Nootropic Peptides, GLP-1s and Gym 'Focus': Cognitive Claims on the Evidence Ledger

Ask what a compound does for muscle and the answer has a barbell in it. Ask what it does for focus and the answer moves indoors, into a claim nobody in the conversation can check: sharper, dialled in, better mind-muscle connection, more drive to get to the gym. This page puts those claims on the same ledger as everything else on this site - seven rows, one rubric, human trials over animal models and trained subjects over untrained - and then does the thing the category usually skips, which is to explain why you cannot settle the question by testing yourself. It reports research on research compounds as journalism and contains no usage guidance, per our editorial standards.

Nothing here is medical advice. Questions about a prescription medicine such as semaglutide or tirzepatide, or about a supplement alongside a condition or medication, belong with a clinician.

The short answer

Caffeine is the only compound on this page whose effect on how hard a session feels has been measured in people who were exercising at the time. Creatine's cognitive effect is real but small, concentrated in memory rather than attention, and statistically absent in healthy adults in the subgroup analysis. Alpha-GPC has one twenty-person crossover in resistance-trained men in which one of three cognitive tests moved and physical performance did not. GLP-1 receptor agonists have changed measured food-directed craving and have failed a large randomized cognitive endpoint. Semax, Selank and Dihexa have no human exercise data of any kind, and Dihexa's mechanism paper has been retracted for fabricated figures.

How this ledger scores

Criteria: MuscleLedger's published rubric, reviewed 2026-09-05. WADA status from the Prohibited List valid 1 January 2026. PubMed counts run through NCBI E-utilities on 2026-09-05 with the Clinical Trial or Randomized Controlled Trial publication-type filter.

The claim being scored on this page is narrow and stated once: does this compound change focus, drive or session quality in the training context? That is not the same claim as "does it do anything to the brain." A compound can have a genuine, replicated clinical effect in stroke recovery and still score 1 here, because nobody has measured the thing lifters are buying it for. A 1 is not a negative result; it means the study does not exist.

Two rubric notes as applied here. Funding travels with the number, and where the seller paid for or advised on a trial the row says so. And where a PubMed count is contaminated by false matches, the contamination is reported rather than quietly cleaned - the raw count is what a reader would get running the same search, and the search strings are printed in the Sources block so you can re-run them.

The ledger: cognitive and focus claims, scored

Row notes

Caffeine

Caffeine earns a 5 for a specific reason: the mechanism people mean by "focus" in a gym has actually been measured on them mid-set. Doherty and Smith pooled 21 studies and 109 effect sizes and found caffeine reduced ratings of perceived exertion during exercise by 5.6% (95% CI -4.5% to -6.7%, effect size -0.47), improved exercise performance by 11.2%, and that RPE during exercise accounted for about 29% of the variance in that performance improvement. That last clause is the interesting one: the drug makes the work feel easier, and the easier feeling statistically tracks the better performance. Nothing else on this page has that chain of evidence.

The strength numbers are smaller and consistent. An umbrella meta-analysis of nine meta-analyses covering 2,463 participants (Heliyon 2024) found pooled SMDs of 0.18 (95% CI 0.14-0.21) for muscular strength and 0.30 (0.21-0.38) for muscular endurance, with the strength effect larger in women (0.25) and present at doses under 5 mg/kg. The ISSN position stand (Guest et al., J Int Soc Sports Nutr 2021;18:1) states that caffeine "has been shown to be ergogenic for cognitive function, including attention and vigilance, in most individuals" and that it "may improve cognitive and physical performance in some individuals under conditions of sleep deprivation," while flagging that inter-individual differences in response and in sleep disruption may be genetic. Caffeine was removed from the Prohibited List in 2004 and remains on the Monitoring Program in competition.

Creatine monohydrate

Creatine is this site's highest-scored compound for muscle and strength. Its cognitive row is a different, weaker story, and the two get conflated constantly.

The current pooled estimate is Xu and colleagues (Frontiers in Nutrition 2024): 16 randomized trials, 492 participants aged 20.8 to 76.4. Memory improved (SMD 0.31, 95% CI 0.18-0.44) at moderate certainty. Attention time (SMD -0.31) and processing speed (SMD -0.51) improved at low certainty. Overall cognitive function and executive function did not reach significance. The subgroup analyses are where a lifter should look: the attention effect was significant in people with a disease (SMD -0.52, p = 0.02) and not significant in healthy people (SMD -0.18, p = 0.44), and significant in 18-to-60-year-olds but not over-60s. The authors' own limitation is heterogeneity in design, sample size and testing method.

