Reading the evidence. Not the marketing.

Every optimization compound has a research literature. Most clinicians and most patients read that literature incorrectly. Animal models are not human trials. Surrogate markers are not clinical outcomes. Statistical significance is not clinical significance. The four questions that separate signal from noise.

I. The evidence hierarchy.

Not all research is equal. Evidence-based medicine organizes study designs into a hierarchy that reflects how much confidence each design can produce. Knowing where a compound sits in that hierarchy is the starting point for every honest clinical conversation. verified [I]

Level I. Systematic reviews and meta-analyses of multiple randomized controlled trials. Highest confidence. The Morton 2018 meta-analysis of protein supplementation and lean mass is an example: pooled data from dozens of trials, consistent signal, high confidence in direction and magnitude. verified

Level II. Individual well-designed RCTs: controlled, blinded, adequate sample size, pre-registered. The TRAVERSE trial (testosterone and cardiovascular events in 5,246 men) and the STEP trials (semaglutide in large populations with defined endpoints) are canonical Level II examples. verified

Level III. Cohort studies and well-designed observational studies. These establish correlation, not causation. They generate hypotheses for RCTs. They do not establish efficacy.

Level IV. Case series and expert opinion. Lowest confidence for establishing efficacy. Useful for generating clinical patterns. Not sufficient to establish that a compound does what it is claimed to do.

Level V (optimization medicine specific). Mechanistic plausibility, animal model data, and accumulated clinical experience. This is where most peptide compounds live. It is a valid evidence tier. It must be labeled correctly. Presenting Level V data as though it were Level II is the single most common evidence error in optimization medicine. inferred from clinical literature landscape

II. The four questions to ask about any compound claim.

Four questions, applied consistently to any study abstract, will produce a more accurate picture of what a compound can reasonably be expected to do than any amount of reading conclusion sentences without context.

Question 1. What was the model?

Human or animal? In vitro or in vivo? Young or old subjects? Sick or healthy? The BPC-157 literature is primarily rodent: robust mechanistic signal, consistent results across multiple labs, but no large controlled human trial. The TRAVERSE trial is 5,246 humans with defined cardiovascular endpoints. These are not equivalent evidence tiers, and treating them as equivalent is not a rounding error. It is a categorical error. verified

Question 2. What was the endpoint?

Surrogate marker (IGF-1 level, telomere length, HOMA-IR) or clinical outcome (all-cause mortality, fracture rate, cardiovascular event, quality-of-life score)? Surrogate markers predict clinical outcomes in theory. Sometimes they do not in practice. Finasteride reduces PSA, a surrogate marker for prostate cancer risk. It may simultaneously increase high-grade prostate cancer incidence, a clinical outcome in the opposite direction. The surrogate and the clinical outcome diverged. That divergence matters clinically in ways that a surrogate-only literature cannot predict. verified [IV]

Question 3. What was the population?

Young healthy men? Elderly frail women? Patients with documented deficiency? Healthy adults without deficiency? Effect size varies dramatically by baseline status, and extrapolating across populations is one of the most consistent sources of error in applied optimization medicine. DHEA supplementation in DHEA-deficient older adults produces meaningful effect on lean mass, libido, and wellbeing in multiple trials. DHEA supplementation in young healthy adults with normal DHEA levels: minimal to no effect. The compound did not change. The population did. The clinical implication changes entirely. verified

Question 4. Was it replicated?

A single positive study is a hypothesis. Three positive studies from independent groups with different methodologies are a pattern worth clinical attention. An unreplicated finding in a high-noise field like longevity research or performance biology should be held lightly, especially when the study had a small sample size or used surrogate endpoints. Ioannidis demonstrated mathematically why most published research findings in small-sample high-noise fields are false. The issue is structural, not fraudulent. The incentives that produce publication bias are institutional, not personal. verified [II]

III. Common evidence errors in optimization medicine.

Three errors appear repeatedly in the optimization compound space. Each has a structural cause. Naming them precisely is the first step toward not making them.

The animal-to-human extrapolation error.

"This compound extended lifespan in mice by 25%" gets cited as evidence for human use. Rodent lifespan studies fail to translate to humans at a high rate. Caloric restriction extends lifespan in nearly every model organism studied: yeast, nematodes, fruit flies, mice, rats. In humans, the data is complex and contested. The biology is not wrong; the extrapolation is. Rodent data is valid as a mechanistic signal and a hypothesis generator. It is not a substitute for human trial data. verified

The deficiency-to-normal extrapolation error.

