What Growth Hormone Actually Does, and Why the Trials and the Gym Disagree

Somebody on Reddit asked whether to start with growth hormone or go the peptide route. Tesamorelin, CJC-1295, ipamorelin. I answered like I knew, because I’d made that call for myself a few months earlier.

What I told him: I’m 44, my own GH pulse isn’t great to begin with, so amplifying a weak pulse didn’t make much sense compared to putting the actual hormone in. Then I went and checked it, and the physiology says the opposite.

That correction turned into a much longer read than I expected, because the deeper I got, the stranger the picture became. The clinical literature and the people using this stuff have spent about forty years answering different questions and then disagreeing about the answers.

What growth hormone does

Start with the strongest evidence that it does anything at all, because it comes from an uncomfortable place. In 1999 two parallel double-blind trials gave high-dose growth hormone to 532 critically ill patients. Mortality was 39% versus 20% placebo in one trial and 44% versus 18% in the other. Both under a thousandth significance. It ended growth hormone use in critical illness.

Sit with that, because it settles one argument permanently. A drug that does nothing cannot double your mortality. Whatever else is true here, growth hormone exerts powerful systemic effects.

It builds collagen, not muscle fibre

In 2010 a Copenhagen group did something the body-composition studies never did. They biopsied muscle and tendon in healthy young men on growth hormone and measured the tissue types separately using isotope tracers, over fourteen days.

Contractile muscle protein synthesis did not move. Tendon collagen synthesis rose about 1.3-fold. Collagen inside the muscle rose nearly six-fold, though that larger number sat just outside statistical significance and should be read as a trend rather than a result.

The dissociation is the finding. Growth hormone’s anabolic signal lands on the collagen matrix and not the muscle fibre. A DEXA scan cannot see that distinction, and neither can a one-rep max.

It replicates in a more useful direction too. Injected directly into the patellar tendon of elderly men, growth hormone raised local collagen synthesis without changing serum IGF-1, which means the effect is local to the tendon rather than spillover from the bloodstream. And in men recovering from two weeks of leg immobilisation, the placebo group lost tendon stiffness while the growth hormone group held it and then increased it during rehab. In those same men, muscle strength recovered identically in both groups. The dissociation again, this time at the functional level.

One caution belongs right here rather than buried later. More collagen is not the same as better tendon. In horses given growth hormone around a tendon injury, the tendon got measurably bigger and weaker, with lower tensile strength and reduced stiffness. Rat rotator cuff repairs came out weaker too. And acromegaly, which is decades of growth hormone excess, produces hypertrophied muscle that is measurably less strong than normal muscle. The mechanism that might help you at eight weeks is the one that makes acromegalic tissue big and useless at twenty years. That is a dose and duration warning, not a contradiction.

It changes what you lose in a deficit

This is the claim with the best evidence in the whole space, and it is not the one people lead with.

The cleanest experiment: healthy subjects fasted forty hours, their growth hormone pharmacologically suppressed, then replaced. Suppressing it raised urea excretion by around half, increased muscle protein breakdown, and lowered fat oxidation, all at once. Restoring growth hormone normalised all three.

That is the lean-sparing-during-a-deficit claim, tested by withdrawal and restoration, which is about as close to causal as this kind of work gets. The clinical version backs it up: in obese subjects on a thirty-eight day restricted diet, the fraction of weight lost as fat was 0.77 with growth hormone versus 0.63 on diet alone. Same weight gone, different composition.

Two honest limits. In that same study the lipolytic effect held for the full four weeks while the nitrogen-sparing effect faded. And one well-conducted study using a different method of blocking growth hormone found no protein effect at all.

Insulin is the switch

Growth hormone is not a general fat-burning agent. It is a lipolytic agent specifically when insulin is low, and this reframes most of the trial literature.

Block growth hormone in fed subjects and lipolysis does not change at all. Block it after a fifty-nine hour fast and the entire fasting-induced rise in lipolysis is abolished. The mechanism is direct: insulin antagonises the exact signalling arm growth hormone uses to release the brake on fat breakdown.

