Predictors of Cancer in Patients With Endogenous Cushing’s Syndrome

Posted on July 21, 2025 by MaryO

We previously reported an increase in overall cancer risk in patients with endogenous Cushing’s syndrome (CS), mainly during the 10-year period following CS diagnosis.

To identify predictors of cancer in patients with CS, we conducted this retrospective nationwide cohort study of patients with CS, diagnosed between 2000 and 2023 in Israel. The cohort comprised 609 patients with CS (age at diagnosis, 48.1 ± 17.2 years; 65.0% women) and 3,018 age-, sex-, socioeconomic status-, and body mass index-matched controls (1:5 ratio).

Patients were grouped according to the occurrence of any malignancy within 10-years after the diagnosis of CS. Cox proportional hazards models, with death as a competing event, were used to identify predictors of cancer development. Independent predictors of cancer development in patients with CS included age ≥60 years (HR 1.75, 95% CI 1.01–2.68), male gender (HR 1.67, 95% CI 1.04–3.05), and adrenal-origin CS (HR 1.66, 95% CI 1.01–2.73). Baseline urinary-free cortisol levels were not associated with cancer development. Patients with ≥4 CS-associated comorbidities had a higher cancer risk (HR 1.76, 95% CI 1.03–3.02; age- and sex-adjusted). The overall 10-year risk of malignancy was twice as high in patients with CS compared to matched controls, with cancer developing, on average, 5 years earlier in patients with CS (62.3 ± 15.0 vs 67.2 ± 12.3 years). Cancer-related mortality at 10-years was twice as high in deceased patients with CS, compared to deceased controls. In conclusion, age ≥60 years at CS diagnosis, male gender, and adrenal-origin CS are independent predictors of cancer diagnosis within 10-years of initial confirmation of CS.

 

Introduction

Prolonged cortisol exposure may promote cancer development and growth (Mayayo-Peralta et al. 2021, Khadka et al. 2023). Epidemiological research showed that extended glucocorticoids use is associated with elevated overall cancer risk (Oh & Song 2020). Recently, several studies suggested that cortisol levels increase cancer risk in patients with endogenous Cushing’s syndrome (CS). A Danish study found higher rates of cancer at the time of CS diagnosis compared to controls (Dekkers et al. 2013). A Swedish study examined comorbidity rates in patients with CS and identified a nonsignificant trend of increased cancer rates in CS compared to the general population, but was probably underpowered for this relatively rare outcome (Papakokkinou et al. 2020). Our nationwide retrospective matched-cohort study, using the Clalit Health Services (CHS) database in Israel (including 609 patients with CS and 3,018 age-, sex-, socioeconomic status- and body mass index (BMI)-matched controls), observed higher rates of all cancer types in patients with CS, with a hazard ratio (HR) of 1.78 (95% CI 1.44–2.20) (Rudman et al. 2024). Elevated cancer incidence was evident in patients with Cushing’s disease (CD) and in patients with adrenal CS. The overall cancer risk remained elevated during the first 10 years that followed CS diagnosis (Rudman et al. 2024). Similarly, a nationwide cohort study from Taiwan investigated the association between endogenous CS and cancer incidence, and reported a standardized incidence ration of 2.08 (95% CI 1.54–2.75) for cancer in patients with endogenous CS (Wu et al. 2025).
Hypercortisolemia and CS-associated comorbidities could drive malignancy development in patients with CS (Rudman et al. 2024, Wu et al. 2025). While it is known that the incidence of diabetes, obesity and insulin resistance is higher in patients with CS than that of the general population (Pivonello et al. 2016, Fleseriu et al. 2021, Reincke & Fleseriu 2023) – all of which are linked to cancer development (Renehan et al. 2008, Ling et al. 2020) – it remains unclear whether these comorbidities specifically contribute to the risk of malignancy within the CS population.
Thus, the aims of the present study were to identify the baseline predictors of cancer development in CS and to test the hypothesis that cumulative cortisol exposure, measured by urinary-free cortisol (UFC), predicts cancer risk in patients with CS.

Methods

Study design and data collection

We conducted a retrospective matched-cohort study using the electronic health record database of Clalit Health Services (CHS), the largest health maintenance organization (HMO) in Israel with over 4.8 million members. The CHS database includes demographic and clinical data, hospital and outpatient clinic diagnoses, medication dispensation, and all laboratory test results conducted at the HMO’s laboratories. All diagnoses and respective dates were identified using the International Classification of Diseases, tenth revision (ICD-10) codes (Supplementary Table S1 (see section on Supplementary materials given at the end of the article)). Weight and height data, and smoking status, were recorded regularly during visits to primary care clinics and in some specialized clinics. 24 h UFC results were collected and the normal reference range for each kit used. As these tests were performed by several different laboratories and devices (in all cases the bioanalytical method used was an immunoassay test), the results were reported as multiples of the upper limit of normal (×ULN). Data were extracted using the CHS research data-sharing platform, powered by MDClone. Importantly, the diagnoses of chronic medical conditions, recorded at the time of CS diagnosis, were validated: any member of CHS who required chronic treatment for a medical condition (e.g., medication for hypertension or diabetes) could only receive his prescriptions if the primary physician has registered the diagnosis, coded according to the ICD-10, in the computerized system. Mortality data were collected from the hospital’s mortality database, which is updated from the Ministry of Interior’s population registry. Data on cancer-specific mortality were obtained from hospital discharge certificates at the time of the hospitalization that ultimately resulted in the patient’s death. The study protocol, including detailed data collection methods, has been previously published (Rudman et al. 2024).

