What I got wrong,
and what it says now

Was

The MycoTOX page sold the test (£398) for damp-building exposure, and the testing pages listed it as “the investigation of choice” for water damage, fatigue and brain fog. The heavy metals post said blood tests miss most metal burden because metals leave the blood within hours or days, credited this to the CDC, and presented hair analysis and organic acids markers as the better way to find it.

Problem

Most mycotoxin exposure comes from food, no study validates a urine level as a diagnosis of mould illness, and the German clinical guideline on indoor mould says the test shouldn’t be used for that (AWMF 2023). No organic acids marker is validated as a sign of mould or metal exposure; one of the “mould” markers mainly reflects coffee and dried fruit. Blood lead is the standard lead test (blood lead has a mean life of about 35 days, and bone stores feed back into it), and urine cadmium reflects long-term burden. When one person’s hair was sent to six labs, results differed more than tenfold for 12 minerals (Seidel 2001).

Now

The MycoTOX page explains what the test measures, why I no longer offer it for mould at home, how to read a result you already have, and what to do instead. The testing pages no longer sell it for this. The heavy metals post sets out the established test for each metal, explains why I don’t use provoked testing, and grades hair analysis and organic acids markers honestly, including the hair test I offer. The HTMA page now carries the same reliability finding, and the EnviroTOX page no longer offers MycoTOX on its own for mould; it says the mycotoxin part of that panel shows exposure only.

Found after the mould post rewrite. Read the MycoTOX page →

Was

The dental exposures post recommended a supplement protocol around dental work (chlorella, vitamin C, NAC, selenium, modified citrus pectin, with doses) and called provoked urine testing (DMSA or DMPS) the most reliable mercury test. The heavy metals post called provoked testing the most sensitive measure of body burden. The HTMA page said oxidation type “determines which dietary approach will produce the best clinical outcome”, and the personalised medicine post said metabolic typing was “not an evidence problem”. The binders post described its one trial as covering “heavy metals”; the Non-Negotiables post gave vitamin D deficiency as below 75 nmol/L; and a medication post credited a 2012 PPI warning to the European Medicines Agency.

Problem

No trial has tested those supplements for reducing harm from dental work. The American College of Medical Toxicology recommends against provoked urine testing, because chelators raise urine metals in anyone and there are no valid ranges (ACMT 2017). No trial shows that diet matched to oxidation type improves outcomes. The zeolite trial tested one lead tracer in 42 people, with authors from the maker. UK deficiency is below 25 nmol/L (10 ng/mL) (SACN 2016). The 2012 warning was the MHRA’s. The dental post also gave an unsourced “19–34%” heart-risk figure and described the chlorhexidine study as placebo-controlled.

Now

The dental post lists what’s worth asking your dentist, grades the supplements as untested with no doses, cites the chlorhexidine study and the American Heart Association review accurately, and explains why I don’t use provoked testing. The heavy metals, HTMA and test bundle pages now point to standard blood or urine tests. The HTMA, products and personalised medicine pages describe oxidation type and metabolic typing as a starting point, not a tested rule. The binders, Non-Negotiables and medication posts are corrected as above.

Found while checking the set after the mould and CoQ10 corrections. Read the dental post →

Was

“The Gap: Cholesterol and Statins”, “Medication Nutrient Depletion” and “Drug–Nutrient Interactions” said statin-related CoQ10 depletion explains muscle pain and fatigue, recommended ubiquinol 100–200 mg daily for anyone on a long-term statin, and recommended checking CoQ10 levels in everyone taking one.

Problem

Statins do lower blood CoQ10, but that it causes muscle symptoms hasn’t been shown. In blinded trials, more than 90% of muscle complaints on statins weren’t caused by the statin (CTT 2022), and trials of CoQ10 for them disagree. No CoQ10 level has been defined as too low on a statin. Doses belong with the evidence that tested them, on the CoQ10 page, not as a blanket recommendation.

