Background
Growth hormone (GH) treatment of patients with GH deficiency (GHD) diagnosed in childhood has historically focused on maximizing adult height. However, this limited goal overlooks the importance of GH for completion and maintenance of somatic and metabolic maturation, including bone mineralization; accrual of lean body mass, with accompanying increases in muscle strength and exercise capacity; and changes in lipid metabolism [
1‐
18]. Thus, there is now consensus that GH replacement is important for those patients with childhood-onset GHD who remain GH deficient after completion of linear growth [
18‐
22].
Pharmacologic GH stimulation testing is generally recommended to confirm the diagnosis of persistent GHD during the childhood-to-adulthood transition, but this procedure requires interruption of GH therapy, is labor intensive, and is logistically challenging, given the scarcity of testing agents now available. In addition, provocative testing is invasive, has the potential for significant side effects, and produces inconsistent results that do not predict treatment response [
19‐
24]. Because of these issues, several European studies have examined clinical and biochemical predictors of persistent GHD [
25‐
30]. However, interpretation of the data is affected by factors such as the retrospective nature of most studies, interstudy differences in diagnostic criteria, and interassay variability. Furthermore, because previous studies have been performed in Europe, where diagnostic and treatment practices differ from US practices, the existing data may not be directly applicable to the largest group of children treated in the USA—those with idiopathic GHD (IGHD). Therefore, this study determined the prevalence of persistent GHD after attainment of adult height in a cohort of US childhood-onset GH-deficient patients during the transition period, with particular focus on those with IGHD, and examined the value of various factors as diagnostic predictors of persistent GHD.
Methods
Patients
This study screened 73 patients at 21 US institutions for entry to a randomized clinical trial of GH effects on bone and body composition in previously treated childhood-onset GH-deficient patients (efficacy and safety data have been reported [
12,
15]). The study was approved by the institutional review boards of participating institutions, and written informed consent was obtained from patients and/or their legal guardians.
Study entry criteria included: age 14–28 years; diagnosis of GHD during childhood/adolescence (either idiopathic or organic [i.e. due to a genetic or structural cause]); GH treatment ≥1 year, completed 6 weeks–5 years before screening; attainment of adult height (height velocity <1 cm/year); no history of spinal or total body irradiation, bone dysplasia, or significant systemic illness. Patients with additional pituitary hormone deficiencies (PHDs) were required to have received stable replacement therapy (thyroxine, glucocorticoids, sex steroids, vasopressin, as needed) for ≥6 months. The US cohort from this international study was selected for the analysis reported here because serum GH, insulin-like growth factor-I (IGF-I), and insulin-like growth factor binding protein-3 (IGFBP-3) concentrations for all US patients were measured at a central laboratory.
Baseline demographic data included etiology and age at diagnosis of childhood GHD, duration of previous GH treatment, presence of additional PHDs, age, and height and weight at retesting.
Assessment of GH secretion
Screening for entry to the adult GH replacement trial included IGF-I and IGFBP-3 measurements followed by GH stimulation testing. A single stimulation test was sufficient for patients with history of multiple PHDs (MPHD); 2 tests were required for patients with history of isolated GHD. Protocol-preferred stimulation tests included insulin tolerance test (ITT), combined arginine/L-dopa test, and glucagon test. However, to represent the breadth of US pediatric endocrine practice, no specific testing protocol was mandated. Patients were eligible to enroll in the GH replacement trial if IGF-I was <1
st percentile for age/sex and peak GH was <5 μg/L. The GH threshold for definition of GHD was specified
a priori in the protocol and is consistent with guidelines for diagnosis of GHD during the transition period [
19‐
21]. Data from all US patients are included in this report, regardless of eligibility for the GH replacement trial.
Laboratory analyses
IGF-I was measured by an IGFBP-blocked radioimmunoassay as described elsewhere
( sensitivity 0.1 μg/L; intra- and interassay coefficients of variation [CV], 1.6% and 6.4%, respectively [
31]). IGFBP-3 was measured by radioimmunoassay (sensitivity 0.13 mg/L; intra- and interassay CV, 1.9% and 9.2%, respectively [
32]). Results were converted to standard deviation scores (SDS) using data for age/sex-matched controls from the same assays. GH was measured using an immunochemiluminometric assay specific for 22-kDa human GH [
33]. All assays were performed centrally at Esoterix Endocrinology, Inc (Calabasas Hills, CA, USA).
