The diagnosis of growth hormone deficiency (GHD) relies on a combination of clinical, auxological, and laboratory criteria. Among these, an insufficient response to stimulation tests for endogenous growth hormone (GH) secretion represents the final step in confirming the diagnosis (1). However, it has long been recognized that stimulation tests face challenges in reproducibility and reliability, including false negatives even in children with genetically confirmed GHD (2, 3). Furthermore, there is considerable inconsistency in threshold values for diagnosing GHD across different countries, ranging from 3.3 ng/mL in Australia/New Zealand (4) to 10 ng/mL in Poland and the USA (5, 6). While in some countries (e.g., USA) the use of recombinant human growth hormone (rhGH) is permitted for idiopathic short stature (ISS), in most countries a diagnosis of GHD is required to access treatment. Given the clinical, economic, and ethical implications of GH therapy (7, 8), refining our understanding of the population currently classified as GHD is essential.

Based on these observations, it has been proposed that among children with unsatisfactory responses to GH stimulation tests, only those with an identifiable anatomical, functional, or genetic cause should be classified as definite GHD (dGHD). In contrast, others should be categorized as “short stature unresponsive to stimulation tests” (SUS) (9) rather than idiopathic GHD. One of the most important aspects of this classification is its role in guiding the future follow-up of these children. For instance, a child diagnosed with GHD has a greater than 5% risk of progressing to combined pituitary hormone deficiency (5) and a significantly increased likelihood of persistent GHD in adulthood (10).

In a previous study, we reported that dGHD and SUS patients exhibit distinct characteristics, with only dGHD patients showing features consistent with the etiopathogenesis of GHD, such as lower pre-treatment IGF-1 levels followed by a greater increase during therapy, higher BMI, and a higher prevalence of positive retesting at the end of treatment (11). As a matter of fact, lower baseline IGF-1 levels, followed by normalization after treatment, are indicative of GHD, reflecting the liver’s GH-stimulated secretion of IGF-1 (12). Moreover, GHD in children is often associated with mild to moderate overweight, which is considered a hallmark feature of the condition (13). Furthermore, organic GHD carries up to a 100% risk of persisting at retesting, compared to approximately one-third in cases without an organic cause (10).

Nevertheless, we found that rhGH supplementation was effective both in dGHD and in SUS in terms of improving near adult height (NAH) (11). Therefore, the SUS category should be distinguished from GHD to avoid labelling children with a condition that may not be definitively confirmed. Based on its poor response to stimulation tests and its positive response to GH treatment, SUS should also be distinguished from ISS and constitutional delay of growth and puberty (CDGP).

Building upon these findings, the present study aimed to further define the clinical and laboratory characteristics of SUS and dGHD patients using a large, multicenter international database of patients treated with rhGH. Specifically, the objective was to determine whether classification into dGHD, SUS, or ISS can predict the NAH of prepubertal children treated with rhGH.

This analysis was performed using the KIGS (Pfizer International Growth Database) observational dataset, which spans from its inception in 1987 until 2012. At the time of analysis, the KIGS dataset contained data on 83,803 children treated with rhGH.

In this study, we included all patients who had undergone a GH stimulation test, initiated rhGH treatment before puberty (Tanner stage B or G <2), and had reached near adult height (NAH). NAH was defined by a height velocity of less than 2 cm/year, an individual growth curve showing asymptotic growth toward adult height, and a bone age of at least 15 years, and treatment with GH for at least 5 years. Children with congenital GHD, neurosecretory dysfunction, Turner syndrome, Prader-Willi syndrome, other syndromes, and chronic renal insufficiency were excluded from the analysis. When neonatal data were available, patients born small for gestational age were also excluded.