Now

Each page says statins lower CoQ10, grades the link to symptoms as mechanistic and the CoQ10 trials as mixed, suggests a time-limited trial with a symptom diary after talking to your prescriber if symptoms persist, and links to the CoQ10 page for the trials and doses. Routine CoQ10 testing is no longer recommended.

Found when the CoQ10 page was corrected on 6 October. Read the CoQ10 page →

Was

Two posts, “Mould, Mycotoxins, and Chronic Illness” and “The Mould Nobody’s Talking About”, said about 24% of people carry HLA-DR genes that stop them clearing mycotoxins, that mycotoxins build up in tissue, that glyphosate worsens mould toxicity, that a sauna “challenge” improves urine testing, that cholestyramine has the strongest evidence, and that recovery needs binders in a “correct sequence”. Several other pages repeated parts of this, including the “Non-Negotiables” post, which recommended binders during weight loss.

Problem

I found no population study behind the 24% figure, only case reports. No human study links glyphosate to mycotoxin clearance, and no evidence supports challenge testing. The only placebo-controlled cholestyramine data is a two-week trial in 13 people (Shoemaker & House 2006), and the German clinical guideline on indoor mould says mycotoxin testing of blood or urine shouldn’t be used for diagnosis (AWMF 2023). Ochratoxin A does stay in the blood for weeks, but “builds up in kidney cells” comes from animal work. For weight loss, pollutant levels in the blood rise while the total in the body falls (Kim 2011), and no trial shows binders help.

Now

One rewritten post at /blog-mould-mycotoxins; the other address redirects to it. It separates what’s strong (damp and mould worsen asthma and breathing) from what isn’t established, grades each claim, starts with fixing the building, and says I sell the MycoTOX test. The audio episode notes, the test bundles, gut testing options and Non-Negotiables pages were corrected to match, and the MycoTOX page now cites the German guideline.

Found while checking pages that repeated claims removed from the MycoTOX page on 6 October. Read the rewritten post →

Was

Presented the test as finding a “mycotoxin syndrome” and “body burden”, said mycotoxins accumulate over years, recommended creatine “loading” to mobilise toxins before the test, and described binders, glutathione and “drainage” as the standard protocol. The table listed toxins the panel doesn’t measure (aflatoxin B1, B2, G1/G2, deoxynivalenol, fumonisins). It also linked the test to post-COVID recovery and mast cell problems, and said EnviroTOX tests heavy metals.

Problem

Most people have some mycotoxins in their body, mostly from food (McKeon 2024), and no published study validates a urine level as a diagnosis of mould illness. Mosaic’s own instructions don’t include creatine loading. The binder and drainage protocols haven’t been tested in trials. EnviroTOX doesn’t include heavy metals.

Now

The page starts with fixing the building, including the new Scottish rules on landlord repairs. It explains that a result shows exposure, not illness, lists the 11 mycotoxins the lab actually measures, follows the lab’s preparation instructions, grades binders and drainage as untested, and states that I sell the test. The EnviroTOX page’s panel list was corrected to match.

Found in a review prompted by an enquiry about mould at home. Read the corrected page →

Was

Titled “The Supplement Every Statin Patient Should Be Taking”, with evidence rated “High”. It said the Caso 2007 trial used 600 mg a day, and that CoQ10 lowers systolic blood pressure by about 11 mmHg. Doses were recommended for statin users and over-50s, a brand was named, and a client example was included.

Problem

Three meta-analyses of CoQ10 for statin muscle symptoms disagree (Kennedy 2020: no benefit; Qu 2018 and Kovacic 2025: some benefit). The largest analysis of statin trials found more than 90% of muscle complaints on statins weren’t caused by the statin (CTT 2022). Caso 2007 used 100 mg a day. The 11 mmHg figure came mostly from open-label studies, and a Cochrane review found no significant effect.

Now

Retitled “CoQ10 and Statins: What the Trials Show” and graded mixed. CoQ10 is now “worth a trial with a symptom diary if muscle symptoms persist, after talking to your prescriber”. Doses appear only in a “How much: what the trials used” block. The blood-pressure claim, the brand and the client example are removed.

