A Growth Prediction Model from Mid-Puberty to Near Adult Height in Adolescents with Idiopathic Isolated Growth Hormone Deficiency Treated with Growth Hormone

Abstract

Introduction

Recombinant human growth hormone (rhGH) treatment of children with idiopathic isolated growth hormone deficiency (IIGHD) typically results in catch-up growth for several years followed by a period of normal growth. The effect of rhGH treatment on late pubertal height gain in adolescents with IIGHD has remained unclear. This study aimed to develop and validate a prediction model for height gain from mid-puberty to near adult height (NAH) in patients with IIGHD, treated with rhGH.

Methods

Data from the Dutch National Registry of Growth Hormone Treatment in Children were used, focusing on 151 patients who received rhGH treatment until NAH. Predictors included age, bone age, Tanner stage, and target height SDS minus height SDS at mid-puberty. Validation was performed in 33 males and 7 females who had a normal GH response in a GH stimulation test at mid-puberty and continued rhGH until NAH.

Results

The model explained 48% of the variance for males (residual SD 4.16 cm) and 18% for females (residual SD 3.64 cm). Validation showed a mean (SD) difference of 1.48 (2.36) cm for males and 3.57 (2.66) cm for females between predicted and attained NAH.

Conclusion

For females, explained variance was insufficient to reliably predict height gain. For GH sufficient males, the model can be used to assess efficacy of continuing or discontinuing rhGH treatment at mid-puberty in future studies.

Introduction

In the Netherlands, approximately 65 children diagnosed with idiopathic isolated growth hormone deficiency (IIGHD) start recombinant human growth hormone (rhGH) treatment annually after comprehensive clinical, laboratory, and radiological evaluation. For children with GHD, rhGH treatment typically results in accelerated growth during the initial years (the phase of catch-up growth), followed by a stable height standard deviation score (SDS) for age and a normal pubertal growth spurt. Most patients attain a normal near adult height (NAH; height velocity [HV] of <2 cm/year over at least 6 months), defined as a height SDS within ±2 SD for the population and within ±1.6 SD of the sex-corrected midparental height (target height [TH]) [1].

Treatment with rhGH typically continues until NAH is achieved, characterized by a HV of less than 2 cm per year. After NAH is reached, GH stimulation testing (GHST) is advised to assess whether the persistence and severity of GHD justifies continued rhGH treatment into adulthood. Studies indicate that 70–80% of patients with IIGHD exhibit a normal GH peak upon retesting, in contrast to those with multiple pituitary deficiencies or organic or genetic causes of isolated GHD [2–5]. This suggests that initial test results may either have been false positives or that GHD may have been transient, with normalization of GH secretion during puberty due to exposure to endogenous sex steroids [6]. The efficacy of rhGH treatment during different growth phases, particularly its contribution to late pubertal height gain in adolescents with IIGHD, has remained unclear [7].

Whereas the majority of existing literature has primarily focused on growth in the first year following start of GH treatment [8–11], our research uniquely emphasizes growth from mid-puberty until NAH. A handful of authors have investigated total pubertal growth (TPG; centimeters height gain from Tanner G2 or B2 until NAH) [12–15]. In a study in adolescents with various conditions, TPG depended primarily on prepubertal factors, such as age and bone age (BA) delay at puberty onset, with only a minor role for GH dose [12].

The aim of this study was to develop and validate a statistical model that predicts the expected height gain from mid-puberty to NAH in patients with IIGHD, undergoing rhGH treatment. This model could serve as a tool to explore whether discontinuing rhGH treatment at mid-puberty, given a normal GH peak during the GHST at mid-puberty, would lead to a lower NAH compared to continuing rhGH treatment until NAH.