At baseline, data on maximum GH peak during the stimulation test, sex, gestational age, weight and length at birth, and midparental height (MPH) were collected. Given that KIGS was an international registry, it seemed unlikely that homogeneous criteria for defining GHD were applied. Therefore, we established the Italian cut-off specifically for this study to ensure consistency and facilitate comparison with the previous report (11). Definite growth hormone deficiency (dGHD) was defined as a peak GH level of <8 ng/dL (14) and the presence of an identifiable genetic, functional, or anatomical cause. This included cases with a genetic diagnosis of isolated GHD (e.g., GH1, GHRHR, RNPC3), multiple pituitary hormone deficiency, or acquired damage (e.g., brain trauma, central nervous system infections, tumors of the hypothalamus or pituitary, cranial or total body irradiation, infiltrative diseases), or the presence of hypothalamic or pituitary abnormalities on MRI. Short stature unresponsive to stimulation tests (SUS) was defined as a peak GH level of <8 ng/dL without any identifiable genetic, functional, or anatomical cause (9). Idiopathic short stature (ISS) was defined as a peak GH level of ≥8 ng/dL. MPH SDS was calculated as follows: (father’s height SDS + mother’s height SDS) ÷ 1.61. This formula corrects for the correlation between parental heights due to assortative mating (15).

At the start of rhGH treatment, we collected data on chronological age, bone age, height SDS, IGF-1 SDS, height velocity (HV) in SDS, weight SDS, BMI SDS, and rhGH dose in mcg/kg/day. We also calculated bone age delay (chronological age − bone age), the prevalence of short stature (height <-2 SDS), and the difference between height and MPH in SDS. Bone age was determined by the treating clinician and/or radiologist through the Greulich-and-Pyle method (16).

In the KIGS registry, height and HV were converted to standard deviation scores (SDS) using height references for healthy children from Prader (17). Weight SDS was calculated using the normal population reference from Freeman (18), and BMI SDS was calculated using the normal population reference from Cole (19).

After 1 year of treatment, at the start of puberty, and at the last visit, we also calculated the changes in height and IGF-1 compared to the start of treatment (delta height and delta IGF-1, both in SDS), as well as the years of treatment duration.

Ethical Committee approval was not required for this study. The General Authorization to Process Personal Data for Scientific Research Purposes (Authorization no. 9/2014) by The Italian Data Protection Authority declared that retrospective archive studies using ID codes, which prevent data from being traced directly back to the data subject, do not need ethics approval (20).

Descriptive statistics were used to describe the data. Continuous variables were presented as medians (10th-90th percentiles) unless noted. Categorical data were expressed as percentages (%). For the pairwise comparisons of continuous variables between groups, the Mann-Whitney U test was used. The chi-square test was used to compare categorical variables between the groups.

Statistical significance was considered for p-values <0.05. To maintain an overall significance level of 0.05, we used the Bonferroni correction method in Tables 1 – 3 , and the bivariate correlation in Table 4 was calculated by Spearman correlation.

For the purposes of interpreting the magnitude of a correlation, r=0.10, r=0.30, and r=0.50 were considered small, medium, and large in magnitude, respectively (21). Furthermore, multivariable linear regression models for predicting height SDS at NAH in the SUS, dGHD, and ISS groups were developed and fitted by least squares and the REG procedure in SAS. A hierarchy of 10 predictive factors was derived using the all possible regression subsets approach, with Mallow’s C(p) criterion employed to order the predictive factors, as described by Weisberg and Cook (22, 23). This method involved evaluating all possible combinations of the candidate predictor variables to identify the subset of predictors that best balanced a model with low bias and a good balance between precision and parsimony. The 10 candidate predictors were: height SDS at rhGH start, birth weight SDS, birth length SDS, max GH peak, treatment duration, sex, rhGH dose, gestational age, age at start of GH, BMI SDS, modeled as individual effects.

An additional analysis was performed to explore interactions between the variables and diagnostic groups. No multiple comparison testing procedure was applied to select the regression coefficients. All statistical analyses were conducted with SAS^® version SAS, version 9.4 (SAS Institute).

The analysis included 3,939 children who had undergone GH stimulation tests, initiated rhGH treatment before puberty, and reached NAH. Of these, 1,401 were classified as having dGHD (36%), 1,300 as SUS (33%), and 1,238 as ISS (31%). All baseline characteristics, data after 1 year of treatment, at the start of puberty, and at NAH are reported in Tables 1 - 3 ; p-values are corrected for multiple comparisons.

This study provides a comprehensive analysis of the differences among prepubertal children treated with rhGH and classified as dGHD, SUS, or ISS. By leveraging a large international cohort, we aimed to identify the most significant predictors of NAH and compare the response to prolonged rhGH therapy across these diagnostic groups.

The main finding of this study is that children with SUS have baseline differences from dGHD, but exhibited a similar response to rhGH treatment, in contrast to the distinct response seen in ISS patients. However, dGHD and SUS differ markedly in clinical, auxological, and biochemical profiles—such as GH peak levels, IGF-1 levels, BMI, and mid-parental height.