Found when grading a published critique of statins against DH’s own pages. Read the corrected page →

Was

“Cohen’s cold studies showed that chronic stress tripled susceptibility”; Kiecolt-Glaser’s wound study described as being in medical students; oxygen readings of 93–94 described with no guidance; a home nebuliser solution with NAC and vitamin C described without a caution; testing described as something that “would have shown” the problem; menstrual synchrony and biophotons offered as evidence of biological communication; extended commentary on the social climate of 2020–21.

Problem

Colds rose from about 27% to 47% across stress levels in Cohen 1991, not threefold. The wound study was in 11 dental students. NHS home-oximetry guidance said to contact a GP or 111 at 93–94% and call 999 at 92% or less. Inhaled NAC can cause bronchospasm. The testing claims are untested. The commentary didn’t serve the health points.

Now

Figures corrected and cited. An NHS oxygen-reading box, a “not a recommendation” caution, “I didn’t test”, and the obvious alternative explanation are added. The testing claims are graded, and the commentary is removed. The post now focuses on what helps: sleep, movement, food, people, naming the stressor, and targeted supplements after testing.

Found in the June–July quality review, and at Stephen’s request. Read the corrected page →

Was

“The Gao et al. (2003) RCT — advanced lung cancer patients — showed significant NK cell activation”. Reishi described as hepatoprotective and recommended for raised GGT and impaired oestrogen detoxification, with no mention of liver injury. Cholesterol (“the same pathway as statins”), cortisol and sleep claims ungraded; minimum ganoderic acid and beta-glucan percentages; two named brands.

Problem

Gao 2003 was an uncontrolled study in mixed advanced cancers, not a randomised trial in lung cancer. The team’s 2005 lung-cancer study found no significant change in NK activity. There are case reports of liver injury with reishi powder, one fatal. A Cochrane review found no effect on cholesterol, HbA1c or blood pressure. The sleep and cortisol evidence is in animals. The percentages had no source.

Now

Each use is graded with its trials named. The liver case reports and stop signs are added, and reishi is no longer suggested for raised liver enzymes. A “How much: what the trials used” block (5.4 g a day of a polysaccharide extract for 8 weeks; no official upper limit) is added, along with a supplier disclosure. The brands and percentages are removed.

Found while correcting the medicinal mushrooms guide. Read the corrected page →

Was

“RCT evidence for VO₂ max improvement in older adults” and “testosterone support through Leydig cell stimulation” for cordyceps. Daily doses for five species. A table rating named brands as “clinical” or “wellness”. No studies were named.

Problem

In the trial usually cited (Chen 2010, 20 older adults), VO₂ max did not change in either group; two exercise thresholds did. The testosterone evidence is in mouse cells and mice, and points both ways. The doses had no trial, population or duration attached. The brand ratings weren’t backed by testing I could cite, and the page didn’t disclose that Detective Health works with mushroom suppliers.

Now

Each species is graded, with its trials named and linked, and doses appear only in the Fungi of the Month editions. The cordyceps claims are corrected. Reishi liver-injury case reports are added. The brand table is replaced by a label checklist, and a supplier disclosure is added.

Found in the June–July quality review. Read the corrected page →

Was

“4× more likely to develop clinical cold with fewer than 6 hours sleep vs 7+ hours (Cohen, SLEEP 2009)”; “3× higher infection rate” with chronic stress (Cohen 1991); “29% fewer sick days” with cold showers (Buijze 2016). Also, that healthcare workers show “remarkable resistance” to infection, and that a smoker’s lung cancer was due to “the terrain”, not the exposure.

Problem

The 4× figure is from Prather 2015, not Cohen 2009, and its confidence interval runs from about 1.1 to 18. Cohen 1991 found colds rising from about 27% to 47% across stress levels, not 3×. Buijze found 29% less sickness absence from work, but no difference in days of illness. Healthcare workers had higher, not lower, COVID-19 infection rates. Smoking is the main cause of lung cancer.

