Incidence and Prevalence
Epidemiological studies on all causes of CS are limited, with the majority of the studies including analysis of patients solely with pituitary CD [5, 7, 8, 17–20]. Only three epidemiological studies [1, 3, 21] have included all causes of CS. Our study is the first epidemiological study that includes patients with all causes of CS in Malta with detailed analysis of all patients.
The incidence of endogenous CS over the 12-year study period was 4.5 cases per million/year, in line with that reported in a study done in Israel [22] and close to that reported in a study done in Vastra Gotaland County in Sweden (3.2 cases per million/year) [3]. However, a lower incidence rate of endogenous CS was reported in other studies [1, 21]. In the study done in New Zealand [1] which included 253 patients with endogenous CS from 4 main endocrinology centres, the reported incidence rate was 1.8 cases per million/year, while in another study done in Denmark [21], where 166 patients with CS were included, the reported incidence rate of CS was 2.3 cases per million/year. Of note, in both studies by Lindholm et al. [21], and Bolland et al. [1], only patients with CS caused by non-malignant diseases were included, with possible underestimation of the overall incidence rate. Further details are shown in Table 5.
Table 5
Summary of previous studies analyzing the annual incidence rate of different causes of CS
Author (year of publication) [reference]
|
Country
|
Period
|
No of patients
|
Incidence Cushing’s Syndrome (/1,000,000/yr)
|
Incidence Cushing’s Disease (CD) (/1,000,000/YR)
|
Incidence Ectopic ACTH (/1,000,000/YR)
|
Incidence Benign Cortisol-Secreting Adenoma (/1,000,000/YR)
|
Incidence ACC (/1,000,000/YR)
|
Prevalence Cushing’s Syndrome (/1,000,000)
|
Comments
|
Etxabe & Vazquez 1994 [17]
|
Spain, Basque Country
|
1975-1992
|
49
|
|
2.4
|
|
|
|
39.1 (CD)
|
CD only
|
Lindholm et al. (2001)[21]
|
Denmark (nationwide)
|
1985-1995
|
166
|
incidence of CS caused by non-malignant causes = 2.3
|
1.2* (1.7)
|
0.1
|
0.1 (M)
1.1 (F)
|
0.6 (M)
0.2 (F)
|
|
*73 proven CD aetiology (total 99 diagnosed CD)
|
Raappana et al. (2010) [5]
|
Finland (northern)
|
1992-2007
|
19
|
|
1.7
|
|
|
|
|
Study on PAs
|
Arnardottir & Sigurjonsdottir (2011) [18]
|
Iceland (nationwide)
|
1955-2009
|
19
|
|
1.5
|
|
|
|
|
CD only
|
Clayton et al. (2011) [19]
|
UK, Stoke-on-Trent
|
1958-2010
|
60
|
|
1.5
|
|
|
|
|
CD only
|
Bolland et al. (2011) [1]
|
New Zealand (nationwide)
|
1960-2005
|
253
|
1.8
|
1.3
|
|
0.3
|
|
79
|
|
Tjörnstrand et al. (2014) [7]
|
Sweden, Vastra Gotaland County
|
2001-2011
|
25
|
|
1.8
|
|
|
|
|
*only patients with CD were included (study on PAs)
|
Broder et al. (2015) [20]
|
USA
|
2009-2010
|
522 (new case 2009)
537 (new case 2010)
|
48.6 (2009)
39.5 (2010)
|
7.6 (2009)
6.2 (2010)
|
|
|
|
|
|
Hirsch et al. (2017) [22]
|
Israel
|
2005-2014
|
85
|
4.5
|
|
|
|
|
|
|
Ragnarsson et al. (2019) [8]
|
Sweden (nationwide)
|
1987-2013
|
390
|
|
1.6
|
|
|
|
|
CD only; 534 patients included but 390 were diagnosed between 1987-2013; the others were diagnosed before 1987
|
Wengander et al. (2019) [3]
|
Sweden, Vastra Gotaland County
|
2002-2017
|
82
|
3.2
|
1.5
|
0.8
|
0.5
|
0.2
|
57
|
|
Current Study (2021)
|
Malta (nationwide)
|
2008-2020
|
35
|
4.5
|
2.3
|
0.5
|
1.5
|
0.3
|
63.9
|
|
A much higher incidence rate was reported in a study from USA [20], with an incidence rate of 48.6 cases per million/year in 2009 and 39.5 cases per million/year in 2010 for all causes of endogenous CS. The patients included in the study were obtained from information from an insurance database using diagnostic codes for CS. However, the medical records were not reviewed for validation of the diagnosis and this together with other methodological aspects likely led to overestimation of the incidence rate of endogenous CS.