Methods

Subjects/Patients

Baseline and follow-up data on growth, BA and laboratory markers during treatment are recorded in the Dutch National Registry of Growth Hormone Treatment in Children. This database, established in 1997, currently contains data from approximately 8,150 children treated with rhGH. Data extracted from the database for this study covered the period from 1998 to 2018. The data are provided by Dutch pediatric endocrinologists and nationwide coverage is 100%. Patients starting rhGH treatment are informed in writing that their pseudonymized data will be entered into the database, unless they choose to opt-out. These data are accessible for scientific studies if permission is granted by the Dutch Growth Hormone Advisory Group. For data collection, informed consent and ethical approval were not required according to Dutch law [16, 17]. This study focused on subjects diagnosed with IIGHD who received rhGH treatment until NAH and were subsequently regarded as GH sufficient after reevaluation at (N)AH.

Inclusion and Exclusion Criteria

Inclusion Criteria

We included participants diagnosed with IIGHD, with a GH peak at diagnosis between 5 and 30 mU/L (1.7–10 µg/L). RhGH treatment was administered for a minimum of 2.5 years and was discontinued at least 1 month prior to reevaluation. Upon retesting at NAH, GH peak was >30 mU/L, or the GH peak was 20–30 mU/L in combination with an IGF-1 SDS >−2. Mid-puberty was defined for males as Tanner stage G3–G4, testicular volume >12 mL, and BA 13–16 years and for females as Tanner stage B3–B4 and BA 11–14 years.

If no retest had been conducted, IGF-1 SDS combined with the clinician’s judgment was used to determine GH sufficiency. If there was no clinician’s judgment but rhGH treatment had been discontinued, patients were included if IGF-1 SDS was above 0 [18]. If no individual SDS was reported from the respective laboratory, IGF-1 SDS was calculated according to a national reference [19].

Exclusion Criteria

Exclusion criteria were the presence of medical or psychological disorders that could influence growth, the use of medication other than rhGH that could influence growth, and dysfunction of other endocrine axes. Additionally, dropouts during treatment due to noncompliance, patient choice, or emigration were excluded.

Measurements

In rhGH treated children, data on growth and puberty are routinely collected at a frequency of 3–4 times per year. For this study, we used measurements at the start of rhGH treatment, 1 year after start, at mid-puberty, and at NAH. Demographic data included sex and parental height. Clinical data included age, height, weight, pubertal stage (including testicular volume), BA, IGF-1 and IGFBP-3 levels, GH peak at diagnosis, GH peak at retest, and rhGH dose.

Testicular volume SDS was calculated using Dutch reference values [20]. For BA assessment, the Greulich and Pyle method was used. A subset of BAs was assessed using BoneXpert^® software [21–24]. The accepted margin for the timing of measurements (laboratory, X-ray, physical examination) relative to height measurement was 6 months.

In this study, we used the conditional TH equation that accounts for assortative mating and parent-offspring correlations, first described by Hermanussen and Cole [25], and adapted to the data of the 5th Dutch Nationwide Growth Study [26]. SD scores (SDS) for height, weight, TH, and BMI were calculated using Growth Analyzer RCT^® based on data from the 5th Dutch Nationwide Growth Study [26] for height and TH and based on the 3rd Nationwide Growth Study 1980 for normative data for weight and BMI [27, 28].

In the Netherlands, GH levels are reported as mU/L or µg/L. In order to use a single unit for the analyses in this study, all values were converted to mU/L. Over the years, different conversion factors have been used by various laboratories. In collaboration with the laboratories and the Dutch Growth Research Foundation, these conversion factors have been mapped and utilized.

Statistical Analysis

Data were analyzed using IBM SPSS Statistics (Version 29.0.1.0) for extraction and collection and R Statistical Software (v4.0.3; R Core Team 2020) for statistical analysis. Separate prediction models were constructed for males and females. Candidate predictors of growth between mid-puberty and NAH were GH peak at start of rhGH treatment, and the following characteristics measured at mid-puberty: age, BA, Tanner stage, rhGH dosage, TH-SDS minus height SDS, BMI SDS and, for males, testicular volume SDS.

Descriptive statistics of continuous variables are presented as mean ± SD unless otherwise specified. A p value <0.05 was considered statistically significant. To avoid a data-driven selection of predictors in the models, only one selection step was applied, removing all variables with p > 0.1.