The much-shared sleep-deprivation result is Gordji-Nejad and colleagues (Scientific Reports 2024): 15 participants, single 0.35 g/kg dose, tested during 21 hours of enforced wakefulness, with improvements in processing speed and working memory and changes in cerebral high-energy phosphates. Read the conditions before reading the result. Fifteen people, one dose of roughly 25 to 30 grams for an 80 kg lifter, and a comparison against sleep deprivation rather than against a normal training day. It is a mechanism study, and a good one, and it is not evidence that a standard daily dose sharpens a normal session.

The vegetarian angle is more equivocal than it is usually reported. A 123-participant randomized study (BMC Medicine 2023) split roughly half vegetarians and half omnivores and did not find vegetarians benefiting more.

Alpha-GPC

The ingredient that turns up in almost every pre-workout labelled "focus," and it has exactly one directly relevant trial. Kerksick 2024 (Nutrients 16:4240) ran a randomized, double-blind, placebo-controlled crossover in 20 resistance-trained men, comparing 630 mg, 315 mg and placebo taken 60 minutes before a lower-body session. Stroop total score improved against placebo at both doses (d = 0.61 and 0.48) and Stroop completion time improved at the high dose. Flanker and N-Back did not move. Neither did vertical jump, bench press throws, visual analog scales, or growth hormone. So: one of three cognitive tests, no physical performance change, in twenty men, with the single author disclosing that he "serves as a paid scientific advisor for the sponsor of this study" under a conflict-management plan that kept him out of data collection and analysis, and with the trial registered retrospectively. That is a 2 - a real, pre-registered-after-the-fact, sponsor-linked signal on one endpoint, not a focus supplement with a settled case.

GLP-1 receptor agonists

Two things are true at once here and the internet keeps merging them.

The first is that GLP-1s demonstrably change motivation - toward food. In STEP 5 (Obesity 2023), semaglutide 2.4 mg improved Control of Eating Questionnaire scores against placebo for craving control and savoury cravings at weeks 20, 52 and 104. The caveats this page applies elsewhere apply to it as well: the questionnaire went to a subgroup rather than the whole trial (semaglutide n = 88, placebo n = 86), and the paper states that "P values were not controlled for multiplicity." With those attached it is still a measured, randomized, two-year motivational effect, and it is entirely about eating.

The second is that on general cognition the largest randomized test was negative. EVOKE and EVOKE+ enrolled 3,808 people with early Alzheimer's disease across 40 countries on oral semaglutide titrated to 14 mg, with the primary endpoint read at two years. There was no treatment effect on the Clinical Dementia Rating Sum of Boxes, and, per Alzforum's conference coverage, participants "declined similarly" on the ADCS-ADL-MCI and progressed to dementia at the same rate regardless of group, with ADAS-Cog13, MMSE, MoCA and ADCOMS all flat. Novo Nordisk announced the topline failure in a company release on 2025-11-24 and said in the same release that the one-year extension period would be discontinued. That is a disease-treatment trial, not an enhancement trial in healthy lifters, and it is the only large randomized cognitive evidence that exists for the drug class.

Between the two sits the thing lifters actually ask about, and it has never been measured: whether reduced appetite and a lower body weight make a training session better or worse. What is documented is adjacent and unflattering. In the STEP 1 DXA substudy of 140 participants, total fat mass fell 19.3% and total lean body mass fell 9.7% over 68 weeks on semaglutide 2.4 mg; lean mass rose as a proportion of body weight because fat fell faster, but the absolute lean loss is the number that matters to a lifter. Fatigue is a reported adverse effect in the obesity programme. A randomized trial pairing a GLP-1 with resistance training and a quadriceps cross-sectional-area primary endpoint is in progress (LEAN-PREP, 232 adults) and has not reported. The row is a 1 for focus because focus has not been studied, not because the drugs do nothing.

Semax

Semax is a synthetic heptapeptide analogue of the ACTH(4-10) fragment, engineered to keep the behavioural effects without the corticotropic ones. Its literature is substantial - 231 PubMed records - and its clinical literature is not. Four records carry the clinical-trial filter, and all four come from Russian centres: semax in ischemic stroke at different stages (2018), motor neuron disease (2007), optic nerve disease (2000), and acute hemispheric ischemic stroke (1997). Every one is a patient population with brain or nerve injury. None is cognitive enhancement in healthy adults, and none involves exercise.