The incidence of benign cortisol-secreting adrenal adenoma is slightly higher (1.5 cases per million/year) than that reported in most studies (0.1-0.5 cases per million/year) but is close to the incidence rate of benign cortisol-secreting adrenal adenomas in females reported by Lindholm et al. [21]. In contrast, we found a considerably higher incidence of CD (2.3 cases per million/year) than that reported in most studies (1.3-1.8 cases per million/year) [1, 3, 5, 7, 8, 18–21]. However, Arnardotir and Sigurionsdottir [18] have also reported a rising incidence from 1.5 to 2.6 per million per year over the last 5 years of their study between 2004 and 2009 whilst Etxabe and Vazquez [17] also found a rising incidence of CD over the years in their study period (from 1.5 to 3.9 per million/year).
The incidence of ectopic ACTH (0.5 cases per million/year) is comparable to the recent study done by Wengander et al. [3] (0.8 cases per million/year), showing a greater incidence than that previously reported in 2001 by Lindholm et al. (0.1 cases per million/year) [21]. On the contrary, the incidence of ACC (0.3 cases per million/year) is consistent to that reported in the previously mentioned studies despite different years of publication [3, 21].
The prevalence of endogenous CS in our study was found to be of 63.9 cases per million. This compares to that of 79 cases per million in New Zealand [1]. However, in a study done by Wengander et al. [3], the estimated prevalence was 57 cases per million. However, it is likely that the prevalence in the latter study was underestimated as patients diagnosed with CS and cured prior to the data collection period, were not included.
Radiological, Biochemical and Blood Count Indices in CS
During analysis of the different subtypes of patients with endogenous CS, there were statistically significant differences between potassium at diagnosis and size of tumour amongst the different subtypes and between benign and malignant causes of CS (Table 2 & 3). Patients with malignant CS had a larger tumour size and lower potassium at diagnosis which in turn correlated with a higher cortisol level. Fan et al. [23] similarly showed a correlation between potassium at diagnosis and cortisol in a group of patients with CD. Using ROC curve analysis, we could establish a cut-off value of potassium at diagnosis predicting a malignant cause for CS.
Few studies have investigated the changes in blood components and serum inflammation-based scores in CS [11–13, 24, 25]. Masri-Iraqi et al. [13] looked into the prevalence of leukocytosis in 26 patients with CD and investigated the changes in leukocytes and neutrophils following biochemical cure. Ambrogio et al. [25] studied 80 CD patients and reported on changes in red blood cells (RBCs) and WBC parameters in active Cushing’s and after surgical remission, while, Marques et al. [11] compared the pre-operative complete blood counts and various serum inflammatory-based scores in 424 patients with pituitary adenomas, of which 70 patients had CD. The only study, which included patients with all causes of endogenous CS, was done in children by Tatsi et al. [12], and investigated the changes in blood components before and after cure. This makes our study, to our knowledge, the first one to investigate the difference in biochemical and inflammatory parameters in adult patients with all causes of endogenous CS in a population-based epidemiological study and hence respecting the usual prevalence distribution of subtypes.
CS is known to be associated with immune system disruption resulting in polymorphonuclear leukocytosis, lymphopenia, low eosinophils and increased release of various cytokines including tumour necrosis factor alpha, interleukin-1 and interleukin-6, which contribute to a low-grade inflammatory state [10]. In our study, we report similar changes in the blood components in patients with all causes of endogenous CS with neutrophilia, lymphopenia, and low eosinophils. Such changes were noted to be significantly different from those after serum cortisol normalization, clearly exposing the role of cortisol in immune system regulation. Furthermore, we report a positive correlation between the parameters indicating severity of CS, and neutrophil count and NLR, and a negative correlation to lymphocyte, eosinophil count and LMR. Other studies have reported similar findings [11, 13], though these studies studied only patients with CD. The monocyte count was not found to be statistically significant in our cohort. Previous reports showed both a higher monocyte count in patients with CD [11] and a lower monocyte count in patients with CS [10] suggesting a complex relationship between hypercortisolism and monocyte count.
In literature, the effect of CS on platelets is not clearly defined with some studies showing a higher level of platelets [26, 27] whilst others show no significant change in patients with CS [28]. Marques et al. [11] found a higher platelet count in patients with CD when compared to non-functional pituitary adenomas and acromegaly, whilst a lower platelet count was associated with more invasive and refractory disease. In our study, the extent of glucocorticoid excess as defined by cortisol level post ODST, was also found to have a negative association with platelet count. Additionally, patients with malignant causes of CS were found to have a lower platelet count than those with benign causes.
