Adequately Adapted Insulin Secretion and Decreased Hepatic Insulin Extraction Cause Elevated Insulin Concentrations in Insulin Resistant Non-Diabetic Adrenal Incidentaloma Patients

Background Insulin-resistance is commonly found in adrenal incidentaloma (AI) patients. However, little is known about beta-cell secretion in AI, because comparisons are difficult, since beta–cell-function varies with altered insulin-sensitivity. Objectives To retrospectively analyze beta–cell function in non-diabetic AI, compared to healthy controls (CON). Methods AI (n=217, 34%males, 57±1years, body-mass-index:27.7±0.3kg/m2) and CON [n=25, 32%males, 56±1years, 26.7±0.8kg/m2] with comparable anthropometry (p≥0.31) underwent oral-glucose-tolerance-tests (OGTTs) with glucose, insulin, and C–peptide measurements. 1mg-dexamethasone-suppression-tests were performed in AI. AI were divided according to post–dexamethasone-suppression–test cortisol-thresholds of 1.8 and 5µg/dL into 3subgroups: pDexa<1.8µg/dL, pDexa1.8-5µg/dL and pDexa>5µg/dL. Using mathematical modeling, whole-body insulin-sensitivity [Clamp-like-Index (CLIX)], insulinogenic Index, Disposition Index, Adaptation Index, and hepatic insulin extraction were calculated. Results CLIX was lower in AI combined (4.9±0.2mg·kg-1·min-1), pDexa<1.8µg/dL (4.9±0.3) and pDexa1.8-5µg/dL (4.7±0.3, p<0.04 vs.CON:6.7±0.4). Insulinogenic and Disposition Indexes were 35%–97% higher in AI and each subgroup (p<0.008 vs.CON), whereas C–peptide–derived Adaptation Index, compensating for insulin-resistance, was comparable between AI, subgroups, and CON. Mathematical estimation of insulin–derived (insulinogenic and Disposition) Indexes from associations to insulin-sensitivity in CON revealed that AI-subgroups had ~19%-32% higher insulin-secretion than expectable. These insulin-secretion-index differences negatively (r=-0.45, p<0.001) correlated with hepatic insulin extraction, which was 13-16% lower in AI and subgroups (p<0.003 vs.CON). Conclusions AI-patients show insulin-resistance, but adequately adapted insulin secretion with higher insulin concentrations during an OGTT, because of decreased hepatic insulin extraction; this finding affects all AI-patients, regardless of dexamethasone-suppression-test outcome.


Introduction
The advances in radiologic examinations have led to more frequent findings of nodules in the adrenal gland, also termed as adrenal incidentalomas (AI) [1]. Radiologic AI diagnosis has opened a new window of action to endocrinologists: On one hand, the clear absence of malignancy has to be proven by adequate imaging, and on the other hand, a check of this nodule's endocrine activity is needed. However, the preponderant part of AI turns out to be benign and seems endocrine inactive [2].
Previous studies have suggested AI to be associated with the Cardiometabolic Syndrome: AI patients are insulin resistant with higher post-glucose-load plasma glucose concentrations and therefore increased prevalence of type-2 diabetes mellitus (T2DM) or glucose intolerance, as well as arterial hypertension and dyslipidaemia, all of which may contribute to the observed, greater cardiovascular risk [3][4][5][6][7][8][9]. Also osteoporosis is more frequent in AI patients [1,10]. Interestingly, some of these disturbances were at least in part reversible after surgical removal of AI [11]. However, despite those numerous studies on insulin resistance in AI, less attention has been paid to insulin secretion, another important predictor of T2DM development, as observed in T2DM-prone offspring of T2DM patients [12][13][14]. We hypothetisized that not only insulin sensitivity, but also insulin secretion might be altered in AI, since glucocorticoids could stimulate insulin release [15,16]. However, assessment of beta-cell function among individuals with different insulin sensitivity is not an easy task, because in non-diabetic humans, insulin resistance is accompanied by a rise in insulin secretion, in order to compensate for the reduced action of insulin on responsive tissues [17,18].
It has been generally accepted for long that a repeated, fasting, post-dexamethasone-suppression-test (DST) cortisol level above 5µg/dL (pDexa5µg/dL) is considered abnormal and brings about diagnosis of overt Cushing syndrome [19], which goes along with signs and/or symptoms specific to overt cortisol excess, such as purple striae, easy bruising, proximal muscle weakness, and plethora [10,19]. More recently, however, a lower level for differentiation between patients with impaired and normal cortisol suppression has been proposed [10,19]. The utmost part of studies during the past decade defined this cut-off level at 1.8µg/dL for hypercortisolaemic subclinical Cushing syndrome, leading to higher incidence of T2DM, hypertension and osteoporosis, but not necessarily pronounced signs and/or symptoms of overt Cushing syndrome [10].
Thus, one of this study's aims is to investigate metabolic parameters in non-diabetic AI patients with sophisticated methods in vivo. We aimed to compare measures of wholebody insulin sensitivity, insulin secretion, glucose appearance and hepatic insulin extraction with those of healthy control subjects (CON). Moreover, beta-cell function was also related to insulin sensitivity to evaluate the ability of the beta-cell to adapt its secretion to changes in insulin resistance. In addition, according to the post-DST-cortisol thresholds of 1.8 and 5µg/dL mentioned above, an in depth analysis was performed by dividing AI patients into 3 subgroups: pDexa<1.8µg/dL, pDexa1.8-5µg/dL, and pDexa5µg/dL. Thus, our study design would also allow the solution to the question of clinicians whether the proven absence of even subclinical Cushing means unchanged insulin sensitivity and/or secretion, or should the mere knowledge of AI existence justify the assumption of altered metabolic parameters, regardless of the DST outcome.