Now

Each figure is cited to the right paper with its actual numbers and uncertainty. Unsourced numbers are removed or graded. A line now says exposure still matters, and the smoking section says plainly that smoking causes lung cancer. The testing suggestions are graded as not trial-tested, and Part Two is linked at the end.

Found in the June–July quality review. Read the corrected page →

Was

“Approximately a 30% increased risk of stroke or coronary artery disease”, cited to Holt-Lunstad 2015. Unnamed “Frontiers in Behavioral Neuroscience 2022” citations for the cortisol and dementia claims. “The highest rates of reported loneliness are in young adults aged 18 to 24”, uncited.

Problem

Holt-Lunstad 2015 is about mortality. The heart and stroke figures are from Valtorta 2016. The 2022 Frontiers item is an editorial with no data. The UK figure is for 16 to 24-year-olds (ONS, 2018).

Now

Each figure is cited to its own study with a link. The cortisol claim is cited to Hawkley & Cacioppo 2010, and dementia to Kuiper 2015. The age figure is cited to ONS. All of it is graded as observational. A section arguing about pandemic policy is replaced by a cited, neutral line.

Found in the June–July quality review. Read the corrected page →

Was

A title and chain of reasoning presenting liver congestion as a cause of back pain; that visceral manipulation “can restore normal valve tone” at the ileocaecal valve; and raised ALT/AST explained by the liver’s position.

Problem

The chain from organ to back pain is a model, not a tested finding, and the valve claim had no source. Raised liver enzymes have many medical causes that need a GP.

Now

Retitled “The Body as Nested Bags: A Model for Thinking About Back Pain”. The page is graded as mechanistic and untested, the valve claim is removed, the ALT/AST section points to the GP, and an NHS box on back-pain red flags is added.

Found in the June–July quality review. Read the corrected page →

Was

“Remove the identified high-reactivity foods completely for a minimum of four to six weeks”, then reintroduce by “eating a significant amount of each food on the reintroduction day”. The test page said 6–8 weeks for Class 2 and 8–12 for Class 3; the allergy comparison post said three weeks. Food IgG was described as causing “chronic, low-grade inflammation” 2 to 72 hours after eating, and anti-gliadin IgA as identifying non-coeliac gluten sensitivity.

Problem

Three different elimination periods on three pages, none matching the one randomised trial, which used 12 weeks (Atkinson 2004). The delayed-inflammation mechanism is untested, and there is no validated blood test for non-coeliac gluten sensitivity.

Now

One protocol, worded the same as in client reports: a 12-week trial, then one food back every 3 days (small portion, normal portion, none), with a symptom diary, and no routine retest. Evidence graded mixed. The mechanism is graded mechanistic and untested, the anti-gliadin claim is removed, and a coeliac test before removing gluten is added. The old blog post now redirects to the test page.

Found when the IgG protocol agreed for client reports was checked against the public pages. Read the corrected page →

Was

Stomach acid “declines progressively from around thirty onwards”; heartburn is “more commonly a sign of insufficient acid”; betaine HCl with a starting dose and a capsule-by-capsule titration “until a mild warmth is felt”; apple cider vinegar before meals; low stomach acid as “the most important single upstream cause of SIBO”.

Problem

In healthy people without gastric atrophy, acid output is preserved or higher with age (Goldschmiedt 1991; Katelaris 1993; Hurwitz 1997). People with reflux have the same or more acid, not less (Gardner 2003). There are no trials of betaine HCl for any symptom; the warmth test has never been validated; and the SIBO claim is untested.

Now

Rewritten as “Low stomach acid: what’s real, what isn’t, and how to check”. Causes are atrophic gastritis, surgery and acid-suppressing drugs; the check is a pepsinogen/gastrin-17/H. pylori blood screen. All doses and the titration are removed, each claim is graded, and PPIs get the same treatment, including rebound on stopping.