Ethical approval

The study was approved by the institutional ethics review board of Rabin Medical Center. As data were collected anonymously and in a retrospective manner, a waiver of informed consent was granted.

Patients and outcome measures

The methods we used for patient selection and matched controls selection have been previously published (Rudman et al. 2024), as the current study is based on the same group of patients with CS and controls. After the initial screening, potential cases with ICD-10 diagnosis of CS had to fulfill at least one of the following criteria: i) 24 h UFC ≥4 ×ULN, ii) 24 h UFC ≥3 ×ULN and surgical intervention to remove a pituitary or adrenal adenoma, and iii) 24 h UFC ≥2 ×ULN and metyrapone, ketoconazole, osilodrostat, cabergoline, or pasireotide treatment. All patients with CS and non-suppressed adrenocorticotropic hormone (ACTH) levels who did not receive pituitary-directed therapy and were diagnosed with a malignancy possibly causative of ectopic CS, including small-cell lung carcinoma, bronchial and thymic carcinoids, medullary thyroid carcinoma, neuroendocrine tumors or pheochromocytoma, were suspected of ectopic CS and were excluded from this study (Rudman et al. 2024). Patients diagnosed with adrenocortical carcinoma before or within 5 years of CS diagnosis were excluded.
All identified cases were individually matched in a 1:5 ratio with age-, sex-, socioeconomic status-, and BMI-matched controls from the general population (CHS members who have never been tested for suspected hypercortisolism). The age of the individually matched controls matched the age of cases ±12 months.
The follow-up period began at the time of CS diagnosis for all cases (newly diagnosed patients with CS) and at the exact same day for each individually matched control. It continued until death, termination of CHS membership, or until the date of data collection (June 30, 2023).
The main outcome measure was the first diagnosis of any malignancy following CS diagnosis, excluding non-melanoma skin cancer. Recurrences of known malignancies were also excluded.

Statistical analysis

The statistical analysis was generated using the SAS Software, Version 9.4, SAS Institute Inc., Cary, NC, USA. Continuous variables were presented by the mean ± standard deviation or median (interquartile range (IQR)). Categorical variables were presented by (n, %). Normality of continuous variables was assessed using the Kolmogorov–Smirnov test. The t-test, Mann–Whitney test, and chi-square test were used for comparison of normally distributed, non-normal, and categorical variables, respectively. The Cox proportional hazard model, with death without malignancy treated as a competing risk, was used to calculate both univariate and multivariate HR; the Fine and Gray methodology for dealing with competing risks was used, both in the cumulative incidence plots and in HR calculations. Baseline variables found to be associated with malignancy in the univariate analysis (with a between-group P-value below 0.05) were incorporated into the multivariate model. The appropriateness of the proportional hazard assumption was assessed visually. Two-sided P-values less than 0.05 were considered statistically significant.

Results

Patient characteristics

From January 1, 2000, to June 30, 2023, a total of 609 patients with CS met the study inclusion criteria (65.0% women, mean age at CS diagnosis of 48.1 ± 17.2 years). All cases of CS were matched with up to five controls based on age, sex, socioeconomic status, and BMI, and amounted to a total of 3,018 controls. Baseline characteristics of all 609 patients and 3,018 controls, with subdivisions according to disease source, are shown in Table 1.

Table 1. Baseline characteristics (at diagnosis/time 0) of patients with Cushing’s syndrome (CS) and matched control of all patients with CS, Cushing’s disease (CD), and adrenal CS.