Study participants
Patients with newly discovered incidentalomas by ultrasound, computer-, and/or magnetic resonance-tomography were admitted between 2000 and 2011 to the endocrine outpatients ward of our department. In total, 217 patients fulfilled the following inclusion criteria: (i) absence of diabetes mellitus and other known diseases such as in liver and/or kidney, or phaeochromocytomas (as far as extractable from the data), as well as (ii) performance of an oral glucose tolerance test (OGTT) with a routine baseline blood analysis including fasting serum cortisol and, (iii) a 1mg dexamethasonesuppression-test (DST) with measurement of serum cortisol at 8:00AM after dexamethasone consumption at 11:00PM the night before. The patients' data were electronically composed using computer-assisted collection, by which all patients' data and diagnoses in the described timeframe were included. The control group (CON) consisted of 25 healthy humans, who were age-, sex-, and body-mass-index-(BMI)-matched (Tab. 1), and did not take any regular medication known to affect insulin sensitivity, -secretion, hyperlipidaemia, and/or hypertension. The data composition as well as the study design and analyses were approved by the local ethics committee of the Vienna Medical University (#1970/2012). Because of the retrospective analysis, no consent was obtained from the patients in any form. The local ethics committee approved this procedure including the waiver of (another) (written) consent.
Oral glucose tolerance test (OGTT). Participants were instructed to arrive at our endocrine outpatients ward in the morning in fasting condition, meaning an at least 10-hour period without consumption of food or beverages except water. Blood was drawn after insertion of a catheter (Vasofix ® ; Braun, Melsungen, Germany) into one antecubital vein at fasting, and 60, 90, and 120min after drinking a solution consisting of 75g glucose (Gluco-Drink75 ® ; Roche Diagnostics, Vienna, Austria) for determination of plasma glucose and subsequent analyses of plasma hormones [20][21][22]. Samples were centrifuged and then either frozen at -80°C or immediately transported to the lab for rapid analyses.

Calculations
Measures of insulin sensitivity, such as the Clamp-like (CLIX), the Matsuda (ISI) and the oral glucose insulin sensitivity (OGIS) indexes, QUICKI, and those of beta-cell function, such as the basal insulin secretion rate and the Insulinogenic Index (IGI) of 0-60min and 0-120min, were assessed as described in details elsewhere [12,[21][22][23][24][25][26][27]. The product of insulin sensitivity with an index of post-hepatic insulin appearance (sometimes termed Disposition Index) and that with C-peptide derived beta-cell function (sometimes termed Adaptation Index) provides figures of the capacity of the beta-cell to adapt its secretion to the changes in insulin resistance. A novel insulin secretion index derived from OGTT C-peptide concentrations, called WHole-Ogtt-SHape-index-C-Peptide (WHOSH_CP), was determined as described elsewhere [28]. Areas under the curve (AUC) were calculated by using the trapezoidal rule. Hepatic insulin extraction (as percentage of the secreted hormone) was estimated as previously described [25]. In addition, for the calculation of endogenous glucose production (EGP), we exploited our recent findings that basal hepatic insulin sensitivity, which was calculated as 100 divided by EGP times fasting insulin secretion, equals ISI-HOMA, the inverse value of HOMA-IR [27]; the rationale of this was confirmed in detail elsewhere [18,29]. Basal endogenous glucose production therefore equals 100 divided by ISI-HOMA times basal insulin secretion and is given in mg·kg -1 ·min -1 .

