Provocative Growth Hormone Testing in Children: How Did We Get Here and Where Do We Go Now?

Abstract

Objectives:

Provocative growth hormone (GH) tests are widely used for diagnosing pediatric GH deficiency (GHD). A thorough understanding of the evidence behind commonly used interpretations and the limitations of these tests is important for improving clinical practice.

Content:

To place current practice into historical context, the supporting evidence behind the use of provocative GH tests is presented. By reviewing GH measurement techniques and examining the early data supporting the most common tests and later studies that compared provocative agents to establish reference ranges, the low sensitivity and specificity of these tests become readily apparent. Studies that assess the effects of patient factors, like obesity and sex steroids, on GH testing further bring the appropriateness of commonly used cutoffs for diagnosing GHD into question.

Summary and Outlook:

Despite the widely recognized poor performance of provocative GH tests in distinguishing GH sufficiency from deficiency, limited progress has been made in improving them. New diagnostic modalities are needed, but until they become available, clinicians can improve the clinical application of provocative GH tests by taking into account the multiple factors that influence their results.

Introduction

Provocative growth hormone (GH) testing (a.k.a. GH stimulation testing) has played a central role in the diagnosis and management of GH deficiency (GHD) for over 50 years. Despite this long history, challenges persist in the performance and interpretation of these tests. In this review, we present the evidence behind their use to contextualize current practice. We demonstrate that the limitations of these tests present at the time of their initial development remain unresolved. We propose approaches to improve clinical practice in the identification of children with GHD.

Growth hormone assays

Accurate measurement of the serum concentration of GH is critical to the interpretation of provocative GH tests. In serum, the GH molecule exists in numerous isoforms, which complicates its measurement. The predominant circulating isoform has a molecular weight of 22 kDa, but a significant fraction (about 10%) circulates as the 20 kDa isoform. Larger isoforms (35 kDa and 45 kDa), dimeric and oligomeric isoforms, as well as smaller fragments derived from the GH1 gene that encodes GH, also circulate in serum [1–3]. Post-translational modifications further contribute to isoform diversity [2]. Clearance of GH molecules, which occurs predominantly renally, is slower for dimeric, oligomeric and 20 kDa GH isoforms, than for 22 kDa GH [2].

Before the advent of recombinant human GH (rhGH), GH assays were based on pituitary-derived GH. The World Health Organization (WHO) Expert Committee on Biological Standardization established the first international standard (IS) for calibration of bovine GH preparations in 1955, using purified bovine pituitary GH extracts; the units arbitrarily assigned to this preparation defined future standards [4]. The first human pituitary-derived GH international reference preparation (IRP), IRP 66/217, was introduced in 1969 [5] (Figure 1A). However, only in 1982 did the WHO establish a human pituitary-derived GH IS validated in an international study (IRP 80/505) [6,7]. Prior to this, local standards of purified human GH (hGH) extracted by various methods and distributed by various agencies had been employed (e.g. Wilhelmi hGH preparations distributed by the National Pituitary Agency, Baltimore, MD) [6,8]. This led to considerable variation in measurement: “The results of [an] international collaborative study showed a wide diversity of estimates of the potency, and content of the ampoules, made by various assay methods and compared with various ‘pure’ and impure pituitary extracts used as local standards in several ‘expert’ laboratories. Such extracts contained various amounts of 22 and 20 kDa proteins, denatured forms, and other impurities from the gland. One must conclude that before the use of the International Standard for human growth hormone and the International Unit it defined, ‘units of growth hormone’ referred to in literature must have been imprecise measures of GH activity or growth hormone proteins [7].”

Figure 1.

Growth hormone measurements in perspective. (A) The evolution of the GHD diagnostic cutoffs in relation to key developments in GH measurements and treatment is depicted. (B) Factors affecting the results of provocative GH testing are shown. These include both test factors and patient factors. *International Standard; ˆmore sensitive than the competitive assays that had previously been used.

In 1993, the first IS IRP using rhGH was introduced (IRP 88/624) [9]. The ability to quantify rhGH in milligrams led to a change in the standard units for GH measurement from mU/L to mcg/L [10]. However, the conversion factor suggested by the WHO Expert Committee on Biological Standardization did not match results of independent assessments using different immunoassays [10,11]. Thus, various conversion factors were adopted, which further complicated comparability of reported GH measurements in research and clinical care [3]. Additionally, even after this standard was developed, different assays continued to use other calibrators, both biologic and synthetic [3,10].

Current consensus guidelines recommend GH assay calibration with IS IRP 98/574, a highly purified (>96%) preparation of the 22kDa rhGH isoform, introduced in 2000 [12–15]. This calibrator does not include the wide range of naturally occurring hGH isoforms. Additionally, with the advent of rhGH, GH immunoassays (the most common GH assay type run by clinical labs today) transitioned from using polyclonal antibodies that targeted naturally occurring pituitary hGH epitopes to monoclonal antibodies targeting epitopes of monomeric 22kDa rhGH [2]. These changes in laboratory techniques mean that contemporary assays have a narrower target than the assays used in the early studies of provocative GH tests, which detected more forms of naturally occurring GH. The fact that 22kDa GH undergoes more rapid clearance than other isoforms presumably also leads to detection of lower concentrations of GH by current, versus earlier, assays.

When comparing historical and contemporary reported GH values, inter-assay differences in immunoassays further complicate interpretation. They can arise from 1) variations in the type of immunoassay (Figure 1B), 2) antibody specificity, 3) matrix used (assay platform and reagents), and 4) interference from GH binding proteins [2,6,10,16]. Adding to measurement variability, in vivo and in vitro bioassays, radioreceptor assays, and mass spectrometry also have been used to assess GH concentrations, [2,6,7,17].

Comparison of GH concentrations in single samples (n=312), each analyzed by 8 distinct immunoassays, demonstrated inter-assay variation as large as 11.4 ng/ml; the coefficient of variation for the three assays that used monoclonal antibodies to measure 22kDA hGH was 24.1 ± 10.2% [16]. Specifically comparing two assays [IMMULITE 2000 (Siemens, Munich, Germany) and BC-IRMA (Beckman-Coulter, Brea, California, USA)], a diagnostic cutoff of 6 ng/dL would yield misaligned categorization (GH deficient or sufficient) between the results of these assays for 29% of the 312 children studied [16]. This study highlights how applying the same diagnostic threshold to measurements by different GH assays commonly can lead to misclassification of patient results. This is a major problem for both clinical practice and research, as it impedes meaningful comparison of results across studies that are performed using different GH assays.

It is generally accepted that newer assays yield lower GH concentrations than older ones. However, it is not possible to approximate equivalent measurements for pre-versus post-rhGH derived standards due to the extent of assay variability. Conversion factors to compare old and new assays based on activity levels that had been assigned to IRPs (e.g. 1 mg=2.6 IU for IRP 80/505 versus 1 mg=3 IU for IRP 98/574), suggested by some consensus guidelines, have been shown to be inaccurate [10,18].