Baseline characteristics, all patients with CS CS (n = 609) Matched controls (n = 3,018) P value CD (n = 251) Matched controls (n = 1,246) P value Adrenal CS (n = 200) Matched controls (n = 991) P value
Age, years, mean (SD) 48.1 (17.2) 47.9 (17.2) 45.7 (17.8) 45.7 (17.8) 51.1 (16.6) 51.0 (16.6)
Sex, no. (%)
 Females 396 (65.0) 1,975 (65.4) 164 (65.3) 818 (65.6) 137 (68.5) 684 (69.0)
 Males 213 (35.0) 1,043 (34.6) 87 (34.7) 428 (34.4) 63 (31.5) 307 (31.0)
Socioeconomic status, no. (%)a
 Low 74 (12.8) 371 (13.0) 34 (14.2) 172 (14.5) 20 (10.5) 99 (10.4)
 Middle 349 (60.6) 1,719 (60.3) 145 (60.7) 713 (60.2) 119 (62.6) 594 (62.9)
 High 153 (26.6) 760 (26.7) 60 (25.1) 300 (25.3) 51 (26.9) 252 (26.7)
Body mass index, Kg/m2, mean (SD)b 30.9 (7.6) 30.0 (6.9) 30.2 (7.4) 29.6 (6.8) 30.8 (7.1) 30.3 (6.3)
Source of hypercortisolism, no. (%)
 Cushing’s disease 251 (41.2)
 Adrenal Cushing’s syndrome 200 (32.8)
 Indeterminatec 158 (25.9)
Smoking status, no. (%)d 0.35 0.77 0.18
 Non-smoker 198 (59.8) 910 (62.6) 87 (64.0) 387 (65.3) 61 (53.0) 327 (60.1)
 Smoker/former smoker 133 (40.2) 544 (37.4) 49 (36.0) 206 (34.7) 54 (47.0) 217 (39.9)
Comorbidities, no. (%)
 Diabetes mellitus 140 (23.0) 396 (13.1) <0.01 55 (21.9) 148 (11.9) <0.01 51 (25.5) 158 (15.9) <0.01
 Hypertension 343 (56.3) 957 (31.7) <0.01 129 (51.4) 338 (27.1) <0.01 129 (64.5) 387 (39.0) <0.01
 Dyslipidemia 258 (42.4) 874 (29.0) <0.01 97 (38.6) 305 (24.5) <0.01 97 (48.5) 339 (34.2) <0.01
 Ischemic heart disease 70 (11.5) 191 (6.3) <0.01 23 (9.2) 67 (5.4) 0.03 25 (12.5) 77 (7.8) 0.04
 Cerebrovascular disease 27 (4.4) 82 (2.7) 0.04 12 (4.8) 35 (2.8) 0.11 10 (5.0) 32 (3.2) 0.21
 Osteoporosis 75 (12.3) 187 (6.2) <0.01 26 (10.4) 57 (4.6) <0.01 28 (14.0) 82 (8.3) 0.02
 Malignancy before CS diagnosis 50 (8.2) 117 (3.9) <0.01 15 (6.0) 48 (3.9) 0.12 20 (10.0) 50 (5.0) 0.01
Cases and controls were individually matched for age, sex, socioeconomic status, and body mass index.
a
Cushing’s syndrome n = 576, controls n = 2,850; Cushing’s disease n = 239, controls n = 1,185; adrenal Cushing’s syndrome n = 190, controls n = 945.
b
Cushing’s syndrome n = 363, controls n = 1,549; Cushing’s disease n = 152, controls n = 644; adrenal Cushing’s syndrome n = 131, controls n = 570.
c
Ectopic ACTH secretion and adrenocortical carcinoma were excluded.
d
Cushing’s syndrome n = 331, controls n = 1,454; Cushing’s disease n = 136, controls n = 593; adrenal Cushing’s syndrome n = 115, controls n = 544.
At baseline, diabetes mellitus, hypertension, dyslipidemia, ischemic heart disease, and osteoporosis were more common among patients with CS (P < 0.01). Smoking rates were similar between the two groups (Table 1). A prior history of malignancy was more prevalent among patients with CS than controls (8.2 vs 3.9%, respectively; P < 0.01) (Table 1).

Predictors of new malignancy in patients with Cushing’s syndrome

Table 2 presents the baseline characteristics of 609 patients with CS, including demographic data, CS etiology, history of malignancy before CS diagnosis, maximal UFC at diagnosis, and CS-associated comorbidities. In a univariate time-to-event analysis of the 10-year cumulative cancer risk, accounting for death as a competing event, we found that age ≥60 years at CS diagnosis, male gender, adrenal-origin CS, hypertension, dyslipidemia, and ischemic heart disease at baseline were all associated with a higher cancer risk. The 10-year cumulative cancer risk, with death as a competing event, stratified by age, sex, and CS etiology is shown in Fig. 1. Prior malignancy, diabetes, and obesity were not associated with an increased risk of malignancy in patients with CS (Table 2). The cohort was divided into three groups based on baseline UFC level (below 5 ×ULN, 5–10 ×ULN, and above 10 ×ULN) with comparison of time to cancer occurrence. Higher UFC levels at the time of CS diagnosis were not associated with cancer development (Table 2 and Fig. 2). In a multivariable Cox regression model (multivariable model 1, Table 2), we found that age ≥60 years at CS diagnosis (HR 1.75, 95% CI 1.01–2.68), male gender (HR 1.67, 95% CI 1.04–3.05), and adrenal-origin CS (HR 1.66, 95% CI 1.01–2.73) were independent predictors of cancer development within 10 years after CS diagnosis. In an additional model (multivariable model 2, Table 2), we observed that patients with ≥4 CS-associated comorbidities at the time of CS diagnosis had a higher risk of cancer (HR 1.76, 95% CI 1.03–3.02), after adjustment for age and sex.

Table 2. Univariate analysis and multivariable regression models for the 10-year cumulative cancer risk in patients with Cushing’s syndrome, accounting for death as a competing event.

