证据评级方法
我们如何评估和评定补充剂研究。
本站上每项成分-健康状况声明都会根据现有同行评审证据的总体情况获得一个字母等级（A 到 F）。本页详细介绍了我们如何得出每个等级——我们的方法是透明且可重复的。
等级定义
Strong Evidence
Multiple high-quality randomized controlled trials (RCTs) or meta-analyses with consistent positive results. Combined sample size exceeds 500 participants across 5 or more independent studies.
示例： Melatonin for sleep onset latency — supported by 20+ RCTs with consistent results.
Moderate Evidence
At least one well-designed RCT showing positive results, supported by additional studies. Results are mostly consistent across the evidence base, with adequate combined sample sizes.
示例： Magnesium for sleep quality — supported by several RCTs with mostly positive outcomes.
Limited Evidence
Preliminary positive findings from small RCTs, observational studies, or mixed results across studies. Evidence is promising but insufficient to draw firm conclusions.
示例： L-theanine for sleep — small positive studies exist but larger trials are needed.
Preliminary Evidence
Only in vitro, animal, case report, or pilot study data available. Or human studies exist but with inconsistent or inconclusive results. More research is needed.
示例： Ashwagandha for hair growth — limited to preclinical and small pilot studies.
Negative Evidence
Fewer than 30% of studies show positive effects, with two or more studies available. The weight of evidence suggests the ingredient does not provide the claimed benefit, or may cause harm.
示例： An ingredient where multiple RCTs show no benefit over placebo.
评分算法
证据等级由四个独立的评分维度计算得出，每个维度对累计分数产生贡献，最终映射为一个字母等级。
| 维度 | 分数范围 | 描述 |
|---|---|---|
| 研究类型质量 | 0–4 | 最高质量研究类型：Meta 分析 (4)、随机对照试验 (3)、对照临床试验/队列研究 (2)、观察性研究 (1)、体外实验/综述 (0) |
| 一致性 | -1 to +1 | >70% 正面结果：+1，<30% 正面结果：-1，其余为 0 |
| 样本量 | -1 to +1 | 总受试者：≥500 (+1)、≥100 (0)、<100 (-1) |
| 研究数量 | -1 to +1 | 研究数量：≥5 (+1)、≥2 (0)、<2 (-1) |
最终等级映射： 得分 ≥6 → A、≥4 → B、≥2 → C、≥0 → D。当正面结果 <30% 且研究数量 ≥2 时强制评为 F。
研究流程
Systematic Search
Identify relevant research from PubMed
For each ingredient-condition pair, we conduct systematic PubMed searches using MeSH terms and title/abstract keywords. We prioritize randomized controlled trials (RCTs), meta-analyses, and systematic reviews, while also including observational studies and pilot trials for emerging evidence.
Paper Screening
Filter for relevance and quality
Retrieved papers are screened for relevance to the specific ingredient-condition relationship. We filter by study type (prioritizing interventional over observational), population (human studies preferred), and publication quality (peer-reviewed journals only).
PICO Extraction
Extract structured study data
From each included study, we extract structured PICO data: Population (sample size, demographics), Intervention (substance, dosage, duration, form), Comparison (placebo or active comparator), and Outcome (primary endpoint, effect size, statistical significance). AI-assisted extraction is validated against source text.
Evidence Grading
Calculate algorithmic grade (A-F)
Our grading algorithm scores each ingredient-condition pair based on four dimensions: (1) highest study type quality, (2) consistency of positive results across studies, (3) total combined sample size, and (4) number of independent studies. The final score maps to a letter grade from A (Strong) to F (Negative).
Publication
Review and publish evidence summaries
Generated evidence summaries undergo compliance review for FDA/FTC adherence. All language uses structure/function claims only. Evidence grades are recalculated automatically when new research is added to the database, ensuring grades reflect the most current body of evidence.
数据来源
局限性
我们的研究方法存在以下已知的局限性，请用户知悉：
- 我们主要在 PubMed 中检索，可能无法涵盖所有相关研究（例如发表在未被索引期刊上的研究）。
- 人工智能辅助的数据提取虽经过验证，但偶尔可能对复杂的研究设计产生误读。
- 我们的评级算法对研究数量和样本量赋予同等权重，这在某些情况下可能无法反映各因素的真实重要性。
- 证据等级反映的是当前的研究状态，随着新研究的发表可能会发生变化。
- 每个人对补充剂的反应各不相同。高证据等级并不保证对每个人都有效。
FDA 免责声明: 这些声明未经美国食品药品监督管理局（FDA）评估。本网站上的产品和信息无意用于诊断、治疗、治愈或预防任何疾病。所展示的证据等级基于我们对已发表的同行评审研究的分析，不构成医疗建议。在开始任何补充剂方案之前，请务必咨询您的医疗保健提供者。