Found when the page was checked against the reflux and stomach acid register. Read the corrected page →

Was

“A 2025 study found that 90.5% of patients with functional abdominal bloating showed significant gut dysbiosis… dysbiosis as a driver of bloating is now well-established.” A section headed “Why testing changes everything”. Low stomach acid as the first driver. No warning signs.

Problem

The study (Akkoyunlu 2025) had 42 people and no control group, so it can’t show dysbiosis is more common in bloating. The 2025 European consensus says tests are unnecessary without alarm signs (Melchior 2025).

Now

The study is cited with its limits. The consensus is given in full, including its first-line treatments. The testing section is now “Where testing fits”, graded not trial-tested. A GP red-flags box, including persistent bloating in women over 50, is added.

Found in the June–July quality pass. Read the corrected page →

Was

“FSH above 40 IU/L on two tests twelve months apart confirms ovarian failure.” Subclinical hypothyroidism “elevates SHBG”. NICE NG23 “updated 2019”. A DUTCH test before and three months after starting HRT as a “minimum clinical standard”, with MTHFR testing. No section on breast cancer.

Problem

Over 45, NICE diagnoses menopause from symptoms without FSH (NG23, updated November 2024); under 40, one FSH above 25 IU/L is needed (ESHRE 2024). Thyroid hormone raises SHBG, so an underactive thyroid lowers it. There is no evidence for DUTCH or MTHFR testing as a standard before HRT.

Now

Diagnosis follows NICE and the 2024 guideline. The SHBG line is removed. DUTCH is described as optional context, graded not trial-tested. A sourced breast cancer section (Collaborative Group 2019) and both halves of the timing evidence (Boardman 2015) are added. The opening example is now labelled a composite.

Found in the June–July quality pass. Read the corrected page →

Was

“A separate trial using a metabolic chamber found the opposite: 24-hour energy expenditure fell by about 4% with no change in intake.” The nutrient-shortfall model (Fulgoni 2024) was cited with no funding note, and the CALERIE result was given without saying who took part.

Problem

The fall was in energy used while sitting quietly in the chamber, not in total 24-hour expenditure (Broskey 2021). Four of the six authors of the nutrient model work for Abbott Nutrition, which sells products for people on restricted diets; readers should know that. And the CALERIE 1 participants did not have obesity, which limits how far that result carries.

Now

“Energy used while sitting quietly in a metabolic chamber fell by about 4%, with no change in eating.” The model is described as a model, with the Abbott affiliation stated, and CALERIE is described as adults without obesity. The rest of the article was rechecked and re-referenced with PMIDs at the same time.

Found when the article was redrafted and the new draft was compared with the live page. Read the corrected page →

Was

“The UK RDA for magnesium is 300mg/day for men and 270mg/day for women — widely considered the minimum to prevent deficiency rather than an optimal intake.”

Problem

The UK figure is a reference nutrient intake (RNI), not an RDA, and it is set to cover about 97.5% of people, so it is not a minimum. The level below which deficiency becomes likely is the lower reference nutrient intake (LRNI).

Now

RNI 300 mg a day for men and 270 mg for women, covering about 97.5% of people; LRNI 190 mg and 150 mg, the level below which deficiency becomes likely (Derbyshire 2018).

Found while checking the page’s intake figures against the source it now cites. Read the corrected page →

Was

An RCT in 46 elderly adults with insomnia in which magnesium glycinate “significantly improved subjective sleep quality, sleep onset, sleep duration, and melatonin levels vs placebo.” Sleep evidence rated 5 stars.

Problem

The trial used magnesium oxide, not glycinate, at 500 mg elemental a day, in people with low magnesium intake. Total sleep time did not change, and sleep was self-reported.

Now

Magnesium oxide, 500 mg elemental a day for 8 weeks, in older adults with low intake: better insomnia severity score, sleep efficiency and time to fall asleep than placebo; total sleep time unchanged; the dose is twice the 250 mg supplemental upper level. Sleep evidence lowered to 3 stars.

Found while checking sources for a forthcoming article on vestibular migraine, which cites the same trials. Read the corrected page →

Was

“Clinically meaningful improvement in PHQ-9 depression scores, comparable in effect size to antidepressant medication in some analyses.” Depression evidence rated 4 stars.