Baseline characteristics Patients (n = 609) Incident cases of cancer (n = 81) Deaths without cancer (n = 40) Univariate analysis Multivariable model 1* Multivariable model 2 (total no. of CS-associated comorbidities, with age and sex adjustment)
Age <60 years (ref) 444 44 10 – – –
Age ≥60 years 165 37 30 2.47 (1.60–3.82) 1.75 (1.01–2.68) 3.18 (2.09–4.86)
Female (ref) 396 42 23 – – –
Male 213 39 17 1.75 (1.13–2.70) 1.67 (1.04–3.05) 1.60 (1.11–2.29)
Low SESa 74 13 5 1.29 (0.65–2.56)
Medium SESa 349 42 21 0.87 (0.52–1.45)
High SES (ref)a 153 22 13 –
Cushing’s disease (ref)b 251 27 21 – –
Adrenal CSb 200 39 8 1.87 (1.14–3.05) 1.66 (1.01–2.73)
Non-smoker (ref)c 198 25 12 –
Smoker/former smokerc 133 21 12 1.25 (0.70–2.23)
Prior malignancy 50 6 14 0.91 (0.40–2.08)
No prior malignancy (ref) 559 75 26 –
Maximal urinary free cortisold
 <5 × ULN (ref) 231 38 18 –
 5–10 × ULN 135 18 7 0.86 (0.49–1.48)
 ≥10 × ULN 70 10 4 0.89 (0.45–1.78)
CS-associated comorbidities
 Obesitye 176 28 17 1.22 (0.71–2.11)
 No obesity (ref)e 187 24 12 –
 Diabetes mellitus 140 21 14 1.28 (0.78–2.10)
 No diabetes mellitus (ref) 469 60 26 –
 Hypertension 343 59 34 2.23 (1.36–3.65) 1.52 (0.83–2.81)
 No hypertension (ref) 266 22 6 – –
 Dyslipidemia 258 44 29 1.81 (1.17–2.81) 0.97 (0.55–1.72)
 No dyslipidemia (ref) 351 37 11 – –
 Ischemic heart disease 70 19 11 2.70 (1.62–4.51) 1.48 (0.81–2.71)
 No ischemic heart disease (ref) 539 62 29 –
 Stroke 27 3 2 1.07 (0.33–3.49)
 No stroke (ref) 582 78 38 –
 Osteoporosis 75 9 9 0.92 (0.46–1.84)
 No osteoporosis (ref) 534 72 31 –
Total no. of CS-associated comorbiditiesf
 0–1 (ref) 295 27 9 – –
 2–3 205 33 15 2.09 (1.36–3.22) 1.40 (0.87–2.26)
 ≥4 109 21 16 3.76 (2.37–5.97) 1.76 (1.03–3.02)
CS, Cushing’s syndrome; SES, socioeconomic status; ULN, upper limit of normal. Bold indicates statistical significance.
*
Included variables: age, sex, source of hypercortisolism, hypertension, dyslipidemia, and ischemic heart disease.
a
n = 576.
b
Ectopic ACTH secretion and adrenocortical carcinoma were excluded.
c
n = 331.
d
Maximal value of urinary-free cortisol divided by the upper limit of normal of the specific assay, n = 436.
e
n = 363.
f
CS-associated comorbidities include obesity, diabetes mellitus, hypertension, dyslipidemia, ischemic heart disease, and osteoporosis.
Figure 1

  1. Download: Download high-res image (201KB)
  2. Download: Download full-size image

Figure 1. The 10-year cumulative cancer risk, with death as a competing event, among patients with Cushing’s syndrome, according to age at diagnosis (A), sex (B), and Cushing’s syndrome etiology (C). CD, Cushing’s disease; CS, Cushing’s syndrome. A full colour version of this figure is available at https://doi.org/10.1530/ERC-25-0059.

Figure 2

  1. Download: Download high-res image (61KB)
  2. Download: Download full-size image

Figure 2. The 10-year cumulative cancer risk, with death as a competing event, among patients with Cushing’s syndrome, according to the maximal value of UFC divided by the upper limit of normal (ULN) of the specific assay used: UFC <5 × ULN (reference), 5–10 × ULN, and ≥10 × ULN. A full colour version of this figure is available at https://doi.org/10.1530/ERC-25-0059.

Univariate analysis and multivariable regression models for the 10-year cumulative cancer risk in patients with CD and adrenal CS are presented in Tables 3 and 4. In the univariate time-to-event analysis of 251 patients with CD, age ≥60 years at CS diagnosis and the presence of dyslipidemia and ischemic heart disease at baseline were associated with higher cancer risk. The multivariable Cox regression model did not identify any significant predicting factors in patients with CD. Tables 3 and 4 also present the univariate analysis for the 10-year cumulative cancer risk in 200 patients with adrenal CS, which showed that age ≥60 years, male gender, and ischemic heart disease were associated with cancer development. In the multivariable model, only age ≥60 years at CS diagnosis (HR 2.66, 95% CI 1.36–5.18) was found to be independently associated with cancer development in patients with adrenal CS (Table 4). UFC levels at the time of CS diagnosis were not associated with new cancer diagnosis in either patients with CD or with adrenal CS (Tables 3 and 4).

Table 3. Univariate analysis and multivariable regression models for the 10-year cumulative cancer risk in patients with Cushing’s disease, accounting for death as a competing event.