Bootstrapping was used to correct for overoptimism. In 1,000 bootstrap samples, prediction models were constructed using the method described above. These models were used to shrink the coefficients in the prediction model and to adjust R² and the prediction error.

Missing Values

For patients without a BA measurement at mid-puberty, the BA was estimated based on the last known BA, as long as it fell within a 2.5-year range. The following calculation was used: (last known BA/corresponding chronological age) × age at anthropometry at mid-puberty.

Missing values in candidate predictors for the prediction model were imputed using fully conditional specification with Bayesian linear regression for variables with a normal distribution and predictive mean matching for other variables. Selection of variables used for the imputation model was done using a built-in function. This function evaluates for each variable to be imputed the correlations with other variables as well as the correlations between the response indicator of the variable and other variables. The number of imputed sets was determined by the proportion of incomplete cases times 100 [29].

Validation of the Equations

To assess their accuracy, the adjusted equations were validated using a prospectively followed cohort that continued rhGH treatment until NAH. Adolescents diagnosed with IIGHD in childhood (GH peak at diagnosis 5–30 mU/L) who started rhGH treatment between 2005 and 2018 and tested GH sufficient (GH peak >20 mU/L) at mid-puberty were included in this prospective multicenter SEENEZ trial. The expected height gain was calculated at mid-puberty and compared with the attained height gain at NAH.

Results

A total of 436 patients were identified. Of these, 285 patients were excluded from the study based on the criteria detailed in Figure 1. This resulted in a final cohort of 151 patients for analysis, of which 87 patients were retested with GH stimulation tests and 64 patients were considered GH sufficient based on an IGF-1 SDS level >0 at NAH.

Fig. 1.

Flowchart illustrating the patient enrollment process.

Baseline Characteristics

Table 1 presents descriptive statistics of the observed data. Data for all candidate predictors were complete in 122 patients (81 males, 41 females). At mid-puberty, BA was missing in 18 patients, GH peak in 11, TH-SDS in 2, and Tanner stage in 1 patient. Compared to the complete cases, the 29 cases with incomplete data grew faster during the first year of GH treatment (delta SDS 0.97 vs. 0.77, p = 0.03) and during the complete treatment period (delta SDS 2.32 vs. 1.91, p = 0.02), had a higher height SDS at mid-puberty (−0.71 vs. −1.12, p = 0.03), and had a higher TH-SDS (−0.45 vs. −0.72, p = 0.02). Based on the 19.2% of incomplete cases, 20 imputed datasets were derived. The GH peak at NAH was not normally distributed and is therefore reported as median (IQR). The mean growth (the outcome to be predicted) in the cohort was 12.0 cm for males (SD 5.6) and 8.6 cm for females (SD 3.9).

Table 1.

Patient characteristics

Sex	n (%)	M	98 (65%)
		F	53 (35%)
TH-SDS	Mean±SD	−0.67±0.64		N = 149
​	​	Start treatment			Mid-puberty			Near adult height
Age, years	Mean±SD	M	7.61±3.32	N = 151	M	14.46±1.19	N = 151	M	16.93±1.71	N = 151
		F	8.78±3.21		F	13.28±1.31		F	15.93±2.15
Height SDS	Mean±SD	−2.80±0.58		N = 151	−1.00±0.95		N = 151	−0.82±0.94		N = 151
BMI SDS	Mean±SD	−0.11±1.20		N = 150	0.03±1.20		N = 151	0.09±1.20		N = 148
GH peak, mU/L	Median (IQR)	15.79 (6.38–21.99)		N = 140	​		​	74.10 (33.00–127.00)		N = 83
BA	Mean±SD	M	5.59±3.23	N = 140	M	13.69±0.83	N = 133	M	16.35±1.07	N = 59
		F	6.72±3.10		F	12.24±0.89		F	14.66±1.13
Tanner stage	n (%)	1	119 (86%)	N = 138	1	–	N = 150	1	–	N = 147
		2	8 (6%)		2	–		2	–
		3	11 (8%)		3	60 (40%)		3	–
		4	–		4	85 (56%)		4	48 (33%)
		5	–		5	6 (4%)		5	99 (67%)
Testicular volume SDS	Mean±SD	−0.3±1.10		N = 87	−0.25±1.10		N = 98	−0.85±0.90		N = 96
GH dose, mg/m2/day	Mean±SD	0.94±0.43		N = 116	0.88±0.18		N = 151	0.0		​

