Documented safety signals

Published findings

• Endotoxin in preparations made from food-grade material

  The FDA has stated it is aware of compounders using food-grade nicotinamide adenine dinucleotide sold by repackagers to make intravenous products, and that ingredients identified as food grade are not suitable for compounding sterile drugs without appropriate processing because of the high risk of contamination with microbes and endotoxins.

  The agency reports adverse event reports following use of injectable NAD+ products including severe chills, shaking, vomiting and fatigue, some requiring medical treatment, and characterises these as consistent with excessive endotoxin levels.

  In one enforcement action an unopened vial from the same lot as a product administered to three patients who developed symptoms was found to contain bacterial endotoxin at 3,360 endotoxin units per millilitre. The product was held to be adulterated.

  FDA reminds compounders to use ingredients suitable for sterile compounding · verified 2026-07-30

• Research-use labelling does not change regulatory status

  The Therapeutic Goods Administration states that a disclaimer describing a product as being for research use only does not, on its own, change a product’s regulatory status, permit importation, or remove advertising or supply obligations.

  The FDA has issued warning letters to companies selling unapproved products falsely labelled for research purposes or not for human consumption, treating the surrounding claims as evidence of intended human use.

  The practical consequence for a purchasing business is that the label on incoming material does not determine what may lawfully be done with it.

  Understanding your responsibilities when importing, compounding and supplying unapproved peptide products · verified 2026-07-30

• Identity errors that a specification check would catch

  No oxidised NAD disodium salt is registered as a distinct substance. The registry number 606-68-8 designates the reduced form, NADH disodium. Because NAD+ is a zwitterion, only a monosodium salt exists for the oxidised form. Material offered as "NAD+ disodium salt" should prompt a request for the registry number, assay and ultraviolet absorbance ratios before acceptance.

  A registry number in wide circulation for GHK-Cu, 89030-95-9, is not a valid CAS Registry Number at all: the published check-digit algorithm admits only 5 for that sequence. A molecular weight of 403.93 circulating for the same compound corresponds to no registered form.

  For AHK-Cu no registry identifier for the copper complex could be verified from any acceptable source, and no UNII exists for it. Incoming material should be identified against a supplier specification and confirmed analytically rather than by registry number.

  FDA Global Substance Registration System · verified 2026-07-30

• Unapproved supply and seizure

  Health Canada has listed compounds covered here among examples of unauthorised injectable peptide drugs it has seized, stating that unauthorised drug products are illegal in Canada and have not been assessed for safety, efficacy or quality.

  Swissmedic has stated of one investigational compound that the substance is still in clinical development and is not authorised internationally, while noting that it is already available on the black market.

  Rapid Alert System notifications record sales bans and market withdrawals in European member states for an unauthorised novel food ingredient covered here.

  Think twice about injecting peptides bought online — unauthorized products can seriously harm · verified 2026-07-30

• Material degradation that is invisible without the right method

  For NAD+, ultraviolet detection at 260 nm alone is not stability-indicating, because the degradation products also absorb at 260 nm. Loss can only be assessed qualitatively by that method, so a material that assays acceptably by ultraviolet may still be substantially degraded.

  The FDA concluded that the bulk substance substantially degrades on exposure to light, moisture, alkaline pH or standard room temperature, and would not be stable under ordinary storage conditions absent multiple compensatory measures.

  For one approved peptide the drug substance is stored frozen while the finished product must not be frozen. Transferring handling instructions between the two is a documented way to get this wrong.

  FDA Briefing Document, Pharmacy Compounding Advisory Committee, nicotinamide adenine dinucleotide review · verified 2026-07-30

Retatrutide adverse events, by trial arm

Figures below are counts posted to the trial registry for the phase 2 obesity trial NCT04881760, reported over baseline through 52 weeks against a threshold of 5 per cent in any group. They describe what happened in that trial, in that population, under supervision. They are not a description of what a compound does to a person.

The trial ran two arms at 4 mg and two at 8 mg that differed only in where escalation began. Those pairs are the most informative comparison in the whole dataset and are kept separate below rather than pooled.

Percentages of each arm. Column labels give the maintenance level and the level escalation started from. Arms of 33 to 35 participants make single events worth roughly 3 percentage points.

