Mechanism is not the same as demonstrated clinical benefit. Laboratory evidence is not human evidence. Animal evidence is not human efficacy. A study becomes useful only when you understand who was studied, what was given, how it was given, what was measured, what happened and what the study cannot tell you.
Ten rules that prevent most evidence mistakes
- Lab evidence is not human evidence. Cells and laboratory systems are useful for mechanisms and candidate effects.
- Animal evidence can support a hypothesis, not establish the same effect in people.
- A plausible mechanism can exist without a meaningful clinical outcome.
- Small or uncontrolled studies carry different weight from well-designed randomised controlled trials.
- Population, route, formulation, duration, dose studied and endpoint can materially change interpretation.
- Biomarkers and surrogate endpoints are not automatically patient-important outcomes.
- One positive study is not scientific consensus. Replication and the wider evidence base matter.
- Null and negative findings matter. WhichPeps keeps them visible.
- A study dose is descriptive. It is not a WhichPeps recommendation for an individual.
- Approval status and evidence strength are different questions. A product can be investigational despite strong trial data, while an approved medicine is approved only for defined products/uses.
What each evidence type can and cannot establish
| Evidence type | What it can tell you | What it cannot establish on its own |
|---|---|---|
| Laboratory / in vitro | Mechanisms, cellular responses, candidate effects. | That the same effect occurs safely or meaningfully in humans. |
| Animal / in vivo preclinical | Whole-organism signals and experimental biology. | Human efficacy, human dose or long-term human safety. |
| Observational human | Associations and real-world patterns. | Causation without stronger design and context. |
| Early/interventional human | Initial safety, PK/PD and outcome signals depending on design. | Broad effectiveness across populations from small or early studies alone. |
| Randomised controlled evidence | Stronger causal comparison when well designed. | Automatic generalisation beyond the studied population, route, formulation or duration. |
| Systematic review / synthesis | Context across multiple studies when methods are robust. | Quality beyond the underlying evidence or immunity from publication bias/heterogeneity. |
How to read a study without being misled
1. Start with the research question
What exact population, intervention, comparator and outcome was the study designed to test?
2. Check the design
Randomised? Blinded? Placebo or active comparator? Prospective or retrospective? Controlled or single-arm?
3. Check the population
Healthy volunteers, people with a diagnosed condition, older adults, athletes or animals are not interchangeable populations.
4. Check the product and route
Trial-grade injected material does not establish equivalence to an unverified vial. Topical evidence does not become injectable evidence.
5. Check the endpoint
Was the study measuring symptoms, function, disease events, weight, a lab biomarker or a surrogate?
6. Check the size and duration
Small studies can miss uncommon harms and produce unstable estimates. Short studies cannot establish long-term outcomes.
7. Read the uncertainty
Confidence intervals, missing data, attrition, multiplicity, subgroup analyses and uncontrolled comparisons can change what the headline means.
8. Look for what failed
Primary endpoints, null results and safety findings matter even when a secondary endpoint or subgroup looks positive.
Three distinctions worth remembering
A registered trial is not a result
A registry record can describe a protocol, recruitment status and planned outcomes before any findings exist. WhichPeps treats “registered”, “completed”, “results posted” and “published” as different states. A trial identifier alone is not evidence that the intervention worked.
Component evidence is not combination evidence
If compound A and compound B each have evidence separately, that does not establish the efficacy, safety, interaction, compatibility or optimal formulation of A + B. Combination claims need direct combination evidence or must remain explicitly uncertain.
A COA is product-quality evidence, not clinical evidence
Identity, assay/content, chromatographic purity, sterility, endotoxin and stability answer different questions. A high HPLC purity percentage cannot by itself establish identity, dose/content, sterility, safety or clinical effectiveness.
How WhichPeps applies this
Compound pages lead with a plain-English summary, then show outcome-level evidence before the deeper study record. Human evidence remains separate from animal/laboratory findings; regulatory status remains separate from evidence strength; negative/null findings remain visible; and source records stay linked. WhichPeps deliberately avoids one overall compound score because evidence can differ materially by outcome, population, route and formulation.