Equations

For males, final predictors of growth between mid-puberty and NAH were TH-SDS minus height SDS at mid-puberty, age, BA, and Tanner stage at mid-puberty. The equation was as follows:

$$\text{Predicted}\text{growth}\text{between}\text{mid}‐\text{puberty}\text{and}\text{NAH}\left(\text{cm}\right)=85.15–1.48\times \text{Tanner}\text{stage}4\text{or}5–2.66\times \text{age}–2.47\times \text{bone}\text{age}+2.43\times \left(\text{TH}\text{SDS}-\text{height}\text{SDS}\right)$$

Percentage explained variance (R²) was 54%. Residual SD was 3.90 cm.

Adjusted for overoptimism, the equation was as follows:

$$\text{Predicted}\text{growth}\text{between}\text{mid}‐\text{puberty}\text{and}\text{NAH}\left(\text{cm}\right)=82.07–1.41\times \text{Tanner}\text{stage}4\text{or}5–2.55\times \text{age}–2.36\times \text{bone}\text{age}+2.33\times \left(\text{TH}\text{SDS}–\text{height}\text{SDS}\right)$$

R ² was 48%, and residual SD was 4.16 cm. In the equation, a value of 1 was assigned for Tanner stage 4 or 5, and a value of 0 was assigned for Tanner stage 3.

For females, the final predictors of growth between mid-puberty and NAH were age and BA at mid-puberty. The equation was as follows:

$$\text{Predicted}\text{growth}\text{between}\text{mid}‐\text{puberty}\text{and}\text{NAH}\left(\text{cm}\right)=42.94–0.68\times \text{age}–2.04\times \text{bone}\text{age}$$

R ² was 34%, with a residual SD of 3.2 cm.

Adjusted for overoptimism, these results were as follows:

$$\text{Predicted}\text{growth}\text{between}\text{mid}‐\text{puberty}\text{and}\text{NAH}\left(\text{cm}\right)=39.85–0.57\times \text{age}–1.72\times \text{bone}\text{age}$$

R ² was 18%, and residual SD was 3.64 cm.

Since the formula in both males and females can predict values below zero in several cases, a minimum threshold of 0 cm was maintained.

Validation of the Models

The validation cohort consisted of 35 males and 8 males from the ongoing prospective SEENEZ trial. As shown in Table 2, the predicted growth from mid-puberty until NAH was 12.73 ± 4.28 cm for males and 11.15 ± 2.31 cm for females. This results in an average predicted NAH of 176.14 ± 4.49 cm for males and 165.53 ± 4.80 cm for females. The difference between the predicted NAH and the attained NAH is 1.48 ± 2.36 cm for males and 3.57 ± 2.66 cm for females.

Table 2.

Validation cohort

​	Males (N = 35)	Range	Females (N = 8)	Range
Predicted growth, cm	12.36±4.22	0–21.84	11.05±2.47	6.39–13.86
Predicted NAH, cm	176.44±4.34	167.47–186.66	165.57±5.19	158.15–173.0
Observed growth, cm	10.92±4.46	2.40–21.10	7.49±3.59	2.70–12.50
Observed NAH, cm	175.00±5.20	164.0–187.7	162.0±5.7	154.5–171.0
Predicted minus observed NAH, cm	1.48±2.36	−3.52; 6.01	3.57±2.66	−0.80; 7.49

Figure 2a displays a scatterplot with predicted growth on the X-axis and observed growth on the Y-axis for males in the validation cohort and Figure 2b shows the same for females in the validation cohort. Figure 2c displays the residuals (predicted minus observed NAH) on the Y-axis against the predicted NAH on the X-axis for males. The residuals remained within an acceptable range and did not display any significant trend in relation to the predicted NAH. In Figure 2c and d, the prediction intervals are shown as horizontal lines at ±1.96 times the adjusted residual SD. For males, these lines are at ±8.15 cm and for females at ±7.15 cm. It is expected that in future observations, 95% of the actual observations lies between these lines. Within the validation cohort of males, all observed values fall within these prediction intervals. For the females, 1 out of 8 exceeds the upper limit by 0.34 cm.