The obvious objection is that a deficit already raises your own growth hormone, so adding more should do nothing. That was tested and it fails. In trained men cycling, growth hormone raised the glycerol response to 716% of baseline against 328% on placebo. The axis is not maxed out.

And the weak link, which belongs in the same breath: in that same study, the tripled fat mobilisation produced no additional fat oxidation during exercise. Fatty acids that leave the fat cell and are not burned go back in. At rest and fasted, growth hormone does raise fat oxidation, by around 29% over twenty-four hours. During exercise, catecholamines already dominate and growth hormone adds nothing.

It takes the fat you cannot see

Nine months of growth hormone in abdominally obese men: total body fat down 9%, abdominal subcutaneous fat down 6%, visceral fat down 18%. Subcutaneous fat is not spared, but it is roughly a third as responsive.

Which fat that is depends entirely on what you came for. Visceral is the metabolically dangerous depot, so an 18% reduction is a real health result that stands on its own. It is also the fat that never shows in a mirror. If your goal is visible leanness, growth hormone works hardest on the tissue you are least interested in, and that probably explains a chunk of the gap between expectation and experience.

Why the trials and the gym disagree

Here is what took me longest to work out. The trial record says no strength gain, modest fat change, no anti-aging case. The strength world has treated this as near-essential for decades. Both of those groups contain people who are not stupid.

I went in assuming the answer was dose, that trials used a fraction of what people take. That turns out to be wrong, at least for the trials that matter.

The athlete trials ran roughly five to ten times a clinical replacement dose, which is the review authors’ own description of their included studies. The most-cited figure for what athletes take traces back to a single 2006 paper whose authors describe their source as unverifiable underground information, with no citation. Every later source, including a medical association report, cites back to that same assertion. When researchers have tried to collect real dosing data, including a hundred-person prospective cohort and a purpose-built bodybuilder study, they came back with none. On a weekly average, the trials may well have dosed harder than whatever people are really doing.

So the dose argument fails. Four others hold up.

The exposure was twenty days. The athletic-performance trials averaged under three weeks, and only three of twenty-seven ran longer than a month. Collagen synthesis rises within fourteen days, but tendon remodelling, the thing that would let you train harder, takes months. A twenty-day trial captures the signal and none of its structural consequence.

The trials fixed training volume by protocol. Four studies added growth hormone on top of resistance training and found no strength advantage. But every subject did the same prescribed program. A recovery benefit expresses itself as tolerating more training. If growth hormone let you train five days instead of three, a trial that mandates three cannot see it. That is a design limitation, not special pleading.

They measured the wrong outcome. In the doping-practice literature, growth hormone is not primarily characterised as a strength or size drug. Those are attributed to other compounds. It is described as a recovery and tissue drug: heal faster, train more often, joints hold up. The trials, almost without exception, measured strength and size. The field spent forty years carefully answering a question the users were not asking.

Growth hormone is essentially never taken alone. In one survey, steroid users were running close to nine compounds at once. A moderate weekly testosterone dose, with no training at all, adds more fat-free mass than growth hormone’s entire measured effect, and unlike growth hormone it produces large dose-dependent strength gains. Someone adding growth hormone to that stack and feeling it work is not imagining things. They are just not in a position to know what the growth hormone specifically did.

What it does not do

Being equally honest in the other direction, because some of this is settled and dressing it up would waste your money.

Strength in healthy adults. Measured directly across multiple trials, including in young trained athletes at around 10% body fat who were urine-tested to confirm they were clean. Six weeks changed neither their body fat nor their maximum strength. This is evidence of absence, not absence of evidence.

Contractile muscle growth. Falsified at the biopsy level. The lean mass that shows up on a scan is substantially extracellular water. In young active adults on supraphysiologic doses for a month, fat-free mass rose 5.3%, while total body water rose 6.5% and extracellular water rose 9.6%. The authors concluded the water accounted for the gain.

Anti-aging. The original 1990 study was twenty-one men pre-selected for low IGF-1, not randomised, and it never measured strength, function, cognition or mortality. The best systematic review in healthy older adults concluded growth hormone cannot be recommended as an anti-aging therapy. The animal evidence runs the other way entirely: mice with less growth hormone signalling live substantially longer, and people with lifelong genetic insensitivity to it get strikingly less cancer and virtually no diabetes.

