Evidence grading
Grades are computed per peptide-outcome pair, never per peptide, from five inputs by a fixed rule. No editorial judgment, no hand-tuning, and no exceptions for any vendor.
The five inputs
Thresholds imported from the grader itself| human-rct | 5 |
|---|---|
| human-observational | 4 |
| animal | 2 |
| preclinical | 1 |
| none | 0 |
| Largest human n ≥ 200 | 3 |
|---|---|
| Largest human n ≥ 50 | 2 |
| Largest human n ≥ 1 | 1 |
| No human study | 0 |
| ≥ 3 independent research groups | 3 |
| ≥ 2 independent research groups | 2 |
| ≥ 1 independent research group | 1 |
| ≥ 90% concordant direction of effect | 3 |
| ≥ 70% concordant direction of effect | 2 |
| ≥ 50% concordant direction of effect | 1 |
| Bias risk One point per flag — industry funding stated, preregistration stated absent, blinding stated absent — floored at −3. A flag is counted only where a record states it. Silence deducts nothing and certifies nothing: deducting for every unreported funding source would subtract the same constant from every peptide in the corpus, which measures how little has been extracted rather than how much bias there is. | 0 to −3 |
From total to letter
The five components sum, floored at zero. The letter is then resolved in order:
- Best study type scoring 1 or less (preclinical or nothing) → E, whatever the total.
- Best study type scoring 2 (animal) → D, whatever the total.
- Total ≥ 11 and replication ≥ 2 → A. Both conjuncts are required: a high total carried by one research group is not replicated evidence, however large the trial.
- Total ≥ 8 → B.
- Otherwise → C.
The first two rules are gates, not scores. Animal evidence cannot reach C by accumulating replication and consistency points, because no quantity of rat studies is a human study.
Replication and consistency are measured over the same studies. Both read the records at the best available study tier — the tier that produced the study-type score — and nothing below it. A pair whose best evidence is one human trial cannot take replication points from three animal groups, and cannot take consistency points from five animal readings either. The two rows in the decomposition beside every grade therefore describe one population, which is what makes the ledger addable by eye.
The consistency ratio needs at least three stated directions before it returns anything; below that it scores 0 rather than awarding full marks for one paper agreeing with itself. The denominator is printed under the row on every peptide page, because a ratio over three readings and the same ratio over ninety are the same number and are not the same evidence.
| Situation | Components | Total | Grade |
|---|---|---|---|
| Two RCTs from separate groups, n = 400, agreeing, preregistered | 5 · 3 · 3 · 3 · 0 | 14 | |
| One industry RCT, n = 300, unblinded, not preregistered | 5 · 3 · 1 · 3 · -3 | 9 | |
| One small human study, n = 24, no replication | 5 · 1 · 1 · 1 · -1 | 7 | |
| Rat studies only, however many | 2 · 0 · 3 · 3 · 0 | 8 | |
| In vitro work only | 1 · 0 · 3 · 3 · 0 | 7 |
What the corpus actually grades
84 graded pairs · 18 with evidence and no grade| Strong | 6 | |
| Moderate | 25 | |
| Limited | 13 | |
| Preliminary | 40 | |
| Insufficient | 0 | |
| Not graded | Pairs holding evidence where no record carries a determined study design. Declined rather than graded E: E says we looked and the evidence is insufficient, and that is a claim about a literature we have not finished reading. | 18 |
| Peptide | Outcome | Records | With a determined design |
|---|---|---|---|
| Bremelanotide | Pigmentation | 2 | 0 |
| GHK-Cu | Inflammation | 2 | 0 |
| Semaglutide | Sleep quality | 2 | 0 |
| Tirzepatide | Mood & anxiety | 2 | 0 |
| Semaglutide | Muscle growth & preservation | 1 | 0 |
| Tirzepatide | Muscle growth & preservation | 1 | 0 |
| Tesamorelin | Neuroprotection | 1 | 0 |
| Tirzepatide | Neuroprotection | 1 | 0 |
| Tirzepatide | Skin appearance & aging | 1 | 0 |
| Thymosin beta-4 | Sleep quality | 1 | 0 |
| Thymosin beta-4 | Tendon & ligament repair | 1 | 0 |
| CJC-1295 | Wound healing | 1 | 0 |
| Ipamorelin | Wound healing | 1 | 0 |
| BPC-157 | Bone & joint health | 1 | 0 |
| Tesamorelin | Wound healing | 1 | 0 |
| BPC-157 | Exercise performance & endurance | 1 | 0 |
| Semaglutide | Fatigue & energy | 1 | 0 |
| Tesamorelin | Immune function | 1 | 0 |
Grade E showing zero does not mean nothing is insufficient. It means the grader has not yet reached a pair it can say that about. Every pair in the table above holds collected evidence and no record in it carries a design the type mapper could determine — mostly reviews, editorials, commentary and papers PubMed never indexed with a publication type. Those pairs used to render nowhere at all, which turned a 0% into a claim about the corpus rather than about our coverage of it.
A grade of E is not a lesser thing to look at, and it is drawn at the same size in the same ink as A everywhere on this site. “There is not much evidence here” is the most common honest answer about peptides and it is usually the one a reader most needs.
A grade is not a safety assessment. It says how much evidence exists for one outcome. It says nothing about side effects, contamination, drug interactions, or whether anyone should take anything.