Previously, loss of pituitary function caused by traumatic

Transcription

1 ORIGINAL ARTICLE Endocrine Care Prevalence of Posttraumatic Growth Hormone Deficiency Is Highly Dependent on the Diagnostic Set-up: Results From The Danish National Study on Posttraumatic Hypopituitarism Marianne Klose, Kirstine Stochholm, Jourgita Janukonyté, Louise Lehman Christensen, Jan Frystyk, Marianne Andersen, Peter Laurberg, Jens Sandahl Christiansen, and Ulla Feldt-Rasmussen Department of Medical Endocrinology (M.K., U.F.-R.), Copenhagen University Hospital, Rigshospitalet, DK-2100 Copenhagen, Denmark; Department of Internal Medicine and Endocrinology (K.S., J.J., J.F., J.S.C.), Aarhus University Hospital, DK-8000 Aarhus, Denmark; Department of Medical Endocrinology (L.L.C., M.A.), Odense University Hospital, DK-5000 Odense, Denmark; and Department of Medical Endocrinology (P.L.), Aalborg University Hospital, DK-9100 Aalborg, Denmark Context: Recent international guidelines suggest pituitary screening in patients with moderate and severe traumatic brain injury (TBI). Predominantly isolated GH deficiency (GHD) was reported in the literature, raising the question of potential methodological bias. Objective: Our objective was to assess the prevalence of GHD in patients admitted in 2008 with TBI, with concurrent assessment of methodological bias. Design and Setting: We conducted a nationwide population-based cohort study at tertiary referral university hospitals. Participants: Participants were Danish patients with a head trauma diagnosis from the Danish Board of Health diagnostic code registry; 439 patients and 124 healthy controls underwent dynamic assessment of GH secretion 2.5 years (median) after TBI. Main Outcome: We evaluated the prevalence of GHD given use of 1) local versus guideline cutoffs, 2) insulin tolerance test (ITT), pyridostigmine (PD)-GHRH or GHRH-arginine (arg) test, 3) single versus repeated testing, and 4) GH assessment by assays with different isoform specificities. Results: The prevalence of GHD was lower by local than by guideline cutoffs (12% vs 19% [PD- GHRH/GHRH-arg, P.001]; 4.5% vs 5% [ITT, P.9]), and by ITT than by PD-GHRH/GHRH-arg (P.006 [local cutoffs]; P.001 [guideline cutoffs]). Only 1% of patients had GHD according to 2 tests. GH assessment by the Immulite or isys assay caused no significant diagnostic differences. Conclusions: The study confirmed a high risk of bias in the management of pituitary testing of patients with TBI and stresses the importance of a proper control group and stringent GH testing including confirmatory testing in cohorts with low a priori likelihood of GHD such as in TBI. Our results question the evidence for newly introduced recommendations for routine pituitary assessment in TBI. (J Clin Endocrinol Metab 99: , 2014) ISSN Print X ISSN Online Printed in U.S.A. Copyright 2014 by The Endocrine Society Received May 31, Accepted October 31, First Published Online November 15, 2013 Previously, loss of pituitary function caused by traumatic brain injury (TBI) was considered rare, accounting for less than 1% of all new cases of hypopituitarism. Recently, TBI-related chronic anterior pituitary hormone deficiency was described as far more frequent, as reviewed by Schneider et al (1), and was suggested to con- Abbreviations: arg, arginine; CI, confidence interval; GCS, Glasgow Coma Scale; GHD, GH deficiency; ITT, insulin tolerance test; OR, odds ratio; PD, pyridostigmine; TBI, traumatic brain injury. doi: /jc J Clin Endocrinol Metab, January 2014, 99(1): jcem.endojournals.org 101

2 102 Klose et al National Prevalence of Posttraumatic Hypopituitarism J Clin Endocrinol Metab, January 2014, 99(1): tribute to fatigue and other common sequelae after TBI. In case of adrenal insufficiency, overlooking the condition could be life-threatening. Thus, recommendations for assessment of pituitary function and replacement after moderate and severe TBI have been introduced (2, 3). A pragmatic limitation to testing only moderate and severe cases was attributed to previously identified risk factors including a low Glasgow Coma Scale (GCS) score, diffuse brain swelling, hypoxia/hypotension, base of skull fractures, increased intracerebral pressure, and axonal injury (4 9) as well as considerations of cost-effectiveness given the high incidence of TBI. The legitimacy of this limited approach was confirmed by some (10) but challenged by others, suggesting more universal testing (11, 12). The finding of mainly isolated deficiencies in TBI patients, and particularly isolated GH deficiency (GHD), has raised the question of potential methodological confounding determined by variable test-retest reproducibility (13), appropriateness of cutoff values in the light of assay heterogeneity (14) and importance of body mass index (BMI)-stratified cutoffs (2). Here we report the prevalence of posttraumatic GHD in a Danish nationwide population of TBI patients, and the impact of common methodological problems, in terms of 1) adequacy of local application of commonly recommended cutoff values from recent international guidelines, 2) differential performance of diagnostic tests, 3) requirement of single or confirmatory testing, and 4) performance of 2 different commercially available GH assays. Patients and Methods Participants We recruited patients identified with a head trauma diagnosis from the Danish Board of Health diagnostic code registry, hospitalized in 2008, aged 18 to 65 years, with total length of hospitalization 24 hours; 2014 patients were identified from a background population of 5.5 million Danish citizens. By retrospective chart review, 856 patients were eligible for inclusion, meeting the following criteria: loss of consciousness, amnesia, or cranial/cerebral imaging abnormalities. Exclusion criteria are shown in Figure 1, outlining the flow from the consecutive cohort to the final cohort of 439 participants. Predisposing factors