"Testosterone supplementation in hypogonadal men improves body composition, bone density, and cardiovascular markers" is a well-supported clinical finding. It becomes an error when transformed into "testosterone is anabolic and will improve my body composition" applied to a eugonadal man with normal testosterone. The effect size in eugonadal men is substantially smaller. The risk-benefit calculus changes. The deficiency literature does not generalize to the non-deficient population without direct evidence, and in many cases that direct evidence does not exist or points in a different direction. verified

The surrogate marker assumption.

"This compound raises IGF-1" does not mean "this compound reduces all-cause mortality." IGF-1 is a surrogate marker. It predicts certain clinical outcomes in certain populations within certain ranges. It is also associated with increased cancer risk at supraphysiologic levels. The clinical outcome it predicts is complex and non-linear. Treating surrogate improvement as equivalent to clinical benefit is what Heneghan et al. called "surrogate outcome substitution," and it is responsible for a substantial portion of medical reversal: the phenomenon by which treatments accepted as standard of care are later found to cause net harm. verified [IV]

IV. The Pivotal evidence tier framework.

Pivotal uses a four-tier internal framework for categorizing compound evidence. It is not a replacement for the Sackett hierarchy. It is an applied translation of that hierarchy to the specific compound landscape we work with. verified [I]

Tier A. Multiple human RCTs, consistent results, clinical outcome endpoints. Examples: semaglutide (STEP trials, large-scale human populations, clinical weight loss and cardiovascular outcomes); testosterone (TRAVERSE and multiple prior RCTs, clinical cardiovascular and body composition endpoints); NR and NMN (multiple human trials on NAD+ biomarker endpoints, emerging clinical data).

Tier B. Human trials with surrogate endpoints, plus strong mechanistic data, plus replicated animal models. A compound with this evidence profile has a credible biological case and some human data, but has not yet been tested in a large controlled trial with clinical outcome endpoints. Examples: BPC-157 (extensive rodent mechanistic literature, clinical observational data in gastrointestinal and musculoskeletal applications); thymosin alpha-1 (multiple human trials in specific immune-compromised populations); SS-31 (Phase II human trial in heart failure with preserved ejection fraction).

Tier C. Strong mechanistic data, replicated rodent models, limited human data. The biological plausibility is well-established. The human evidence is early or narrow. Examples: Epithalon (human fibroblast data and Soviet-era clinical literature, not replicated in modern Western trial formats); Dihexa (rodent models for cognitive function, early human trials); MOTS-c (human biomarker data emerging from exercise physiology research).

Tier D. Mechanistic hypothesis, in vitro data only, or a single unreplicated rodent study. Most novel peptides appearing in research chemical markets fall here. This is not necessarily a permanent designation. It is the current evidence state. It should inform both clinical caution and patient communication.

V. How to discuss evidence tiers with patients.

Every compound recommendation should include a one-sentence evidence statement delivered before the recommendation is accepted. Not buried in a disclosure form. Spoken, clearly, as part of the clinical rationale. The format is simple.

For a Tier A compound: "This compound has multiple large human trials showing [specific clinical outcome] in [population similar to yours]." For a Tier C compound: "This compound has strong animal model data and early human biomarker data, but no large controlled human trial yet. The mechanistic case is solid. The clinical confirmation is still building."

The patient who receives this framing is an informed participant. The patient who does not is a passive recipient. Shared decision-making requires documenting the evidence tier, the patient's understanding of it, and the clinical rationale for proceeding despite incomplete evidence where applicable. This is not a legal formality. It is the clinical standard for any decision made under uncertainty.

The question to ask: "Given what I have told you about the evidence, does this make sense for your goals?" A patient who answers yes with full information is a partner. A patient who answers yes because they trust the clinician without understanding the evidence is not informed consent. It is borrowed confidence, and borrowed confidence does not hold when outcomes diverge from expectations.

VI. Applying evidence literacy to this note series.

The operator notes in this series cite peer-reviewed literature for every claim that has it. Verified tags indicate a paper has been directly referenced and the claim is traceable to the cited source. Inferred tags indicate mechanistic extrapolation from established biology: the reasoning is disclosed, the certainty is lower.

Not every claim in these notes is Level I evidence. Nor should it be. Optimization medicine operates at the frontier of human performance biology. The frontier is not fully mapped by RCTs, and waiting for RCTs before forming any clinical position would mean practicing medicine that is decades behind the biological evidence. The obligation is different: know which claims are Level I, which are Level III, and which are Level V. Prescribe accordingly. Adjust as the evidence matures.

THE PIVOTAL PROTOCOL is an intelligence and education layer, not a prescriber. The frameworks described here are derived from the cited literature and from Pivotal's own protocol design history. Every clinical decision belongs to a licensed physician with full knowledge of the case. Begin a conversation. Do not begin self-administration from a website.