GET score assessment and definition
GET score items were selected according to evidence available in the literature [
2,
3], and their weighting was defined arbitrarily following extensive discussion in the study group. The GET score was designed to cover a range between 0 and 100 points, composed from clinically measurable parameters. It was intended that GH-untreated patients with AGHD should be positioned approximately in the middle of the range (with a mean of ~ 50 points and a standard deviation [SD] of ~ 20 points) and that the range should allow the measurement of treatment effects. Fifty percent of the GET score points were generated from HRQoL parameters and 50% from physical measurements of somatic parameters.
The Short-Form Health Survey 36 (SF-36), one of the most commonly used generic instruments for measuring HRQoL, covers eight HRQoL elements assessing physical and psychological health [
6]. Previous research has established the relationship between the EuroQol five dimensions questionnaire (EQ-5D), a generic five-item instrument providing a simple descriptive profile and a single index value for health status, and a tool used to measure HRQoL in patients with AGHD [
7,
8]. An increase in the SF-36 and EQ-5D visual analogue scale (VAS) score reflects an improvement in self-perceived health. The SF-36 and the VAS component of EQ-5D (EQ-5D-VAS) have both been used previously in patients with AGHD [
7,
9]. The QoL-Assessment of GHD in Adults (QoL-AGHDA), a disease-specific, need-based measure [
10], developed based on in-depth interviews with adult patients with GHD is also a recognised measure for the assessment of QoL. However, restricted licence use did not permit use of this tool in our study.
The HRQoL parameters of the GET score comprised the SF-36 score [
7] (20 points) and the EQ-5D-VAS (20 points), together with the disease-related days off work (10 points).
Details on the allocation of the GET score points from SF-36 and EQ-5D-VAS are given in Additional file
1: Table S1. As the SF-36 covers eight dimensions, the arithmetic mean of the score points from each dimension was taken and included into the GET score (an example is shown in Additional file
1: Table S2). Based on data from Saller et al., [
11] > 30 disease-related days off work during the previous 6 months generated a score of 0 points, and < 4 disease-related days off work generated a score of 10 points (Additional file
1: Table S1).
The somatic parameters comprised bone mineral density (BMD) (20 points), waist circumference (10 points), low-density lipoprotein cholesterol (LDL-C) (10 points), and body fat mass (10 points). Details on the allocation of the GET score points for the somatic parameters are provided in Additional file
1: Table S3.
Dual-energy X-ray absorptiometry (DXA) is the gold standard for BMD measurement [
3,
12]. As patients’ ages spanned more than five decades, the z-score was selected as the most suitable parameter for measuring BMD. The most pronounced effect of GH substitution on BMD is detectable at the lumbar spine [
13], hence this was the measuring site for the GET score. Based on published data [
14], DXA BMD lumbar spine z-score ≤ −2 was assigned a score of 0 points, and a z-score ≥ 0 was assigned a score of 20 points.
Waist circumference reflects visceral fat accumulation and is established as a key criterion for the diagnosis of metabolic syndrome and as an independent cardiovascular risk factor [
15]. When including this parameter in the GET score, individual variance, risk threshold, and published data from patients with AGHD with rather small therapeutic effects had to be considered [
16,
17]. Therefore, waist circumference ≥99 cm in females / ≥113 cm in males scored 0 points, and waist circumference ≤ 80 cm in females / ≤94 cm in males scored 10 points.
Based on the baseline values and the therapeutic effects of GH substitution on LDL-C in patients with AGHD [
18,
19], LDL-C ≥ 3.98 mmol/L (154 mg/dL) scored 0 points, and ≤2.59 mmol/L (100 mg/dL) scored 10 points.
Using a Tanita scale, body fat mass can be assessed with body impedance analysis. Based on data from Rosenfalck et al. [
20], body fat mass percentage ≥ 44.1% scored 0 points, and ≤21.5% scored 10 points.
