Jung-woo Chae’s lab at Chungnam National University built its calculator from the charts of 91 Korean children. All of them were short and prepubertal, and all were treated with growth hormone at a single center between July 2020 and December 2023. The result is called GrowCast, and it works the way a mortgage calculator does. You enter a child’s age, sex, height and weight, the father’s height, one blood protein and the dose, and it draws the curve the child is predicted to follow. It is the newest entry in a long and oddly specific genre of medical research: the model that promises to tell a family whether the shots will pay off before they sign up for years of injections given six or seven times a week.
The paper went online on July 21 in Value in Health, which is neither a pediatrics journal nor an endocrinology journal. It belongs to health economics, the discipline whose job is deciding which treatments are worth paying for, and a tool that predicts which children will get the most out of an expensive drug sits comfortably in its pages.
The treatment results from Korea are modest. Over a mean of 619 days, roughly 20 months, the children’s mean height percentile rose from 1.26 at baseline to 9.16 at follow-up, which left them still near the bottom of the chart, just less so. The simulated two-year medians ran from the 4th percentile for the low-response profiles to the 31st for the high. The strongest predictor of a bigger response was body-mass index, with heavier children growing more, rather than any marker of hormone deficiency. Lower levels of the binding protein IGFBP-3 and a shorter father also tracked with more growth. The authors write that GrowCast is not yet a validated dosing or prescription tool and that larger prospective, multicenter studies have to come before routine clinical use, which is a fair caveat for something built retrospectively from one hospital’s records.
That caution is warranted, because the genre GrowCast joins has been running for more than twenty years and has a sponsor problem. The field is full of calculators, each refining the last: the Cologne prediction model, the KIGS model built on Pfizer’s international growth database, and the OPTIMA study, which tested whether an early prediction could set the dose. Some of the biggest of them draw on registries kept by the companies that sell the hormone.
The clearest recent example ran in the April issue of Hormone Research in Paediatrics. A team led by Andrew Dauber mined two patient registries for the clinical features that separate good responders from poor ones among children with idiopathic short stature (ISS). In 207 patients, younger children and those furthest below their target height responded best in the first year, and higher doses tracked with more growth.
The registries, NordiNet IOS and ANSWER, belong to Novo Nordisk, which sells the growth hormone Norditropin. The study was funded by Novo Nordisk, and three of its authors are Novo Nordisk employees who hold the company’s stock. The paper itself states that the funder “was involved in decisions relating to the study design, execution and analysis, and manuscript conception, planning, writing, and decision to publish.” Dauber discloses consultancy and grant funding from Novo Nordisk, Pfizer and BioMarin. None of that makes the regression wrong. It does tell you whose question was being answered.
Across these models that question is remarkably consistent. It is never whether to treat. It is how to sort: who will respond, and how to tune the dose so that the response varies less. A 2009 randomized trial in 153 children, most with growth hormone deficiency and 43 with ISS, found that dosing by predicted responsiveness cut the spread of outcomes around a height target by 32%. A later trial put the premise to a harder test in children with ISS alone. This open-label randomized study of Pfizer’s Genotropin was planned to run four years and enrolled 316 children, and it compared a formula-driven, target-based dose with plain weight-based dosing. The two-year analysis was where the formula was supposed to show it could reach the target with less variability. It did not.
The models spend much less effort on what the hormone does to the rest of a healthy child’s body. In a 2012 randomized study of short prepubertal children, left ventricular mass rose over two years of treatment in the ISS group as well as in children with true hormone deficiency. The change was already visible at three months and appeared regardless of dose, although the researchers found no impairment of heart function.
Set the dosing arithmetic aside and an older question comes back, the one Joyce Lee’s team at the University of Michigan asked in 2006. Writing in Archives of Pediatric and Adolescent Medicine, three years after the FDA approved growth hormone for idiopathic short stature in 2003, they ran the numbers. A child could cost as much as $100,000 to gain an average of two inches over five years of treatment, which came to more than $52,000 an inch. Roughly 400,000 American children aged 4 to 15 were eligible, and treating all of them would cost about $40 billion. The money bought height, and nobody had shown that it bought anything else. In the team’s words, “currently there is no research that demonstrates that it improves the quality of life for children with ISS.”
It helps to remember what the diagnosis means. Idiopathic short stature is the label given to a child who is short and in whom doctors, having looked, found nothing wrong. In other words, a healthy child sitting below a line on a chart. The prediction models don’t change that. They just get more precise about the centimeters.
GrowCast’s makers concede that their tool isn’t ready to set a dose, and they’re right to. What it can already do is rank short children by how much height a drug is likely to add, which is useful information for whoever pays for the drug. Joyce Lee’s question, whether any of it leaves the child better off, has now outlasted twenty years of increasingly precise calculators, and none of them was built to answer it.
Sources
- News-Medical: New model predicts growth hormone therapy response in children with short stature
- Dauber et al., Hormone Research in Paediatrics (NordiNet IOS and ANSWER predictors)
- Newswise: Cost for growth hormone therapy doesn’t measure up
- Counts et al., Hormone Research in Paediatrics, 2013 (Genotropin ISS trial)
- Kriström et al., JCEM, 2009
- Nygren et al., Clinical Endocrinology, 2012
- Jung et al., European Journal of Endocrinology, 2009 (OPTIMA)