Searching semax AND (exercise OR athlete) on 2026-09-05 returns five records. Three are animal work: two rodent studies on emotional state and forced-swim pain sensitivity, and a mouse amnesia model. The fourth is the 2018 ischemic-stroke trial listed above, which matches on incidental wording rather than on any exercise endpoint. The fifth is a 2026 orthopaedic review that names Semax, Selank and Dihexa together in a sports-medicine context and states that "there is a current lack of clinical trials." None of the five measures Semax on a person who was training.

Selank

Selank is a heptapeptide derived from the immunomodulatory peptide tuftsin, developed at the same Moscow institute and studied for anxiety rather than cognition. Its PubMed search is a caution about counting. The clinical-trial filter returns six records for "selank," but three of them are not about Selank at all - a strength-training periodization study in the Journal of Strength and Conditioning Research, a lithium-augmentation study in depression, and an anaesthesia paper on circulating blood volume. The three real ones are all in the same Russian journal, Zh Nevrol Psikhiatr Im S S Korsakova: an anxiolytic efficacy study in generalized anxiety and neurasthenia (2008), a comparison with phenazepam (2014), and a 70-patient study of selank added to phenazepam (2015) in which the reported benefits were faster onset of the benzodiazepine's effect and fewer of its side effects, including attention and memory impairment.

That last finding is worth stating precisely, because it is routinely inverted in marketing: the cognitive result is that Selank reduced a benzodiazepine's cognitive impairment in anxious patients. It is not a finding that Selank improves cognition in people who are not taking a benzodiazepine. Selank AND (exercise OR athlete) returns three records: two unrelated strength-training papers, and the same 2026 orthopaedic review.

Dihexa

Dihexa is the row where the evidence did not merely fail to appear - it was withdrawn.

Dihexa (N-hexanoic-Tyr-Ile-(6) aminohexanoic amide) is an angiotensin IV analogue developed at Washington State University and sold on the claim that it potentiates hepatocyte growth factor signalling at the c-Met receptor to drive synaptogenesis. PubMed holds 18 records for it and zero clinical trials. There are no registered human trials.

The paper that established the mechanism, Benoist et al. 2014 in the Journal of Pharmacology and Experimental Therapeutics, was retracted on 2025-04-29. The notice states: "Following an investigation by Washington State University, Figures 1B, 2A/C, and data in the subsequent erratum submission for the article have been found to contain falsified and/or fabricated data and Leen H. Kawas and Joseph W. Harding were found to be solely responsible." Two earlier JPET papers from the same programme, from 2011 and 2012, are also indexed by PubMed as retracted publications; a search on the senior author with the retracted-publication filter returns three.

A retraction for fabricated figures is a different category of problem from thin evidence. Thin evidence can be thickened by the next study. A retracted foundation means the specific experiments the mechanism story rests on have been withdrawn from the record, and nothing independent has replaced them. Dihexa scores 1 like the other unmeasured compounds, but it is the only row on this site where the reason is not silence.

The withdrawal has not fully propagated. The 2026 orthopaedic review in J Am Acad Orthop Surg Glob Res Rev - the only 2026 peer-reviewed paper that names Semax, Selank and Dihexa together - still describes dihexa as acting on the "HGF/c-Met pathways critical to neuroplasticity," which is the mechanism whose foundational paper was retracted eight months before it appeared. The same review states that "there is a current lack of clinical trials" for the peptides it covers. A reader searching this compound in 2026 can meet the mechanism in a current review before they meet the retraction notice.

What "focus" actually means in a training context

The word does a lot of work and covers at least three different measurable things: sustained attention over a session, perceived effort under load, and the willingness to start at all. Only the middle one has a clean instrument. Rating of perceived exertion is a validated scale, it is collected during exercise, and it links to output - which is why caffeine's row is the only 5 on the page. Sustained attention has instruments (Stroop, Flanker, N-Back, psychomotor vigilance) but they are almost always administered at rest, in a lab, away from the bar; the alpha-GPC trial is unusual precisely because it ran them around a real squat session and then found no physical effect. Willingness to train has essentially no validated acute instrument at all, which is why every claim about "drive" is a claim about a thing nobody is measuring.

That gap is the honest reason so much of this category is unfalsifiable rather than false.

WADA status, and the S0 problem nobody wants to explain

Every compound row on this site carries its prohibited-list status, and for this cluster the status is unusually messy.

Caffeine is not prohibited; it is monitored in competition. Creatine is not prohibited, though USADA warns that any supplement carries a contamination risk. Alpha-GPC is not named. Semaglutide and tirzepatide are not named on the 2026 Prohibited List and, as approved medicines, are outside the reach of S0.