Regarding the long-term response to hormonal treatment, all three patient cohorts benefited from rhGH administration, resulting in an overall reduction in the prevalence of short stature. The prevalence of short stature at the start of treatment was similar between the SUS and ISS groups (93%), but lower in the dGHD group (82%). However, by the end of rhGH treatment, the prevalence of individuals achieving normal height was comparable between the dGHD and SUS groups, with 74% reaching a normal height, compared to only 65% of ISS patients who achieved this outcome. Patients with dGHD showed a higher NAH, achieving a median height of -1.06 SDS, compared to -1.28 SDS in SUS, with both groups significantly outperforming ISS patients, who reached -1.59 SDS. This difference does not appear to be related to rhGH dose, as children with dGHD experienced the greatest improvement in height SDS both during the first year of treatment (+0.74 SDS vs. +0.68 in SUS and +0.59 in ISS), while receiving significantly lower rhGH dose compared to SUS and ISS, and by the end of treatment (+2.08 SDS vs. +1.80 in SUS and +1.51 in ISS), when receiving higher rhGH doses. However, both SUS and ISS patients achieved a NAH closer to their genetic potential: the differences from MPH were -0.05 SDS in SUS and -0.13 SDS in ISS, compared to -0.34 SDS in dGHD. Ultimately, the multivariate analysis revealed that, when controlling for factors such as sex, birth weight, height at the start of treatment, GH peak at stimulation tests, treatment duration, and GH dose, the difference in NAH between dGHD and SUS was minimal, at only 0.01 SDS. On the contrary, ISS patients had a significantly lower NAH, with a difference of -0.13 SDS compared to dGHD and -0.12 SDS compared to SUS, further underscoring their lower overall growth potential in response to rhGH treatment, as already reported (24, 25). Nevertheless, in our cohort, the response of ISS children to rhGH treatment was found to be better than what has been reported in previous studies (26, 27). When considering bone age, it was found to be homogeneous throughout all categories at baseline; however, delay persisted during treatment and at NAH solely in the dGHD group. This is consistent with previous findings of bone age progression being more rapid in non-GHD patients treated with rhGH and that bone age progression in GHD patients reaches a plateau after an initial increase with replacement hormonal therapy (28, 29).

In our study, the most significant positive predictors of response to rhGH therapy, in terms of NAH, were height at the start of treatment and mid-parental height, particularly in the ISS group. It is well established that height is a polygenic trait, with normal adult height determined by the interaction of multiple factors. While nutritional, endocrinological, and social factors are important, up to 80-90% of adult height is genetically determined (30). Major gene mutations can lead to GHD or skeletal dysplasia, but polymorphisms, copy number variations, and milder mutations also play a critical role in regulating height (31). Therefore, it can be hypothesized that in children with ISS, who do not have a genetically determined GH deficiency, other inherited polygenic factors are likely key determinants of adult height, explaining the observed correlation with MPH.

It is noteworthy that SUS patients exhibited a similar response to rhGH treatment despite having distinct characteristics from children with dGHD, both at baseline and throughout treatment.

While both groups, by definition, had a GH peak <8 ng/mL, children with dGHD exhibited a significantly lower peak compared to those with SUS (2.70 vs. 5.15 ng/mL). It is well-known that GH stimulation tests suffer from issues with reproducibility, potentially yielding false positives in cases of syndromic short stature not related to GH deficiency (e.g., SHOX deficiency) (32), or false negatives in genetically confirmed GHD (3, 33). Nevertheless, Coutant et al. have reported cases of severe GHD due to identifiable genetic or anatomical abnormalities, or combined pituitary hormone deficiency (CPHD), where peak GH levels in stimulation tests were consistently <5 ng/mL, lower than in ‘idiopathic’ GHD cases (34). In addition to their intrinsic reproducibility issues, GH stimulation tests may also fail in children with otherwise normal growth, reflecting temporary or functional insufficiency of GH secretion rather than true GHD. Such conditions include CDGP, neurodevelopmental disorders such as ADHD (35, 36). These scenarios likely contribute to the high rate of failed stimulation tests reported in normally growing prepubertal children, with previous studies documenting failure rates ranging from about 23% to as high as 75% when conventional cut-offs are applied (37). This highlights the considerable risk of misclassification when relying solely on these tests.