What the table shows that a summary would not

• The response is not monotonic, and how escalation was approached mattered more than where it ended. The two 8 mg arms differed only in starting level and reported nausea at 17.1 per cent against 60.0 per cent, and vomiting at 5.7 per cent against 25.7 per cent. The 60.0 per cent figure is the highest rate anywhere in the trial, above the 45.2 per cent at 12 mg. The phase 2 diabetes trial reproduces the pattern, with nausea peaking at 41.7 per cent in its fast-escalated 8 mg arm against 19.6 per cent at 12 mg.
• Serious adverse events show no dose relationship at all. The rate was 4.3 per cent on placebo and 3.2 per cent at 12 mg; the highest rate in the trial, 6.1 per cent, was in a 4 mg arm. The diabetes trial reads the same way, 6.7 per cent on placebo against 4.3 per cent at 12 mg. A pooled relative risk of 1.46 spans 1. The gastrointestinal gradient is real; extending that intuition to serious events is not supported.
• Discontinuation attributable to adverse events was low across every arm, at most 3.2 per cent. The largest single cause of non-completion in the trial was withdrawal by participants in the placebo arm, at 18.6 per cent, which distorts naive comparisons of completion rates.
• Allodynia — pain from a stimulus that would not normally cause it — appeared in 6.5 per cent of the 12 mg arm and in no other arm. Phase 3 releases record the related term dysesthesia across all arms, rising with dose and reaching 20.9 per cent in one 12 mg group.
• One death occurred in the trial, in a 4 mg arm. The registry states no cause and no causality assessment. The same arm lists drowning and a renal carcinoma among its serious events, but the registry does not link either to the death and no link is inferred here.

Presentations that warrant urgent medical assessment

Recorded because they appeared as serious events in these trials and because delay is the modifiable factor. Acute pancreatitis, acute cholecystitis, acute kidney injury, cardiac failure, cerebrovascular accident and a prolonged QT interval were each recorded.

• Severe or persistent abdominal pain, particularly pain radiating to the back, with or without vomiting — the presentation of acute pancreatitis.
• Pain in the upper right abdomen, fever or jaundice — the presentation of acute cholecystitis.
• Vomiting that prevents fluid intake, reduced urine output, dizziness on standing — dehydration progressing toward acute kidney injury, which was recorded as a serious event in this trial.
• Chest pain, palpitations, fainting, breathlessness, or sudden one-sided weakness or speech difficulty.

Retatrutide holds no marketing authorisation in any jurisdiction. It has no approved product information, no dispensing pharmacist and no pharmacovigilance route outside a registered trial. Nothing on this page substitutes for medical assessment.

Lean mass loss during weight reduction

A substantial fraction of the weight lost on incretin therapies is lean tissue rather than fat. This is measured, consistent across trials, and it is the best-evidenced modifiable risk in this whole section.

• A tirzepatide body-composition substudy at 72 weeks reported 74 per cent of the weight lost as fat mass and 26 per cent as lean mass, with the proportions consistent across sex, age and weight-loss tertile. The placebo arm of the same substudy split 75 to 25. The ratio was a property of losing weight, not of the drug.
• A semaglutide substudy at 68 weeks reported a lean body mass reduction of 5.26 kg alongside a fat mass reduction of 8.36 kg, against 1.83 kg and 1.37 kg on placebo. Expressed as a share of each arm’s own weight loss, lean mass was around 36 per cent on semaglutide and around 56 per cent on placebo. The placebo group lost a higher proportion of its weight as lean tissue than the treated group did.
• Two measurement facts deflate the headline percentages, and both are frequently omitted. Adipose tissue itself carries a fat-free component of roughly 15 to 20 per cent of its mass, so a portion of measured lean loss is obligatory and is not muscle. And skeletal muscle is only about half of total fat-free mass. A figure describing fat-free mass overstates muscle loss on both counts.
• A systematic review of 35 randomised trials across four incretin agents reported a median of 28.3 per cent of total weight loss attributable to muscle-based indices. It also noted that no included study reported objective physical function outcomes.
• Restricting to trials achieving at least 10 per cent weight loss, pooled fat-free mass loss was 33.3 per cent of total weight loss for incretin therapy, against 7.7 per cent for diet combined with exercise.

