How we changed our mind
A scoring system is only credible if it actually fires. This page lists every meaningful tier change or score reversal in the SupplementScore database over the past twelve months, with the triggering evidence and the editorial reasoning. Some of these reversals make us look more rigorous; some make us look more cautious; one or two make us look like we got it wrong the first time. All of them are here so you can see how the methodology actually moves.
Promotions
Triggering evidence: A 2024 NIH NIDDK-supported meta-analysis of 27 randomised trials (PMID 38688104) on psyllium for cardiovascular and metabolic endpoints. The combination of cholesterol effect, glycaemic effect in pre-diabetes, and consistent gastrointestinal benefits pushed the composite score from 70 to 74.
Why we changed: The earlier T2 placement reflected mostly older single-trial evidence with mixed funding. The new meta-analysis is large, recent, public-funded, and PubMed-indexed — satisfying the tier-1 citation gate cleanly. Replication condition was already met across the older trial base.
Triggering evidence: A 2025 network meta-analysis comparing standardised saffron extracts head-to-head with sertraline and citalopram in mild-to-moderate depressive symptoms found saffron approached SSRI effect sizes at 28 mg/day.
Why we changed: The mood evidence base is now too consistent across enough independent groups to keep saffron at T3. We did not promote to T1 because the trials are short and the effect-size confidence intervals overlap broadly with placebo at the lower bound.
Triggering evidence: A 2024 WHO-supported meta-analysis (PMID 39641338) of zinc trials in respiratory infections in adults clarified the dose-response relationship and showed clear duration-of-symptom reduction at 75 mg+ daily started within 24 hours of symptom onset.
Why we changed: Earlier guidance was justifiably cautious because of mixed older trials and the GI-tolerance issues at high doses. The clearer dose threshold from the 2024 analysis lifts the score for short-course acute use specifically. Long-term high-dose zinc remains capped because of copper-deficiency concerns.
Triggering evidence: The 2024 Xu systematic review (PMID 39070254) of 16 RCTs and 492 participants confirmed cognitive benefits in adults aged 18–60, particularly in women and disease-state populations.
Why we changed: Creatine was already T1 on muscular endpoints. The cognitive evidence raised the efficacy subscore (e:4 → e:5) and added a cognition tag. This is the kind of within-tier change that often slips past readers but matters for how the supplement is recommended.
Demotions
Triggering evidence: The 2024 O'Keefe review (PMID 39617283) consolidated multiple RCT signals showing pharmaceutical-dose omega-3 (1.8–4 g/day) increases atrial fibrillation risk by ~50%, while dietary intake at ~650 mg/day from food is associated with lower AFib risk.
Why we changed: Standard-dose omega-3 (250–1,000 mg combined EPA+DHA daily) remains T1 for cardiovascular and inflammatory endpoints. The high-dose use case (specifically targeting hypertriglyceridaemia or post-MI risk reduction) is split off to T2 because the AFib safety signal is a real deduction from the safety subscore. Both supplement entries now exist in parallel.
Triggering evidence: Audit revealed the osteoarthritis protocol citation set was recommending doses (3,000+ mg/day) that exceed the NIH ODS upper limit and have produced hepatotoxicity in case series.
Why we changed: The efficacy evidence at the protocol dose is real but the safety subscore drops materially when the recommended dose exceeds a tolerable upper limit. We added an explicit liver-monitoring instruction to the entry and demoted the tier accordingly. This is the kind of change a "promote on efficacy alone" methodology would miss.
Triggering evidence: A 2025 audit of the citation gate found nine T1 entries whose gate-satisfying study turned out, on closer reading, to have undisclosed manufacturer involvement that pushed the funder classification from public to industry-mixed.
Why we changed: These were our errors. The original tier assignments did not apply the funder classification stringently. Re-classified the funder type, demoted the tier accordingly, and tightened the editorial pipeline so that funder-type review is now a separate explicit step rather than implied. Names of the nine entries are in _archive/score-changes.csv.
Safety-driven changes
Triggering evidence: Multiple regulatory advisories (Australian TGA, European EFSA, several Asian MFDS) on raw / improperly processed ho-shou-wu hepatotoxicity case series.
Why we changed: Newly added during the May 2026 adaptogen expansion. We chose to include rather than exclude (the supplement is widely sold in TCM contexts) but capped at T3 with explicit liver-monitoring guidance and a recommendation to use only properly processed material from reputable suppliers. This is the safety-first inclusion pattern we apply when a supplement is widely used but carries a real, documented risk.
Triggering evidence: FDA position that BPC-157 is not eligible for compounded use; widespread grey-market distribution; absent human clinical trial efficacy evidence.
Why we changed: Newly added during the May 2026 longevity-novel-compound expansion. BPC-157 has a large preclinical literature and meaningful market presence, but no human RCT efficacy evidence and an ambiguous regulatory status. Capped at T4 by the citation-gate methodology — preclinical-only does not qualify for any higher tier regardless of the strength of the animal data.
What this exercise teaches us
Three lessons from the past year of changes:
First, the funder-policy step needed strengthening. The 2025-08 batch of nine demotions was caused by us originally applying funder classification too loosely. The editorial pipeline now treats funder-type review as an explicit separate step rather than something that emerges from the literature read.
Second, "efficacy went up, so promote" is not always the right move. The omega-3 high-dose split is a case where new efficacy evidence at a specific dose was paired with new safety concerns at the same dose. The composite score machinery correctly produces a demotion in that situation rather than a promotion. We trust that more now than we did at the start of the year.
Third, the safety-first inclusion approach for ho-shou-wu and BPC-157 is the right pattern when a supplement is widely used but carries real documented risk or has insufficient human evidence. Better to include the entry with explicit cautions than to omit it and leave readers to find their information from a less careful source.
How to use this page
If you want to see the full audit trail of every score change at the row-level (not just the headline ones above), the data lives in _archive/score-changes.csv in the public repository, with one row per change including the triggering PMID, the responsible reviewers, and the score-formula recalculation. We refresh this page annually; meaningful interim changes appear on the homepage "What changed this week" feed in the meantime.