for study participation included older age (P.01), longer hospital stay (P.01), admission to a neurosurgical ward (P.003), and more severe ICD-10 codes (Table 1). A healthy control group was recruited by newspaper advertisement or web-based recruiting ( (n 124) for establishment of local cutoff values for the dynamic tests used. Patients and controls were well matched for gender and BMI, but patients were older (P.01), and had higher waist circumference (P.04) (Table 1). For further validation of the insulin tolerance test (ITT) cutoff value, biochemical data were retrospectively collected from 39 consecutive patients with multiple pituitary hormone deficiency submitted to ITT at Rigshospitalet during the period when the Immulite GH assay was used. The study was approved by the local ethical committee (J.nr. H-B ) and the Danish Data Protection Agency (J.nr ). All participants or their closest relatives gave their written informed consent before enrollment. Methods All patients and controls underwent dynamic assessment of GH release. In patients, testing was performed at 1 of 4 participating centers. All patients were scheduled for pyridostigmine (PD)-GHRH test. Furthermore, all patients geographically belonging to 1 of 2 centers performing ITT routinely were scheduled for dual testing with the PD-GHRH test and ITT performed in random order. A GHRH-arginine (arg) test was performed in all cases of contraindications. PD-GHRH and GHRH-arg tests will be referred to as combined tests. Single testing was performed in one-third of patients planned for dual testing, equally due to contraindications (54%), and noncompliance (46%). GH assessment was performed (range ) years after the trauma. Testing started between 8:00 and 10:00 AM after an overnight fast. ITT was performed by administering insulin iv ( U/kg, Actrapid; Novo Nordisk) to induce nadir blood glucose 2.2 mmol/l and adequate hypoglycemic symptoms. Blood was collected for analysis of GH at 15, 0, 15, 30, 45, 60, 75, and 90 minutes. During the PD (120 mg at 60 minutes, Mestinon; Meda) plus GHRH (1 g/kg iv at 0 minutes; Ferring) test, GH was sampled at 75, 60, 0, 20, 30, 45, 60, and 90 minutes. During GHRH (1 g/kg iv at 0 minutes) plus arg (0.5 g/kg [maximum 30 g], infused from 0 30 minutes) test, GH was sampled at 15, 0, 30, 45, 60, 75, and 90 minutes. Premenopausal women were tested 8 weeks off estrogen-containing contraception. Three physicians (M.K., J.J., and L.L.C.) and one technician performed all tests. Serum samples were stored at 80 C until centralized analysis. Full time profiles were measured by Immulite assay; all samples from each subject were measured in the same assay series, and all samples were assayed by the same technician within a few batches. Peak values were thereafter reanalyzed by the IDS-iSYS GH assay. Assays First, GH was analyzed by chemiluminescence immunoassay (DPC Immulite 2000; Siemens; calibrated against World Health Organization International Standard 98/574). Lower and upper limits of quantification were 0.05 and 40 g/l, respectively, with analytic sensitivity of 0.01 g/l. The local intra-assay coefficients of variation were 8% and 9% at concentrations of 2.6 and 6 g/l, respectively. Second, GH was analyzed by a novel monoclonal chemiluminescence immunoassay (IDS-iSYS; Immunodiagnostic Systems), calibrated against International Standard 98/ 574. Lower and upper limits of quantification were 0.04 and 100 g/l, with analytical sensitivity of 0.01 g/l (15). The local intra-assay coefficient of variation was 8.3% at a concentration of 0.23 g/l. Diagnostic criteria for GHD Local cutoff values were calculated from healthy controls. Peak GH was log-transformed before analyses to reduce skewness. Cutoff values were then defined by the 2.5th percentile. For further validation, the optimal cutoff to separate patients with

3 doi: /jc jcem.endojournals.org 103 multiple pituitary hormone deficiency with 95% specificity from healthy controls was calculated for the ITT. We compared the prevalence of GHD defined by local cutoff values with the prevalence by published cutoff values in recent guidelines (ie, peak GH 3 g/l in response to ITT or below BMI-stratified cutoffs in response to the GHRH-arg test, defined as peak GH 11 g/l at BMI 25 kg/m2, peak GH 8 g/l at BMI kg/m2, or peak GH 4 g/l at BMI 30 kg/m2) (2, 3, 16). The BMI-stratified cutoff values for the GHRH-arg test were applied to the PD-GHRH test based on previous suggestions of common cutoff values (17, 18). Statistical analyses Paired and unpaired prevalence comparisons included calculation of 95% confidence intervals (CIs) for the differences. Categorical data were compared by 2 test. Between-group comparisons of normally distributed continuous data were analyzed by Student s t test, otherwise by Mann-Whitney U test. Logistic regression analyses were conducted to analyze the relation be- tween insufficient test responses and predictive or confounding factors, including TBI severity, comorbidity, and body composition. Results were presented as odds ratios (ORs) with 95% CIs. Assay agreement was evaluated by Bland-Altman plot with limits of agreement defined as mean difference 2 SD. Observations with missing values were excluded from analyses. A twosided P.05 was considered significant. Statistical analyses were performed by SAS version 9.1. Results Local cutoff values for the ITT, PD-GHRH, and GHRH-arg test Calculations of local cutoff values were based on peak GH from 124 healthy individuals submitted to ITT (n 117), PD-GHRH (n 92), or GHRH-arg (n 41) tests. Median peak GH decreased with increasing BMI indepen- Figure 1. Flow chart outlining the flow from the consecutive cohort including all patients aged 18 to 65 years hospitalized at a Danish hospital in 2008, registered with a TBI diagnosis and total length of hospitalization ⱖ24 hours, to the final cohort of 439 patients undergoing dynamic GH assessment.