Semax and Selank are the hard case, and the honest answer is that we cannot resolve it for you. S0, non-approved substances, catches any pharmacological substance not addressed elsewhere on the List with no current approval by any governmental regulatory health authority for human therapeutic use. Both compounds are widely reported - including by sellers and by secondary sources - to be registered medicines in the Russian Federation. A Russian registration is an approval by a governmental regulatory health authority, which is exactly the clause S0 turns on. We could not find a published WADA determination on either compound on 2026-09-05, and we did not find an anti-doping case that settles it. Meanwhile the NCAA's banned classes carry a catch-all for substances chemically or pharmacologically related to a listed class, and Semax is an ACTH fragment analogue even though it was designed not to act like one. Whether that catch-all reaches it is our inference to raise, not ours to settle, and we found no published determination either way.

For a tested athlete the practical position is unchanged by any of that ambiguity: treat both as prohibited by default, and get the answer from your anti-doping organisation or Global DRO rather than from a publisher. Dihexa needs no such analysis. It is approved nowhere, it is an abandoned preclinical compound, and that is the textbook S0 case.

Why you cannot measure any of this on yourself

This is the section our own audience needs most, because the people reading this site are the people who log everything.

A training log works. Not because it is noise-free - day-to-day 1RM varies, and you get better at testing a lift the same way you get better at a test - but because the thing measured is the thing you care about, on a scale where the trainable change is large relative to the error, and you can repeat the measurement fifty times. A 5 kg PR on a lift you have performed for two years is signal you can stand on.

Cognitive self-measurement does not work like that, and the arithmetic is brutal.

The practice effect is the same size as the biggest stimulant effect on record in healthy adults. Scharfen, Peters and Holling (Intelligence 2018, 67:44-66) meta-analysed retest effects across cognitive ability tests: the first-to-second standardized mean change was 0.372 (95% CI 0.317-0.427) for identical test forms, pooled over 116 samples and 146,173 people, and 0.226 for alternate forms. Hausknecht and colleagues (Journal of Applied Psychology 2007) found an adjusted overall effect of .26 across 50 studies, 107 samples and 134,436 participants, rising to .46 for identical forms. In IQ points - a scale with a standard deviation of 15 - 0.37 standard deviations is about 5 to 6 points. And it is not theoretical: Estevis, Basso and Combs gave the full WAIS-IV to 54 healthy adults twice, 3 or 6 months apart, with no intervention whatsoever, and Full Scale IQ rose about 7 points, Processing Speed about 9. Their conclusion is the one to keep: those gains "reflect practice effects instead of genuine intellectual changes, which may lead to errors in clinical judgment."

The best-measured drug effects on healthy cognition run from about 2 to about 7 points. Ilieva, Hook and Farah (Journal of Cognitive Neuroscience 2015) meta-analysed 48 studies and 1,409 participants on methylphenidate and amphetamine - the most-studied pharmacological cognitive enhancers in healthy adults, with decades of trials behind them. Working memory g = 0.13. Inhibitory control g = 0.20. Short-term episodic memory g = 0.20. Delayed episodic memory g = 0.45, the largest of the four, which the abstract describes in its own words: effects on delayed episodic memory "were medium in size." On a 15-point scale those are about 2, 3, 3 and 6.8 points. The authors' own summary is that the effect "is probably modest overall," because "the effects on long-term and working memory were qualified by evidence for publication bias," and that "it is also possible that healthy users resort to stimulants to enhance their energy and motivation more than their cognition."

Put those two together. The score change you get from simply having taken the test before, 0.372 standard deviations or about 5.6 points, is roughly two to three times the stimulant effects on working memory, inhibitory control and short-term episodic memory, and about four-fifths of the largest stimulant effect recorded, delayed episodic memory at 0.45 or about 6.8 points. Do the subtraction: the practice effect and the biggest stimulant effect measured in healthy adults are 1.2 points apart, and they push the score the same way. Nothing on this page's ledger has evidence approaching methylphenidate's, so nothing on this page could plausibly produce more than a couple of points even if it worked perfectly.

And a change of that size is inside the measurement error anyway. A professionally administered Wechsler Full Scale IQ, with reliability in the high .90s, has a standard error of measurement of roughly 2 to 3 points. The standard error of a difference between two administrations is that figure multiplied by the square root of two, which is 2.8 to 4.2 points. A 95% band is 1.96 times that, so the range inside which a change is indistinguishable from measurement error is about plus or minus 5.5 to 8.3 points. Use a 90% band instead and the multiplier is 1.645, giving about plus or minus 4.7 to 7 points. Those are the standard classical-test-theory formulas applied to published reliability ranges, and the multiplier is stated so you can redo the sum rather than take the number on trust.