Initial studies

The clinical need for provocative GH testing arose from the early recognition that GH secretion varies throughout the day in an unpredictable pattern, with the highest pulses occurring during deep sleep. This makes random assessment of GH levels unhelpful for diagnostic purposes outside of the newborn period [19,20]. In the first 15 days of life, random GH measurement may be useful due to a GH surge at birth and the developmental lack of sleep-entrainment of GH secretion (see [17] for a thorough review). Although provocative testing was deemed necessary, concerns arose about whether pharmacologically stimulated GH levels reflect normal function [21]. In parallel with pharmacological stimulation, tests that sample GH levels in a physiologic state were also developed. The testing protocols employed in the initial pediatric studies of the more commonly used pharmacological stimuli are summarized in Table 1. The subjects, GH assay and standards used in the initial pharmacologic and physiologic studies are presented in Table 2. Side effects, considerations, and limitations for each of the pharmacologic agents are shown in Table 3.

Table 1.

Protocols used in initial studies of provocative GH tests

Test	Study	Protocol
ITT	Kaplan et al, 1965 [23]	0.1 units/kg insulin with GH levels drawn at 0, 30, 60, and 90 min. after insulin injection
	Kaplan et al, 1968 [24]	0.1 units/kg IV regular insulin after overnight fast followed by 30 min. of bedrest with samples drawn at 0, 15, 30, 45, 60, 90, 120 min. after insulin injection from blood drawing IV
Arginine	Parker et al, 1967 [25]	0.5 g/kg L-arginine monochloride (30g in adults) over 30 min with samples drawn at -30, 0, 15, 30, 45, 60, 90, 120 min. from blood drawing IV 0.1 u/kg IV regular insulin (0.05 units/kg if suspicion for GHD), with samples drawn at 0, 20, 40, 90 min. from blood drawing IV
Glucagon	AvRuskin et al, 1968 [26]	0.03 mg/kg (max 1.0 mg) IM glucagon after two hour tolbutamide test with samples drawn at 0, 15, 30, 45, 60, 90, 120 min.
	AvRuskin et al, 1971 [27]	0.03 mg/kg (max 1.0 mg) IM glucagon following overnight fast with samples drawn at 0, 15, 30, 45, 60, 90, 120 min. and patients kept supine starting 30–60 min. before testing (both with and without preceding 2 hour tolbutamide test)
	Weber et al, 1970 [28]	1 mg IV glucagon following 14h overnight fast and IV placement with samples drawn at -15, 2, 5, 10, 20, 30, 40, 50, 60 min.
Levodopa	Hayek and Crawford, 1972 [30]	10 mg/kg oral levodopa with samples drawn every 30 min. from -30–120 min from a blood drawing IV
Clonidine	Gil-Ad et al, 1979 [31]	0.15 mg/m2 oral clonidine over 30 min. after IV placement with samples drawn every 30 min. from 0–180 min.
GHRH	Grossman et al, 1983 [33]	200 mcg synthetic human pancreatic GRF-40 over 30 min. after overnight fast, IV placement (at 0815h), and supine bedrest for 30 min, with samples drawn every 15 min. from 0–120 min.
	Laron et al, 1984 [34]	1 mcg/kg synthetic GHRH 1–44 after overnight fast & 30 min. bedrest with samples drawn at 0, 5, 10, 15, 30, 45, 60, 90, 120 min.

Table 2.