4 104 Klose et al National Prevalence of Posttraumatic Hypopituitarism J Clin Endocrinol Metab, January 2014, 99(1): Table 1. Characteristics of Patients and Healthy Controls a Patients Eligible for Inclusion Participants Nonparticipants Healthy Controls n Gender, men, n (%) 292 (67) 298 (72) 78 (63) Length of hospitalization, d, median (range) 4.0 (1 217) d 3.4 (1 256) Primary department of admission, n (%) Neurosurgery 184 (42) d 116 (28) Neuromedicine 66 (15) 71 (17) Orthopedic 158 (36) 177 (43) Other 31 (7) 53 (12) CT abnormalities, n (%) e Commotio cerebri 187 (43) 204 (49) Commotio cerebri, only 104 (24) d 130 (31) Vault or facial fracture 44 (10) 28 (7) Base of skull fracture 48 (11) 45 (11) Diffuse or focal brain injury 77 (18) 66 (16) Traumatic subdural hemorrhage 67 (15) 46 (11) Other diagnoses of intracranial bleeding or cerebral edema 68 (16) 58 (14) GCS, n (%) GCS (69) GCS (31) Age at TBI, y 42.4 (14.6) d 39.7 (15.2) Age at endocrine assessment, y, mean (SD) 44.9 (14.6) d 38.6 (14.2) BMI, kg/m 2, mean (SD) b 26.0 (4.4) 25.3 (5.6) Waist 102 cm (men) or 88 cm (women), n (%) c 119 (27) d 22 (18) Tests performed, n (%) ITT (single test) 29 (7) 0 (0) PD-GHRH (single test) 200 (45) 4 (3) GHRH-arg (single test) 37 (8) 1 (1) ITT plus PD-GHRH 156 (36) 79 (64) ITT plus GHRH-arg 15 (3) 31 (25) ITT plus PD-GHRH plus GHRH-arg 0 (0) 9 (7) a Weight was measured with patients wearing light indoor clothing without shoes. Height was measured barefoot. Waist circumference was ascertained by tape measure to the nearest 0.5 cm with the subject in a standing position and was measured midway between the lower rib and the iliac crest during expiration. b BMI missing in 2 patients. c Waist circumference missing in 8 patients and 1 control. d P 0.05, compared with nonparticipants and controls, respectively. e By ICD-10 codes. dent of test (Table 2). However, the lower limit of the reference interval was BMI-independent for the ITT, although BMI-dependent for the combined tests (Table 2). Thus, a common cutoff was calculated for ITT, whereas BMI-stratified cutoffs were calculated for the combined tests. The local cutoff for ITT was 2.6 g/l, whether defined by the 2.5th percentile of healthy controls or the optimal combination of sensitivity (97%) and specificity (99%) from receiver operating characteristic analysis. The BMIstratified cutoffs for the PD-GHRH and GHRH-arg tests are given in Table 2. The proportion of lean controls above or below the age of 40 years failing a cutoff of 11 g/l did not differ comparing the PD-GHRH ( 40 years, 1 of 37 [2.7%]; 40 years, 1 of 18 [5.6%]) and GHRH-arg test ( 40 years, 0 of 13 [0%]; 40 years, 1 of 9 [11%]). Impact of local vs guideline-derived cutoff values The prevalence of GHD defined by local cutoffs was 9 of 200 (4.5%) in patients tested by ITT and 48 of 407 (11.8%) in patients tested by a combined test. Defined by guideline cutoffs, the prevalence remained unchanged at 5.0% for the ITT (P.9), whereas it increased to 18.9% for the combined test (P.001) (Figure 2A). Similarly, the false-positive rate of GHD in healthy controls increased to 4% by ITT and up to 20% in overweight and obese healthy controls by combined tests by application of guideline cutoffs (Table 2). The prevalence of GHD in patients with moderate or severe TBI (ie, GCS 13, n 133) was 6% (ITT, n 65) and 8.5% (combined test, n 117) by local cutoff values, but 6% (ITT) and 16% (combined test) by guideline cutoffs. The prevalence of GHD in patients with moderate or severe TBI did not differ from the total cohort (P.47).