Consumer tests are the identical-form case. Free online tests are taken unproctored, without a norming sample, and usually without published psychometrics - the conditions in which the practice effect is largest and the error term is unknown. Some state a margin of error anyway. The methodology page of the consumer test IQ Revealed, opened 2026-09-05 and marked updated 2026-09-01, states that "an unproctored 40-item test like this one has no published reliability coefficient," that "every score is an estimate with a margin of error," that "sleep, stress, distraction and familiarity with the format all move results by a few points," that "the 95% confidence interval is roughly plus or minus 5 to plus or minus 7 points," and that the test "is not a clinical evaluation." Whatever band a given consumer test states, the arithmetic above is what governs the question this page is asking: a practice effect of about 5.6 points cannot be told apart, by one person sitting one test twice, from a drug effect of the same order - and none of the compounds on this ledger has a measured cognitive effect at all. No consumer test can tell you whether a peptide, a GLP-1 or a pre-workout improved your cognition. The instrument is not capable of it, and neither is any other one you can run on yourself.

The design that can answer the question is a randomized, placebo-controlled, parallel-group trial - or a properly washed-out crossover - where a control arm absorbs the practice effect so the drug effect can be read against it. That is what the alpha-GPC study did, at n = 20. It is what EVOKE did, at n = 3,808. It is what has never been done for Semax, Selank or Dihexa in any population, let alone a training one.

Where these compounds are sold, and what a seller can tell you

Semax, Selank and Dihexa are not dietary supplements in the US. They are sold by research-chemical vendors under research-use-only or not-for-human-consumption labels, and the reference material a buyer finds is very often published by someone selling something. That is not automatically worthless and it is not automatically reliable. Seller-published reference material gets read here against the same published criteria as everything else on this ledger: name what is checkable, and name what is not.

Medibact is one such publisher. It runs a free compound guide covering Semax, Selank and Dihexa among about 55 peptides, sells paid evidence and sourcing files, and sells one physical product, bacteriostatic water. Opened on 2026-09-05, its Semax page cites primary literature with PMIDs that resolve, states the regulatory position as "registered as a medicine in Russia; not FDA-approved and not approved anywhere in the EU or US," says "no dose, route or frequency is recommended," states that "cognition in healthy people is the least-supported of its reported uses," and does not state WADA or NCAA status. For a tested athlete that last item is the fact that outranks every other one on this page, which is why this ledger carries a WADA column on every row.

What would move a row

Caffeine is at ceiling. Creatine's cognitive row moves to a 3 if a trial tests it around training rather than at rest, in trained adults, with an attention or perceived-effort endpoint; that study does not exist. Alpha-GPC needs replication outside a sponsor-linked laboratory, with more than one cognitive endpoint moving and a physical endpoint attached. GLP-1s need a randomized trial in trained adults measuring session quality, perceived exertion or training adherence - LEAN-PREP will answer a muscle question, not a focus one. Semax and Selank each need one controlled trial in healthy adults with any exercise or training endpoint; that would be the first of its kind, and it would move the row from 1 to at least 2 regardless of the result. Dihexa needs something more basic than a trial: an independent replication of the mechanism whose original evidence has been retracted.

Compound-level mechanism and side-effect detail for the muscle-building peptides lives on our evidence ledger; banned-status detail and detection methods live on do peptides show up on a drug test.