Initial studies of GH evaluation in pediatric populations

GH test	Study	Study population	Patient ages (years)	Number of females	Pubertal status	Peak GH in children with suspected GH sufficiency (ng/mL)	Peak GH in children with suspected GHD (ng/mL)	GH assay	GH standard
ITT	Kaplan et al, 1965 [23]	n = 43 2 “normal children” 10 “hypopituitary dwarfism” 21 “severe growth retardation without laboratory evidence of hypopituitarism” 1 CAH 2 GH excess 1 gonadal dysgenesis 3 hyperthyroidism 2 untreated hypothyroidism	not reported	not reported	not reported	not reported	<1	RIA using polyclonal antibodies	Wilhelmi HGH (HS-475A)
	Kaplan et al, 1968 [24]	n=144 81 short stature (HtSDS <−2) - 20 CSS - 22 “primordial dwarfism” (normal growth rate) - 9 “XO gonadal dysgenesis” - 5 “delayed adolescence” - 5 “maternal deprivation” - 9 “psychosocial dwarfism” - 11 miscellaneous 53 hypopituitarism 10 control	1–1/12 to 16–1/12	32 of 81 - 2 of 20 - 10 of 22 - 9 of 9 - 2 of 5 - 1 of 5 - 1 of 9 - 7 of 11 not reported 5 of 10	prepubertal	controls: range 11.0–28.4; mean 12.4 ng/mL CSS: range 3.2–24.3	range <1.0–5.0 (only one 5.0, next highest value 2.2)	chromato-electrophoretic RIA using polyclonal antibodies [104]	Wilhelmi HGH (HS-612A, HS-705)
Arginine	Parker et al, 1967 [25]	n=84 23 short stature - 13 “possible pituitary dwarfs” 49 controls 12 children with DM	1–7/12 to 15 4 to 45 not reported	2 of 23 21 of 49	pre- & post-pubertal (about 1/2 of controls post-pubertal)	range 5.2–85.0; mean 21.0 (95% CI 5.2–85.0 ng/mL) (controls)	n/a	competitive double antibody RIA using polyclonal antibodies [107]	Wilhelmi HGH (HS 612A)
Glucagon	AvRuskin et al, 1968 [26]	n=56 23 “normal children” 5 genetically short children 4 genetically tall girls 12 girls with 45X 8 individuals with hypopituitarism 4 individuals with untreated acquired hypothyroidism	8 to 18 (“normal children”) remainder not reported	13 of 23 remainder not reported	not reported	not reported	not reported	not reported	not reported
	AvRuskin et al, 1971 [27]	n=48 19 children with height within 2SD of mean for age 15 genetically short M with height <−2 SD for age 7 genetically tall F with height >+2 SD for age 7 F with hypopituitarism	means F 10.3±0.5 M 10.6±0.8 8.6±0.8 11.6±0.6 10.0±1.1	9 of 19	prepubertal and pubertal	means 14.3±3.9 in normal F 17.1±3.2 in normal M 19.5±6.8 in short M 27.4±6.6 in tall F	mean 1.1±0.2	competitive double antibody RIA using polyclonal antibodies [108]	not reported (original report of assay used Wilhelmi NIH-GH-HS 612B)
	Weber et al, 1970 [28]	n=45 23 normal weight 12 obese 10 insulin dependent diabetic	3–4/12 to 12–10/12 8–7/12 to 14–9/12 1–10/12–11	14 of 23 (9 of 13a) 5 of 12 (3 of 7a) 7 of 10	not reported	range 5.5–18.8 <0.5–6.3 1.1–9.2	n/a	electrophoretic double antibody RIA using polyclonal antibodies [109]	not reported (original report of assay used Wilhelmi NIH-GH-HS 705)
Levodopa	Hayek and Crawford, 1972 [30]	n=10 7 CSS 3 GHD (arginine or insulin stimulated peak GH <7 ng/ml)	5 to 13 12 to 20	not reported	prepubertal (except 2 GHD patients)	range 3.9–23.0	range 0.5–1.9	competitive double antibody RIA using polyclonal antibodies [108]	not reported (original report of assay used Wilhelmi NIH-GH-HS 612B)
Clonidine	Gil-Ad et al, 1979 [31,32]	n=25 18 CSS 7 GHD (by insulin, arginine or levodopa provocative testing)	6 to 17 3 to 16	6 of 18 3 of 7	not reported	mean 34.4±4.5 range 16.0–80	range (min not reported)-5.5	double-antibody RIA (CAE-Sorin, Saluggia, Italy)	reported as provided by National Pituitary Agency
GHRH	Grossman et al, 1983 [33]	n=10 3 children with GHD 1 adult with GHD 6 "normal subjects”	5 to 15 29 not reported	2 of 3 1 of 1 not reported	not reported	range 15–148	range 12–43	RIA	IRP 66/217
	Laron et al, 1984 [34]	n=11 8 patients with GHD (by insulin, arginine or clonidine provocative testing) 3 CSS	5–10/12 to 20 13–9/12 to 14–5/12	3 of 8 0 of 3	not reported	range 18–100	range 0.5–22	not reported	not reported
Exercise	Buckler, 1972 [41]	n=73 8 health men 65 healthy individuals 4 suspected hypopituitarism	28 to 48 8 to 48 (n=28 >18) 13 to 40	26 of 73	not reported	range 5 to >40 for 8–18y 3 to >40 for >18y	range 2–7	competitive double antibody RIA using polyclonal antibodies [110]	IRP 66/217
	Keenan et al, 1972 [42]	n=27 22 healthy short children 3 features of genetic condition 2 with GHD	3–9/12 to 15–4/12 4–10/12 to 12–9/12 6–6/12 to 12–9/12	5 of 22 1 of 3 1 of 2	pre-pubertal (2 Tanner 2)	range 4.2–42.5	<1.0	competitive double antibody RIA using polyclonal antibodies [107]	Wilhelmi HGH (HS-1147-BC)
ICGH	Plotnick et al, 1975 [19]	n=18 12 healthy children 6 health adults	5–3/12 to 18–5/12 22 to 28	4 of 12 0 of 6	prepubertal (n=7) and pubertal (n=5)	range of 24h ICGH pre-pubertal 4.07–6.50 pubertal 4.75–9.38	n/a	competitive double antibody RIA using polyclonal antibodies [107]	not reported (original report of assay used Wilhelmi NIH-GH-HS 612B)
	Zadik et al, 1990 [49]	n=119 119 children with normal height and weight and growth velocity	7 to 18	16 of 36 T1 26 of 43 T2–3 18 of 40 T4–5	36 T1 42 T2–3 40 T4–5	mean T1 4.4 ±1.2 T2–3 5.5±2.1 T4–5 5.8 ±1.6	n/a	competitive double antibody RIA using polyclonal antibodies [107]	not reported (original report of assay used Wilhelmi NIH-GH-HS 612B)
	Spiliotis et al, 1984 [50]	n=45 7 short children 22 healthy children 16 children with GHD (peak GH <7 ng/ml)	7.4 to 15.15 5.0 to 17.9 3.5 to 20.6	1 of 7 6 of 22 3 of 16	prepubertal and pubertal	range of 24h ICGH 3.1–12.2	range of 24h IC-GH <0.6–2.9	RIA	not reported
	Richards et al, 1987 [51]	n=67 19 normal stature 13 GHD 4 NSD 23 healthy short 8 miscellaneous	means 10.7 ± 0.7 11.2 ± 0.5 11.2 ±1.2 11.3 ± 1.1 12.4 ± 0.9	not reported	not reported	mean 24h ICGH 5.67–6.73 mean 12h ICGH 6.86–8.59	mean 24h IC-GH 1.65 mean 12h IC-GH 1.79	not reported	reported as provided by National Pituitary Agency
	Lanes, 1989 [52]	n=59 10 normal stature 35 short with CDG 7 GHD 7 NSD	6 to 12 7 to 13 means 9.1±3.5 10.3±3.1	2 of 10 9 of 35 1 of 7 1 of 7	prepubertal	overnight ICGH (7–9h) means 4.4±1.8 normal stature 4.2±1.8 CDG	overnight IC-GH (7–9h) mean 1.6±0.2	double antibody RIA (Diagnostic Products Corp, Los Angeles, CA, USA) [107]	not reported

ITT = insulin tolerance test (ie. insulin-induced hypoglycemia); RIA = radioimmunoassay; CAH = congenital adrenal hyperplasia; HtSDS= height standard deviation score; CSS = constitutional short stature; F = female; M = male; ICGH = integrated-concentration of GH; NSD = neurosecretory dysfunction; CDG = constitutional delay of growth;

^aglucagon testing alone

Table 3.

Side effects, specific considerations and limitations for different pharmacologic provocative GH tests

Test	Side Effects	Special Considerations	Limitations
ITT	• hypoglycemia	• secretagogue administered as IV bolus or subcutaneous injection  • must monitor blood glucoses closely and have IV dextrose readily available should severe hypoglycemia develop  • requires physician availability in case the need for resuscitation arises	• generally contraindicated in children <2 years of age or with a history of seizures
Arginine	• facial flushing during infusion  • nausea and vomiting	• secretagogue administered as 30-minute infusion  • extravasation of the arginine infusion burns tissues so IV placement should be verified before starting the infusion
Glucagon	• nausea and vomiting  • late hypoglycemia  • fatigue  • headache	• secretagogue administered as intramuscular or subcutaneous injection  • provokes rise in cortisol  • can be conducted during assessment for hyperinsulinism at the end of a fasting study [103]	• long testing duration of 3–4 hours  • often used in children <2 years of age instead of insulin given relatively safer profile
Levodopa	• nausea and vomiting  • headache  • fatigue  • vertigo	• secretagogue administered as oral medication	• may not be readily available in pediatric centers
Clonidine	• hypotension  • somnolence	• secretagogue administered as oral medication  • test is performed with patient lying recumbent  • vitals are monitored closely throughout the test  • normal saline should be available for infusion to support blood pressure if it drops	• may not be safe in children with some cardiac conditions
GHRH	• flushing	• secretagogue administered as infusion over 1 minute	• cannot detect hypothalamic GHD