Fig. 2.

Scatterplots for males (a, c) and females (b, d), illustrating observed versus predicted growth and residuals versus predicted AH. In figures a and b, the dotted line represents the null line, and the solid line represents the fit line. In figures c and d, the dotted line shows the prediction intervals (±1.96 times the adjusted residual SD). a Males – observed vs. predicted growth from mid-puberty until NAH (cm) in 35 males (r = 0.86, p= <0.001). b Females – observed vs. predicted growth from mid-puberty until NAH (cm) in 8 females (r = 0.65, p = 0.08). c Males – residuals (predicted minus observed NAH) vs. predicted NAH (cm) in the male validation cohort (r = −0.12, p = 0.51). d Females – residuals (predicted minus observed NAH) vs. predicted NAH (cm) in the female validation cohort (r = 0.04, p = 0.93).

To further investigate potential significant trends, we examined the relationship between the residuals and the final predictors included in the equation, using the same method as Blum et al. [30]. Table 3 summarizes these relationships, highlighting that all absolute correlations were below 0.23.

Table 3.

Correlations between residuals and included predictors in the validation cohort

​	Males		Females
	Pearson correlation	significance (2-tailed)	Pearson correlation	significance (2-tailed)
Tanner 4 or 5	0.000	0.998	​	​
Age	−0.118	0.513	0.131	0.780
BA	−0.230	0.197	−0.057	0.903
TH-SDS – height SDS	−0.035	0.847	​	​

Discussion

We developed a prediction model to estimate growth from mid-puberty until NAH, using data of 151 patients with IIGHD who demonstrated sufficient GH secretion after rhGH treatment was discontinued. The equation shows that the model explains 48% of the variance for males (residual SD 4.16 cm) and 18% for females (residual SD 3.64 cm). The model is intended to serve as a tool to predict growth while continuing rhGH treatment in midpubertal patients who are no longer fulfilling a diagnosis of GH deficiency. This is the first prediction model that focuses on the final stages of puberty, when only minimal additional height gain can be expected from rhGH treatment in GH sufficient adolescents. A link to use the prediction model online can be found in the online supplementary material (see https://doi.org/10.1159/000547488).

Most previous studies have concentrated on prediction models of growth within the first year after the start of rhGH treatment [8–11, 31], yet some have looked at TPG [12–15]. Lonero et al. [32] developed a prediction model for NAH in children with GHD, emphasizing the importance of early diagnosis and the role of TH. The model suggests that starting rhGH treatment before puberty yields the most growth benefits, with less added values during the final stages of puberty. Ranke et al. [12, 13] investigated factors influencing TPG in adolescents with GHD using data from the KIGS database. They found that TPG in IIGHD could be explained by age, age minus BA and height minus TH at puberty onset, and mean rhGH dose during puberty, with algorithms explaining 66% of variance. The authors’ analysis indicated only a moderate dependence of TPG in IIGHD on rhGH dose, suggesting that TPG primarily depends on prepubertal growth outcomes rather than rhGH dose [13]. Ranke et al. focused on puberty onset, while our model focused on mid-puberty. Although we used a similar predictive modeling approach, the lack of data on age at puberty onset in our cohort precludes our ability to make a direct comparison. In our study, rhGH dosing was also not a significant predictor, which aligns with the findings by Ranke et al. [12, 13].

To evaluate the model’s predictive accuracy, we looked at the proportion of variance explained by the model, denoted as R². Higher R² values indicate better model performance, with 100% being the best possible score, scaled from 0 to 100%. Another indicator is the SD of the residuals (SDres), which represent the differences between observed and predicted heights. A lower SDres indicates more accurate predictions, with 0 being the ideal value, measured in centimeters or SDS [9, 33, 34]. The reliability and utility of our prediction model, with an R² of 48% for boys and 18% for girls, is lower for girls compared to other predictive models. This may be due to the insufficient amount of data available for girls, making it challenging to develop a robust model for the female group.