Statistical analyses
Statistical analyses were performed using the SAS software system (SAS Institute, Inc, Cary, NC). Because stimulated GH values were not normally distributed, the nonparametric Wilcoxon test was used to evaluate differences between GH-deficient vs. non–GH-deficient patients with respect to number of additional PHDs, serum IGF-I/IGFBP-3, age at original diagnosis, weight, and body mass index (BMI; kg/m2). The difference in peak GH among patients with 0, 1, ≥1, or ≥2 PHDs was examined using the nonparametric Kruskal-Wallis test. Relationships between peak GH and potential explanatory variables were assessed using Spearman correlation coefficients (rs). Summary data for continuous variables are presented as mean ± SD unless otherwise noted.
Calculation of sensitivity, specificity, positive predictive value, and negative predictive value
Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated to determine the utility of clinical and laboratory variables as screening tests for persistent GHD (defined as peak GH response <5 μg/L). Screening variables included etiology of childhood GHD (organic vs. idiopathic), age at childhood diagnosis, number of additional PHDs, and study entry values for weight, BMI, IGF-I, and IGFBP-3. Continuous variables (age, weight, BMI, IGF-I, and IGFBP-3 SDS) were tested to determine cut-off values predictive of GHD. Patients with values beyond the cut-off were classified as having a positive screening test (screen) for GHD. Patients with a positive screen who had maximum GH < 5 μg/L were designated as true positive (TP); patients with a positive screen who had peak GH ≥ 5 μg/L were designated false positive (FP); a negative screen accompanied by peak GH ≥ 5 μg/L was defined as true negative (TN); a negative screen with peak GH < 5 μg/L was defined as false negative (FN).
The following additional definitions were used: sensitivity (of the screening test), represent the probability of a positive screen among patients with GHD (i.e. proportion of GH-deficient patients correctly identified by the screen, calculated as TP/[TP + FN]); specificity, the converse of sensitivity, represents the probability of a negative screen among non–GH-deficient patients (proportion of non–GH-deficient patients correctly identified by the screen; TN/[TN + FP]); PPV, is the probability of GHD among patients with a positive test (proportion of patients with positive screen who were GH deficient; TP/[TP + FP]); NPV, is the probability of being non-GH deficient among patients with a negative screen (proportion of patients with negative screen who were non–GH deficient; TN/[TN + FN]). These calculations were determined for all patients (organic and idiopathic combined) and repeated separately for patients with IGHD.
Discussion
Since the early 1990s the role of GH in many physiologic processes in adulthood has become clearer, and the importance of GH replacement for GH-deficient adults is well established [
19‐
21,
34]. Many studies have demonstrated deficits in somatic and metabolic maturation in GH-deficient individuals untreated during the transition period [
1‐
4,
6‐
17]. However, the determination of precisely which patients require ongoing GH therapy has been less clear, as many patients treated for childhood GHD do not fulfill diagnostic criteria for adult GHD after completion of linear growth. This finding may reflect a number of factors, including differences in diagnostic criteria for GHD in childhood
vs. adulthood, lack of reproducibility of GH stimulation tests, and perhaps sex steroid–mediated maturational changes in hypothalamic control of GH secretion during puberty [
23,
35‐
37]. Consequently, retesting GH secretion in adolescents and young adults with childhood-onset GHD is generally recommended [
19‐
22]. However, such testing requires interruption of GH therapy, and the results vary by protocol, secretagogue, and GH assay; lack reproducibility; and do not predict treatment response [
23]. Furthermore, the increasingly limited availability of many agents for which GH stimulation testing protocols are established (e.g. arginine, GH-releasing hormone, L-dopa) leaves few options other than ITT, which requires physician presence because of the risk of complications such as seizures as a result of significant hypoglycemia [
24,
33]. Therefore, this study aimed to provide a rational basis for GH stimulation retesting in US patients by examining factors predictive of persistent GHD in a cohort of 73 patients with history of childhood-onset GHD who underwent centralized measurements of IGF-I, IGFBP-3, and GH after completion of childhood treatment. Because of limited published information, particular attention was focused on factors predictive of persistence in patients with history of IGHD, the most common form of childhood GHD treated in the USA.