To calculate a GET score, the first step is to calculate the overall SF-36 GET score points by taking the average of all eight SF-36 GET score points based on the transformed SF-36 domain scores (Additional file
1: Table S1). The second step is to add the GET score points for the remaining HRQoL parameters – EQ-5D-VAS and disease-related days off work (Additional file
1: Table S1). The third step is to look up the GET score points for the somatic parameters using the GET score points as shown in Additional file
1: Table S3. The addition of all components sums up to the final GET score. An example of a calculation of GET score is provided in Additional file
1: Table S4. If individual parameters are missing, the score is calculated without these parameters, but adjusted accordingly (Additional file
1: Table S5). For example, BMD has a weighting of 20%; the maximum score achievable without BMD would be 80. If a patient achieved a determined score of 67 without BMD, adjustment of the determined score would be 67/80*100, resulting in a final GET score of 83.75 (Additional file
1: Table S5). A minimum number of parameters giving a total weighting of ≥70% is required to determine the adjusted GET score, otherwise the GET score is set to missing.
Proof of concept study
Study design
GH-treatment-naïve patients with AGHD, defined according to GH Research Society criteria [
3], under the care of endocrinologists, were enrolled into a prospective, observational, non-interventional, multicentre proof of concept study.
The indication and clinical decisions regarding GH replacement (Norditropin® [somatropin, recombinant human GH], Novo Nordisk A/S, Denmark) were made by the treating physician according to usual clinical practice. GH-treated patients were compared with patients in whom no treatment was initiated; the decision not to initiate GH replacement was taken jointly by the patient and the physician. The study was performed in accordance with the Declaration of Helsinki [
21]. Ethical permissions were obtained from the Ethical Commission of the Chamber of Physicians of the German Federal State of Hessia. Informed consent was obtained from all study participants.
The study recruitment period was originally planned for 24 months, but extended to 36 months due to limited recruitment. Participation commenced at visit 1, when baseline data were collected and GH treatment was initiated in the treatment group. Interim follow-up visits (visits 2–4) were planned for approximately every 6 months, but occurred at varying intervals, and the participants’ involvement concluded at visit 5. If the patient prematurely discontinued participation, the last interim visit became the final visit. Duration of follow-up was calculated as days between first and last visit.
The inclusion criteria for data analysis were availability of baseline demographic data (gender, date of birth), information about GH therapy for treated patients and at least one of four follow-up visits.
The GET score was calculated, and if there were too few parameters to provide a total weighting of ≥70%, the GET score was set to missing. In the proof of concept study, IGF-I concentrations were measured mainly as a parameter for plausibility, verifying whether GH had or had not been administered. IGF-I was assessed centrally using the iSYS automated chemiluminescent IGF-I assay (Immunodiagnostic Systems Ltd., Boldon, UK). The assay employs two monoclonal antibodies and is calibrated against WHO International Standard 02/254 (National Institute for Biological Standards and Control, Hertfordshire, UK) [
22].
Statistical analysis
In observational studies, clinical practice is reflected in missing values, missing visits and fewer untreated controls than treated patients, thereby providing unbalanced data; therefore, a repeated measures model was found to be the most appropriate method to analyse the available data. The study sample size was determined by the ability to recruit patients within the study period. By using the repeated measures multiple regression model for the GET score analysis, correlation of data within the individual patient were taken into account when patients were observed at several visits over time within the study period. Any overall differences between the mean GET score of the two groups in the full study period could be detected. Due to the ageing of the patients over the study period, deterioration over time could potentially occur in the parameters included in the GET score, therefore untreated controls versus treated patients were evaluated.
The model included treatment group (control and treated), visit, and the interaction term between visit and treatment as explanatory variables. Gender, age and treatment duration were also included in the model to adjust for potential differences in patient characteristics in the two groups. The overall difference in GET score between control and treated groups in the full study period was estimated by least squares means (LSM). Missing data were handled by the repeated measures model when evaluating the GET score and were considered missing completely at random. Descriptive statistics were applied for all parameters and data are presented as mean ± SD, unless otherwise stated. Statistical analysis was performed using SAS v9.4 (SAS Institute Inc., Cary, North Carolina, USA).