Higher BMI, lower IGFBP-3 and shorter paternal height predicted greater height gain — and the resulting web tool is explicitly not yet validated for prescribing.

Growth hormone treatment for idiopathic short stature has an uncomfortable property: it works, on average, by a modest amount, and the variation between children is large enough that averages are close to useless for an individual family deciding whether years of near-daily injections are worthwhile.

A study in Value in Health from Chungnam National University in Daejeon attempts to narrow that uncertainty, modelling growth trajectories in treated children and identifying which baseline characteristics predict a larger response.

The lead researcher is Professor Jung-woo Chae of the College of Pharmacy at CNU.

What was measured

The study is a retrospective, single-centre analysis of 91 prepubertal Korean children with idiopathic short stature — 41 boys and 50 girls — treated with recombinant human growth hormone between July 2020 and December 2023. The drug was given subcutaneously six to seven times a week, with doses adjusted during routine care.

Idiopathic short stature is a diagnosis of exclusion: a child is well below the expected height for their age with no identified cause, no growth hormone deficiency, no underlying disease. It is the category where treatment is most contested, because the children are not deficient in the hormone they are being given.

The mean treatment period was 619 days, give or take 307 — a spread wide enough to matter, since height gain accumulates with exposure. Mean height percentile rose from 1.26 at baseline to 9.16 at follow-up.

Growth trajectories were modelled with a Gompertz nonlinear mixed-effects model driven by cumulative growth hormone exposure, which estimated a total growth-response parameter of about 17.2 percentile points — with substantial variation between patients.

The three predictors

Three baseline characteristics were associated with a greater response: higher body mass index, lower IGFBP-3, and shorter paternal height. BMI showed the strongest association.

“Three baseline characteristics were associated with a greater predicted response,” Chae said, “with BMI showing the strongest association”.

Each has a plausible rationale, and the plausibility runs in different directions.

IGFBP-3 is a binding protein for insulin-like growth factor 1, the mediator through which growth hormone acts, and it is used clinically as a marker of growth hormone activity. A child with lower IGFBP-3 has more headroom in that axis, so a larger response to supplementation is what the physiology would predict.

Shorter paternal height is more likely statistical than biological. Percentile gain and absolute height gain are not the same quantity, and a child short because their parents are short may simply have more room to climb the percentile scale.

BMI being the strongest predictor is the one that most needs care in how it is read, which is addressed below.

The tool

From the model the team built GrowCast, a web tool that generates predicted height and percentile trajectories from a child’s age, sex, height, weight, paternal height, IGFBP-3 level and growth hormone dose.

Two-year simulations produced median predicted height percentiles ranging from 4.0 per cent for lower-response profiles to 31.0 per cent for higher-response profiles — a spread that is the point of the exercise, since it is the difference between a treatment that moves a child a little and one that moves them substantially.

“GrowCast enables clinicians to compare modeled treatment scenarios,” Chae said, and to “communicate potential growth trajectories with patients and families”.

Communication is the honest framing. A family weighing years of injections against a projected percentile gain is judging proportionality, and currently doing so with population averages; a projection specific to their child is better input even if imperfect.

The limits, which are substantial

The authors state plainly that GrowCast is not yet a validated dosing-prescription tool, and that larger prospective multicentre studies are needed before routine clinical use. Several features of the study explain why that caution is appropriate.

Ninety-one children at one centre is a small basis for a predictive model with three predictors, and the figure a reader should hold onto is not the 17.2 percentile points but the substantial variation around it. A model fitted on 91 patients and evaluated on the same 91 patients will describe them better than it describes anyone else — which is what external validation exists to test, and what has not been done here.

The design is retrospective, which means doses were adjusted during routine care rather than by protocol. Clinicians raise doses in children who are responding poorly and hold them in children doing well, so cumulative exposure is partly a consequence of response as well as a cause of it, and a model driven by cumulative exposure inherits that entanglement.

There is no control group. Prepubertal children grow without treatment, and the rise from the 1.26th to the 9.16th percentile is a change in a treated cohort rather than a change relative to what would have happened anyway. Percentile gain does partly adjust for expected growth, which is why percentiles were used, and it does not substitute for an untreated comparison.

The cohort is also entirely Korean and prepubertal, and both matter for transfer: growth reference charts are population-specific, and the pubertal growth spurt is the period in which most height is gained and most treatment decisions become urgent.

Why BMI deserves particular caution

The strongest predictor is the one most open to misreading, and it should be said explicitly that this is not a finding about what children should weigh.

In a cohort of children selected for being very short, BMI varies for reasons that include nutrition, body composition and the simple arithmetic of dividing weight by the square of a small height. Higher BMI in this context is a statistical association with treatment response in 91 children, not an instruction, and nothing in the study tested changing a child’s weight or examined what would happen if anyone tried.

The broader context is worth stating too, because the release does not. Whether growth hormone should be prescribed for idiopathic short stature at all remains genuinely contested among paediatric endocrinologists. Trials generally show gains in final adult height of a few centimetres for several years of daily injections at considerable cost, and the psychosocial benefit that is often the implicit justification has been hard to demonstrate. A model that identifies which children gain most is useful on either side of that argument — it could target treatment to those likeliest to benefit, or it could show how many are being treated for little.

None of this is a reason for any family to start, continue or stop a treatment, and decisions of that kind belong with a child’s own paediatric endocrinologist. What the study offers is a sharper way of describing the uncertainty, which is a real contribution to a decision that is currently made with very little individual information.

The paper was published online on 21 July 2026. The release credits the Chungnam National University Evaluation Team, does not give the full author list, and states neither funding sources nor conflicts of interest.

---

Notes

1. “Longitudinal Modeling and Simulation of Growth Hormone Efficacy in Children With Idiopathic Short Stature.” Value in Health, 2026. DOI: 10.1016/j.jval.2026.05.006.
2. Chungnam National University. “Chungnam National University researchers develop model for predicting growth hormone therapy responses in children with short stature.” EurekAlert!, 5 October 2026. https://www.eurekalert.org/news-releases/1146307.

Photo by CDC on Pexels.com