as 'equivalent' to clinical HLA typing, and cited an uncited $100–300 clinical cost figure.
What changed: Reworded throughout to make clear this screening relies on tag SNPs, is not equivalent to clinical HLA typing, and that a negative result should not, by itself, be used to rule out celiac disease. Removed the unsourced cost figure and added a citation on the tag-SNP method's reduced reliability outside people of European descent.
2026-08-01 · HLA-DQ2 and HLA-DQ8: What Your Celiac Disease Genes Mean
What was wrong: A core section — 'The Greater Than 99% Rule' — taught a false inference: it stated the >99% negative-predictive-value rule for 'HLA-DQ2 and HLA-DQ8 testing' without ever distinguishing clinical HLA typing from this product's tag-SNP proxy, so a reader would reasonably conclude a negative 23andMe-derived result also carries that certainty. A later section ('What About Gluten Sensitivity?') and one FAQ answer repeated the same unqualified claim. Separately, the article and one other FAQ answer stated DecodeMyBio's screening provides 'the same high negative predictive value as a clinical HLA test.' The article's own description field (feeding its meta description, Open Graph tag, and Article JSON-LD) was missed in this pass — it promised readers 'what a negative result rules out,' the exact framing the rest of the correction retracted.
What changed: Restructured 'The Greater Than 99% Rule' so the >99% figure is explicitly attributed to clinical HLA typing, with the tag-SNP proxy introduced as a distinct, less-certain method (detailed later in the article). Reworded the 'What About Gluten Sensitivity?' passage and the remaining FAQ answer the same way. Reworded the two 'same high negative predictive value' instances to state the tag-SNP approach is informative but not equivalent to clinical HLA typing, and added a citation on its reduced reliability outside people of European descent. The description field was corrected in a later pass to say the article explains 'why a negative result does not rule out celiac disease on its own,' matching the rest of the correction.
2026-08-01 · At-Home Pharmacogenomic Testing
What was wrong: The article claimed consumer DNA chip results have '99.5%+ concordance with clinical assays,' a figure with no supporting source, repeated an uncited '$250–$2,000+' clinical pricing figure in several places, and separately stated a negative celiac screening result 'effectively rules out celiac with greater than 99% negative predictive value.'
What changed: Removed the unsourced concordance percentage. Replaced the repeated clinical-cost figures with links to our sourced pharmacogenomic testing cost breakdown ($100–$2,000+, matching that article's own cited range). Reworded the celiac passage to state a negative tag-SNP result should not, by itself, be used to rule out celiac disease.
2026-08-01 · What to Do With Your 23andMe Raw Data
What was wrong: The article and FAQ claimed 23andMe's genotyping arrays are 'typically >99% concordance' accurate (once in the body, once in the FAQ), and stated raw data files contain '600,000 to 700,000' variants as a fixed figure (twice in the body, once in the FAQ) — both without a source establishing a precise number.
What changed: Removed the unsourced concordance percentage in both instances, and reworded the variant count in all three instances as an approximate, chip-version-dependent range. Note (2026-08-11): this page was later merged into /learn/upload-23andme-raw-data-guide and now redirects there. The URL above is kept as the accurate record of where the correction was made.
2026-08-01 · How To Upload 23andMe Raw Data for Pharmacogenomic Analysis
What was wrong: The article stated raw data files contain 'typically 600,000 to 700,000 data points for 23andMe v5, or around 700,000 for AncestryDNA' as a precise figure.
What changed: Reworded to note the count is commonly cited in that range but varies by chip version.
2026-08-01 · DNA Test for Medication
What was wrong: The FAQ claimed genotyping accuracy is '>99%' for tested variants, without a source.
What changed: Removed the unsourced percentage; the FAQ now states genotyping is generally reliable for the variants tested.
2026-08-01 · 23andMe and Antidepressant Response
What was wrong: The article body and one FAQ answer claimed consumer genotyping arrays have 'typically >99% concordance,' without a source.
What changed: Removed the unsourced percentage in both places.
2026-08-01 · Use Your Raw Data for Pharmacogenomic Medication Insights
What was wrong: The article body claimed '>99% concordance rates' plus an uncited 'A 2020 study in Clinical Pharmacology & Therapeutics confirmed high concordance' between consumer arrays and clinical-grade testing — a study reference with no author, PMID, or link. The FAQ (rendered twice, in-body and in the FAQ schema) repeated the '>99% concordance' figure.