Our finding that 100% of US patients with history of organic GHD had persistent GHD confirms previous European reports [
26,
27,
30,
38,
39]. Similarly, we found a very high prevalence of persistent GHD in patients with ≥1 additional PHD (96% PPV) [
25,
29,
40‐
42]. Thus it appears that despite potential differences between US and European physicians with regard to diagnosis and treatment of childhood GHD, the key factors associated with its persistence appear consistent across these geographies. The single patient with an additional PHD (TSH) who did not fulfill the study definition of GHD may nevertheless have a partial GH secretory defect because peak GH response to arginine/L-dopa was 9.0 μg/L. Other studies have concluded that such patients may have a milder form of GH “insufficiency” [
29,
43‐
45]. As GH is usually the first anterior pituitary hormone affected by pathological insults, there is a biological rationale to suspect that patients with ≥1 additional PHD will likely have persistent GHD [
46,
47]
.
Organic etiology of GHD and presence of additional PHDs reflect the severity of hypothalamic-pituitary dysfunction, so it is not surprising that severe GHD persisted in almost all such patients; provocative GH retesting thus appears unnecessary in patients with organic disease [
29,
38,
39,
42]. Instead, GH potentially could be continued uninterrupted through the transition period (with appropriate dosage adjustment) to avoid the adverse changes in body composition, lipid profile, and cardiac function that may develop following discontinuation of GH [
1‐
4,
6‐
17]
. Furthermore, patient care could potentially be improved by providing the family with a clear expectation at the initiation of childhood treatment, of the likelihood that GH treatment will be required in adulthood.
Although only half of our patients with MPHD had a childhood diagnosis of organic disease, some patients whose MPHD was labeled “idiopathic” may, in fact, have had an undiagnosed genetic disorder. This is suggested in other studies by the presence of mutations in genes encoding pituitary transcription factors, most commonly
PROP1, in up to half of patients with an original diagnosis of idiopathic MPHD [
37,
48‐
51]
. Furthermore, up to one-quarter of children with isolated GHD may have detectable genetic defects [
49,
52,
53]. Thus, genetic studies should be obtained whenever possible in any patient with MPHD or early-onset isolated GHD, because presence of a mutation would obviate the need for GH stimulation retesting after childhood treatment, and allow such patients to continue replacement therapy uninterrupted. Similarly, although our study did not include magnetic resonance imaging (MRI) assessment, MRI anomalies have been reported as a significant predictor of persistent GHD during transition [
27,
37,
41,
54], and certain MRI findings may indicate a genetic basis for hypothalamic-pituitary disorders [
55‐
57]
.
In contrast to those with organic hypothalamic-pituitary dysfunction, patients with childhood IGHD present a substantial diagnostic dilemma, and prior studies have not evaluated predictive factors for persistent GHD in this specific population. Moreover, as idiopathic patients represent the majority of recipients of childhood GH treatment in the USA [
58‐
60]
, they constitute the bulk of the clinical load for US pediatric endocrinologists. Therefore, our study specifically examined factors predictive of persistent GHD in this subgroup. Only about one-third of idiopathic patients (36%) retested as GH deficient; this was true for even fewer patients with isolated IGHD (17%). The low rate of persistent GHD in our US idiopathic cohort is similar to the rates reported in Belgian, British, and French studies, in which 15%–24% patients with childhood isolated IGHD remained GH deficient when retested [
26,
40,
61]. However, our results differ notably from those of an Italian study in which 52%–65% of young adults with isolated IGHD were GH deficient on retest
, likely reflecting the fact that about one-third of patients in the Italian study had severe childhood GHD [
39].