Pharmacological GH stimuli

Insulin

In 1963, Roth et al demonstrated that insulin-induced hypoglycemia provoked a rise in GH concentrations in healthy adults [22]. Kaplan et al assessed children’s serum GH response to insulin administration [23]. The absence of a rise in GH concentration after insulin-induced hypoglycemia [i.e. insulin tolerance test (ITT)] in children with “hypopituitary dwarfism” in contrast to the rise observed in children without GHD, was considered confirmatory for GHD. The authors noted that the diagnostic value of the test was limited, however, because some children with normal fasting GH levels (defined as 1–12 ng/ml) still had reduced ITT responses. A larger follow-up study similarly found differing GH responses between children with hypopituitarism and healthy controls; GH response in all other study sub-groups was similar to controls (Table 2) [24]. HGH treatment (2–4 mg 3 times weekly for 6–12 months) led to growth of 9–15 cm/year in 20 patients (100%) with fasting and peak GH concentrations <2 ng/ml, but no change in growth rate in 6 of 8 patients (75%) with peak GH concentration >3 ng/ml (mean 12 ng/ml) [24]. Based on their data, the authors proposed that GH concentrations of 7 ng/ml or more while fasting or after ITT were highly suggestive of normal GH production, while concentrations 3–5 ng/ml were more consistent with possible GHD. They noted an overlap between GH peaks in patients with GHD and non-GHD short stature that precluded identifying a normal response range. Stimulated GH concentrations in 6 of the 20 children (30%) with constitutional short stature (CSS) did not exceed their cutoff of 7 ng/ml. However, they justified this cutoff by stating that 16 patients (100%) identified by their criteria as having isolated GHD responded to hGH treatment, compared with 7 of the 8 patients (88%) with GH levels above the proposed threshold who did not “show a sustained growth acceleration” to hGH treatment.

Arginine

Parker et al compared GH responses to arginine and ITT in children, defining non-responders as those with peak GH levels below the 95% confidence interval (CI) [25]. None of the control subjects who underwent both tests were non-responders. The authors concluded that arginine was useful in diagnosing GHD without the risk of symptomatic hypoglycemia present with the ITT. Responses in patients with diabetes mellitus (DM) similar to those in controls demonstrated that the arginine-stimulated GH response was not insulin mediated. Five of the 23 children (22%) with short stature were non-responders to insulin but not arginine, and 3 of 23 (13%) were non-responders to arginine but not insulin. The authors hypothesized that the two agents may stimulate GH secretion via different mechanisms or that arginine may be more potent in provoking a GH response, but there was no evidence that one test was better than the other.

Glucagon

In the first study evaluating children’s GH response to glucagon, no significant rise in GH concentrations was observed in those with hypopituitarism in contrast to the rise seen in children without hypopituitarism [26]. Glucagon was administered after tolbutamide (an agent that induces hypoglycemia) with no explanation provided for this. A follow-up study replicated these findings, including assessment of 4 normal height boys and 3 short (height −2 SD below the mean) boys without tolbutamide to confirm that glucagon alone stimulated GH secretion [27]. Concurrently, Weber et al assessed peak GH concentration following glucagon administration in hospitalized children undergoing studies of insulin secretion [28]. They observed a rise in GH levels after glucagon administration independent of stress (a known trigger of GH release) and preceding administration of tolbutamide (n=8), arginine (n=1), or saline (n=4). Following isolated glucagon administration, a rise in GH concentration, defined as at least 2 ng/mL above baseline, was seen in 10 of 20 (50%) children suspected to have GH sufficiency, with lower peak GH levels observed in obese children. An increase in GH concentration following intravenous catheter placement (observed in 9 of 20 (45%)) was not considered a response, but rather a result of stress. Similar to arginine, GH rise in children with type 1 DM demonstrated that glucagon’s effect was insulin independent [29].

Levodopa

Following the observation that levodopa stimulated GH release in patients with Parkinson disease, GH response to levodopa was studied in children [30]. The effectiveness of levodopa was shown by the absence of rise in GH concentration in those with GHD (peak 0.5–1.9 ng/mL) compared to a rise in 4 of 7 patients (57%) with CSS (peak 10–23 ng/dL). Of the remaining patients with CSS, 2 had a stress-related rise in GH concentration upon intravenous catheter placement (12.1 and 17.0 ng/dL) and one responded to ITT.

Clonidine

Gil-Ad et al first investigated children’s GH response to clonidine [31,32]. GH concentrations rose in all children with CSS but not in those with GHD, except one child with GHD whose GH level increased from 2.5 to 5.5 ng/ml. The disparate GH responses between the groups, absence of increase in prolactin levels (used as a proxy measure for stress), and lack of correlation between GH response and glucose concentration, were proposed as evidence that clonidine stimulated GH release and could be used to assess for GH sufficiency in children.

GHRH

Grossman et al demonstrated that the administration of synthetic human pancreatic GH releasing factor (hpGRF) 40 [analogous to GH releasing hormone (GHRH)] stimulated GH release in 4 patients with GHD who failed ITT (peak GH concentration 12–43 ng/ml with synthetic hpGRF versus 1–5 ng/ml with ITT) [33]. This suggested a hypothalamic origin to GHD in those patients. Subsequently, the finding that synthetic GHRH 1–44 stimulated a rise in GH concentration to 21 ng/ml and 22 ng/ml in 2 of 8 patients (25%) with GHD, comparable to the peak GH levels of 18 ng/ml, 22 ng/ml, and 100 ng/ml in 3 patients with CSS (and normal ITT), suggested that GHD may be hypothalamic or pituitary in origin. This established a role for synthetic GHRH provocative GH testing [34].

Mechanisms of stimulation

These early studies helped elucidate how each agent provokes GH release through varied mechanisms, which may contribute to the inter-agent differences in provoked peak GH concentrations. In ITT, the insulin-induced hypoglycemia (reduction of about 40–50% in fasting glucose levels and emergence of neuroglycopenia symptoms) leads to GH release via suppression of somatostatin (an inhibitor of GH release) and stimulation of adrenergic receptors (promoters of GH release) [20,35]. Similarly, clonidine-mediated GH secretion occurs via alpha 2-adrenergic receptor agonism that provokes GHRH release and inhibits somatostatin release [35]. Arginine inhibits somatostatin release [35]. Levodopa administration leads to an adrenergic response that indirectly provokes GH secretion by stimulating alpha-adrenergic mediated GHRH release; beta-adrenergic stimulation also occurs, which inhibits GH secretion, an effect that can be blocked by simultaneous administration of a beta-blocker [35]. GHRH administration directly stimulates somatotropes to secrete GH; endogenous somatostatin dampens this effect, but addition of arginine or pyridostigmine blocks somatostatin [20,35]. The mechanism of glucagon-mediated GH release remains unclear; blood glucose fluctuations and stimulation of adrenergic receptors by noradrenaline secretion that occurs following glucagon administration may contribute but are not the primary or sole mediators of the observed GH secretion [36–40].