What the evidence says reduces it

• A randomised trial in 160 older adults reported lean mass falling 2 per cent with resistance training and 3 per cent with combined training, against 5 per cent with aerobic exercise alone, at a matched weight loss of around 9 per cent. Hip bone density followed the same order. This is the cleanest single demonstration, because weight loss was equivalent across arms and only the exercise differed.
• Pooled estimates of the size of that benefit vary with the size of the dataset, and the most quoted figure is the most favourable one. Six trials in older adults yielded resistance training preventing 93.5 per cent of the lean mass loss caused by caloric restriction. Twenty-five trials in 1,608 participants yielded a moderate protective effect. A moderator analysis across 65 trials in 2,537 participants found that resistance training combined with caloric restriction still reduced lean mass significantly. The direction is consistent — training shifts the ratio of what is lost — but the strongest version of the claim rests on the smallest dataset.
• Duration qualifies it further. In the 25-trial analysis the protective effect on fat-free mass reached significance only in interventions of five months or less; at six months and beyond it did not.
• A review of 52 studies reported that 81 per cent of energy-restriction-only groups lost 15 per cent or more of their weight as fat-free mass, against 39 per cent of groups combining energy restriction with exercise.
• A meta-analysis of 24 randomised trials reported that higher-protein energy-restricted diets preserved 0.43 kg more fat-free mass than standard-protein diets at equivalent energy intake, and better preserved resting energy expenditure.
• The protein finding is conditional, and the condition is the part usually dropped. Trials that raised intake only to around 1.0 to 1.1 g per kg of body weight, without resistance training, repeatedly found nothing. Comparisons of 1.7 against 0.9 g/kg over 12 weeks in older adults, 1.02 against 0.86 g/kg in a factorial trial that missed its target, and 1.4 against 0.8 g/kg in postmenopausal women without exercise all returned no significant difference in fat-free mass. The trials that did show an effect used roughly double the lower arm alongside training: 2.4 against 1.2 g/kg with training six days weekly, and 2.3 against 1.0 g/kg in trained participants.
• A meta-regression reported that lean mass gains from resistance training were impaired once an energy deficit reached approximately 500 kcal per day, while strength gains were maintained. This is directly relevant given the size of the deficit these compounds induce.

Whether it matters functionally

Lean mass is a proxy. What it is a proxy for is strength, mobility and physical capacity. Where those were measured directly, they were preserved or improved — but they were almost never measured.

• A 12-month cohort with serial scanning reported lean mass falling around 3 kg by month seven and then stabilising, while grip strength rose by approximately 4 kg and the prevalence of sarcopenic obesity fell from 49 per cent to 33 per cent.
• Over 52 weeks, relative muscle strength changed by +1.0 per cent on liraglutide alone and +3.3 per cent when combined with exercise, against −7.8 per cent on placebo. The placebo group, losing less weight, lost more relative strength.
• A thigh-muscle imaging analysis found muscle volume reductions consistent with population-based expectations for that degree of weight loss, while muscle fat infiltration improved beyond what weight loss alone predicts. Muscle quality improved even as muscle volume fell.
• In a heart-failure trial, six-minute walk distance improved by an estimated 20.3 m over placebo.
• Against this: no pivotal obesity trial of any of these compounds prespecified an objective strength or physical-function endpoint. Across all 33 registered retatrutide trials and 278 outcome measures there is no grip strength, gait speed, short physical performance battery or exercise-capacity endpoint. The only function-related outcomes are questionnaires.
• A counter-signal exists in older adults with type 2 diabetes, a group largely excluded from the obesity trials. Retrospective data there report declining grip strength and accelerated sarcopenia, in contrast to the shorter randomised trials. The most defensible synthesis is that lean soft tissue change is not a reliable predictor of strength change in either direction.

What happens after stopping is unmeasured

The magnitude of weight regain after discontinuation is well established. A pooled analysis of eight trials in 2,372 participants reported regain of 2.20 kg for liraglutide and 9.69 kg for semaglutide and tirzepatide; an extension study recorded participants regaining 11.6 percentage points, roughly two thirds of what had been lost. It is frequently asserted that regained weight returns disproportionately as fat, which would make repeated cycles progressively worse for body composition. No published human study has measured this. Neither of the two randomised withdrawal trials carried a body-composition endpoint, and the pooled analysis covered only body weight, waist circumference and body mass index. The assertion is physiologically plausible and has no dataset behind it in this literature.