Cushing’s disease baseline characteristics Patients (n = 251) Incident cases of cancer (n = 27) Deaths without cancer (n = 21) Univariable Multivariable model 1 Multivariable model 2 (total no. of CS-associated comorbidities, with age and sex adjustment)
Age <60 years (ref) 196 16 8 – – –
Age ≥60 years 55 11 13 2.72 (1.27–5.81) 1.83 (0.65–5.18) 3.54 (1.81–6.92)
Female (ref) 164 14 14 – – –
Male 87 13 7 1.79 (0.85–3.80) 1.61 (0.73–3.56) 1.22 (0.68–2.18)
Low SESa 34 5 5 1.29 (0.42–3.99)
Medium SESa 145 13 12 0.81 (0.32–2.00)
High SES (ref)a 60 7 4 –
Non-smoker (ref)b 87 6 6 –
Smoker/former smokerb 49 8 8 2.59 (0.91–7.39)
Prior malignancy 15 26 15 0.66 (0.09–5.13)
No prior malignancy (ref) 236 1 6 –
Maximal urinary-free cortisolc
 <5 × ULN (ref) 133 15 14 –
 5–10 × ULN 70 6 3 0.78 (0.31–2.00)
 ≥10 × ULN 40 6 4 1.48 (0.57–3.79)
CS-associated comorbidities
 Obesityd 69 7 10 0.80 (0.31–2.09)
 No obesity (ref)d 83 10 6 –
 Diabetes mellitus 55 5 6 0.89 (0.34–2.35)
 No diabetes mellitus (ref) 196 22 15 –
 Hypertension 129 17 17 1.67 (0.76–3.64)
 No hypertension (ref) 122 10 4 –
 Dyslipidemia 97 15 13 2.31 (1.08–4.96) 1.64 (0.58–4.61)
 No dyslipidemia (ref) 154 12 8 – –
 Ischemic heart disease 23 5 4 2.71 (1.03–7.08) 1.29 (0.43–3.86)
 No ischemic heart disease (ref) 228 22 17 – –
 Stroke 12 2 1 2.34 (0.53–10.30)
 No stroke (ref) 239 25 20 –
 Osteoporosis 26 24 15 1.14 (0.35–3.66)
 No osteoporosis (ref) 225 3 6 –
Total no. of CS-associated comorbiditiese
 0–1 (ref) 136 11 6 – –
 2–3 74 10 8 2.19 (1.13–4.26) 1.48 (0.72–3.04)
 ≥4 41 6 7 3.69 (1.78–7.66) 1.72 (0.73–4.02)
CS, Cushing’s syndrome; SES, socioeconomic status; ULN, upper limit of normal. Bold indicates statistical significance.
a
n = 239.
b
n = 136.
c
Maximal value of urinary free cortisol divided by the upper limit of normal of the specific assay; n = 243.
d
Cushing’s disease, n = 152.
e
CS-associated comorbidities include obesity, diabetes mellitus, hypertension, dyslipidemia, ischemic heart disease, and osteoporosis.

Table 4. Univariate analysis and multivariable regression models for the 10-year cumulative cancer risk in patients with adrenal Cushing’s syndrome, accounting for death as a competing event.

Adrenal Cushing’s syndrome baseline characteristics Patients with CS at risk (n = 200) Incident cases of cancer (n = 39) Deaths without cancer (n = 8) Univariable Multivariable model 1 Multivariable model 2 (total no. of CS-associated comorbidities, with age and sex adjustment)
Age <60 years (ref) 134 20 1 – – –
Age ≥60 years 66 19 7 2.12 (1.13–3.96) 2.66 (1.36–5.18) 2.70 (1.35–5.42)
Female (ref) 137 17 3 – – –
Male 63 22 5 2.97 (1.58–5.58) 1.81 (0.94–3.51) 3.11 (1.73–5.57)
Low SESa 20 3 0 0.62 (0.17–2.28)
Medium SESa 119 22 3 0.73 (0.36–1.44)
High SES (ref)a 51 13 5 –
Non-smoker (ref)b 61 12 3 –
Smoker/former smokerb 54 10 2 0.86 (0.38–1.99)
Prior malignancy 20 4 2 1.03 (0.38–2.81)
No prior malignancy (ref) 180 35 6 –
Maximal urinary free cortisolc
 <5 × ULN (ref) 98 23 4 –
 5–10 × ULN 65 12 4 0.84 (0.42–1.69)
 ≥10 × ULN 30 4 0 0.53 (0.18–1.52)
CS-associated comorbidities
 Obesityd 64 15 4 1.68 (0.75–3.74)
 No obesity (ref)d 67 10 3 –
 Diabetes mellitus 51 12 3 1.47 (0.75–2.89)
 No diabetes mellitus (ref) 149 27 5 –
 Hypertension 129 29 7 1.73 (0.84–3.58)
 No hypertension (ref) 71 10 1 –
 Dyslipidemia 97 20 7 1.20 (0.65–2.25)
 No dyslipidemia (ref) 103 19 1 –
 Ischemic heart disease 25 10 4 2.65 (1.31–5.34) 1.51 (0.68–3.32)
 No ischemic heart disease (ref) 175 29 4 – –
 Stroke 10 1 1 0.58 (0.08–4.35)
 No stroke (ref) 190 38 7 –
 Osteoporosis 28 2 2 0.32 (0.08–1.33)
 No osteoporosis (ref) 172 37 6 –
Total no. of CS-associated comorbiditiese
 0–1 (ref) 82 14 1 – –
 2–3 76 15 2 1.24 (0.62–2.48) 0.95 (0.44–2.02)
 ≥4 42 10 5 2.46 (1.20–5.05) 1.21 (0.51–2.85)
CS, Cushing’s syndrome; SES, socioeconomic status; ULN, upper limit of normal. Bold indicates statistical significance.
a
n = 190.
b
n = 115.
c
Maximal value of urinary-free cortisol divided by the upper limit of normal of the specific assay; n = 193.
d
n = 131.
e
CS-associated comorbidities include obesity, diabetes mellitus, hypertension, dyslipidemia, ischemic heart disease, and osteoporosis.