Physiologic GH tests

Exercise

Buckler measured 8 adult males’ GH responses to various exercise states using bicycle ergometer and found that GH concentrations peak 25–30 minutes after the start of intense exercise [41]. GH levels drawn at 25–30 minutes post-exercise were then assessed in 65 healthy individuals, with adequate response defined as greater than 10 ng/ml [41]. This cutoff was met by 88% of females and 90% of males. This finding, combined with the absence of GH response to both exercise and pharmacological stimulus in patients suspected to have GHD, supported the authors’ conclusion that exercise testing could be used to demonstrate GH sufficiency.

In the first dedicated pediatric study, subjects were instructed to walk briskly for 15 minutes and then walk rapidly or run up and down stairs for 5 minutes [42]. After the exercise, GH concentrations exceeded 7 ng/ml in 20 of 25 children without known GHD (80%) but neither of the 2 patients (0%) with known GHD. The authors stated that the similarity between GH response to exercise and to arginine and clonidine supported the utility of exercise-stimulated GH testing, but recognized that the inability to regulate the amount of effort that each child exerted could limit its use. Later protocols used bicycle ergometers or treadmills to allow for standardized incremental workloads, with heart rates measured to confirm exertion, but the test still performed poorly [43,44]. Despite standardized workload, 20 of 30 normally growing children (67%) failed to exhibit post-exercise GH concentrations >10 ng/ml [43]. Exercise-induced GH peaks of 168 “normal children” and 25 children with GHD (grouped based on peak GH concentration ≥10 and <10 ng/mL respectively on arginine-ITT testing; ages 2–17 years, 64 females) were later compared to define expected error in GH response [45]. Peak GH concentrations between the two groups overlapped, with 20 children (12%) in the normal group failing to reach GH concentrations of at least 10 ng/mL and 3 children (12%) in the GHD group “passing” with peak GH concentrations >10 ng/mL after exercise [45]. The combined challenges of obtaining the cooperation of younger children, achieving optimal workload, individual variability in GH response to exercise (which is affected by the degree of one’s physical fitness [41]), and a low positive predictive value, caused this test to be deemed unreliable [20,43,46,47].

Secretory profiles

Plotnick et al examined 24-hour GH secretion in healthy children and young adults instructed to engage in their normal activity [19]. Blood was collected every 30 minutes from a withdrawal pump to derive an integrated concentration of GH (ICGH) [19,48]. They identified a pattern of GH secretion wherein baseline secretion was augmented by secretory peaks that occurred throughout the day, with larger peaks at night. Higher levels were observed in children than in adults, with mean ICGH in prepubertal children, pubertal children, and adults of 5.10±0.96 ng/ml, 7.37±1.70, and 3.47±1.67, respectively. A larger pediatric-specific study assessed ICGH by the same technique and evaluated for variation based on pubertal status [49]. A statically significant increase in mean ICGH levels with increasing Tanner stage was observed. A trend of higher ICGH in females than males was not statistically significant.

Spiliotis et al then found that ICGH in children with height <1st percentile and growth velocity <4 cm/year, but peak GH >10 ng/ml on provocative GH testing, demonstrated lower mean ICGH compared to control children (both short and normal stature) [50]. Mean peak pulse amplitude and mean number of pulses for these children were between those of children with GHD and controls. In 6 of 7 of these children (86%) with discordant results between ICGH and provocative testing, GH treatment resulted in growth velocity increases comparable to those of patients with GHD. The authors suggested that these children had GH neurosecretory dysfunction (NSD). Although all the patients with discordant results were prepubertal with bone age (BA) delayed by at least 2.3 years, the authors argued that they were distinct from those with constitutional delay of growth (CDG) who failed provocative testing and were thought to have partial GHD. The authors simultaneously cautioned that these tests can be misleading because one third of those with GHD diagnoses had peak GH concentrations >10 ng/ml during sleep. A greater interest in assessing spontaneous GH secretion, particularly during sleep, when GH peaks are highest, followed.

Richards et al found that overnight 12-hour ICGH measurements correlated closely with 24-hour ICGH measurements (correlation coefficient 0.963, p<0.001) in children with both normal and abnormal GH status [51]. Lanes later reported low overnight GH levels (ICGH <3 ng/ml) in 12 of 45 (25%) of normally growing children, which overlapped with the values seen in patients with NSD or GHD [52]. Donaldson et al, demonstrated significant night to night variation in overnight ICGH [53]. These results and other similar findings led to questions about the utility of overnight ICGH [15,47,53]. Poor reproducibility in the same individual and high burden made the test undesirable; in the United States, it largely disappeared from clinical use once idiopathic short stature became an approved indication for rhGH, obviating the need to diagnose NSD.

Key takeaways

Regardless of the provocative test used, overlap between the profiles of children with known GHD and children suspected to have GH sufficiency precluded defining “normal” from “abnormal” thresholds, a challenge that remains today [15,18]. Differences in measurement methodologies (Table 2) undoubtedly contributed to the wide range of GH concentrations observed in suspected GH-sufficient children. Further complicating the ability to define reference ranges for provocative GH tests, each provocative agent produced different peak GH concentrations, pointing to a possible need for agent-specific GH norms.

An adequate response to one type of provocative GH test, but not another, may be due to differential functioning of the various pathways that regulate GH secretion. Because GHD may be hypothalamic or pituitary in origin, it is important to consider that a child with GHD that is hypothalamic in origin can retain normal somatotropes and may therefore falsely “pass” provocative GH testing. However, they may have reduced endogenous secretion, as seen in NSD. Additionally, the response to a particular agent may be limited by endogenous negative-feedback pathways and/or the GH reserve in somatotropes at the time of testing [20,35,54]. The observation in 3 of 10 of these early studies that peak GH levels were achieved after intravenous catheter placement and before stimulus administration, points to the stress of venipuncture as one mechanism by which GH reserves may be depleted at the time of testing [24,29,30]. Indeed, a recent study found that 11 of 76 children (14%) being assessed for GHD demonstrated adequate GH concentration of >7 ng/mL upon intravenous catheter placement and not with ITT performed 30 minutes later [36].

Despite the acknowledged “continuity between absent and normal GH secretion during childhood” [47] demonstrated over 50 years ago, arbitrary cutoffs have long guided pediatric endocrinologists evaluating GH function [20,55]. In these early studies (excluding those of GHRH), children with a pre-test diagnosis of GHD based on structural pituitary abnormalities or primary hypopituitarism mostly demonstrated peak GH levels less than 2–3 ng/mL. This is reflected in the practice of using this cutoff to define severe GHD. However, given the evolution in GH assays and standards, these values cannot be translated directly to contemporary practice other than to conclude that “very low peak GH levels” are indicative of GHD, as per recent guidelines [15]. After the introduction of rhGH made GH treatment more widely available, higher provoked GH thresholds to define GHD were accepted by consensus on what constituted GH sufficiency (Figure 1A) [47,56,57]. The rising bar for defining GH sufficiency partly reflected an attempt by clinical trials to capture even mild GHD, which could be more readily treated after rhGH resolved the GH scarcity of the pituitary GH era [15,55]. As shown, available data for healthy children’s GH response to provocative tests poorly supports use of these thresholds, particularly given that modern cutoffs do not adequately reflect the fact that newer GH assays yield lower GH concentrations [15].