Joint benefit. Contradicted in direction. Joint pain is one of the most commonly reported adverse effects, not a perk.

Sleep

Better sleep is one of the top reasons people try this, and it was the first thing the guy on Reddit mentioned. The evidence is worse than thin. Some of it is invented.

The physiology is real and it is where the idea comes from: in men, roughly 70% of daily growth hormone output happens during early sleep, and both deep sleep and growth hormone fall steeply with age on the same timeline. But the arrow runs the other way. Researchers blocked the GHRH receptor in healthy young men, removing about 93% of the nocturnal growth hormone pulse, and slow-wave sleep did not change. Deep sleep drives the hormone. The hormone has not been shown to drive deep sleep.

The only controlled sleep-lab data on exogenous growth hormone comes from deficient patients on replacement, and delta activity dropped about 30% versus placebo. Lighter sleep, not deeper. For healthy adults using it non-medically there is no polysomnography data at all.

The peptides are worse. There is no human sleep data for ipamorelin, CJC-1295 in either form, sermorelin or tesamorelin. What exists is data on neighbouring compounds that contradict each other: native ghrelin increased deep sleep, hexarelin at the same receptor family decreased it, GHRP-6 moved a different stage entirely, and GHRH helped men while impairing sleep in young women. You cannot average that into a class effect.

And in the one population where researchers measured both, patients reported feeling more rested while their EEG showed sleep getting lighter. Subjective sleep and measured sleep can move independently.

Four studies that do not exist

While researching this I kept hitting very specific, very confident citations. A 2019 study in Endocrine showing ipamorelin raised stage 3 sleep 23% on polysomnography. A sixteen-week sermorelin trial showing 34% more slow-wave sleep. A 2015 Clinical Endocrinology ipamorelin trial showing 6.2% visceral fat loss. A 2021 Aging Cell sermorelin trial showing, also, 6.2% visceral fat loss.

None of them exist. No PMID, no authors, no DOI, no trace anywhere except pages selling the product. Two of them report the identical effect size, which is what templated content looks like.

I’m putting this in its own section because it is the most useful thing I can hand you. This market manufactures evidence. If you read a number precise enough to sound trustworthy, search the study, and if the only results are shops, the study is not real.

What nobody has measured

I went looking for data on recovery between training sessions, because that is the specific thing lifters report. Creatine kinase, soreness, force restoration after damaging exercise, tolerated training volume.

There is none. Not weak data, not conflicting data. Nobody has given people growth hormone, made them train hard, and measured any of it. Not once.

That absence is not evidence the effect is fake. It is evidence the question was never asked, which after forty years is remarkable on its own. The closest human proxies both point the same direction, tendon stiffness preserved through disuse and quadriceps strength recovering better after ACL reconstruction in a small trial, but both are rehabilitation from injury rather than recovery between sessions.

If you are using growth hormone for recovery, you are operating on mechanism and experience. That is not nothing. It is also not a study.

Growth hormone versus the peptides

Back to the question I was asked. My reasoning was that at 44 my own pulse isn’t strong enough to be worth amplifying, so I’d rather put the hormone in directly.

The hidden assumption was that a weak pulse means a weak pituitary. That isn’t how the axis ages. Output falls roughly 14% per decade after 30, but what fails is the signal, not the gland. The hypothalamus sends less GHRH and more somatostatin, which is the brake. The pituitary keeps its capacity, and that is exactly what secretagogues act on. An oral secretagogue in adults aged 64 to 81 took IGF-1 back into the young-adult range. My argument was backwards.

A narrower version survives: agents working only through the GHRH pathway do get somewhat blunted with age, and combining pathways restores the response. That argues for a different secretagogue strategy, not against secretagogues.

The evidence behind each compound is wildly uneven, and that gets flattened everywhere.

Ipamorelin is the one people discuss most confidently and the one with the least behind it. Its published human trials were intravenous, for recovery of bowel function after surgery, and stopped before phase 3. Every percentage you see quoted for it is extrapolated from mechanism, animal work, or a clinic’s own numbers.