Statistical analyses
All data are given as means±SEM. Before further analysis, the distribution of the variables was tested by visual examination for marked non-normality and/or the Kolmogorov-Smirnov test, yielding that every variable was normally distributed. Comparisons between two, or more than two groups, were done by using two-tailed unpaired Student's ttests, or ANOVA with post hoc least significant difference (LSD) tests, respectively. Linear methods were used for correlation analyses using Pearson's correlation coefficient r, or if logarithmic, by Spearman's method. Differences were considered statistically significant at p-values≤0.05. Statistical analyses were performed using SPSS ® (SPSS Inc., Chicago, IL) computer software.

Results
All AI patients combined (Tab. 1+Fig.1) Anthropometric characteristics, such as age, BMI, and sex, as well as liver and kidney parameters (transaminases and creatinine) were not different between AI and CON (Tab. 1). Fasting glucose concentrations were slightly, but significantly, higher in AI by 4mg/dL, and glucose intolerance was six-fold higher in AI (each p<0.04). Serum concentrations of total, LDLand HDL-cholesterol were comparable between both groups, while triglycerides were 31% higher in AI. AI showed higher OGTT glucose, insulin, and C-peptide concentrations ( Figure  1A-C), which resulted in increased AUCs (Tab. 1). Basal (i.e. hepatic) insulin sensitivity, calculated from QUICKI, was unaffected, whereas whole-body insulin sensitivity was clearly reduced in AI, as displayed by CLIX ( Figure 1D), ISI, or OGIS (Tab. 1), the latter reflecting glucose clearance (each p<0.001).
Post-hepatic insulin-related indices, IGI and the Disposition Index, were elevated in AI. On the other hand, those derived from C-peptide (fasting beta-cell function, the Adaptation Index and WHOSH_CP) were similar between AI and CON. This fact is also reflected by a 15% reduction in AI of hepatic insulin extraction, while fasting EGP was comparable.

The three AI subgroups
Results are shown in Tab. 2 and Figure 2. Sex and BMI were similar among all 3 AI subgroups and CON, but age was higher in pDexa1.8-5µg/dL, when compared to pDexa<1.8µg/dL and CON (p<0.03). PDexa<1.8µg/dL showed lower high-density lipoprotein-(HDL)-cholesterol and higher triglyceride concentrations. Basal cortisol was highest in pDexa5µg/dL, when compared to both pDexa<1.8µg/dL and pDexa1.8-5µg/dL (p<0.001). Basal cortisol was different and rose, whereas suppression of fasting cortisol by DST fell among the subgroups (p<0.001). Fasting glucose concentrations were slightly higher by 5mg/dL in pDexa<1.8µg/dL than CON (p<0.05). PDexa<1.8µg/dL and pDexa1.8-5µg/dL showed 6-to 7-fold higher glucose intolerance. Circulating concentrations and AUCs of glucose, insulin, and C-peptide during OGTT (Figure 2A-C) were mostly higher in each AI subgroup, when compared to CON. Whole-body insulin sensitivity and glucose clearance, as determined by CLIX ( Figure 2D), and OGIS (Tab. 2) were lower in pDexa<1.8µg/dL and pDexa1.8-5µg/dL, when compared to CON (each p<0.006), whereas also pDexa5µg/dL displayed lower insulin sensitivity, with regard to ISI (p<0.02). Fasting insulin secretion appeared higher in pDexa5µg/dL, and fasting beta-cell function in both pDexa<1.8µg/dL and pDexa5µg/dL. The Insulinogenic Index at 0-60min and 0-120min, as well as the Disposition Index were elevated in all 3 subgroups (each p<0.008 vs. CON), whereas no differences were found between all AI subgroups and CON with regard to the Adaptation Index. However, pDexa<1.8µg/dL had a slightly, but significantly, lower Adaptation Index than pDexa1.8-5µg/dL and pDexa5µg/dL (p<0.04). WHOSH_CP was 55% higher in pDexa5µg/dL. Again, hepatic insulin extraction was lower by 13-16% in each AI subgroup than in CON (p<0.004), while fasting EGP was comparable among all subgroups and CON.
In order to provide a more subtle assessment of beta-cell function in AI subgroups, we calculated the associations of CLIX with Insulinogenic Index (IGI, 0-60min) and Disposition Index (DI) in CON, which significantly (each p<0.04) follow the formulas: IGI = 83.0 -28.7 x log e (CLIX) and DI = 30.4 -8.5 x log e (CLIX). When applying these formulas to AI subgroups, we obtained lower calculated than measured results of both Insulinogenic and Disposition Indexes, as depicted as gray symbols in Figure 2E+F. The differences between observed and calculated values of both Insulinogenic Index (r=-0.452, p<0.001) and the Disposition Index (r=-0.451, p<0.001) were very closely negatively associated with hepatic insulin extraction ( Figure 2G+H), which was positively related to measures of insulin sensitivity, such as CLIX (r=0.415, p<0.001) (Figure 2i), OGIS (r=0.645, p<0.001), ISI (r=0.649, p<0.001), and QUICKI (r= 0.516, p<0.001).

