The Use of IGF-I to Monitor Long-Acting Growth Hormone Therapy-Timing is an Art….

Recombinant growth hormone (GH) is used to treat children and adults with growth hormone deficiency (GHD). Numerous studies confirmed efficacy and safety within approved indications, but it remains an invasive and expensive treatment. Adherence to treatment is important, and might be compromised by the need for daily injections. Long-acting growth-hormone (LAGH) preparations, allowing for less-frequent injections, hold the promise to improve adherence (1). It might be premature to expect that LAGH automatically will result in better adherence to, and, in consequence, better efficacy of GH therapy (2). Some recent studies in adults do not suggest daily injections are the key problem for many patients (3, 4). However, LAGH certainly is a preferred option for individual patients, and might be particularly important in children to improve adherence (3).

Concerns were raised that the GH profile following LAGH injection does not mimic the physiologic secretion profile. While the observation certainly is correct, it has been pointed out that the pharmacokinetics following daily GH injection have no similarity to endogenous secretion profiles either (5). Concerns were also raised regarding differences in pharmacodynamic (PD) profiles with daily GH vs LAGH. Indeed, the clinician must be aware of a fundamental difference in this respect: Concentrations of insulin-like growth-factor 1 (IGF-I), the most frequently used PD marker of GH, increase in the beginning of daily GH treatment, but after a steady state on stable dose is reached, fluctuations during the 24 hours following the GH injections are negligible. In contrast, on injection of LAGH, IGF-I concentrations increase over a few days, reach a peak (which can exceed the reference interval), and finally decline to trough concentrations before the next injection. While PD profiles over the injection interval become similar after some LAGH injections, they continue to exhibit much greater excursions compared to daily GH. Therefore, even if the same total GH dose is administered, and assuming the same “total IGF-I exposure” is reached, the temporal pattern of circulating IGF-I concentrations remains hugely different between daily and long-acting GH. Based on available data from clinical studies, this difference seems not to affect efficacy, and no new side effects or risks have been identified for LAGH so far (5). Whether the greater fluctuation of IGF-I, that is, higher peak and lower trough concentrations, might be associated with long-term risks is speculation. Observation periods are still short, and several aspects deserve attention in future studies (1).

However, the striking difference in the PD profile causes a very practical problem for physicians used to interpreting IGF-I concentrations during treatment with daily GH: When after LAGH injection should blood samples be drawn? Which time point is best to represent peak or average IGF-I exposure, or to adjust the dose? What to do if a patient’s blood sample cannot be drawn at the ideal time point? Difficult questions—and answers will differ between LAGH products (1). A recent position statement from the Growth Hormone Research Society therefore called for studies that “take into account the pharmacokinetics and pharmacodynamics of each product in order to gauge the optimal timing of IGF-I measurement for each individual LAGH product” (5).

In this respect, the study by Kildemoes and colleagues (6) in this issue of the Journal of Clinical Endocrinology and Metabolism provides urgently needed information for one of the LAGH preparations. Somapacitan, a noncovalent albumin-binding GH compound, is at an advanced stage of development (1). Data on efficacy and safety are available for children and adults (7, 8). In earlier studies with high sampling frequency, nonlinear PD models were developed specific for somapacitan. Using this information, Kildemoes and colleagues now present data on the development and assessment of linear models for predicting mean and peak IGF-I concentrations from samples taken at any time point during the dosing interval. This is important to allow physician and patients flexibility with sampling for IGF-I measurements. While retaining the demographic characteristics from the original studies, they performed a full parametric simulation for all participants at all doses, ending up with 39 200 IGF-I profiles. Notably, only simulated profiles with an average IGF-I SD score (SDS) greater than 4 over the whole dosing interval were excluded, but peak IGF-I greater than 4 SDS—as occasionally seen in LAGH clinical trials—were included. IGF-I profiles were found to depend on dose, with higher doses associated with higher fluctuations. In adults, low body weight was also associated with greater fluctuations, but this was not seen in models for children.

The authors state that their linearized models provide “a reliable prediction of the weekly mean and/or peak IGF-I SDS” but make clear this applies only if samples are taken at steady state. Early in treatment it is less reliable, but also after dose adjustments, reliable predictions can be made only after at least one dosing interval at the new dose has passed. To put “reliability” of predictions in a clinical context, it is helpful that the authors provide 90% prediction intervals (PIs) for samples taken on different days in Table 1. For example, while the 90% PI for mean IGF-I in adults is as high as ± 0.6 SDS with sampling on days 0 or 7, it is only ± 0.2 SDS on days 4 or 5 after. Prediction of peak IGF-I SDS from a single blood sample is best on day 2 (± 0.1 SDS), but poor on day 0 or any day after day 5 (> ± 0.7 SDS). The acceptable degree of uncertainty obviously needs to be decided by clinicians, but the information provided by Kildemoes and colleagues is key to include IGF-I in an educated decision.

The authors also discuss limitations of their model. First, it was developed based on retrospective data. While it looks appropriate from a model-building point of view, it would be reassuring to see data confirming its reliability from upcoming, prospective observational studies. This would also allow us to establish how reliability might be affected by differences between IGF-I assays and reference intervals, and by dose-to-dose variations in bioavailability in a real-world scenario.

The most important aspect for successful clinical implementation of optimal monitoring of LAGH treatment in our view relates to practicality: Few physicians will appreciate the need to use sophisticated formulas, and to look for “slope of variables” presented in this article. The data presented suggest that—if blood sampling at defined time points is feasible (for the patient and the physician)—the use of samples taken on day 4 or 5 after administration is appropriate to get an estimate of mean IGF-I. In the clinical context, the 90% PI of ± 0.2 SDS seems acceptable for most situations. However, to give more flexibility regarding sampling time points, and to make dose adjustments in LAGH treatment as convenient as we know it from daily GH, the algorithms should be transferred to easy-to-use, electronic tools. Ideally, physicians should need to enter only the dose, weight, time point of sampling, and IGF-I to obtain the appropriate interpretation. We acknowledge this is not any easy task. Differences in IGF-I assays have to be taken into account, the device needs formal approval, and it needs to be adjusted for each individual LAGH product. However, with LAGH, timing is an art—and manufacturers need to help the artist!