The degree of cortisol secretion is associated with diabetes mellitus and hypertension in patients with nonfunctioning adrenal tumors

In recent years the management of patients with incidentally discovered adrenal masses (adrenal incidentalomas, AI) has become a topic of growing interest [1, 2]. The importance of AI in the clinical practice is due to the accumulating evidence that in up to 50% of AI patients may have a mild autonomous cortisol secretion (MACS), defined by the presence of hypercortisolism in the absence of the classic signs and symptoms of cortisol excess [2]. The interest in this condition is due firstly to the fact that the prevalence of this hidden cortisol excess in the adult population is not negligible, being estimated to reach 2% in adults over 60 years of age [3]. Secondly, MACS has been found to be associated with increased mortality, mainly due to a higher risk of cardiovascular events [4‐9]. This is partially explained by the increased frequency in MACS patients of cardiovascular risk factors, diabetes mellitus (DM) and hypertension (HT) [10, 11]. A third argument is the improvement of DM and HT in MACS patients after adrenalectomy [12‐15].

Interestingly, some recent data have suggested that there is an increased risk of cardiovascular events (CVE) even in patients with nonfunctioning adrenal tumors (NFAT) [16‐19]. Accordingly, the mortality risk associated with the presence of NFAT seems to be comparable to that in AI patients with MACS [20], and adrenalectomy seems to be beneficial on HT and DM even in patients with NFAT [14]. Finally, a recent study by our group shows that a not negligible percentage of patients with AI without MACS who undergo unilateral adrenalectomy because of the size of the adenoma may experience a post-surgical hypocortisolism, which should not be expected in the absence of preoperative MACS [21]. In this study, cortisol after 1 mg overnight dexamethasone suppression test (F-1mgDST) levels measured before surgery with a cut-off set at 1.2 µg/dL (33 nmol/L)—and not 1.8 µg/dL (50 nmol/L) as suggested by current guidelines [1]—could completely rule out the risk of postoperative hypocortisolism. Moreover, although the small sample size did not allow us to draw firm conclusions, in this cohort AI patients with F-1mgDST < 1.2 µg/dL (33 nmol/L) tended to have a better metabolic profile than those with F-1mgDST ≥ 1.2 µg/dL [21]. As the occurrence of hypocortisolism after the removal of a unilateral adrenal adenoma may be considered as evidence of some degree of excessive cortisol secretion prior to surgery, these findings collectively suggest that the F-1mgDST cut-off for defining the absence of hypercortisolism in AI patients should be lowered to at least 1.2 µg/dL (33 nmol/L). This idea is supported by previous data showing that F-1mgDST with a cut-off set at about 1.4 µg/dL (39 nmol/L) has the highest accuracy in predicting cardiovascular risk [16, 22] or the incidence of diabetes mellitus [17] in AI patients.

Therefore, our hypothesis is that a subgroup of the so-called “NFAT” patients may still exhibit a hidden hypercortisolism. Therefore, the aim of the present study was to evaluate in a large group of NFAT patients: i) the possible association between metabolic consequences known to be associated with hypercortisolism, such as HT, DM, obesity (OB), dyslipidemia (DL) and CVE, and parameters of hypothalamic–pituitary–adrenal (HPA) axis activity (ACTH and F-1mgDST); ii) the cut-off of the HPA axis activity parameters able to identify NFAT patients with metabolic consequences.

The results of the ROC curve analysis assessing the associations between F-1mgDST and the explored outcomes (i.e. HT or DM or HT plus DM or DL or OB or CVE) are shown in Table 1. The presence of HT, DM and HT plus DM, but not of OB, DL and CVE, was significantly associated with F-1mgDST levels. No significant associations were found between these outcomes and ACTH levels (data not shown). The cut-off with the best accuracy in identifying patients with either HT or DM (AUC 0.604, 95% CI 0.560–0.649, sensitivity 60.2%, specificity 56.0%) or HT plus DM (AUC 0.611, 95% CI 0.545–0.675, sensitivity 60.4%, specificity 69.6%), was found to be 1.2 µg/dL (33 nmol/L).

We, then, compared the clinical characteristics of patients with F-1mgDST < 1.2 µg/dL with those of patients with F-1mgDST 1.2–1.79 µg/dL (33–49.4 nmol/L). As shown in Table 2, the latter group showed an older age, lower ACTH levels and a higher prevalence of HT, DM, HT plus DM and CVE than the former group.