The 10-year cancer risk in patients with Cushing’s syndrome vs controls

In the 10 years following CS diagnosis, 81 (13.3%) patients with CS were diagnosed with cancer and 40 (6.6%) died without malignancy, compared with 206 (6.8%) and 152 (5.0%) controls, respectively. Similar to the previously reported risk for the entire follow-up period (Rudman et al. 2024), the overall 10-year risk of malignancy, calculated with death as a competing event, was twice as high in patients with CS than in matched controls (HR 2.01; 95% CI, 1.55–2.60) (Supplementary Fig. S1).
The mean age of cancer development in patients with CS was 62.3 ± 15.0 years, compared with 67.2 ± 12.3 years in controls (P < 0.01). The risk of cancer across different subgroups (patients with CS vs controls) is shown in Fig. 3 and Table 5. The number of cases for each type of cancer in patients with CS and controls is shown in Supplementary Table S2.
Figure 3

  1. Download: Download high-res image (222KB)
  2. Download: Download full-size image

Figure 3. The 10-year cancer risk in subgroups of the entire cohort (cases vs matched controls). Cases and controls were individually matched for age, sex, socioeconomic status, and body mass index.

Table 5. The 10-year cancer risk in subgroups of the entire cohort (cases vs matched controls).

Subgroup Cushing’s syndrome Individually matched controls HR 95% CI
Patients Incident cases of cancer Deaths Patients Incident cases of cancer Deaths
Age
 <60 444 44 10 2,200 84 30 2.67 1.85–3.85
 ≥60 165 37 30 818 122 122 1.55 1.08–2.24
Sex
 Females 396 42 23 1,975 117 89 1.83 1.29–2.61
 Males 213 39 17 1,043 89 63 2.25 1.54–3.28
Socioeconomic status
 Low 74 13 5 371 14 17 5.03 2.37–10.68
 Middle 349 42 21 1,719 116 90 1.83 1.28–2.60
 High 153 22 13 760 73 36 1.52 0.94–2.45
Smoking status
 Smoker/former smoker 133 21 12 544 46 34 1.84 1.09–3.10
 Non-smoker 198 25 12 910 66 48 1.79 1.13–2.83
Comorbidities
 Obesity 176 28 17 698 45 62 2.52 1.57–4.04
 No obesity 187 24 12 851 70 46 1.58 0.99–2.51
 Diabetes mellitus 140 21 14 396 45 75 1.35 0.80–2.25
 No diabetes mellitus 469 60 26 2,622 161 77 2.13 1.58–2.86
 Hypertension 343 59 34 957 106 126 1.59 1.16–2.19
 No hypertension 266 22 6 2,061 100 26 1.71 1.07–2.72
 Dyslipidemia 258 44 29 874 100 100 1.53 1.07–2.17
 No dyslipidemia 351 37 11 2,144 106 52 2.15 1.47–3.13
 Ischemic heart disease 70 19 11 191 30 49 1.89 1.06–3.35
 No ischemic heart disease 539 62 29 2,827 176 103 1.88 1.41–2.52
 Stroke 27 3 2 82 7 17 1.50 0.39–5.74
 No stroke 582 78 38 2,936 199 135 2.02 1.56–2.63
 Osteoporosis 75 9 9 187 23 32 0.98 0.45–2.10
 No osteoporosis 534 72 31 2,831 183 120 2.15 1.64–2.83
Cases and controls were individually matched for age, sex, socioeconomic status, and body mass index.
Among 487 cases and 2,411 controls with an attainable follow-up period of at least 10 years, 52 patients with CS and 184 controls died (from any cause) during the 10 years that followed CS diagnosis. Eight (15.4%) patients with CS died due to malignancy, compared with 12 (6.5%) patients in the control group (P = 0.04).
During the 10-year follow-up after CS diagnosis, 27 out of 251 patients with CD (10.8%) were diagnosed with malignancy, compared to 71 (5.7%) controls. Among 200 patients with adrenal CS, 39 (19.5%) were diagnosed with cancer, compared to 79 (8.0%) controls. The 10-year risk of overall malignancy was higher in patients with CD (HR 1.92, 95% CI 1.23–3.00) and in patients with adrenal CS (HR 2.63, 95% CI 1.79–3.87), compared to controls (Supplementary Fig. S1). The number of cases for each specific cancer type in patients with CD and adrenal CS and their individually matched controls is elaborated in Supplementary Table S2.