While this indicates potential for improvement, our observations still provide valuable insights, particularly for boys. In our study, the SDres was 4.16 cm for boys and 3.2 cm for girls. Our model offers similar precision for individual estimates compared with previous reports. For example, the KIGS model shows an R² of 45% in IIGHD patients and an SDres of 1.72 cm in the first year of GH treatment, 30–40% for the second to fourth year response with an SDres of 0.95–1.19 cm, and 65–68% in TPG with an SDres of 2.9–4.5 cm. The Gothenburg model only reported an SDres of 0.19 SDS for first-year growth. The Cologne model shows an R² of 89% and an SDres of 0.93 cm/year for first-year growth and 53–72% upon validation, but in this model, the early growth response is used as one of the predictors (no SDres reported) [9, 10, 14, 30, 32, 35].

Various studies have shown a higher incidence of IIGHD in boys than girls, with a consistent male predominance, which may be due to social or cultural factors [36–40]. One limitation of our study is the small number of girls, which led to a lower explained variance and greater deviation in the validation cohort. We considered using height SDS values instead of centimeters, so that a combined model could be made for both sexes. We decided not to do this not only because of practical usability but also because of lack of ethnic-specific reference ranges and potential errors from speculative extrapolation of the distribution of the normal range [30]. Additionally, missing data in our study are due to the dependence of contributions from many different pediatric endocrinologists and centers. We tried to impute the missing values as accurately as possible, using state-of-the-art methods [41]. For example, we set clear boundaries for imputation, such as using a narrow range for the missing BAs. Furthermore, the formula in its current form can also predict growth below 0 cm. We intentionally chose not to use a log-transformed formula due to its reduced applicability in clinical practice. It is essential for the equation to be practical for everyday use in clinical settings.

Challenges in developing a prediction model for growth in the latter part of puberty include the need for extensive patient data (e.g., age at puberty onset, birth weight, and previous year’s HV) and the narrow timing of prediction models, which may not align with the real-world irregularity of patient visits. One of the strengths of our study is that mid-puberty has a broad range of age and Tanner stages and can typically be captured during regular check-up. Additionally, the necessary data for our prediction model are easily accessible in a standard electronic patient record, ensuring that the model can be easily implemented in clinical practice without requiring extensive additional data collection. The model’s validation in an independent cohort enhances its strength and applicability. The decision to avoid a data-driven selection of predictors, applying only one selection step, further strengthens the model’s methodological validity. This approach minimizes the risk of overfitting and ensures that the model remains generalizable to different patient populations across countries.

Wit et al. reviewed prediction models for rhGH treatment in short children, including patients with GHD, and described three key models for first-year HV, creating an adapted table outlining the goals and requirements for clinically relevant prediction models [9]. The prediction model we developed largely meets the criteria for an ideal prediction model, as it has been validated with an independent cohort of the same patient group, explains as much as possible the variability in treatment response, is based on readily available variables, and is easy to use in clinical practice [9, 35, 42]. Ranke et al. [43] highlighted the necessity of balancing growth outcomes with costs and long-term risks, advocating for optimized, individualized treatment plans to improve patient outcomes and reduce unnecessary long-term treatments. Our prediction model aligns well with this objective, serving as a desired solution from a clinical perspective.

In a subset (64/151) of our cohort, IGF-1 SDS was used to determine GH sufficiency based on the pediatric endocrinologists’ judgment. If there was no clinician’s judgment but rhGH treatment had been discontinued, the patient was included if IGF-1 SDS was above 0. If the individual SDS was not known, it was calculated using Dutch reference ranges [19]. Clinically, an IGF-1 SDS > −1 is generally considered sufficient. Postma et al. [44] found limited variation in IGF-1 concentrations across laboratories but increased variability when converted to SDS, highlighting the need for standardization in IGF-1 assessment. Fava et al. identified an optimal IGF-1 cutoff value of −1.4 SDS (95% CI [−1.94; 0.77]) correctly classifying 85.1% of their study population of 97 subjects with childhood-onset GHD at a median age of 17.39 years [18]. This indicates that we are operating within a safe zone and can be reasonably confident that our patients were GH sufficient.

