Growth Hormone in Healthy Older Adults: One Study, and What Came After It

In healthy adults with an intact growth hormone axis, the controlled evidence does not show that growth hormone works as an anti-aging treatment. The systematic review that settles the question found small changes in body composition — lean mass up, fat mass down — with no demonstrated improvement in functional outcomes such as strength or exercise capacity, and a significantly higher rate of adverse events among treated participants 2. That conclusion rests on randomised controlled trials. The contrary impression rests substantially on one 1990 report of 21 men, twelve of whom received growth hormone and nine of whom received nothing at all 1. This article is about the distance between those two bodies of evidence, and about a distinction — replacement versus supplementation — that the entire field turns on.

What the 1990 study reported

The study enrolled men over 60 whose serum IGF-1 concentrations sat below the range typical of healthy young adults. Twelve received human growth hormone for six months. Nine received no treatment and served as controls. At six months the treated group showed an increase in lean body mass and a decrease in adipose tissue mass, together with increases in skin thickness and lumbar vertebral bone density. The authors observed that the magnitude of these changes was comparable to the changes that accumulate over a decade or more of aging 1.

That is an accurate description of the finding, and the measurements were made and reported in good faith. What follows is not an accusation of error. It is a description of what a study of this design can and cannot establish, which is a separate question from whether it was conducted carefully.

The four limitations, stated precisely

None of these is hidden, exotic, or a matter of interpretation. They are the standard structural limits of the design that was used.

1. Size. Twenty-one participants in total, in groups of twelve and nine. At that scale a single unusual responder moves a group mean substantially, and the confidence interval around any estimate is wide.
2. No placebo control. The nine controls were untreated rather than blinded — no placebo injection, no concealment. Everyone involved knew who was receiving the hormone, including the people taking the measurements.
3. Duration. Six months. Aging is measured in decades. A six-month movement in a surrogate measure says nothing about that measure's trajectory over years, and nothing at all about health outcomes.
4. Surrogate endpoints. Lean mass and fat mass are measurements, not benefits. Nobody experiences their lean body mass. People experience whether they can climb stairs, carry a load, or recover from a fall. None of that was measured.

The people closest to the work said so. The journal that published the study subsequently cautioned against the extrapolation being drawn from it, and the investigators themselves objected to their results being read as a case for anti-aging treatment. The paper was a preliminary physiological observation that generated a hypothesis. It was received as a conclusion.

How 21 participants became an industry

The reading that spread was simple: growth hormone reverses aging. That is not what the paper said, but it is a short step from "changes comparable in magnitude to those of aging" to "reverses aging", and the step was taken almost immediately. Within a few years a commercial sector existed on the strength of it, citing a single unblinded six-month study as though it were a demonstrated clinical effect.

It is worth pausing on the arithmetic. Twelve men received the drug. Not twelve hundred. Not a hundred and twenty. Twelve. The entire comparison group was nine men who received nothing. A result of that size, in a design of that type, is a reason to fund a proper trial. It is not a reason to treat anybody.

The controlled trials that followed

The proper experiment was eventually done. A randomised controlled trial published in 2002 administered growth hormone, sex steroids, or both to healthy aged women and men, with blinding and placebo control, and measured function alongside composition 3. Body composition moved in the expected direction. The functional picture was considerably less impressive than the composition data implied, and adverse effects were common enough to be a headline result rather than a footnote.

The decisive document is the 2007 systematic review, which pooled the randomised controlled trials of growth hormone in the healthy elderly 2. Its conclusions are narrow and consistent. Treated participants gained lean body mass and lost fat mass, by small amounts. There was no demonstrated improvement in functional outcomes — not in strength, not in exercise capacity. And adverse events were significantly more frequent in treated participants than in controls.

• Soft tissue oedema — swelling from fluid retention, the most consistently reported effect.
• Arthralgia — joint pain.
• Carpal tunnel syndrome — median nerve compression, plausibly a consequence of the same fluid retention.
• Gynaecomastia — breast tissue development in men.
• Impaired fasting glucose and raised diabetes risk — growth hormone antagonises insulin action, so this is a mechanistic expectation rather than a surprise.