Sensitivity analyses

Due to possible bias in individuals with a genetic predisposition to cancer, and in patients at increased risk due to prior cancer treatment, we excluded all patients with prior history of cancer (50 cases and 117 controls). Following this exclusion, patients with CS still exhibited a higher 10-year cancer risk (HR 2.12, 95% CI 1.62–2.77).
Patients with adrenal cancer diagnosed before or within 5 years of CS diagnosis were excluded from the study. However, as it is possible that adrenal cancer was either not recorded properly or unrecognized at the time of CS diagnosis, we performed an analysis of the risk of malignancy excluding all adrenal cancer cases and found no change in the 10-year risk of overall cancer (HR 1.92, 95% CI 1.48–2.50).
The diagnosis of CS patients in the study included an ICD-10 coding of the diagnosis and laboratory evidence of hypercortisolism (and test date) in all cases. However, because in many cases the diagnostic and treatment process are lengthy, there is a possibility of information bias caused by patients included in the database who were diagnosed before the time period included in the study. Therefore, we performed a sensitivity analysis excluding the first year of the study, without any change in the 10-year risk of malignancy among CS patients (HR 1.98, 95% CI 1.51–2.58).

Discussion

Patients with CS have higher morbidity and mortality (Gadelha et al. 2023, Loughrey et al. 2024), and it has been recently established that CS is associated with an increased cancer risk (Rudman et al. 2024, Wu et al. 2025). However, predictors of a new cancer diagnosis have not been studied. In this nationwide retrospective study, the 10-year cancer risk in 609 patients with CS was twice as high as in 3,018 matched controls. Importantly, the 10-year risk was notably higher in patients with CD (HR 1.92, 95% CI 1.23–3.00) and in those with adrenal CS (HR 2.63, 95% CI 1.79–3.87), compared to controls. Furthermore, the risk of cancer was higher in patients with CS, regardless of age and sex. On average, cancer development in patients with CS occurred at an age that was 5 years younger than that of controls who developed cancer (62.3 ± 15.0 vs 67.2 ± 12.3 years, respectively).
Our study is the first to identify predictors of new cancer diagnosis in patients with CS. A multivariate regression model showed that age ≥60 years at CS diagnosis (HR 1.75, 95% CI 1.01–2.68), male gender (HR 1.67, 95% CI 1.04–3.05), and adrenal-origin CS (HR 1.66, 95% CI 1.01–2.73) were identified as independent predictors of cancer development within 10 years. In addition, we found that patients with ≥4 CS-associated comorbidities at the time of CS diagnosis had an increased risk of cancer (HR 1.76, 95% CI 1.03–3.02; adjusted for age and sex). Interestingly, diabetes and obesity were not associated with malignancy development in patients with CS. Importantly, we found no association between UFC levels at the time of CS diagnosis and cancer development rates.
CS most commonly affects young women, a population not inherently at high risk for malignancy, with the exception of breast cancer (National Cancer Institute, Surveillance, Epidemiology, and End Results (SEER) Program, December 2024. https://seer.cancer.gov/statfacts/html/aya.html). Our study demonstrates that young patients and female patients with CS are at an increased risk of cancer, as compared with matched controls from the general population. However, within the group of patients with CS, we found age and sex disparities in malignancy risk: men and elderly patients (over 60 years of age) showed a higher cancer risk (Table 2). Advanced age is a universal risk factor for cancer (Campisi 2013), and patients with CS are no exception. Previous studies found that male patients with CS are more susceptible to metabolic derangements than female patients (Liu et al. 2015, Broersen et al. 2019), a difference that likely results from gender disparity in response to glucocorticoid receptor activation (Bourke et al. 2012).
In addition, our study found that CS of adrenal origin is associated with a higher risk of malignancy, as compared with CD, after adjustment for age, sex, and significant CS-related comorbidities. Notably, patients with a history of adrenal cancer or ectopic CS were excluded. This finding is difficult to explain, since most studies have found that patients with CD present with higher UFC levels (Berr et al. 2015, Rubinstein et al. 2019, Schernthaner-Reiter et al. 2019) and a longer delay in diagnosis (Rubinstein et al. 2019, Schernthaner-Reiter et al. 2019) compared to those with adrenal CS. One potential explanation for this observation is that adrenal adenomas may be linked to a higher incidence of malignancy, as studies have shown that cancer mortality is increased with autonomic cortisol secretion, with malignancy being the most common cause of death in patients with mild autonomous cortisol secretion (Patrova et al. 2017, Deutschbein et al. 2022). Another conceivable explanation stems from previous research that reported higher rates of non-adrenal malignancies in patients with bilateral adrenal tumors and autonomous cortisol secretion (Kawate et al. 2014), suggesting a possible genetic predisposition in patients with adrenal adenoma that may contribute to the development of overall cancer.
Interestingly, in our study, patients with adrenal CS had a history of malignancy at a higher rate than their individually matched controls at the time of CS diagnosis (Table 1). In contrast, no difference in the rate of malignancy was found between patients with CD and controls. Although it is possible that a prior history of malignancy contributed to the higher risk of cancer observed in patients with adrenal CS, we did not find that a prior malignancy predicted subsequent cancer risk in this population when we analyzed our cohort of patients with adrenal CS (Table 4).
