Use this structure for new articles. Articles report what studies did and found and do not tell readers what to do. Read the Disclaimer first.
> A summary of published research, not medical advice. See the Disclaimer.
[One sentence: what it is. No benefit language.]
**Best-supported effect:** [effect] in [population]. [Verdict with its evidence grade, e.g. "Probably helps (moderate evidence)"].
**Most-cited weak claim:** [claim]. [Verdict].
How we grade evidence · Last reviewed: [Month YYYY]
<div class="bw-know">
<p class="bw-know-h">What you need to know</p>
<p>[One paragraph of 5-7 sentences. Only facts already in this article's tables, text and references: the best-supported finding with its evidence grade, the main weak or untested claims, the most relevant harms, interactions or at-risk groups, and the doses and durations the studies used. Past tense for trial results, numbers where the article has them, no new sources and no advice.]</p>
</div>
## What it helps (n) {#what-it-helps}
| Effect | Verdict | What was measured | Studied in |
|---|---|---|---|
[Verdict in bold, from the phrase table on the Evidence Levels page, including the size word, "a lot" (large), none (moderate or unclear) or "a little" (small), and the evidence grade in brackets: (strong evidence), (moderate evidence), (weak evidence) for Limited, (very weak evidence) for Very limited. What was measured: Hard outcome, Measured, Self-reported or Lab marker. Put {.verdict-table} on the line right after the last table row.]
[2-4 sentences per row: number of trials or participants, duration, result in plain units, the problems that set the grade (verdict first, e.g. "May help (weak evidence): small trials that disagree"), citations [1][2].]
## What it doesn't help (n) {#what-it-doesn-t-help}
| Effect | Verdict | What was measured | Studied in |
|---|---|---|---|
[No-effect verdicts from either evidence section go here: Makes little or no difference, Probably makes little or no difference, or May make little or no difference, each with its evidence grade. This section exists only when it has rows. Counts are table rows; anchor ids stay fixed. What it helps holds Helps and Probably helps verdicts, including size words. Weak benefit, unclear, unknown and worse verdicts stay in their existing section.]
## Claimed but unsupported or contested (n) {#claimed-but-unsupported-or-contested}
| Claim | Verdict | What was measured | Studied in |
|---|---|---|---|
[Per row: what trials found, and where the claim comes from by source type (a mouse study, one small trial). Never a brand or a person. End with one sentence on what hasn't been studied.]
## How it might work {#how-it-might-work}
[2-3 plain sentences: what researchers think could explain the effects, and whether most of the evidence comes from people or from animals and lab cells.]
### Theory 1: [Plain name of the idea]
[Step-by-step cause and effect in plain language. Name the exact molecule, receptor, hormone, store, enzyme or cell at each step (for example "phosphocreatine in muscle", not "energy stores"). Say which steps are shown only in animals or cells. Cite every sentence [n].]
**Evidence for**
- [One finding per bullet, saying whether it was in people, animals or lab cells, cited [n].]
**Evidence against or unclear**
- [Null results, small or short studies, steps never tested in people, cited [n].]
[Repeat for every main mechanism in the literature, usually 3 to 7 theories.]
**Overall:** [One or two cited sentences on how well these mechanisms are shown in people.] A mechanism shown in people does not by itself show a benefit.
[Outcome and safety findings go in the effect, side effect and risk sections, not here. Full sentences, no jargon left unexplained.]
## Side effects (n) {#side-effects}
| Side effect | Verdict | How often vs placebo | Source | Studied in |
|---|---|---|---|---|
[Verdict in bold, using the harm phrases: "Happens (strong evidence)", "Probably happens (moderate evidence)", "May happen (weak evidence)", "Unknown (very weak evidence)" (case reports, postmarketing reports or animal data only), "No harm found (strong evidence)", "Probably no harm (moderate evidence)", "May not increase it (weak evidence)", "Unclear: studies disagree (weak evidence)". Grade harms like benefits: one trial under 500 people is "May happen (weak evidence)", self-reported symptoms stop at "Probably happens (moderate evidence)", and "Happens (strong evidence)" needs the Strong rules. How often vs placebo: numbers when reported ("44% vs 16%"), otherwise "Not reported" or "Case reports only". Source: citation numbers [3][9]. Put {.verdict-table} on the line right after the last table row.]
[At most 1-2 sentences under the table, only for facts the table can't hold.]
[Studied in, for Side effects and Interactions: copy population text only from this page, linked to the row or its citations. Otherwise leave the cell blank and add the row to the gap list. A study design alone is not a population. Existing cell text stays intact.]
## Who was excluded or at risk
This lists what studies and labels report. A pharmacist or prescriber can say whether it applies to a specific person.
- [Factual bullets, cited: "Trials excluded people with...", "Case reports describe... in...", "The label warns...".]
- Not studied in: [pregnancy, children, kidney disease, ...]
[Never "don't take if", "avoid" or "should".]
## Interactions (n) {#interactions}
This lists what studies and labels report. A pharmacist or prescriber can say whether it applies to a specific person.
| Combined with | What happened | Verdict | Source | Studied in |
|---|---|---|---|---|
[Sources: labels, case reports, interaction trials and pharmacology reviews. Verdict in bold, same harm phrases as Side effects, graded like benefits: "May happen (weak evidence)" for a single small interaction or PK study, "Happens (strong evidence)" only when the Strong rules are met, "Unknown (very weak evidence)" for case reports or label cautions with no measured study, "May not increase it (weak evidence)" when a single small study found no effect. Put {.verdict-table} on the line right after the last table row.]
[If nothing is documented, skip the table and write this one sentence: "No interactions with drugs or supplements are documented in labels, case reports or interaction studies found for this article."]
## What the studies used
These are the amounts used in the studies. They are not recommendations.
[Forms, doses, timing and durations, past tense, as ranges, cited. "Trials used 3-5 g/day for 4-12 weeks [1]."]
## Regulatory status
[Required for Research Peptides & Compounds and Prescription Drugs; place it just before References. State approval status (e.g. FDA/EMA), approved uses if any, prescription or legal status, and sport bans (WADA), each cited to the regulator's own page, plus "Checked: [Month YYYY]". No sourcing, buying or administration details. For research compounds, say whether human data exist at all.]
## References
1. Author A, et al. Title. Journal. Year. PubMed link (design, size, funding if declared)
[6-10 sources for the effect sections. How it might work, Side effects, Who was excluded or at risk and Interactions add their own sources, numbered after these. Each link opened and its title checked. In the text, cite as [1] or [3][9]; the site turns each number into a link to its reference, with a link back.]
House rules: no "take", "should", "optimal" or "recommended"; no "treats", "cures" or "prevents"; no emoji, colours, badges, product names or affiliate links; in prose, lead with the verdict, then the reason; give numbers instead of "research suggests". The Evidence Levels page explains the ratings.