That list is why the risk-benefit arithmetic does not work in a healthy population. Where a treatment produces a measurable benefit, side effects are weighed against it and the calculation can come out either way. Where the pooled trials find no demonstrated functional benefit, there is nothing on the other side of the scale to weigh them against.

Lean mass is not the same as muscle

There is a technical reason the composition results and the functional results diverge, and it recurs throughout this literature. Dual-energy X-ray absorptiometry, the standard method for measuring body composition, divides the body into fat mass, bone mineral, and lean mass. Lean mass is defined by subtraction: everything that is neither fat nor bone. That includes muscle. It also includes water.

Growth hormone causes fluid retention — reliably, and to a degree sufficient to produce the oedema sitting at the top of the adverse-event list. Fluid retained in tissue is registered as lean mass. So a reported increase in lean body mass following growth hormone administration is a composite of some increase in tissue and some increase in body water, and the measurement cannot separate them.

This is why the functional endpoints carry more weight than the composition endpoints, and it explains why the two diverge in exactly this literature. Water does not generate force. If a measured lean-mass increase were predominantly contractile tissue, strength would be expected to follow it. In the controlled trials it largely does not 23. The composition endpoint is measuring something real. It is simply not measuring the thing that would matter.

Replacement and supplementation are different questions

This is the distinction everything above depends on, and it is routinely collapsed. Growth hormone replacement in adults with diagnosed growth hormone deficiency is legitimate, evidence-based endocrinology. Adult growth hormone deficiency is a defined clinical entity, arising most often from pituitary disease, pituitary surgery, cranial irradiation, or traumatic brain injury. It carries a documented symptom burden and documented metabolic consequences, and its management is covered by an Endocrine Society clinical practice guideline setting out who should be tested, how the diagnosis is established, and how treatment is monitored 4.

Growth hormone supplementation in a person whose growth hormone axis is intact is a different intervention with a different evidence base, and the two are not transferable. Restoring a deficient hormone to a normal concentration and pushing a normal concentration higher are not the same physiological act — no more than treating anaemia and giving iron to someone with normal haemoglobin are the same act. The trials summarised above were conducted in the second population, and the answer they returned is the answer described above.

The guideline is equally specific about how deficiency is established, and this is where the most consequential misreading occurs. Adult growth hormone deficiency is diagnosed by provocative testing: a stimulus is administered and the pituitary's growth hormone response is measured against a defined cut-off 4. It is not diagnosed from a symptom list. Fatigue, reduced muscle mass, central adiposity, and low mood are the presenting features of a long list of common conditions, and of ordinary life. Nor is it diagnosed from age. Growth hormone secretion declines across the adult lifespan in essentially everyone; that decline is a normal physiological trajectory, not a pathological state, and it is not what the deficiency diagnosis refers to.

Holding those two categories apart resolves most of the confusion in this area. Evidence drawn from deficient patients, who do improve on replacement, is routinely cited in support of use in people with normal axes, where the controlled trials found what they found 23. The citation is valid for the first population and invalid for the second, and what separates them is a laboratory result rather than a description of how somebody feels.

What would change the picture

Nothing here rules out some useful effect in some healthy older population the trials to date have not identified. Absence of demonstrated benefit is not proof of absence of benefit, and a systematic review describes the trials that exist rather than the trials that could be run 2. What would change the conclusion is straightforward to specify: adequately powered randomised trials, blinded and placebo-controlled, with functional and clinical endpoints as primary outcomes, run long enough for falls, fractures, independence and mortality to become observable, and with adverse events counted against whatever benefit is found.

That trial has not been done, and the adverse-event profile makes it a difficult one to justify ethically. Until it is, the honest statement of the evidence is the review's own: small composition changes, no demonstrated functional improvement, significantly more harm 2.

The general lesson

The 1990 paper is the clearest available example of a specific failure mode: a small, preliminary, unblinded study read as a conclusion, cited for decades in preference to the larger controlled trials that superseded it, and used to underwrite a commercial category. Those trials were eventually done and are freely available. They say something considerably more modest, and the gap between them and the original is the gap between a hypothesis and an answer.

The practical instruction generalises. When a claim rests on a single study, establish four things: the design, the number of participants, whether there was a placebo, and whether the endpoint was a measurement or an outcome. For the study that built this field the answers are unblinded, twenty-one, no, and a measurement. Everything downstream of it should be read in that light.