Sermorelin is not FDA-approved, whatever a clinic page says. Approved in 1997 for growth failure in children, withdrawn in 2008 for business reasons. Everything available now is compounded.

And underneath all of it, no head-to-head trial exists. Not one, for any endpoint, between any secretagogue and growth hormone. When someone tells you a peptide gets you 70% of what growth hormone does, there is no study that number could have come from.

On the idea that these are self-limiting

A line that shows up constantly: secretagogues are self-limiting, negative feedback stops you going too high, so you cannot really get into trouble. It is false, and it is used to justify skipping bloodwork.

Feedback modulates, it does not cap. Tesamorelin’s own FDA label tells prescribers to consider stopping the drug for persistent IGF-1 above three standard deviations. You do not write that instruction for a drug that cannot push someone too high. And in tesamorelin’s trials patients still ended up with IGF-1 above the normal range and still developed diabetes at elevated rates, despite an intact feedback loop. The argument that pituitary feedback keeps you safe is weaker than it sounds.

The DAC question, briefly

I’ve always read that the DAC version is a no-go for everybody. The mechanics of why the two versions get dosed completely differently are in a separate post, so I won’t repeat them. Two things aren’t in that one.

The usual objection is that DAC’s sustained elevation flattens your pulses and desensitises the receptor. That was directly tested on this compound: blood sampled every twenty minutes across twelve overnight hours, before and a week after a dose. Pulse frequency and magnitude came back unchanged, so the pulses themselves survive.

I used to stop reading there, and that was a mistake. The same data contains the part people are actually worried about: trough growth hormone rose more than sevenfold, with mean GH up 46% and IGF-1 up 45%. The pulses are intact, but they now sit on a floor that never comes back down. The paper’s own title says it plainly — pulsatile secretion persists during continuous stimulation. That is not the objection answered, it is the objection measured. The pulse survives. The off-period does not.

So the reasons to be careful with DAC are not really about losing your pulses. They are about never being off: you cannot stop it quickly, the floor stays raised for as long as it is in you, and total IGF-1 exposure is higher for it.

CJC-1295 with DAC reached Phase 2 in HIV patients with visceral obesity, 192 people enrolled, and the trial was halted on 17 July 2006 after a participant in Argentina died of a heart attack. The investigating physician concluded he had silent coronary disease and that it was unrelated to the drug. The company closed the program and never restarted it, and that Phase 2 was never published, so the safety data has never been available outside the company. A competing GHRH analog in the same class ran its Phase 3 straight through and was approved in 2010.

Blood sugar, and the test I was getting wrong

Growth hormone raises blood sugar, through the same lipolysis described above: free fatty acids go up and interfere with insulin signalling. In treated adults it usually shows up as a small shift, and the risk concentrates in people who are older, heavier, and have been on longer.

I’ve been checking fasting glucose twice a week because of this. Having looked at it properly, I was leaning on it too hard and running it too often, and skipping the tests that actually catch this early.

Fasting glucose is a late marker. Insulin resistance builds for years while your pancreas compensates and the number reads normal the whole time. What catches it earlier is fasting insulin and HbA1c. The fix isn’t to drop fasting glucose, GH can push it up on its own, it’s to run those two alongside it every few months instead of leaning on glucose alone. Tesamorelin’s own safety signal was HbA1c, not fasting glucose: about 5% of patients crossed 6.5% versus 1% on placebo. That’s what I’m changing.

The claim that peptides sidestep this is not established. A two-year trial of an oral secretagogue showed reduced insulin sensitivity and a sustained rise in fasting glucose. For CJC-1295, ipamorelin and sermorelin nobody has measured it. That is an absence of data, not a finding of safety.

IGF-1, and why the Z-score is the number

The raw value on its own is close to useless. IGF-1 falls steeply with age and differs by sex, so the same number can be unremarkable at 25 and high at 60. What you want is the Z-score: how many standard deviations you sit from the median for your age and sex. Normal runs roughly negative 2 to positive 2, and every clinical target is written that way.

You also cannot compare IGF-1 across labs. When researchers ran eight commercial assays against the same healthy population, the reference ranges came out meaningfully different. A jump between two labs might be the assay rather than you. Pick one and stay with it.