Apart from the presence of additional PHDs, the strongest independent predictor of persistent GHD in our idiopathic cohort was the finding of IGFBP-3 below -2.0 SDS, which had 100% PPV for persistent GHD. In contrast, a subnormal IGF-I value (i.e. <-2.0 SDS) was not prognostically helpful in those with history of IGHD, as only half of such patients retested as GH deficient. However, an extremely low IGF-I (<-5.3 SDS) provided 100% PPV; in addition, the combination of IGF-I SDS below -2.0 and young age at original diagnosis of IGHD was strongly predictive of persistent GHD. Our finding of lack of predictive power of subnormal IGF-I contrasts with the good concordance between IGF-I and peak GH reported in European studies [
25‐
27,
55], perhaps reflecting the typically greater severity of GHD in European children, differences in agents and diagnostic cut-points used for GH testing, and time between discontinuation of GH and retesting (as GHD may manifest after increasing time off treatment [
43,
44]). Furthermore, IGF-I secretion is controlled by other factors in addition to GH, such as nutritional status and sex steroid milieu [
32,
62,
63]. Perhaps more importantly, IGF-I may provide a good screen for GH
sufficiency, as 100% of idiopathic patients who had IGF-I > -1.6 SDS were GH sufficient on retest (100% NPV for GHD). Patients with IGF-I SDS values above this level after discontinuation of GH treatment could be spared the invasive process of GH stimulation retesting after completion of childhood therapy, as all would be expected to be GH sufficient, and instead could be followed clinically.
The other useful predictor of persistent GHD in the idiopathic cohort was age <4 years at original diagnosis (specificity 97%, PPV 89%), likely reflecting the fact that growth failure occurs earlier in children with more severe GHD [
29]. Consequently, families of children who are very young at initial diagnosis of IGHD should be forewarned of the likelihood of its permanence.
This study has a number of potential limitations. First, no direct comparison of GH stimulation test results at the time of childhood diagnosis versus results on retest in the present study could be made because initial testing was performed at the individual institutions and not at a central laboratory. For the same reason, we were unable to assess the predictive value of a number of other clinically relevant parameters, such as pretreatment IGF-I, height SDS, height velocity, or height gain in response to childhood treatment. Second, the single cut-point of 5 μg/L defined in the protocol to represent the threshold for GH deficiency irrespective of the testing agent used, may be considered to lack precision; a subsequent study in patients with
adult-onset GHD (conducted after our study was designed and implemented) indicates that different diagnostic thresholds are appropriate for different agents [
33]. However, evidence for the appropriateness of this approach is lacking for patients in the transition period, as noted by consensus statements from endocrine societies [
19‐
21]. Third, because our study population comprised patients screened for aGH replacement trial, the cohort may represent the more severe end of the US childhood GHD spectrum, and persistent GHD may be less likely in milder cohorts. Nevertheless, our finding that only 17% of patients with history of isolated IGHD had persistent GHD is consistent with European data for this subgroup. Fourth, IGF-I assays have substantial interlaboratory variability, so the very low IGF-I SDS values predictive of persistent GHD in our study may not be applicable to IGF-I measured elsewhere. Fifth, obesity is associated with blunted GH response to stimulation, even in non–GH-deficient individuals [
64], leading to potential bias toward overdiagnosis of GHD. Thus the peak GH threshold of 5 μg/L used for diagnosis of GHD in this study may be inadequately stringent for obese patients (BMI > 30 kg/m
2) [
20]. Nevertheless, as all obese patients in this study had additional PHDs, misdiagnosis due to obesity-related blunting of GH secretion seems unlikely. Finally, it is acknowledged that no single study can provide comprehensive guidelines for the broad range of patients treated and followed in different clinical settings, and assessment should be individualized for each patient.
Competing interests
This study was sponsored by Eli Lilly and Company (Indianapolis, IN). In compliance with the Uniform Requirements for Manuscripts, established by the International Committee of Medical Journal Editors, the sponsor did not impose any impediment, directly or indirectly, on the publication of the results of this study.
Authors’ contributions
CAQ and JJC conceived the objectives questions and analyses reported in this manuscript; CAQ coordinated the study and manuscript development, and drafted the manuscript; AJZ and CCL participated in the design of the analyses and performed the statistical analyses; DMB, CH, LL, DRR, and ET revised the manuscript for intellectual content. All authors read and approved the final manuscript.