Moreover, F-1mgDST values differed among patients with HT plus DM, HT or DM and patients with no comorbidities. In particular, levels of F-1mgDST were significantly higher in patients with HT (1.23 ± 0.36) as compared with patients without HT (1.11 ± 0.39, p < 0.001) and in patients with DM (1.28 ± 0.34) than in patients without DM (1.14 ± 0.38, p < 0.001), even after adjusting for age (data not shown). Additionally, levels of F-1mgDST were significantly higher in patients with HT plus DM (1.29 ± 0.34) and HT or DM (1.21 ± 0.36) compared to those without any comorbidity (1.09 ± 0.39), p = 0.0001 and p = 0.0008 respectively (Fig. 1).

Finally, we assessed whether the reported associations between F-1mgDST 1.2–1.79 µg/dL and HT, DM, HT plus DM, and CVE were still statistically significant after adjusting for potential confounding factors. As reported in Table 3, F-1mgDST 1.2–1.79 µg/dL was found to be statistically associated with the presence of HT and DM even after adjusting for age, gender, OB, DL and DM (for HT) or HT (for DM), and with the presence of HT plus DM even after adjusting for age, gender, OB and DL. At variance, the presence of CVE was only statistically associated with age and not with gender, BMI, active smoking habit and F-1mgDST levels, even though for F-1mgDST levels a tendency toward a statistically significant association was perceivable.

Evidence from the literature suggests that some so-called “NFAT” may still present a partially autonomous cortisol secretion [27]. To date, no data have been available on the possibility that the degree of cortisol secretion in NFAT could be associated with the metabolic consequences known to be related with hypercortisolism, such as HT, DM, OB, DL and CVE. We, therefore, aimed to evaluate in a group of NFAT patients: (i) the possible association between HT, DM, OB, DL and CVE and ACTH and/or F-1mgDST, as estimates of cortisol secretion; (ii) the cut-off values of these parameters to identify NFAT patients with metabolic consequences.

We found that the presence of HT, DM and HT plus DM, but not of OB, DL and CVE was significantly associated with F-1mgDST levels after adjusting for age and comorbidities that could have biased the results. The cut-off with the best accuracy in identifying patients with either HT or DM or HT plus DM was found to be 1.2 µg/dL (33 nmol/L), although the low AUC values suggest a low accuracy of the associations found.

However, the present data could be of interest as they suggest that among AI patients so far diagnosed as not to be affected by hypercortisolism [1], some individuals might, in fact, have a partially autonomous cortisol secretion, which could be reflected in a higher prevalence of cardiovascular and metabolic consequences. These findings are consistent with previous data suggesting that not only in patients with MACS but also in some patients with NFAT, the removal of the adrenal mass could lead to the amelioration of HT and DM [14, 20]. In addition, the incidence of DM was suggested to be increased in patients with NFAT compared with patients without AI [17] and mortality to be similarly increased in NFAT and MACS patients [20]. Finally, the idea that some patients with NFAT may have a partially autonomous cortisol secretion is also reinforced by the notion that a post-surgical condition of hypocortisolism may anyway occur in up to 29% of patients with NFAT who underwent surgery for adenoma size [21, 28]. The lack of association of cortisol secretion with OB and DL is not surprising, as interventional studies suggest that cortisol hypersecretion in MACS has a negative effect particularly on HT and DM, whereas DL and OB are relatively less influenced [14]. At variance, the lack of the association between F-1mgDST and CVE after adjusting for age and other confounders may be due, at least in part, to the reduced sample size compared with the whole number of patients included in the study.

However, the present study also sought to answer the question of whether parameters of HPA axis activity could help in distinguishing NFAT patients with metabolic consequences. We found that ACTH levels were not associated with the presence of comorbidities in NFAT. The finding that ACTH is not a reliable index of the degree of cortisol secretion in AI was somewhat expected since even in patients with MACS, ACTH levels could not be used as the sole parameter of cortisol excess [29], probably because in MACS patients the degree of cortisol hypersecretion may not be sufficient to completely suppress the circadian ACTH secretion [30]. The same issue was therefore predictable to occur also in NFAT patients.