Problem

The trial was open-label with no placebo: people knew when they were taking magnesium, so part of the effect may be expectation. The comparison with antidepressants is not supported by the trial.

Now

Open-label randomised crossover, 126 adults, magnesium chloride 248 mg elemental a day for 6 weeks: PHQ-9 improved by 6 points against 6 weeks of no treatment. A placebo-controlled trial is the next step. Depression evidence lowered to 2 stars.

Found while checking sources for a forthcoming article on vestibular migraine, which cites the same trials. Read the corrected page →

Was

“Several meta-analyses support magnesium as a prophylactic intervention. The evidence is strong enough that the European Headache Federation includes magnesium in its migraine prevention guidelines.”

Problem

I could not find the European Headache Federation guideline the page cited. The placebo-controlled trials are mixed, and they tested specific forms and doses the page did not name.

Now

Mixed trial evidence, named: one trial of 600 mg a day as trimagnesium dicitrate cut attacks by 42% against 16% on placebo (Peikert 1996); another found no benefit and stopped early (Pfaffenrath 1996). The 2012 American Academy of Neurology and American Headache Society guideline graded magnesium “probably effective” (Level B). Glycinate and malate have not been tested for migraine.

Found while checking sources for a forthcoming article on vestibular migraine, which cites the same trials. Read the corrected page →

Was

“I wouldn't routinely go above 400mg/day without a clinical reason and RBC magnesium monitoring.”

Problem

It contradicted the limit I use in client reports: 250 mg of elemental magnesium a day from supplements, the EU upper level for supplemental magnesium (Scientific Committee on Food, 2001), set on diarrhoea risk.

Now

250 mg elemental a day from supplements by default, with the UK Expert Group on Vitamins and Minerals guidance level of 400 mg named. Food doesn't count towards either. Higher only for a specific reason and a set period.

Found in the same source check, by comparing the page with my client report templates. Read the corrected page →

Was

Figures of 70% and 62% for the share of UK adults falling short on magnesium, credited to the National Diet and Nutrition Survey.

Problem

I could not trace those figures to the survey.

Now

Derbyshire 2018 (NDNS years 1–6): average intake from food about 91% of the reference intake for men and 80% for women; about 14% of men and 12% of women aged 20–59 below the lower reference intake. The analysis was funded by the Health & Food Supplements Information Service, and the page says so.

Found in the same source check. Read the corrected page →

Was

A supplement-of-the-month page on ashwagandha that recommended KSM-66 or Sensoril doses and said it may support thyroid conversion, with no mention of liver injury.

Problem

Case series from the US Drug-Induced Liver Injury Network and Iceland (Björnsson 2020) and from India (Philips 2023, including three deaths in people with existing liver disease) document liver injury. Denmark has banned products containing ashwagandha. A supplement page that leaves out documented harm is not balanced.

Now

Liver injury is the first caution on the page, with the warning signs and both case series. The Danish ban is named and sourced. The thyroid caution now says ashwagandha can raise thyroid hormone levels and to tell your GP if you take thyroid medication. A line notes there is little safety information beyond a few months.

Found while writing an article on coffee add-ins, which covers ashwagandha and had to agree with this page. Read the corrected page →

Was

The trial “enrolled 31 Japanese adults over 50”, at 2.4 g a day for 12 weeks, measured on the MMSE.

Problem

The published trial randomised 30 adults aged 50 to 80, at about 3 g a day for 16 weeks, measured on the revised Hasegawa Dementia Scale. The lead author worked for Hokuto Corporation, a mushroom producer, and the page did not say so.

Now

Correct size, dose, duration and scale; scores better than placebo from week 8 and falling within four weeks of stopping; the lead author’s employer named. The page also now carries the same disclosure of my commercial relationships with mushroom suppliers as the turkey tail page.

Found in the same coffee add-ins check, by reading the trial abstract on PubMed. Read the corrected page →