An unresolved question is the efficacy and cost-efficiency of rhGH treatment during the latter half of puberty in patients diagnosed with IIGHD in childhood. Given that most adolescents show a normal GH peak upon retesting, discontinuing rhGH after mid-puberty may have little to no negative effect on attained AH. Only a few retrospective and prospective studies reported that early GH retesting can identify those who no longer need treatment, reducing side effects and costs. However, the lack of solid data on the potential negative effect on attained AH if treatment is discontinued at mid-puberty means that physicians worldwide continue rhGH treatment until NAH is achieved. Addressing this gap in knowledge is crucial for optimizing treatment protocols and healthcare resource allocation. Future research should focus on the efficacy and cost-efficiency of rhGH treatment during the latter half of puberty in GH sufficient patients. Therefore, implementing our model in future studies and clinical decision support systems, and familiarizing healthcare providers with its use, will enhance its practical application, ultimately improving the management of rhGH treatment in patients with IIGHD.

Conclusions

This study developed a prediction model for height gain in adolescents with IIGHD during rhGH treatment in the final stages of puberty. For females, explained variance was insufficient to reliably predict height gain. For GH sufficient males, the model can be used to assess efficacy of continuing or discontinuing rhGH treatment at mid-puberty in prospective studies and to facilitate shared decision-making regarding treatment continuation at mid-puberty.

Acknowledgments

The authors wish to thank the patients and their families for their invaluable contributions. We extend our gratitude to the healthcare professionals involved, including the treating physicians; Debbie van Vliet-de Groot, research nurse; and Gladys Zandwijken, former data manager of the Dutch Growth Research Foundation. We specially express our gratitude to the late Marie-José Walenkamp, pediatric endocrinologist, who wrote the study protocol for the prospective and retrospective studies and initiated and coordinated this project until her untimely passing. We are grateful for her great diligence in managing this research project for many years. We are deeply grateful to our coauthor Ardine Reedijk, whose dedication and contributions to this manuscript were invaluable, and who sadly passed away during the submission process.

Statement of Ethics

The study was conducted in accordance with the Declaration of Helsinki, and the Medical Ethics Review Committee of Erasmus Medical Centre (MEC-2021-0671, data of approval 18-10-2021) concluded that the study falls outside the scope of the Medical Research Involving Human Subjects Act (WMO). Patients starting rhGH treatment are informed in writing that their pseudonymized data will be entered into the database, unless they choose to opt-out. These data are accessible for scientific studies if permission is granted by the Dutch Growth Hormone Advisory Group. For data collection, informed consent and ethical approval were not required according to Dutch law.

Conflict of Interest Statement

J.M.W. was a member of the journal’s Editorial Board at the time of submission. The other authors declare no conflicts of interest.

Funding Sources

This study was funded by ZonMw (project number 837004021). The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

Author Contributions

Conceptualization: J.V., M.A.J.R., J.M.W., A.M.J.R., R.D.O.E., and D.C.M.K. Methodology: M.A.J.R., J.V., J.M.W., A.M.J.R., R.D.O.E., and D.C.M.K. Validation and formal analysis: M.A.J.R. and J.V. Resources: A.M.J.R. Data curation: J.V., A.M.J.R., and M.A.J.R. Writing – original draft preparation: J.V. Writing – review and editing: J.V., M.A.J.R., J.M.W., D.C.M.K., and E.L.T.A. Visualization: J.V. Supervision: D.C.M.K. All authors have read and agreed to the published version of the manuscript.

Funding Statement

This study was funded by ZonMw (project number 837004021). The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

Data Availability Statement

The datasets presented in this article are not readily available because data were obtained from the Dutch National Registry of Growth Hormone Treatment in Children. Further inquiries can be directed to the corresponding author.