Studies to establish reference ranges

Zadik et al evaluated healthy children’s GH responses to clonidine, insulin and arginine, alongside 24-hour ICGH [49] (Table 4). Taking the higher peak GH concentration if two tests were performed, the 5^th percentile for peak provoked GH concentration was found to be 10 ng/ml, driven by the higher concentrations seen in response to clonidine. Statistically significant pubertal and sex differences were observed with clonidine testing: mean peak GH concentrations increased with advancing pubertal status and were higher in prepubertal males than females and in Tanner 4–5 females than males. While the authors proposed using paired stimulation and a cutoff of 10 ng/ml to define GH sufficiency, the appropriateness of this commonly used cutoff should be questioned based on the variability in peak GH concentrations with stimulus, sex, and pubertal status.

Table 4.

Studies of provocative GH tests that sought to establish normative data for peak provoked GH levels

Study	Provocative tests	GH assay & standard	GH peak (ng/ml)
Zadik et al, 1990 [49] children with normal weight, height and growth velocity n = 119 ages 7–18y	• clonidine (0.15 mg/m2 orally), n=66 • arginine (0.5g/kg IV), n=23 • insulin (0.1units/kg IV), n=19 n=29 only clonidine n=18 insulin and clonidine (testing done on separate days) n=22 arginine and clonidine (testing done on separate days) n=2 insulin or arginine only	competitive double antibody RIA using polyclonal antibodies [107] standard not reported (original report of assay used Wilhelmi standard NIH-GH-HS 612B)	mean±SD [range] | 5th %ile • clonidine 21.0±10.7 [3.0–43.0] | 4.5 T1  M 16.9±6.7 (n=12)  F 12.8±5.1 (n=12) T2–3  M 17.8±7.7 (n=10)  F 18.3±9.0 (n=10) T4–5  M 26.5±12.0 (n=10)  F 35.5±5.1 (n=11) • arginine 13.1±6.1 [2.6–27.0] | 4 • insulin 14.2±6.3 [3.6–27.0] | 3.6
Rochioccioli et al, 1993 [58] children with height <−2 SD, slowed growth velocity, delayed BA, and absence of known etiology n = 3143 ages 3–16.5y	• arginine (0.5/kg IV), n=625 • clonidine (25 pg/m2 orally), n=330 • insulin (0.1 units/kg IV), n=198 • ornithine (12g/m2 IV), n=162 • insulin + arginine (doses as above), n=203 • clonidine + betaxolol (dose as above + 0.25 mg/kg orally), n=2003 • levodopa (125 mg for body weight (BW) <15 kg, 250 mg for BW 15–30 kg, 500 mg for BW > 30 kg), n=685 • glucagon + propranolol (1mg IM glucagon 2h after 1 mg/kg oral propranolol), n=442 • glucagon + betaxolol (doses as above), n=815	double antibody RIA (CIS-ORIS, Gif-sur-Yvette, France) standard not reported	mean | 5th %ile | %<5 ng/ml | %<10 ng/ml • arginine: 10.2 | 0.2 | 45 | 69 • clonidine: 11.5 | 0.5 | 37 | 64 • insulin: 11.8 | 0.5 | 32 | 65 • ornithine: 14.2 | 0.8 | 25 | 47 • insulin+arginine: 14.3 | 0.6 | 27 | 53 • clonidine+betaxolol: 15.7 | 0.8 | 23 | 48 • levodopa: 19.3 | 1.8 | 13 | 31 • glucagon+propranolol: 20.8 | 2.1 | 16 | 29 • glucagon+betaxolol: 21 | 2.8 | 10 | 29
Ghigo et al, 1996 [59] children with height −3.7 to +1.2 SD, growing well with normal BMI, IGF-I levels, and concordance between BA and CA n = 472 ages 3.8–17.4y	• exercise (bicycle ergometer), n=33 • insulin (actrapid 0.1units/kg IV), n=59 • arginine (0.5 mg/kg IV), n=79 • clonidine (150 mg orally), n=69 • levodopa (125 mg BW <15 kg, 250 mg BW 15–30 kg, 500 mg BW > 30 kg), n=55 • glucagon (1mg IM), n=40 • pyridostigmine (PD) (60 mg orally), n=53 • GHRH (1 mcg/kg IV), n=134 • GHRH + PD (doses as above, PD 60 min before GHRH), n=94 • GHRH + arginine (doses as above), n=81 • n=461 M, 236 F	non-competitive immunoradiometric assay using monoclonal antibodya (HGH-CPK, Sorin, Saluggia, Italy) standard not reported	mean±SD [range] | %<7 ng/ml | %<10 ng/ml • exercise: 12.7±1.1 [3.0–28.3] | 15.1 | 36.4 • insulin: 13.2±1.2 [2.7–46.4] | 23.7 | 49.1 • arginine: 16.7±1.2 [0.5–48.4] | 12.6 | 32.9 • clonidine: 13.1±1.8 [3.8–86.0] | 10.1 | 23.2 • levodopa: 13.0±1.1 [1.9–40.0] | 23.6 | 36.4 • glucagon: 16.9±1.9 [1.9–49.5] | 10.0 | 35.0 • PD: 13.5±1.10 [2.5–35.0] | 15.1 | 35.8 • GHRH: 28.8±2.0 [2.7–102.7] | 8.9 | 14.9 • GHRH+PD: 47.8 ±1.9 [19.6–106.0] | 0 | 0 • GHRH+arginine: 61.8±2.8 [19.4–120.0] | 0 | 0

BW=body weight; M=males; F=females; BA=bone age; CA=chronological age;

^anon-competitive immunoradiometric assay is more sensitive than the competitive RIA

A large-scale retrospective study subsequently compared 9 provocative GH tests in children with short stature (5473 tests) [58]. Distribution of patient characteristics across tests was reportedly not different based on age, sex, and BA; 2330 (74%) underwent 2 different tests and the remainder 1 test. There were significant differences in maximum and mean peak GH concentrations among all tests studied (Table 4). Mean peak GH concentration could vary two-fold between two tests. The proportion of children meeting commonly used cutoffs for GHD differed by as much as 30% between tests. Even with combined insulin and arginine provocative testing, 27% of peak GH concentrations were less than 5 ng/ml, compared to 32% with insulin alone and 45% with arginine alone. No difference in mean peak GH concentrations was found based on BA, used as a surrogate for pubertal status. Whether or not there were sex differences in peak GH concentrations was not reported.