In our study, we also observed a significant inverse correlation between GH peak and NAH, as previously documented (38). This suggests that individuals with lower GH peaks during stimulation tests may respond better to rhGH therapy. Notably, this correlation between GH peak and NAH was also evident in ISS patients, confirming prior concepts positioning ISS as a transitional state between GHD and GH insensitivity (39, 40). The homogeneity of this correlation throughout all categories of patients supports the hypothesis that SUS may represent an intermediate category between dGHD and ISS. This interpretation is further supported by the reduced rate of confirmation of pathological GH secretion upon retesting in SUS patients (11).

Compared to SUS patients, children with dGHD also had lower IGF-1 levels at diagnosis (-2.50 vs. -1.91 SDS). They also exhibited the best response to rhGH treatment after 1 year, with the highest increase in IGF-1 levels (+1.80 SDS vs. +1.51 in SUS and +1.21 in ISS, respectively). IGF-1 levels are known to be lower in organic GHD compared to idiopathic GHD (12), and our group previously demonstrated that dGHD patients have lower IGF-1 levels at diagnosis compared to SUS (11). Although IGF-1 levels can be influenced by various factors (such as malnutrition, pubertal stage, and age) (41), it is clear that they are highly dependent on GH secretion. It is important to note that confounding factors in our cohort were minimized, as all patients were prepubertal, and IGF-1 values were reported as SDS adjusted for sex and age. Moreover, ISS patients had higher IGF-1 levels, within the normal range, consistent with the different etiologies of short stature.

BMI in dGHD patients was higher compared to SUS patients, both at the start (0.20 vs. -0.35 SDS) and at the end of treatment (0.34 vs. 0.02 SDS), though still within the normal range. Previous studies have suggested that GHD in children is often associated with mild to moderate overweight (13). However, other research has reported that BMI tends to be within the average range, with no significant differences observed between organic and idiopathic GHD (42). From a metabolic perspective, GH acts as an anabolic hormone, promoting lipolysis, lipid uptake in skeletal muscle, and insulin synthesis (43). Rather than focusing solely on BMI, attention should be given to fat distribution patterns. However, our study did not assess body composition, so it is unclear whether the relative increase in BMI observed in dGHD patients was due to an increase in lean mass or fat mass (44).

With regard to birth weight and length, dGHD patients had a history of higher birth weight and length (-0.40 vs. -0.67 SDS, and -0.19 vs. -0.51 SDS, respectively). It is well known that, despite GH being the key hormone for maintaining postnatal growth, it plays a negligible role in prenatal growth (45, 46); even in cases of congenital GH deficiency — excluded from this study due to their specific characteristics — birth length is not affected (47).

Compared to the dGHD group, children with SUS had a lower MPH (-1.20 vs. -0.65 SDS) and a lower difference between final height and MPH (-0.05 vs. -0.34 SDS). Children with ISS had an even lower MPH (-1.33 SDS) and a lower difference between final height and MPH (-0.13 SDS). This suggests that at least some SUS and ISS patients may fall under the category of ‘familial short stature’ (FSS), suggesting that an autosomal dominant form of short stature might coexist in this subgroup of patients (48). It is often observed that children with short stature, initially diagnosed with GHD, may actually have genetic variants associated with growth, but rarely involving GH secretion or function (49, 50). In fact, the GH-IGF-1 axis has recently been found not to be the sole determinant of chondrogenesis and growth, which are also greatly influenced by a complex interplay of hormonal, paracrine, extracellular, and intracellular factors (51). The low GH peak observed during stimulation tests in SUS could thus be viewed as an epiphenomenon rather than the underlying cause. Indeed, the SUS category may include a variety of patients whose short stature, rather than being caused by definite GH deficiency, may be due, for example, to growth plate disorders. The latter are frequent amongst children with a “GHD” diagnosis, and have an excellent response to rhGH therapy (50). It is, in fact, also known that response to rhGH treatment is not specific for GHD, as children with monogenic causes of short stature different from GHD and even children with ISS experience height gain with replacement therapy (50, 52). The role of genetic testing in shaping the modern diagnostic approach to pediatric short stature is undeniable (53–55). Unfortunately, the period during which the KIGS registry was utilized does not reflect the current era, with its availability of advanced panels and clinical tests.