5 doi: /jc jcem.endojournals.org 105 Table 2. Local and Guideline Cutoff Points for the ITT, PD-GHRH, and GHRH-arg Tests a Peak GH, g/l Local ( g/l) Guideline ( g/l) Median Impact of diagnostic test Both by guideline and local cutoff values, the prevalence of GHD was higher when using a combined test compared with ITT in patients as well as controls (Figure 2A). An insufficient response to ITT was observed equally frequently in patients and controls, whereas an insufficient response to combined tests was more frequent in the patients both by local (P.001) and guideline cutoffs (P.007) (Figure 2A). The higher prevalence by combined tests was not related to pretest patient selection in terms of post-resuscitation GCS score, length of hospitalization, loss of consciousness, amnesia, CT abnormalities by registered ICD-10 codes, treatment with opioids, or BMI. However, patients tested by combined tests had higher waist circumference and, by contraindication to ITT, included all patients with pre- or posttraumatic seizure disorders (Table 3). Defined by guideline cutoffs, patients diagnosed as GHD by a combined test had higher BMI (mean 27.7 vs 25.6 kg/m 2 ), larger waist circumference (99.5 vs 89.8 cm), and older age (50.6 vs 45.8 years) (all, P.001). Increased waist circumference remained independently prognostic for the diagnosis of GHD after adjustment for age and TBI severity (OR 2.8 [ ]). This was also true when assessed by PD-GHRH and GHRH-arg tests separately, Reference Interval b Cutoffs c 95% CI Cutoffs, g/l ITT (n 117) , 3.0 All Normal weight (n 75) d Overweight (n 25) Obese (n 17) PD-GHRH test (n 92) Normal weight (n 55) d , e 4 Overweight (n 21) , e 19 Obese (n 16) , e 19 GHRH-arg test (n 41) Normal weight (n 22) d , Overweight (n 7) d , Obese (n 12) , Abbreviation: FPR, false-positive rate in healthy controls by application of guideline cutoffs. a The cutoffs are based on peak GH from stimulation tests performed in 124 healthy individuals. Due to the significant impact of BMI, cutoffs for the PD-GHRH and GHRH-arg tests are BMI-stratified. Cutoffs commonly recommended in guidelines are given for comparison, together with the false-positive rate, showing the percentage of healthy controls that would fail the respective GH stimulation test by application of these guidelinederived cutoffs. b Reference intervals are given as 2.5th to 97.5th percentiles. c Cutoff for the ITT was defined as mean 1.96SD, whereas those for the PD-GHRH and GHRH-arg tests were defined as the 2.5th percentiles. False-positive rates are given as the percentage of healthy subjects with a peak GH below those cutoffs commonly recommended in consensus guidelines. d Ceiling effect caused by upper limit of detection of 40 g/l. e BMI-related cutoffs not available from the literature; thus, cutoffs are adapted from the GHRH-arg test based on previous definition of a common cutoff points for the 2 tests. FPR, % whereas no such relations were observed for ITT. The prognostic effect of waist circumference did not remain after application of local cutoffs (OR 1.0 [ ]). Dual testing by ITT and a combined test in 169 patients and 108 controls allowed for direct test comparison. As in the total cohort, the prevalence of GHD was higher by a combined test compared with ITT, reaching significance only by guideline cutoffs (Figure 2B). Impact of dual testing Patients undergoing dual testing had similar gender, age, and BMI distribution, but fewer had increased waist circumference compared with the total cohort (P.011). By local cutoffs, the prevalence of GHD did not differ from the total cohort (ITT, 7 of 169 [4%] vs 9 of 200 [4.5%] [P.87]; combined tests, 13 of 169 [8%] vs 48 of 407 [11.8%] [P.15]). Only 1% of the patients (0% of controls) had an insufficient response to both tests. Similar results were recorded by guideline cutoffs (Figure 2B). Impact of GH assay Assessment of GH assay was based on peak GH from 94 controls and 439 patients. The correlation between methods was excellent in both patients and controls (Figure 3, A and B). Method agreement is illustrated in Figure 3C, including

6 106 Klose et al National Prevalence of Posttraumatic Hypopituitarism J Clin Endocrinol Metab, January 2014, 99(1): Discussion Figure 2. The prevalence of insufficient test responses in the total cohort (A) and in the subgroup undergoing dual testing (B). A, Prevalence of insufficient test responses to either ITT or PD-GHRH/GHRH-arg (ie, combined tests) in the total cohort of TBI patients (black columns) and healthy controls (white columns), respectively, as defined by either local or guideline-derived cutoffs. Whiskers indicate the 95% CI. GHD was more frequently diagnosed in TBI patients tested by a combined test as compared with ITT, and even more so if guideline cutoff values were applied instead of local cutoffs. The results from healthy controls illustrate the high falsepositive rate resulting from application of guideline-derived cutoffs, which was significantly above the generally accepted 2.5% for the combined tests (P.02). *, P.005 compared with patients. B, Prevalence of insufficient test responses in the subgroups of patients (black columns, n 169) and controls (white columns, n 117) undergoing dual testing, as defined by either local cutoff values or guideline-derived cutoff values. Confirmed insufficiency was defined as a concordant positive result to both the ITT and a combined test. Whiskers indicate the 95% CI. peak GH