Prospective assessment of GH response to one or more provocative tests in normally growing children was then conducted due to persistence of inadequate reference data [59]. Only provocative tests using GHRH produced statistically different (higher) mean peak GH concentrations than other tests (Table 4). Among the 78 children who underwent two tests, the proportion of peak GH levels <7 ng/ml and <10 ng/ml were 3% and 10%, respectively. This excluded testing with GHRH combined with arginine or pyridostigmine, for which no peak GH concentration was <7 ng/ml. The only significant difference in peak GH concentration based on pubertal status was seen with arginine: prepubertal 14.8±1.4 ng/ml versus pubertal 19.2±2.1 ng/ml. No sex differences were detected. The authors concluded that, aside from testing using GHRH combined with either arginine or pyridostigmine, GH provocative tests are “unsuitable” to discriminate between children with and without GHD due to the extensive variability in responses. Supporting this statement was the fact that the lowest peak GH levels in response to individual tests (excluding GHRH combined with arginine or pyridostigmine) in their group of normally growing children ranged from 0.5–3.8 ng/ml (mean 2.1 ng/ml).

The variation in peak GH concentrations across these studies is highlighted in Table 4. The proportion of normally growing children who could be falsely identified as having GHD on the basis of testing alone in the studies of normally growing children using commonly assigned cutoffs is striking [49,59]. The inter-study variation underscores the potential effects of using different assays and standards. The intra-study variability speaks to the importance of recognizing the lack of comparability among different agents, and conceivable response differences depending on sex and pubertal status.

Effects of patient factors on response to provocative GH tests

The aforementioned studies did not adequately assess differences related to sex, pubertal status, age, and BMI. Many included more males than females, did not report on or assess the impact of pubertal status, and either included only normal weight children or did not comment on BMI. Subsequent investigations have failed to resolve these issues with the exception that adults are no longer included in pediatric studies now that differences in GH secretion between these two populations are well established [18,47].

Sex hormone effects

The fact that GH concentrations increase with sex steroid administration was known as early as 1966 [60–63]. Normalization of suboptimal peak GH concentrations in short children (most with delayed BA) after puberty or with sex steroid priming was first demonstrated in both boys and girls in 1979 [64], although it had been observed in boys as early as 1968 [65]. After this was repeatedly demonstrated in short children [66–68], the effect of sex steroid priming was assessed in healthy children without short stature (weight within +2 SD for age and normal BA for chronological age) [44]. Response to exercise on a treadmill (day 1) and arginine-ITT (day 2; 0.5 g/kg arginine IV over 30 minutes followed by 0.1 mcg/kg regular insulin intravenously at 60 minutes) was assessed in 88 subjects (ages 4–20 years, 41 females). GH levels were measured by polyclonal radioimmunoassay (Hazelton Biotechnologies Vienna, VA, USA; standard not reported). Mean peak GH concentrations increased with pubertal stage. Using a cutoff of 7 ng/ml for GHD, 11 of 18 (61%) of the prepubertal subjects (without estrogen priming) met criteria for GHD, compared to 7 of 16 (44%) for Tanner stage 2, 2 of 18 (11%) for Tanner stage 3 and 0 of 32 (0%) for Tanner stages 4 and 5. Peak GH concentrations also rose with advancing BA. A subset of prepubertal children (5 females 8.0±2.3 years, 6 males 10.0±1.7 years) underwent repeat testing 6 weeks apart, first without priming and then after 2 days of oral ethinyl estradiol (40 ug/m² divided TID with meals). Estrogen administration led to increased peak GH concentration: range 1.9–20.3 ng/ml without versus 7.2–40.5 ng/ml with priming. The investigators interpreted the rise in GH concentrations with advancing pubertal state and the high rates of apparent GHD without sex steroid priming compared to with priming, as evidence that children with Tanner stage <4 should undergo sex steroid primed testing to minimize the risk of false positive results. However, it remained to be shown that these unprimed lower peak stimulated GH levels were not consistent with functional deficiency.

At the time of publication of the above study, some children with CDG were known to have lower GH levels than those without delay, with normalization of GH levels after puberty [69,70,71,72]. Subsequently, 50 boys aged 10.1–16.3 years (mean 13.3±1.4 years) with CDG and suboptimal provoked GH peak (<10 ng/ml) without sex hormone priming, but normal peak after priming, were followed without GH treatment until they reached their adult heights [73]. Adult heights reached or exceeded predicted and mid-parental heights in all but 3 of 50 boys (6%). While other studies demonstrated that sex hormone priming led to fewer inadequate responses on provocative GH testing [74], this study provided longitudinal evidence to bolster prior studies supporting priming. Accordingly, it was proposed that sex steroid priming is important for distinguishing between CDG and GHD in prepubertal children of pubertal age [15,18]. A similar study assessing girls has not been performed; results have been extrapolated to girls [15]. The use of sex steroid priming outside of this context remains controversial. Those who oppose the practice argue that it may lead to false negative results by augmenting the endogenous GH response in a non-physiologic manner [62,75,72]. When used, wide variation in how priming is implemented exists [18,75]. Martinez et al proposed different cutoffs based on whether or not provocative GH testing was preceded by estrogen priming [76], but this suggestion has not been adopted in society guidelines.

Obesity effects

Blunted response to provocative GH testing in obese children was first reported in the 1970s [77]. Subsequent studies found an inverse relationship between BMI and GH levels [78–81] and showed that GH concentrations in obese children increase with weight reduction [82]. BMI-adjusted provoked GH cutoffs have not been established in children, even though there is recognition that obese children may be misdiagnosed with GHD given their blunted responses [83,84]. A recent effort to re-examine this issue retrospectively investigated the relationship between BMI and (unprimed) peak GH response to ITT (0.1 units/kg) and either clonidine (0.125 mg/m²) or levodopa (500 mg for body weight (BW) >30 kg; 250 mg for BW 15–30 kg; 125 mg for BW <15 kg) in prepubertal children (ages 3–11 years) with idiopathic GHD (peak GH <5 ng/mL; n= 114, 9 obese), idiopathic partial GHD (peak GH 5–10 ng/ml; n=325, 17 obese) or organic GHD (GHD secondary to pituitary damage; n=21, 2 obese) [84]. Peak GH concentration decreased by 1.06 ng/ml for every 1 kg/m² increase in BMI among children with complete or partial idiopathic GHD (GH assay not reported). A statistically significant difference was seen in clonidine test results, with mean peak GH of 0.9 ng/mL in children with BMI >95th percentile for age compared to 5.5 ng/mL in individuals with BMI between the 5th and 95th percentile for age. Receiver operating characteristic curve analysis resulted in suggested cutoffs for diagnosing GHD in obese children of 5.57 ng/ml overall, or 8.75 ng/ml for partial idiopathic GHD, 1.78 ng/ml for complete idiopathic GHD and 1.7 ng/ml for organic GHD. Despite these data indicating that thresholds of 7 or 10 ng/mL for GH sufficiency in obese children are inappropriate, adjustment for BMI in interpreting provocative GH test results is not commonly performed.