Discussion
This retrospective study was undertaken in a large cohort of more than 200 non-diabetic patients with adrenal incidentaloma diagnosis to investigate whole-body insulin sensitivity and its relation to insulin secretion, as well as hepatic insulin extraction and endogenous glucose production, by applying advanced index calculation and sophisticated methods to OGTT data. AI patients underwent a 1mg dexamethasone-suppression-test for diagnosis of (subclinical) Cushing syndrome.
This study's major results in non-diabetic AI are: (i) reduced insulin sensitivity with higher circulating concentrations of glucose, insulin, and C-peptide during OGTT, (ii) increased insulin secretion and beta-cell function, (iii) diminished hepatic insulin extraction, (iv) negative association of insulin sensitivity with peripheral insulinaemia, but (v) C-peptide-derived Adaptation Index comparable to that of CON that shows a normal capacity of compensating for insulin resistance by augmenting insulin release.
By dividing AI according to post-dexamethasonesuppression-test cortisol threshold levels of 1.8 and 5µg/dL, three subgroups were created: pDexa<1.8µg/dL, pDexa1.8-5µg/dL, and pDexa5µg/dL. In these subgroups, we found: (vi) insulin resistance clearly present in pDexa<1.8µg/dL and pDexa1.8-5µg/dL; (vii) beta-cell function elevated, hepatic insulin extraction reduced, and the Adaptation Index still comparable to that of CON; (viii) that the adjustment of insulinderived beta-cell indexes from relationships in CON measured values in AI subgroups to be higher than expected, and (ix) the differences between measured and calculated levels tightly and negatively correlated with hepatic insulin extraction; finally, (x) that post-DST-cortisol concentrations were positively associated with fasting beta-cell function and the Adaptation Index.  controls (CON, n=25, o) and adrenal incidentaloma patients (AI, n=217, •). Differences were statistically analyzed by using Student's t-test: *, p<0.05.

Whole-body insulin sensitivity
OGTT-derived indexes of whole-body insulin sensitivity showed a clear reduction in AI, the subgroups pDexa<1.8µg/dL and pDexa1.8-5µg/dL, and a borderline decrease in pDexa5µg/dL. Surrogate indexes however did not agree: in fact, according to ISI, also the latter group exhibited a significant insulin resistance. This impairment, which was frequently shown in AI previously [3,[5][6][7][8][9], is the core of the Cardiometabolic Syndrome and contributes to higher cardiovascular disease risk and complications proper of these patients [4]. Of note, the degree of insulin resistance observed in this study in non-diabetic AI can be seen as impressive: healthy, non-diabetic subjects with a BMI on average in the overweight, but not obese range, are expected to display an insulin sensitivity of ~7mg·kg -1 ·min -1 from the clamp-test, as others and ourselves have previously shown [22]. This expected result was seen only in the healthy overweight controls, but not the matching AI, whose insulin sensitivity was on average even below the threshold of pronounced insulin resistance of <5mg·kg -1 ·min -1 [22]. Our finding would correspond to an obesity-like insulin sensitivity [22], meaning that all our AI patients should be considered at risk for disease and mortality, as if they were several kilograms heavier. This interesting outcome could not only contribute to explain their higher cardiovascular risk [4], but -to our surprise -affects all AI patients, also the pDexa<1.8µg/dL, and should be therefore spread over any AI patient, regardless of the DST outcome.
Moreover, the possible presence of liver insulin resistance can be assessed by QUICKI and EGP [29,30], which were comparable to those of CON in all AI patients combined and in every subgroup. This means that in fasting condition, these patients behave normally in terms of insulin sensitivity, but show their impairment only in dynamic conditions after a glucose load. Since, in general, non-diabetic subjects still have unaltered fasting EGP, regardless of presence or absence of insulin resistance [30,31], this study confirms the non-diabetic state of our patients.