One more thing that changes how you read your own numbers: IGF-1 plateaus with dose while fat loss appears to keep climbing. IGF-1 is a good exposure marker and a poor proxy for whether the lipolytic effect is scaling.

On suppression, and the comparison I was going to make

I wanted to draw the obvious parallel: growth hormone is to your pituitary what testosterone is to your HPTA. After reading the literature I dropped it.

Suppression is real. But with testosterone, suppression starves an organ. Your testes shrink without LH, which is why recovery is slow and sometimes incomplete. Nothing atrophies in the growth hormone axis. Your liver resumes making IGF-1 whenever growth hormone returns, and in the studies we have suppression lifts within hours to days.

The caveat, and it is a caveat rather than a reassurance: nobody has studied what happens to the axis after years of above-normal growth hormone in a healthy adult. The comforting data come from replacement doses, largely in children, over short periods. “No evidence of permanent shutdown” is true. “Proven safe long term” is not.

The one place the comparison works is diagnostic. While you’re on, your IGF-1 tells you nothing about your own production, the same way total testosterone tells you nothing on TRT. Getting a real baseline means being off long enough to have one, which is an argument for taking it before you start anything.

The risk side

The everyday profile is mostly one thing wearing different hats: fluid retention. Across eighteen randomised trials it showed up as swelling, aching joints and carpal tunnel symptoms, with soft-tissue edema in 44% of growth hormone subjects against 1% on placebo. One trial had to cut its dose mid-study because of how often these appeared.

I’m dealing with this one now. It’s in my hands, mostly at night, enough that it wakes me up. Carpal tunnel symptoms, though not carpal tunnel the way people usually mean it. It eased off when I brought my dose down, which a structural problem in the wrist wouldn’t do. Worth knowing the difference, because if you look up your own symptoms you land straight on surgical pages for something that may settle once exposure comes down.

Glucose deterioration is progressive. At the low end it reads as impaired fasting glucose. In one trial at a genuinely high dose over twenty-four weeks, around 17% developed frank diabetes. This gets worse with time on, not better.

Duration appears to matter more than dose. In acromegaly, how long the growth hormone excess lasted predicts cardiac damage, including biventricular enlargement, diastolic dysfunction and valve disease, better than how high the level was. The largest mortality cohort in treated patients found no relationship to daily or cumulative dose. If you are running this for years rather than blocks, duration is the variable to think hardest about.

On cancer, both popular versions are wrong. People with naturally higher IGF-1 get modestly more cancer in large observational studies, and genetic studies support a causal link for colorectal specifically while coming out inconclusive for prostate and breast. Effect sizes are on the order of 10 to 20% higher odds per standard deviation, not a doubling. That evidence studies people whose IGF-1 was set high for life by genetics, which is not the same as an adult raising it with a drug, and the largest follow-up of treated patients found no significant excess mortality in the low-risk group. The unambiguous part: growth hormone is contraindicated in active malignancy. Not a discussion point, a contraindication on the label.

The legal reality, as of July 2026

Treat this as a snapshot. Somatropin and tesamorelin are FDA-approved, both for narrow indications. Sermorelin’s approval was pediatric and the product left the market in 2008. CJC-1295 and ipamorelin have never been approved for human use.

The compounding picture is mid-rewrite. FDA has been moving peptides onto and off its restricted category, and removal doesn’t by itself authorise compounding. If the status of a specific compound matters to your decision, check it against FDA directly rather than against me.

Growth hormone also has its own federal criminal statute. Distributing it for anything outside an approved indication carries up to five years, and FDA reads “distribution” to include writing the prescription. That is a different legal category from the peptides, none of which carry an equivalent law. I’m not a lawyer and this isn’t legal advice, but if you didn’t know that, you should.

So how do you decide?

The useful move isn’t ranking the compounds. It’s working out which thing you care about, because that sorts the field faster than any pharmacology argument.

That last row should decide it. Both routes need IGF-1 with a Z-score and your metabolic markers, checked on a schedule, from the same lab, starting from a baseline you take before you touch anything. If you’re not going to do that, the honest answer isn’t growth hormone or peptides. It’s neither, not yet.