Other modulators of peak GH levels

Endogenous factors that regulate the GH-IGF-I axis may impact GH response to provocative testing. Stimuli for GH secretion include GHRH, ghrelin (high during fasting and low after feeding) and stress [8,55]. Inhibitory signals for GH secretion include somatostatin, IGF-I (via negative feedback on somatotrophs and the hypothalamus), free fatty acids, and leptin (via somatostatin) [55]. Conditions that lower hepatic IGF-I production, such as malnutrition or Laron syndrome (GH insensitivity due to mutations in the GH receptor), reduce the IGF-I negative feedback, leading to higher baseline GH levels. Chronic under-nutrition can be associated with reduced GH secretion [85]. Hypothyroidism and Cushing’s disease are associated with blunted response to provocative GH testing [8,47]. Exogenous glucocorticoids and chronic inflammation also influence regulation of GH secretion [8,47,86]. Finally, variations in GH levels according to child age (independent of pubertal status) have been reported [24,47].

Future directions

Despite attempts to establish best practices for execution and interpretation of provocative GH tests [15,18], these tests poorly differentiate GHD from GH sufficiency, which continues to complicate their clinical application. Given the multiple patient and test characteristics that affect results and the high frequency of low peak GH concentrations on testing normal children, requiring an ‘inadequate’ response to two provocative tests to diagnose GHD was proposed as early as 1967 and accepted as standard by 1974 [8]. Similarly, testing after an overnight fast to avoid glucose suppression of GH secretion has been recommended and implemented since at least 1999 [87]. These measures have done little to account for the many factors that contribute to the variability of provoked GH concentrations (Figure 1B). Adjusted norms to account for variation in GH response to provocative testing based on sex, age, puberty, and BMI are not available. In the 55 years since their introduction, limited progress has been made in improving the clinical application of provocative GH tests due to the multiplicity of factors that determine test results.

Identification of new GH secretagogues that enable provocative GH testing with better sensitivity and specificity are always hoped for, but this has not yet been realized. Macimorelin acetate, a synthetic oral ghrelin receptor agonist, was approved in Europe and the U.S. for provocative GH testing in adults, after testing demonstrated high sensitivity and specificity with better reproducibility than other secretagogues [88]. It remains unclear how children would respond to this agent as studies have not been performed [86]. GH-releasing peptide 2 (GHRP2) is a newer GH secretagogue that has been evaluated for use in children and has been shown in adults to provide the added benefit of allowing for assessment of the hypothalamic-pituitary-adrenal axis [89,90]. However, a pediatric study of this agent demonstrated overlap in peak GH concentrations between GHD and GH sufficient children, similar to existing agents [89].

Perhaps it is time to finally move away from provocative GH tests and focus on new diagnostic approaches. Discovery of various genetic etiologies of GHD promises to make genetic testing increasingly useful in diagnosing GHD [91]. Additionally, the amino-terminal propeptide of C-type natriuretic peptide, a growth plate paracrine regulator, and collagen X biomarker, a degradation product of endochondral ossification, have been shown to predict height velocity and could become useful biomarkers to assess for GH sufficiency [92–94].

In the meantime, as consensus guidelines have stated, the use of a uniform threshold for clinical interpretation of provocative GH test results is not justified by the available data [15,57]. Indeed the 2016 guidelines published by the Pediatric Endocrine Society state: “By modern immunometric methods and standards, 10 [ng/mL] is just below the mean response obtained to most provocative tests in normally growing children, whose 5^th percentile lies below 5 [ng/mL] for most tests” [15]. If provocative GH tests are going to continue to be used, agent- and assay-specific reference ranges cannot be ignored. The need for standardization of GH assays remains critical [13,15,18,95]. Use of antibody-independent mass spectrometric methods is a promising approach for more precise measurement of GH concentrations [1,96]. Studies to establish normative ranges using contemporary assays are required. These will have to be more comprehensive than prior studies, with study designs that allow for detection of differences based on BMI, sex, age, and pubertal status.

While provocative GH tests continue to be used in the absence of new diagnostic tools, immediate practical improvements in the clinical application of provocative GH testing can be achieved. The importance of thoroughly assessing auxologic data in the evaluation of short stature, especially changes in growth over time with attention to the family’s growth and puberty patterns, cannot be overemphasized [18,86,97]. Sitting height to standing height ratio reference charts, such as those newly developed for the pediatric population in the United States, can be used to screen for disproportion due to skeletal dysplasia [97,98]. Once non-endocrine and other endocrine causes of short stature have been excluded, auxologic data can be combined with other available data, such as IGF-I levels and bone age, to guide the need for provocative GH testing [15,18,86,99]. It is worth noting that some children with hypopituitarism maintain normal growth velocity despite obvious GDH, often in the setting of having undergone resection of tumors found in the hypothalamic-pituitary region, but also in some congenital conditions associated with pituitary dysfunction [100]. As such growth velocity is insufficient to clarify GH status in certain pediatric populations. Clinical decision rules can be applied to use clinical and laboratory data to inform the need for provocative GH testing and decrease the likelihood of obtaining false positive results [101]. A validation study of a clinical decision rule under which provocative GH testing was reserved for only children with growth velocity <−1 SD and IGF-I level <−2 SD demonstrated that application of the rule would lead to a two-thirds reduction in unnecessary provocative GH tests [102]. Implementation of protocols designed to detect a GH response to the stress of venipuncture should also be introduced to identify children who have sufficient GH concentrations in response to venipuncture stress, but not the testing agent [36]. Tests could be standardized to minimize delay in GH sampling after venipuncture (or start sampling if there is a delay in stimulus administration) to improve test performance. In the absence of adequate reference data, rather than relying on cutoffs that are ill-supported by available data, clinicians can consider patient features (i.e. pubertal status and BMI), as well as characteristics of GH assays used, to interpret test results alongside other clinical data.

Conclusion

The poor sensitivity and specificity of provocative GH tests noted since the 1960s have not improved. Complex data on the performance of these tests have been over-simplified to derive diagnostic cutoffs that can be used in clinical care. This undoubtedly results in misdiagnosis and likely over-treatment. When interpreting test results, clinicians should consider the various factors that influence circulating and provoked GH levels as applies to the individual patient, as well as issues related to the measurement of GH concentrations. Ultimately, new diagnostic modalities are needed to improve diagnosis of GHD.

Abbreviations

BA

bone age

BW

body weight

CDG

constitutional delay of growth

CI

confidence interval

CSS

constitutional short stature

DM

diabetes mellitus

GH

growth hormone

GHD

growth hormone deficiency

GHRH

growth hormone releasing hormone

hGH

human growth hormone

hpGRF

human pancreatic growth hormone releasing factor

ICGH

integrated concentration of growth hormone

IRP

International Reference Preparation

IS

International Standard

ITT

insulin tolerance test

NSD

neurosecretory dysfunction

rhGH

recombinant human growth hormone

WHO

World Health Organization