Insulin secretion
Insulin secretion in general rises since the very beginning of the appearance of the condition of insulin resistance, aiming for compensation of reduced action to insulin [32]. Insulin secretion in AI patients is comparable to that of healthy control subjects, i.e. adequately adapted to the relative degree of insulin resistance, as shown by the C-peptide-derived Adaptation Index, not different from CON in every subgroup. On the other hand, the indexes based on post-hepatic insulin levels were higher than CON and tightly and negatively related to hepatic insulin extraction. From this, it follows that higher OGTT insulin concentrations observed in AI were due to a higher post-hepatic insulin release associated to a lower extraction, proper of insulin-resistant states [33,34]. The reasons for this lower insulin extraction by the liver are still obscure: it may be the interplay of insulin resistance [33,34] and/or glucocorticoid excess, which may also lead to this phenomenon [35].
Another, interesting finding of this study was the positive association of post-DST-cortisol to both fasting beta-cell function and the Adaptation Index. Of note, in healthy people, the dexamethasone-suppression-test examines the remaining cortisol release in the adrenal cortex following short-term suppression of hypothalamic ACTH secretion. Thus, the post-DST-cortisol level can be regarded as adrenal excess production, most likely autonomously. Exposure of beta-cells to glucocorticoids resulted in beta-cell expansion and enhanced insulin secretion, but also a blunted C-peptide secretion upon stimulation [36], which was found in previous studies of ourselves in glucocorticoid-treated, non-diabetic, renal transplant patients [25], and healthy humans [16]. Interestingly, in this study, we did not only observe increased insulin secretion due to reduced extraction in AI, but also alterations in the OGTT C-peptide shape, which was borderline higher in AI, and increased in pDexa5µg/dL, as determined by the novel WHOSH_CP-index [28]. Another evidence for a rather slight stimulation of insulin secretion by cortisol overproduction seems the weak, but significantly positive association of post-DST-cortisol concentrations with Adaptation Index during OGTT and beta-cell function at fasting.
Another issue that might be associated with AI is the more frequent co-appearance of phaeochromocytomas. As far as possible by using the database, we have excluded all patients with phaeochromocytoma. In this context it is of note that phaeochromocytoma-derived catecholamines not only worsen insulin sensitivity, but also insulin secretion via alpha-2adrenoceptors in beta-cells, which may contribute to diabetes mellitus development [37,38].

Limitations
The major drawbacks of this study were on the one side the retrospective analyses with known disadvantages, and, on the other side, the measurement of metabolite and hormone concentrations at only two time-points within the first OGTT hour. However, none of the calculated parameters and indexes was greatly affected thereby, so that the main question seems to be solved. In a clinical setting with a high number of patients to be dealt with, less frequent OGTT blood sampling seems justified and unavoidable, owing to limited time slots in outpatients' routine treatment.

Conclusions
Patients with adrenal incidentalomas show insulin resistance, but adequately adapted insulin secretion with higher insulin concentrations during a glucose challenge, due to a decreased hepatic insulin extraction. These findings affect all AI patients, regardless of the outcome of the dexamethasone-suppressiontest so that AI diagnosis seems to bring about high likelihood of metabolic alterations involved in the Cardiometabolic syndrome.
electronic composition of the patient's data collection. In addition, the authors thank the colleagues in the Endocrinology Division's and Central Laboratory of the Vienna Medical University for precise metabolite and hormone analyses.