concentrations 15 g/l, ie, range considered of clinical significance for the diagnosis of GHD. A cross-tabulation of test results defined according to local cutoff values for the Immulite and isys assay, respectively, showed concordant test outcomes in 546 of 555 tests (98%). Discordance was as often caused by an insufficient response when analyzed by Immulite as by the isys assays (P.5). The present study is the first of its kind to assess national prevalence of posttraumatic GHD, in direct comparison with a large healthy control group. The study has illustrated that the prevalence of GHD was highly dependent on choice of local or guideline-derived cutoffs, choice of diagnostic test, and use of either single or confirmatory testing. These results were independent of whether Immulite or isys GH assays were used. Diagnostic thresholds To assess the appropriateness of local application of guideline-derived cutoff values, a gender- and BMI-matched control group was included. We observed an unacceptably high false-positive rate, both for the ITT and the combined tests, illustrating the high risk of misclassification when applying generally accepted diagnostic cutoffs from international guidelines (2, 3). The use of standard cutoffs is challenged by vastly different assays for hormone measurements, and thus, establishment of local diagnostic test-specific cutoff values are mandatory (14), although rarely used because they are laborious to establish. Inclusion of 124 healthy controls allowed calculation of local test and assay-specific cutoffs. Aiming for a high specificity, the cutoff for ITT was defined at 2.6 g/l as compared with the standard of 3 g/l. For the GHRH-arg test, fairly similar cutoffs were identified in the nonobese subjects, whereas a 25% lower cutoff was identified in obese subjects. For the PD-GHRH test, the cutoff was halved in overweight and obesity, as compared with the adapted guideline cutoff values from the GHRH-arg test. Similar cutoff values for the PD-GHRH and the GHRH-arg tests have been suggested (17, 18), but definite BMI-related cutoff values for the PD-GHRH test are not available from the literature. Our data indicated that direct adaptation of BMI-related cutoffs from the GHRH-arg to the PD-GHRH test was

7 doi: /jc jcem.endojournals.org 107 Table 3. Prognostic and Confounding Factors in the Groups of Patients Subjected to ITT and PD-GHRH or GHRH-arg tests (ie, Combined Tests) ITT (n 200) Combined Test (n 407) Gender, men 128/200 (64) 274/407 (67) Age at time of testing, y 43.2 (14.7) 45.1 (14.7) (SD) TBI severity GCS score 13 65/197 (33) 117/401 (29) Hospitalization, d 3 (1 171) 4 (1 347) Loss of consciousness 152/176 (86) 303/359 (84) Amnesia 177/197 (90) 339/373 (91) CT abnormalities 138/200 (69) 287/407 (71) Base of skull fracture 43/135 (32) 80/281 (29) Comorbidity Opioid treatment 11/200 (5.5) 31/407 (7.6) Epilepsy 23/407 (5.7) Body composition BMI, kg/m (4.1) 26.0 (4.4) Normal weight 99/200 (50) 183/405 (46) Overweight 71/200 (35) 156/405 (38) Obese 30/200 (15) 66/405 (16) Waist (cm) 88.5 (13.1) 91.5 (14.4) a 102 cm (men); 88 cm (women) 38/197 (19) 109/401 (27) a Data are given as mean (SD), median (range), or n (%). a P.05, as compared with patients tested with ITT. inappropriate. Although our data did not allow a thorough test comparison, it is noteworthy that 3 of 10 obese controls tested by PD-GHRH and GHRH-arg test failed both tests at the commonly used cutoff of 4 g/l. A limitation to the locally defined cutoff values for the combined tests, and the GHRH-arg test in particular, is inclusion of relatively few controls. The cutoffs for the GHRHarg test in nonobesity were, however, similar to those published from larger cohorts (16). Differential weighting of sensitivity and specificity may explain the observed difference between local and guideline-derived cutoffs. Recommended cutoffs for the GHRH-arg test (16) were established to avoid overlooking deficiency, at the expense of specificity, which was only 75% in overweight and obese subjects. Although this approach is reasonable in patients with high a priori risk of GHD, eg, patients with multiple pituitary hormone deficiencies, it remains unreasonable in low-risk populations with 25% risk of being misclassified as GHD. The recent reports of a high risk of predominantly isolated forms of GHD in TBI patients has raised suspicion of overt misclassification by applying a diagnostic program tailored for high-risk populations. Available GH assays are generally recognized to yield different results. To minimize interassay variability, new criteria for standardization and evaluation of GH assays were recently established (14). One new commercial assay, the IDS-iSYS, fulfilled the novel criteria and was reported to yield slightly lower GH concentrations than the widely used Immulite assay (15). However, head-to-head comparison of the isys and Immulite GH assays in the present study showed excellent method agreement with no significant diagnostic difference. Thus, we did not retrieve lower concentrations measured by isys as compared with Immulite, the reason for which remains unclear. Test performance The diagnostic set-up included different dynamic tests, allowing for test comparison. In the total cohort, GHD was recorded double or thrice (depending on cutoff) as often in patients tested by a combined test, compared with ITT. This difference was not caused by pretest patient selection by TBI severity, medical