Where I land

I’m staying on it. What the evidence supports, connective tissue and holding muscle through a deficit and visceral fat, lines up with what I’m using it for. The things it demonstrably doesn’t do were never what I was after.

What I’m changing is the bloodwork. Adding fasting insulin and HbA1c on top of fasting glucose, and easing off the twice-a-week checks, because the test I was leaning on couldn’t catch the thing I was watching for on its own. And I’m retiring the argument I gave on Reddit, because it was backwards.

The part worth taking past growth hormone: I had a story that sounded like physiology and I’d never checked whether it was true. It took an afternoon to find out it wasn’t, and I’d been saying it out loud to strangers who were making decisions. Most of what circulates in this space is plausible-sounding mechanism repeated until it feels like evidence. Some of it, it turns out, is studies that were never run at all.

Read more

• CJC-1295 with DAC vs without: why one name covering two molecules produces completely different schedules.
• HGH IU to mg: the unit conversion that confused me for longer than I’d like to admit.
• Tracking bloodwork over time: why the trend beats any single draw, which is the whole argument behind the Z-score section.
• Before you buy peptides: what a certificate of analysis does and doesn’t tell you. Relevant here, because purity testing can’t detect endotoxin.
• Lab price tracker: what the markers in this post cost if you order them yourself.

Frequently asked questions

Sources

The reason this post reads the way it does is that these four findings do not agree with either the forum position or the “GH does nothing” position, and I would rather show the work than ask anyone to take my word for it.

• Takala 1999. High-dose GH in critically ill patients roughly doubled mortality, 39% against 20% in one arm and 44% against 18% in the other. It is a grim result and it settles one thing cleanly: a drug that does nothing cannot do that. GH is not inert. Two parallel trials, 532 patients between them. N Engl J Med 1999;341:785–92.
• Doessing 2010. Collagen synthesis rose while myofibrillar synthesis did not. Being precise about the numbers matters here: the tendon collagen result is a 1.3-fold increase at p=0.02, which is solid. The muscle collagen figure of 5.8-fold is p=0.06, which is a trend, and reporting it as a significant finding, as it often is second-hand, overstates it. J Physiol 2010;588:341–51.
• Nørrelund 2001. Suppress GH during a 40-hour fast and protein breakdown rises by roughly half while fat oxidation falls; restore GH and both return to baseline. That is the lean-sparing role shown causally, in the catabolic state where it actually operates. Eight subjects, each their own control across four conditions. Diabetes 2001;50:96–104.
• The trials that shaped the “GH does nothing” consensus have two structural problems. Mean duration was about 20 days, and the four GH plus resistance-training studies fixed training volume by protocol, which means a recovery benefit had no mechanism by which to show up even if one existed. The review states the duration directly — a mean of 20 days across the studies that dosed for more than one day, ranging from 4 to 84, with only three running past 30 days. It pools 27 study samples and 303 people who received GH, for 13.3 person-years of treatment in total. Ann Intern Med 2008;148:747–58.
• Ionescu & Frohman 2006. The CJC-1295 pulsatility study, and worth reading past the headline. Pulse frequency and magnitude were unchanged, but trough GH rose 7.5-fold, mean GH 46% and IGF-1 45%. Preserved pulses on a permanently raised floor, which is a different thing from an untouched rhythm. J Clin Endocrinol Metab 2006;91:4792–7.
• What has never been measured: inter-session recovery in humans. Not weak evidence, not mixed evidence. It has not been an outcome variable. The thing lifters have claimed for forty years has never been tested.
• A dead end worth recording: the argument that athlete trials used too little GH does not hold, because they ran roughly 5 to 10 times replacement dosing, at or above credible real-world use. The gap is duration, not dose.

This is my own experience and general education, not medical advice, and nothing here is a recommendation to use any of these compounds. CJC-1295 and ipamorelin are not approved for human use. Somatropin and tesamorelin are approved only for specific medical indications, and growth hormone is contraindicated in active malignancy. Regulatory status described here is current as of July 2026 and is actively changing. Talk to a qualified professional before you start, stop, or change any medication, supplement, or protocol, and before acting on any lab result.