Since all endogenous causes of CS were included in our study, we could report the differences in WBC indices and ratios between the various subtypes and between benign and malignant causes of CS (Table 2 & 3). Neutrophilia, lymphopenia and low eosinophils were significantly more pronounced in malignant causes of CS and correlated with serum and urine cortisol levels. This further confirms in vivo, the in vitro results reporting a strong correlation between cortisol and the inhibition of lymphocyte proliferation [29]. Moreover, using ROC curve analysis, we attempted to establish useful cut-off values (post-ODST cortisol and 24-hour urinary cortisol) which can help predict malignant causes of CS, which to our knowledge is a novel aspect in the diagnostic workup of CS.
Over the past years, there has been extensive research on the role of different blood components and serum inflammation-based scores on the long-term outcome of different cancers. In a review by Bugada et al. [30], various inflammation-based scores have been mentioned as potential preoperative prognostic markers in cancer patients. An increase in NLR, caused by high neutrophils and low lymphocytes, have been found to be related to the degree of cancer aggressiveness and adverse prognosis in many tumours including lung cancer, breast cancer, thyroid cancer, gastrointestinal cancers, urologic cancers, gynaecological cancers, glioblastoma multiforme and craniopharyngioma [30–35]. Other changes in blood components in malignancy include thrombocytosis, and thus a high PLR, and elevated monocytes which are also associated with poor prognosis in various tumours including gastric cancer, oesophageal, pancreatic, colorectal cancers and urologic cancers [30, 33]. However, the role of eosinophil count in malignancy is still uncertain as several studies that have shown both good and poor prognosis with eosinophilia [36, 37].
Different cancers cells from various tissues such as prostate, bladder, breast, kidney and pancreas have been shown in vitro to produce cortisol and suppress CD8+ T-lymphocyte proliferation suggesting a role of cortisol in carcinogenesis [29]. In our study we found that malignant causes of CS had a statistically significant higher production of cortisol with a significant greater suppression of lymphocytes and a higher NLR. Extending the in vitro results [29], we can hypothesize on a possible role of cortisol hypersecretion in the tumorigenesis of malignant causes of CS through its effect on immunomodulation. In malignant cancers, polymorphonuclear leukocytosis has been attributed to the increased production of cytokines and growth factors such as granulocyte-colony stimulating factors and vascular endothelial growth factor (VEGF) that promote neutrophil proliferation from the bone marrow leading to neutrophilia [30, 38]. Furthermore, chronic hypercortisolism, promotes the release of polymorphonuclear leukocytes (PMN) from the bone marrow, decreases apoptosis and prolonging their half-life, and decreases the expression of endothelial adhesion molecules on the surface of the neutrophils, thus reducing extravasation from the blood into peripheral tissues [10, 13]. The lymphocyte suppression observed in both malignancy and CS, is due to alteration in the CD4+ helper/CH8+ suppressors ratio and apoptosis of mature T-lymphocytes [10, 30, 39]. Hence, a greater degree of inflammation in malignant and aggressive cancers can possibly stimulate the secretion of a higher level of cortisol to inhibit the tumour-induced immune response, producing more profound changes in serum inflammatory markers, similar to those observed in malignant CS. This process leads to a negative series of events that build on and reinforce each other in the tumorigenesis process.
In our study, a NLR >3.9 was 100% sensitive in predicting malignant CS. In literature, several retrospective studies on different types of tumours, have used various cut-offs, ranging between 2.5 and 5, above which NLR was associated with higher mortality and decreased disease-free survival [30]. Our data thus suggests that the NLR can also be used in patients with CS to predict severity of disease, differentiating between benign and malignant causes, and prognosis.
The relatively small number of patients with CS included, represents a limitation in our study. In Malta, there is one NHS Hospital to which all patients are referred for evaluation, diagnosis and treatment. Hence, we are confident that all patients with suspected endogenous CS in the Maltese Islands were included. While we report a sensible incidence of ACC and ectopic CS, we cannot rule out that some of these patients may have been missed. It is possible that some patients with malignant CS, due to an atypical clinical picture and rapid course of disease, were never referred for an endocrine assessment, with the true incidence rate of both ACC and ectopic CS being somewhat underestimated. This relatively small number of patients may have provided insufficient statistical power to detect significant differences between different subtypes (benign vs malignant causes and pre- vs post-treatment) and certain blood indices. Another limitation is that the study is retrospective and the blood tests were done routinely and not for the study. Additionally, the patients may have had underlying undiagnosed conditions which may have influenced the biochemical and blood count indices.