What is not established

An absence of published harm is not evidence of safety, particularly where the studies that would detect it have not been run. The following are gaps in the evidence base rather than reassurances.

• No completed and published randomised trial of twelve months or longer exists for NAD+ precursor supplementation, so long-term safety data does not exist.
• Circulating NAD+ concentration rises reproducibly across trials, but three independent meta-analyses report that this does not translate into measured metabolic or muscle outcomes. A biomarker response is not a clinical outcome.
• One risk-of-bias assessment rated none of the twelve included NMN trials at low risk of bias, and two separate author groups have independently used the language of exaggeration to describe how this field reports itself.
• For retatrutide, no public chemistry, manufacturing and controls package exists, no pharmacopoeial monograph exists, and no appearance, solubility, storage or stability data has been published in any acceptable source.
• For AHK-Cu there is no coordination-chemistry measurement, no stability study, no analytical method and no human study of any kind. Exactly one primary study exists and it is ex vivo.
• For several approved products every numeric specification limit is withheld from the public assessment record as commercially confidential, so no acceptance criterion can be quoted.

Handling precautions by compound

Drawn from each compound page, so this list stays in step with the underlying record.

• Semaglutide

  • The active substance is stored frozen and the finished products refrigerated and protected from light. No occupational exposure limit or personal protective equipment guidance is published in any acceptable source, so containment is determined by the receiving facility under its own assessment.
  • The precursor peptide is produced in a yeast strain, which is not a host for mammalian viruses, and no raw materials or excipients of human or animal origin are used. The assessment record concludes there is no risk of contamination with mammalian viruses and no TSE concern.

• Tirzepatide

  • The drug substance is hygroscopic, which implies moisture-protective handling, and is stored frozen between -25 and -10 degrees Celsius.
  • The finished product requires light protection, delivered by the autoinjector and marketing pack rather than by the molecule, and must not be frozen.
  • No excipients of animal or human origin are used. No occupational exposure limit or personal protective equipment guidance is published in any acceptable source, so containment is set by the receiving facility under its own assessment.

• Orforglipron

  • Handled as a hygroscopic, light-sensitive pharmaceutical solid. Containment and environmental controls are determined by the receiving facility under its own occupational assessment.

• Retatrutide

  • The Therapeutic Goods Administration states that describing a product as being for research use only does not change its regulatory status, permit importation, or remove supply and advertising obligations.
  • The FDA has issued warning letters to companies selling unapproved products containing retatrutide that were labelled for research purposes or not for human consumption, treating website claims as evidence of intended human drug use.
  • The developer states that retatrutide is an investigational molecule that cannot be legally sold or marketed for human use.

• GHK-Cu

  • No REACH registration or harmonised classification was located for this substance, which is consistent with cosmetic-only use. Occupational controls are therefore set by the receiving facility under its own assessment rather than by a harmonised hazard classification.

• AHK-Cu

  • No official safety source, REACH registration or harmonised classification was located for this substance. Occupational controls are set by the receiving facility under its own assessment, and cannot be inferred from data published for a related peptide.

• NAD+

  • A very hygroscopic solid. Stored desiccated and protected from light and moisture. The FDA concluded that it will not be stable under ordinary storage conditions unless multiple compensatory measures are implemented.
  • The FDA has stated that ingredients identified as food grade are not suitable for compounding sterile drugs without appropriate processing, because of the high risk of contamination with microbes and endotoxins, and has reported adverse events following use of injectable NAD+ products including severe chills, shaking, vomiting and fatigue, some requiring medical treatment, consistent with excessive endotoxin levels.
  • The FDA recommended against including this substance on the list of bulk drug substances that may be used in compounding under section 503A, on the basis of its instability rather than its characterisation.
  • Endotoxin is the documented failure mode where food-grade material has been used to prepare sterile products. In one enforcement action an unopened vial was found to contain bacterial endotoxin at 3,340 to 3,360 endotoxin units per millilitre, and the product was held to be adulterated. Material intended for a sterile application should be qualified for endotoxin and bioburden against a specification, and food-grade or dietary-supplement-grade material should not be assumed suitable.
  • The FDA has stated that manufacturers and repackagers should clearly identify on the label any ingredient intended for use in foods or dietary supplements, which is the practical control against grade substitution downstream.