treatment with opioids, or BMI, whereas patients tested by combined tests had higher waist circumference. Despite application of BMIstratified cutoffs, BMI and waist remained significantly related to increased risk of an insufficient test outcome for combined tests, but not ITT. The increased prevalence of GHD in patients tested by combined tests could thus be caused merely by a higher rate of abdominal adiposity, in line with previous observations by Makimura et al (19) and Colao et al (20). Although increased fat mass impaired the stimulated GH response to all types of tests, as also described by many others (17, 21 23), the ITT-defined cutoff value for GHD was seemingly unaffected by adiposity. False low peak GH caused by occasional cases of falsely absent GH responses to ITT (13, 24) could erroneously decrease cutoff values and give the impression of no relationship to BMI. Still, the rate of false-positive results during ITT remains unknown. However, local 5th and 10th percentiles from healthy controls of 3.0 and 4.3 g/l in normal-weight subjects and of 3.0 and 3.8 g/l in overweight or obese subjects, respectively (data not shown), supported a higher robustness of the ITT cutoff regarding the impact of BMI. Higher robustness was also indicated by Biller et al (25). In an obese cohort, they reported 95% specificity for GHD by a cutoff of 3.3 g/l for ITT, ie, similar to general recommendations, but 95% specificity for the GHRH-arg test by a cutoff of 1.5 g/l, ie, substantially lower than recommended. The difference in test performance was in keeping with available literature, generally showing lower prevalence of posttraumatic GHD in studies using ITT, as summarized by Kokshoorn et al (26). BMI-related cutoffs were not generally accepted at the time most of the original studies included in the review were conducted. This may explain the higher prevalence of GHD found in studies using the GHRH-arg test (6, 27, 28) and other BMI-dependent tests and may also have contributed to an overestimation of the

8 108 Klose et al National Prevalence of Posttraumatic Hypopituitarism J Clin Endocrinol Metab, January 2014, 99(1): Figure 3. A and B, Correlation plots illustrating the relationship between peak GH as measured by isys and Immulite 2000 in healthy controls (A) and patients (B), respectively. C, Method agreement is illustrated by the Bland-Altman plot by the difference between peak GH concentrations measured by Immulite 2000 and isys plotted against their average concentration. Data includes peak GH concentrations 15 g/l from healthy controls and patients. The solid line indicates the mean difference, which estimates the average bias of one assay relative to the other. The dotted lines indicate the 95% limits of agreement. The mean between-assay difference was 0.04 g/l, with limits of agreement 1.7 g/l. pooled prevalence of posttraumatic hypopituitarism in the recent meta-analysis (1). Most included patients were tested by PD-GHRH test, which was chosen due to its merits in terms of robustness (ability to differentiate between normality and GHD) and safety (no need for close supervision) (18, 29). Such requirements are essential in this particular population due to its size, low a priori risk of GHD and lack of validated treatment efficacy. Ghigo et al (29) identified the PD- GHRH test as a reliable tool for diagnosing GHD in patients aged 20 to 40 years. Healthy adults aged 40 to 65 years were not included in the study, whereas some above the age of 65 displayed low GH responses. For comparison, the GHRH-arg test produced high GH responses above the age of 65 and because of its effectiveness in even very elderly subjects, the GHRH-arg test was preferred for elderly patients with suspected GHD (30). In the present study, only individuals below the age of 65 years were included. The proportion of lean controls above and below the age of A04 failing the guideline cutoff of 11 g/l did not differ comparing the PD-GHRH and GHRH-arg test. Thus, differences in BMI and lack of 40- to 65-yearold controls in the study by Ghigo et al (29) may have discredited the PD-GHRH test. However, this issue merits further investigation. Single vs confirmatory testing In 169 patients and 117 controls submitted to confirmatory testing by 2 alternative tests, up to 14% were insufficient by one, whereas only 1% of patients and none of the controls were insufficient by both. Lower prevalence was anticipated upon confirmatory testing in keeping with the observation that studies using confirmatory tests in general reported lower prevalence of posttraumatic hypopituitarism (4, 10, 31). Previous guidelines recommended 2 abnormal GH tests to establish the diagnosis of isolated adult-onset GHD, whereas 1 test was sufficient in cases of additional pituitary deficiencies (32, 33), illustrating that the diagnostic accuracy is dependent on the patient population studied (34). This requirement of confirmatory testing has been omitted in recent guidelines. Furthermore, moderate and severe TBI, and aneurysmal subarachnoid hemorrhage were added to conditions where GH testing and treatment should be considered (2, 3). Their entrance into the guidelines was based on evidence from studies mainly perform-