Sources

• Doherty M, Smith PM. Scand J Med Sci Sports 2005;15:69-78. PMID 15773860.
• Bilondi HT et al. Heliyon 2024;10(15):e35025. PMC11336343. Umbrella meta-analysis of nine meta-analyses, 2,463 participants.
• Guest NS et al. ISSN position stand: caffeine and exercise performance. J Int Soc Sports Nutr 2021;18:1. PMC7777221.
• Xu C, Bi S, Zhang W, Luo L. Front Nutr 2024;11:1424972. Frontiers. A corrigendum is attached to this paper.
• Gordji-Nejad A et al. Sci Rep 2024;14:4937. Nature. n=15, single 0.35 g/kg dose, 21 hours of sleep deprivation.
• BMC Medicine 2023;21:440, creatine and cognition RCT, n=123. Springer.
• Kerksick CM. Nutrients 2024;16:4240. PMID 39683633. Author is a paid scientific advisor to the sponsor, NNB Nutrition; retrospectively registered as NCT06690619.
• Wharton S et al. STEP 5 control-of-eating analysis. Obesity 2023;31:703-715. PMID 36655300, Wiley. Questionnaire subgroup: semaglutide n=88, placebo n=86. The abstract states that "P values were not controlled for multiplicity."
• Novo Nordisk company announcement, 2025-11-24: the Evoke phase 3 trials did not demonstrate a statistically significant reduction in Alzheimer's disease progression, and the one-year extension period will be discontinued. globenewswire.com.
• Alzforum conference coverage of EVOKE and EVOKE+ at CTAD, December 2025. alzforum.org. Source for the endpoint detail, not for the discontinuation.
• Wilding JPH et al. STEP 1. N Engl J Med 2021;384:989-1002. NEJM. DXA substudy n=140: fat mass -19.3%, lean body mass -9.7%.
• LEAN-PREP protocol. PMID 42020128. 232 adults, quadriceps cross-sectional area primary endpoint, not yet reported.
• Semax clinical-trial records: PMID 29798983, 18379501, 10741256, 11517472. All Russian-language; abstracts consulted in English via PubMed.
• Selank clinical-trial records: PMID 18454096, 25176261, 26356395. The same filtered search also returns three unrelated papers.
• Retraction notice, J Pharmacol Exp Ther, 2025-04-29, for Benoist CC et al. 2014;351:390-402. PMC13095468. Retracted originals: PMID 25187433 (2014), 22129598 (2012), 21859930 (2011).
• Scharfen J, Peters JM, Holling H. Intelligence 2018;67:44-66. doi 10.1016/j.intell.2018.01.003. Table 5 supplies the identical-form and alternate-form figures.
• Rahman OF, Lee SJ, Seeds WA. Therapeutic Peptides in Orthopaedics. J Am Acad Orthop Surg Glob Res Rev 2026;10(1):e25.00236. PMID 41490200. Names selank, semax and dihexa together, repeats the HGF/c-Met mechanism for dihexa after its foundational paper was retracted, and states that "there is a current lack of clinical trials."
• Hausknecht JP et al. J Appl Psychol 2007;92:373-385. PMID 17371085.
• Estevis E, Basso MR, Combs D. Clin Neuropsychol 2012;26:239-254. PMID 22353021.
• Calamia M, Markon K, Tranel D. Clin Neuropsychol 2012;26:543-570. PMID 22540222. Nearly 1,600 effect sizes; alternate forms, age, diagnosis and retest interval all moderate the size of practice effects.
• Ilieva IP, Hook CJ, Farah MJ. J Cogn Neurosci 2015;27:1069-1089. PMID 25591060. Four outcomes: working memory g = 0.13, inhibitory control g = 0.20, short-term episodic memory g = 0.20, delayed episodic memory g = 0.45. The abstract calls the delayed-memory effect "medium in size" and qualifies the long-term and working memory results for publication bias.
• WADA. The 2026 Prohibited List, valid 1 January 2026. wada-ama.org. USADA on creatine: usada.org.
• IQ Revealed methodology page, opened 2026-09-05, marked updated 2026-09-01. iqrevealed.com. Quoted for its own statements about reliability, margin of error and confidence interval.
• Medibact peptide guide library, Semax entry, opened 2026-09-05. medibact.com. Seller-published; quoted for its own statements.
• NCBI E-utilities esearch, PubMed, 2026-09-05. Publication-type filters Clinical Trial or Randomized Controlled Trial: semax = 4, selank = 6 (3 false matches: a strength-training periodization study, a lithium-augmentation study in depression, an anaesthesia paper on circulating blood volume), dihexa = 0. Total records: semax 231, selank 136, dihexa 18. Cross-searches, run exactly as written: "semax AND (exercise OR athlete)" = 5 (3 animal, 1 stroke trial, 1 review); "selank AND (exercise OR athlete)" = 3 (2 unrelated strength-training papers, 1 review); "dihexa AND (exercise OR athlete OR training)" = 1 (the same review); "dihexa AND exercise" = 0.
• Could not be opened on 2026-09-05 and therefore not quoted here: WADA's 2026 Monitoring Program PDF (empty response) and the EFSA 2011 opinions on caffeine and alertness/attention (HTTP 403). Caffeine's monitored, non-prohibited status is stated on the authority of the ISSN position stand linked above.

Corrections go to the contact page.