On the other hand, the present data show that a cut-off of F-1mgDST levels set at 1.2 µg/dL (33 nmol/L) is the threshold with the best accuracy in identifying patients with metabolic consequences of a possible cortisol excess. Again, the low AUC levels suggest that this association should be taken cautiously. However, it is of interest that, the same threshold of F-1mgDST was found to be predictive of hypocortisolism after the removal of an adrenal mass with a 100% sensitivity [21]. In general, the idea that the currently used F-1mgDST cut-off of 1.8 µg/dL (50 nmol/L) may not be fully reliable in identifying AI patients with possible hypercortisolism is reinforced by previous data. Indeed, in AI patients the best accuracy for predicting cardiovascular risk and insulin resistance was obtained by using a cut-off of cortisol after two days low dose dexamethasone suppression test set at 1.4 µg/dL (39 nmol/L) and 1.1 µg/dL (30 nmol/L) respectively [16]. Moreover, in a previous study by our group the F-1mgDST cut-off with the best compromise between sensitivity and specificity for predicting cardiovascular events in AI patients was found to be as low as 1.5 µg/dL (41 nmol/L) [22].

However, the potential utility of the F-1mgDST cut-off of 1.2 µg/dL (33 nmol/L) in identifying NFAT patients with metabolic consequences of cortisol excess is clearly debatable, given the low diagnostic accuracy (AUC ~ 0.6 for HT, DM and HT plus DM). This is probably because other variables may influence the relation between a slight hypercortisolism and its metabolic consequences in AI patients, such as the production of cortisol precursors with biological effects, the fluctuations in cortisol secretion [27] and the possible interindividual differences in cortisol sensitivity [31, 32].

This study has some limitations. Firstly, its retrospective and cross-sectional design prevents us to draw clear conclusions on the causal effect of cortisol hypersecretion on the outcomes considered (i.e. HT, DM, OB, DL and CVE). Indeed, longitudinal data (although retrospective) would have been of greatest interest, if a long-term follow-up (5 to 10 years) would have been performed. Unfortunately, since these patients were affected with benign NFAT, many of them have been discharged after about 2 years of follow-up, as suggested by the available guidelines [1], preventing us to have reliable long-term longitudinal data. Moreover, although the analyses have been corrected for confounding variables such as age, the age variable still may have an impact on the occurrence of HT and DM. Therefore, the lack of an age- and gender matched control group is a clear limitation of the study. However, given the well-known negative effects of hypercortisolism on these metabolic disorders, a possible causal role of a partial cortisol autonomy may be proposed even on the basis of the present cross-sectional and not-controlled study. Secondly, we do not have data regarding on the circadian rhythm of cortisol, but several data showed a low reliability of midnight salivary cortisol for the diagnosis of MACS and, therefore we do not expect a better performance of this parameter of HPA axis activity in this milder context. In addition, the fact that the available guidelines still suggest that only hypertensive or with unexplainable hypokaliemia AI patients should be screened for hyperaldosteronism [1], prevented us to have data on the aldosterone secretion in many of our patients. Therefore, we cannot exclude that some of these patients were affected by a mild hyperaldosteronism, that could have influenced the cardiometabolic complications.

Thirdly, we do not have data on blood dexamethasone levels to demonstrate the full reliability of F-1mgDST levels as a parameter of cortisol secretion. Finally, we cannot exclude that the high exclusion rate (from about 2500 patients with adrenal masses to 615 NFAT patients) could have determined a selection bias. However, since the main part of patients with adrenal masses has been excluded as affected by neoplastic disease and given, the anyway large sample size, in our opinion, an important confounding effect of the patient selection is unlikely.

Despite these limitations, this is the first study attempting to assess the possible presence of some degree of cortisol hypersecretion (autonomy) in a large sample of patients with what was defined "NFAT" and how to identify it.

In view of the higher cardiometabolic risk in NFAT patients with F-1mgDST 1.2–1.79 µg/dL than in those with F-1mgDST < 1.2 µg/dL (33 nmol/L), it is possible to hypothesize that the term NFAT could be not adequate to represent all AI patients defined so far as having a “non-functioning” adrenal tumour, as already suggested by others [27].

Even though still preliminary, the present data may, therefore, encourage large longitudinal studies aimed to better understand if some NFAT patients may have a relatively increased autonomous secretion of cortisol and/or of its precursors, how to individuate them and how this possible hypersecretion could influence the occurrence of cardio-metabolic complications even in relation to the interindividual difference in cortisol sensitivity.