9 doi: /jc jcem.endojournals.org 109 ing single testing and included studies not using currently required BMI-stratified cutoffs. Given the present data and those from other very recent studies also using more stringent diagnostic criteria (8, 10, 35, 36), the evidence for including TBI and SAH has been weakened and the indication for routine GH and pituitary assessment in these patient populations needs reconsideration. Study population Patients were recruited from the national background TBI population and, thus, included fewer patients with moderate to severe TBI than most of the original studies. In general, studies including less severely injured patients tended to report lower prevalence of posttraumatic GHD (4, 8, 10), in line with identification of risk factors for posttraumatic hypopituitarism, eg, low GCS score, diffuse brain swelling, hypoxia/hypotension, base of skull fracture, increased intracranial pressure, and axonal injury(4 7, 9). In the present study, GHD was unrelated to trauma severity, and different distribution of TBI severity was thus unlikely to cause significant bias. Although baseline data were available for all patients, outcome measures were unavailable for nonparticipants. A weakness of the present study, and studies in general where inclusion depends on the patient s wish to participate, was lack of background and outcome information from nonparticipants. Thus, whether nonparticipants matched participants to a degree allowing for extrapolation to the general background population remains largely unknown. Although planned as unselected, included patients were older with more severe TBI as compared with nonparticipants. Such selection bias may ultimately result in a prevalence of posttraumatic GHD different from an otherwise eligible background TBI population. Given the previous identification of trauma severity as predictor of posttraumatic hypopituitarism (4 7), the present results more likely overestimated than underestimated the true prevalence. Conclusion A low prevalence of posttraumatic hypopituitarism was demonstrated in an a priori unselected TBI population by applying stringent criteria including confirmatory testing. The prevalence was, however, highly dependent upon management of testing for GHD, and the study confirmed a high risk of bias in pituitary testing of patients with TBI. The results stress the importance of a proper control group and stringent GH testing, particularly in cohorts with a low a priori likelihood of GHD such as TBI. Based on our data, together with data from other more recent studies, also supporting a lower prevalence than observed in the original studies, general recommendations concerning testing for hypopituitarism in TBI should be considered premature. Acknowledgments We thank laboratory technicians Casper Kok, Anette Vinding, Kirsten Rasmussen, Susanne Andersen, and Anne-Mette Christensen for their excellent technical assistance. We are also grateful to Immunodiagnostic Systems for kindly providing the isys GH kits. Address all correspondence and requests for reprints to: Marianne Christina Klose, MD, Department of Medical Endocrinology, PE2131, Copenhagen University Hospital, Rigshospitalet, Blegdamsvej 9, DK-2100 Copenhagen. klose@rh.dk. The study was funded by unrestricted grants from The Danish Agency for Science, Technology, and Innovation; The Reseach Council of the Capital Region of Denmark; The Lundbeck Foundation; Novo Nordisk; The A.P. Moller Foundation for the Advancement of Medical Science; Arvid Nilsson s Foundation; Christenson-Ceson s Foundation; Axel-Muusfeldt s Foundation; and Else and Mogens Wedell-Wedellsborg s Foundation. Immunodiagnostic Systems provided isys GH kits free of charge but had no influence on results or their interpretation. Disclosure Summary: The study received an unrestricted grant from Novo Nordisk, and Immunodiagnostic Systems provided isys GH kits free of charge, but had no influence on results or their interpretation. The authors have nothing else to declare. References 1. Schneider HJ, Kreitschmann-Andermahr I, Ghigo E, Stalla GK, Agha A. Hypothalamopituitary dysfunction following traumatic brain injury and aneurysmal subarachnoid hemorrhage: a systematic review. JAMA. 2007;298: Ho KK. Consensus guidelines for the diagnosis and treatment of adults with GH deficiency II: a statement of the GH Research Society in association with the European Society for Pediatric Endocrinology, Lawson Wilkins Society, European Society of Endocrinology, Japan Endocrine Society, and Endocrine Society of Australia. Eur J Endocrinol. 2007;157: Molitch ME, Clemmons DR, Malozowski S, Merriam GR, Vance ML. Evaluation and treatment of adult growth hormone deficiency: an Endocrine Society clinical practice guideline. J Clin Endocrinol Metab. 2011;96: Klose M, Juul A, Poulsgaard L, Kosteljanetz M, Brennum J, Feldt- Rasmussen U. Prevalence and predictive factors of post-traumatic hypopituitarism. Clin Endocrinol (Oxf). 2007;67: Kelly DF, Gonzalo IT, Cohan P, Berman N, Swerdloff R, Wang C. Hypopituitarism following traumatic brain injury and aneurysmal subarachnoid hemorrhage: a preliminary report. J Neurosurg. 2000; 93: Bondanelli M, De Marinis L, Ambrosio MR, Monesi M, Valle D, Zatelli MC, Fusco A, Bianchi A, Farneti M, degli Uberti EC. Occurrence of pituitary dysfunction following traumatic brain injury. J Neurotrauma. 2004;21: Schneider M, Schneider HJ, Yassouridis A, Saller B, von Rosen F, Stalla GK. Predictors of anterior pituitary insufficiency after traumatic brain injury. Clin Endocrinol (Oxf). 2008;68: Kokshoorn NE, Smit JW, Nieuwlaat WA, et al. Low prevalence of