Several limitations of this study should be noted.

First, as an observational study, this study has several potential limitations. Since this is a retrospective cohort study, without randomization, confounding and biases cannot be entirely eliminated, and the determination of causation is limited. Enrollment bias is possible, though sites were expected to enroll all eligible patients treated with rhGH. Follow-up bias over the duration of KIGS is a potential limitation as well; estimates presented in the manuscript are based on patients with non-missing data. Furthermore, only patients who remained in the registry until reaching NAH were included, whereas dropouts—who may have had less favorable outcomes—were not captured. Therefore, our results should be interpreted as conditional on persistence in follow-up until NAH.

Additionally, although the dataset includes a large number of patients, data on certain variables (such as IGF-1 levels, e.g.) were incomplete for some individuals, which may affect the accuracy of our results and, to some extent, cause some overlap between the three groups. Furthermore, we did not have information on the use of sex steroid priming before GH stimulation tests, nor on the presence of ADHD or other neurodevelopmental disorders, or the use of psychoactive medications. These unmeasured factors may have influenced GH test responses in some children. In addition, our analyses applied Italian diagnostic criteria and cut-offs to all patients in the KIGS database. While this approach ensured consistency across the cohort, it may not fully reflect the heterogeneity arising from country-specific diagnostic and treatment practices. Lastly, it must be acknowledged that the lack of genetic analyses does not allow us to establish a definitive pathogenetic mechanism for short stature in all categories of patients. All these limitations should be considered when interpreting our findings.

However, this is the first large (almost 4,000 children who started rhGH treatment in pre-puberty) multi-centric study comparing dGHD, SUS, and ISS children who underwent treatment with rhGH, and, even without genetic confirmation, it shows that the three groups have distinct clinical, auxological, and laboratory characteristics. Furthermore, the extended timeframe of the study has allowed long follow-up times and thus identification of predictive factors for height SDS at the end of rhGH treatment across the three categories of patients.

These results underscore the need to reconsider the use of the GHD label for children without clear-cut evidence of deficiency. It is not merely a semantic issue to use the definition of SUS instead of “idiopathic GHD”. Firstly, categorizing a case as idiopathic GHD still results in a diagnosis of GHD, even without identifying a specific cause. This diagnosis would be based on stimulation tests, which are not highly reliable, and on characteristics inconsistent with true GHD (e.g., normal IGF-1 levels, normal BMI, and a lower target height). Moreover, the follow-up approach differs significantly, as in true GHD, the risk of persistent deficiency and the development of other pituitary deficits is much higher (56). In our opinion, this study further emphasizes the need for a correct terminology when performing the differential diagnosis of short stature, which must be supported by clinical, auxological, and laboratory findings but should ultimately be confirmed through genetic testing. Indeed, even though short stature is unlikely to have a monogenic origin, knowledge on its genetic causes is continuously improving, with rates of genetic diagnosis in “idiopathic” short stature as high as 46% with whole exome sequencing (WES) analysis (57). While it may be argued that such genetic testing may be an economic burden, so are stimulation tests and rhGH therapy, which should thus be offered to children who can most benefit from them (7). Furthermore, retaining these children under the GHD umbrella risks inappropriate expectations, unnecessary follow-up for pituitary dysfunction, and a misleading clinical narrative for families. Our findings strengthen the case for adopting more accurate terminology and classification strategies—grounded in biology, not just GH peak thresholds—in line with modern principles of precision medicine. We hope that in the future this may aid in offering both clinicians and patients’ families a more accurate prediction of current and future height with and without hormone supplementation.

In conclusion, this study highlights the importance of classifying children with growth disorders into distinct diagnostic categories (dGHD, SUS, and ISS) to better understand their response to rhGH therapy and predict their growth outcomes. Our findings support the clinical value of rhGH treatment across different diagnostic groups, including patients with GHD of unknown etiology. All three groups—SUS, dGHD, and ISS—showed a positive response to therapy, with a general reduction in the proportion of individuals presenting with short stature at NAH. Although children with dGHD and SUS had almost the same response to rhGH in terms of NAH, they differ in baseline characteristics and may have different follow-up. When an identifiable anatomical or functional cause for GHD is not found, we believe that genetic testing should be considered in all children with SUS.