What changed: Removed the unsourced percentage and the unverifiable study reference in the body; the accuracy claim is now qualitative. Removed the percentage in both FAQ instances.
2026-08-01 · Celiac Disease vs. Gluten Sensitivity
What was wrong: Across the intro description, several body passages (including the 'What Genetic Testing Can Tell You' section and the paragraph introducing DecodeMyBio's own screening), a promotional callout, and two FAQ answers, the article stated or implied that a negative result from DecodeMyBio's consumer tag-SNP screening — not just clinical HLA typing — rules out celiac disease with 'over 99% confidence,' without distinguishing the two test types.
What changed: Reworded every instance so the >99% figure and 'essentially ruled out' language are attributed specifically to clinical HLA typing. Everywhere DecodeMyBio's own tag-SNP screening is mentioned, added that it is not equivalent to clinical HLA typing and that a negative result there should not, by itself, be used to rule out celiac disease.
2026-08-01 · 23andMe Didn't Shut Down: What to Do With Your DNA
What was wrong: A 'Celiac Screening' passage stated that a negative result from the tag SNPs in 23andMe data 'rules out celiac with greater than 99% certainty,' without noting that a tag-SNP proxy is not the same as clinical HLA typing.
What changed: Reworded to state the >99% figure describes confirmed celiac patients carrying the genes (the basis for clinical HLA typing's reliability), and that this product's tag-SNP result is informative but should not, by itself, be used to rule out celiac disease.
2026-07-31 · Is Pharmacogenomic Testing Worth It?
What was wrong: The article stated that clopidogrel non-response is associated with CYP2C19 ultrarapid metabolizers. That is backwards: ultrarapid metabolizers activate clopidogrel more, not less.
What changed: Corrected to CYP2C19 poor metabolizers, who convert less of the drug to its active form and are the group CPIC flags for reduced clopidogrel effectiveness.
2026-07-31 · Best MTHFR Supplements
What was wrong: Three inline PubMed IDs were wrong. One of them pointed at an unrelated paper rather than the study being described.
What changed: Corrected to Prinz-Langenohl 2009 (PMID 19917061) for L-methylfolate in C677T homozygotes, Wilson 2012 (PMID 22277556) for riboflavin and blood pressure, and Papakostas 2010 (PMID 20595412) for SAMe. The same Prinz-Langenohl citation was corrected in the methylation testing guide.
2026-07-31 · 23andMe and Antidepressant Response
What was wrong: Aripiprazole dosing guidance was attributed to CPIC. CPIC has no aripiprazole guideline.
What changed: Re-attributed to the Dutch Pharmacogenetics Working Group (DPWG), which is the body that actually publishes CYP2D6 aripiprazole recommendations.
2026-07-31 · MTHFR and Pregnancy
What was wrong: The recommendation against routine MTHFR testing was attributed to the American Academy of Family Physicians (AAFP) in three places across two articles.
What changed: Re-attributed to the American College of Medical Genetics and Genomics (ACMG), which issued that recommendation. Corrected in this article's body and FAQ, and in the MTHFR supplements article.
2026-07-31 · Why Codeine Doesn't Work for Some People
What was wrong: This article cited an FDA drug safety communication URL for codeine and tramadol that no longer resolves — the page returned a 404.
What changed: Replaced with the FDA's current Codeine Information page, which covers the same safety communications.
2026-07-01 · What to Do With Your 23andMe Raw Data
What was wrong: The article contained outdated statements about 23andMe's bankruptcy proceedings and the status of the company.
What changed: Corrected to reflect that the sale of 23andMe's assets to TTAM Research Institute completed in July 2025 and the service continues to operate. Note (2026-08-11): this page was later merged into /learn/upload-23andme-raw-data-guide and now redirects there. The URL above is kept as the accurate record of where the correction was made.
2026-07-01 · 23andMe Shutting Down? What to Do With Your Data
What was wrong: The article described the sale of 23andMe's assets to TTAM Research Institute as pending. The acquisition had already completed in July 2025.
What changed: Updated the article to reflect the completed acquisition, current ownership, and the continued availability of accounts and raw data downloads.
Last reviewed: August 2026 · Vytautas Jazbutis