10 110 Klose et al National Prevalence of Posttraumatic Hypopituitarism J Clin Endocrinol Metab, January 2014, 99(1): hypopituitarism after traumatic brain injury: a multicenter study. Eur J Endocrinol. 2011;165: Bavisetty S, Bavisetty S, McArthur DL, et al. Chronic hypopituitarism after traumatic brain injury: risk assessment and relationship to outcome. Neurosurgery. 2008;62: ; discussion van der Eerden AW, Twickler MT, Sweep FC, et al. Should anterior pituitary function be tested during follow-up of all patients presenting at the emergency department because of traumatic brain injury? Eur J Endocrinol. 2010;162: Wilkinson CW, Pagulayan KF, Petrie EC, et al. High prevalence of chronic pituitary and target-organ hormone abnormalities after blast-related mild traumatic brain injury. Front Neurol. 2012;3: Tanriverdi F, Unluhizarci K, Kelestimur F. Pituitary function in subjects with mild traumatic brain injury: a review of literature and proposal of a screening strategy. Pituitary. 2010;13: Hoeck HC, Jakobsen PE, Vestergaard P, Falhof J, Laurberg P. Differences in reproducibility and peak growth hormone responses to repeated testing with various stimulators in healthy adults. Growth Horm IGF Res. 1999;9: Clemmons DR. Consensus statement on the standardization and evaluation of growth hormone and insulin-like growth factor assays. Clin Chem. 2011;57: Manolopoulou J, Alami Y, Petersenn S, et al. Automated 22-kD growth hormone-specific assay without interference from Pegvisomant. Clin Chem. 2012;58: Corneli G, Di Somma C, Baldelli R, et al. The cutoff limits of the GH response to GH-releasing hormone-arginine test related to body mass index. Eur J Endocrinol. 2005;153: Procopio M, Maccario M, Savio P, et al. GH response to GHRH combined with pyridostigmine or arginine in different conditions of low somatotrope secretion in adulthood: obesity and Cushing s syndrome in comparison with hypopituitarism. Panminerva Med. 1998;40: Andersen M, Hansen T, Støvring R, et al. The pyridostigminegrowth-hormone-releasing-hormone test in adults. The reference interval and A Comparison with the insulin tolerance test. Endocrinol Metab. 1996;3: Makimura H, Stanley T, Mun D, You SM, Grinspoon S. The effects of central adiposity on growth hormone (GH) response to GH-releasing hormone-arginine stimulation testing in men. J Clin Endocrinol Metab. 2008;93: Colao A, Di Somma C, Savastano S, et al. A reappraisal of diagnosing GH deficiency in adults: role of gender, age, waist circumference, and body mass index. J Clin Endocrinol Metab. 2009;94: Cordido F, Dieguez C, Casanueva FF. Effect of central cholinergic neurotransmission enhancement by pyridostigmine on the growth hormone secretion elicited by clonidine, arginine, or hypoglycemia in normal and obese subjects. J Clin Endocrinol Metab. 1990;70: Rasmussen MH, Hvidberg A, Juul A, et al. Massive weight loss restores 24-hour growth hormone release profiles and serum insulin-like growth factor-i levels in obese subjects. J Clin Endocrinol Metab. 1995;80: Ghigo E, Mazza E, Corrias A, et al. Effect of cholinergic enhancement by pyridostigmine on growth hormone secretion in obese adults and children. Metabolism. 1989;38: Fisker S, Jorgensen JO, Orskov H, Christiansen JS. L-arginine and insulin-tolerance tests in the diagnosis of adult growth hormone deficiency: influence of confounding factors. Clin Endocrinol (Oxf). 1998;48: Biller BM, Samuels MH, Zagar A, et al. Sensitivity and specificity of six tests for the diagnosis of adult GH deficiency. J Clin Endocrinol Metab. 2002;87: Kokshoorn NE, Wassenaar MJ, Biermasz NR, et al. Hypopituitarism following traumatic brain injury: prevalence is affected by the use of different dynamic tests and different normal values. Eur J Endocrinol. 2010;162: Aimaretti G, Ambrosio MR, Di Somma C, et al. Residual pituitary function after brain injury-induced hypopituitarism: a prospective 12-month study. J Clin Endocrinol Metab. 2005;90: Schneider HJ, Schneider M, Saller B, et al. Prevalence of anterior pituitary insufficiency 3 and 12 months after traumatic brain injury. Eur J Endocrinol. 2006;154: Ghigo E, Aimaretti G, Gianotti L, Bellone J, Arvat E, Camanni F. New approach to the diagnosis of growth hormone deficiency in adults. Eur J Endocrinol. 1996;134: Ghigo E, Goffi S, Nicolosi M, et al. Growth hormone (GH) responsiveness to combined administration of arginine and GH-releasing hormone does not vary with age in man. J Clin Endocrinol Metab. 1990;71: Agha A, Rogers B, Sherlock M, et al. Anterior pituitary dysfunction in survivors of traumatic brain injury. J Clin Endocrinol Metab. 2004;89: Consensus guidelines for the diagnosis and treatment of adults with growth hormone deficiency: summary statement of the Growth Hormone Research Society Workshop on Adult Growth Hormone Deficiency. J Clin Endocrinol Metab. 1998;83: Lissett CA, Thompson EG, Rahim A, Brennan BM, Shalet SM. How many tests are required to diagnose growth hormone (GH) deficiency in adults? Clin Endocrinol (Oxf). 1999;51: Reid MC, Lachs MS, Feinstein AR. Use of methodological standards in diagnostic test research. Getting better but still not good. JAMA. 1995;274: Klose M, Brennum J, Poulsgaard L, Kosteljanetz M, Wagner A, Feldt-Rasmussen U. Hypopituitarism is uncommon after aneurysmal subarachnoid haemorrhage. Clin Endocrinol (Oxf). 2010;73: Gardner CJ, Javadpour M, Stoneley C, et al. Low prevalence of hypopituitarism after subarachnoid haemorrhage using confirmatory testing and with BMI-specific GH cutoff levels. Eur J Endocrinol. 2012;168(4):