In this study, we found no association between the cumulative exposure to excess glucocorticoids (measured as UFC levels) and the development of malignancy (Fig. 2), but we did identify an association between the total number of CS-related comorbidities and cancer risk (adjusted for age and sex) (Table 2). Previous studies have similarly shown no correlation between the degree of hypercortisolism and the presence of CS-related comorbidities in patients with CS (Schernthaner-Reiter et al. 2019), including diabetes and obesity (Giordano et al. 2014, Bavaresco et al. 2024). Those findings support the hypothesis that individual sensitivity to glucocorticoids varies across tissues, such that UFC levels do not always correlate with symptom burden or comorbidities. Patients who are more sensitive to excess cortisol may experience a broader range of CS-associated comorbidities. Several genetic mutations and alterations have already been identified as causes of variation in cortisol sensitivity, including the genes encoding the human glucocorticoid receptor (NR3C1) (Chrousos et al. 1982, Riebold et al. 2015, Laulhé et al. 2024), the chaperone protein that regulates proper folding of the glucocorticoid receptor (HSP90) (Riebold et al. 2015), and the nuclear protein that modulates glucocorticoid receptor actions (NR2C2) (Zhang et al. 2016). In addition, mutations in glucocorticoid response elements (Vandevyver et al. 2013), variations in RNA-binding to the glucocorticoid receptor (Lammer et al. 2023), and epigenetic changes (Paes et al. 2024) may also play a role in inter-individual differences in response to cortisol excess.
In the univariate model we have performed, the total number of CS-related comorbidities was associated with cancer development, and the risk of malignancy increased with the number of comorbidities. However, after adjustment for age and sex, the HR was significantly moderated (mainly due to a strong correlation between age and comorbidity) but remained graded. We find this observation to support the concept that cancer is a CS-associated comorbidity, and suggest that patients with CS (especially older men with adrenal CS) suffering from multiple disease-related comorbidities require closer follow-up and a rigorous age-adjusted cancer screening, in accordance with guidelines for the general population.
We have previously reported an increased risk of genitourinary, thyroid, and gynecological cancers in patients with CS (Rudman et al. 2024). A Taiwanese national cohort study reported that liver (27.7%), kidney (16.7%), and lung (13.0%) cancers were the most common cancers among patients with CS (Wu et al. 2025). Despite the small absolute number of cases in each cancer type in this study, we found that the incidence in patients with CS was higher across all cancer groups, except for malignant melanoma. One might think that patients with CS underwent more imaging and laboratory tests, and therefore more cases of low-risk cancers (e.g., clinically insignificant prostate or thyroid cancer) were diagnosed in patients with CS than in controls. However, as we have shown, the overall 10-year malignancy-associated mortality was twice as high in patients with CS compared to controls, indicating that malignancies in this group were clinically significant.
Surgery for CS, especially for CD, improves some but not all comorbidities (Dekkers et al. 2013, Terzolo et al. 2014, Papakokkinou et al. 2020, Puglisi et al. 2024). Improvement of comorbidities with medical therapy have been noted in several clinical trials (Fleseriu et al. 2012, 2022, Petersenn et al. 2017); however, there are no prospectively collected data on the risk of cancer in these patients treated long-term. A retrospective study examining the course of several CS-related comorbidities showed that the risk of cancer in patients with CS who did not achieve remission was higher compared to the risk of cancer for patients in remission, yet these analyses did not reach statistical significance, partly due to the limited sample size (Papakokkinou et al. 2020).
In order to successfully identify predictors of cancer in patients with CS, this research of an uncommon outcome (malignancy) in patients with a rare disease (CS) required a long-term follow-up of a large, population-representative cohort, paired with well-matched control group. Matching for socioeconomic status is another strength of this study, as its impact on morbidity has recently been demonstrated in several studies (Ebbehoj et al. 2022, Claudel & Verma 2024).
However, this study has limitations. Missing data prevented us from determining the specific CS etiology in some patients. Correct classification of all cases with an indeterminate diagnosis (as either CD or adrenal CS) would have allowed us to improve the power of subgroup analysis of patients with CD and adrenal CS; however, we had very strict criteria for determining the etiology of CS. Not all data regarding socioeconomic status, BMI, and smoking status were available. In addition, the impact of hypopituitarism and overreplacement of glucocorticoids in patients with CD could not be assessed.
Since the control group was drawn from the general population, ascertainment bias cannot be ruled out, as it is likely that patients with CS underwent more physician-initiated imaging and laboratory tests, and therefore more cases of cancer could have been diagnosed in patients with CS than in controls. However, we consider this bias to be unlikely for most cases of aggressive cancer, especially given our long follow-up period.
While this nationwide study includes a relatively large sample size, we acknowledge that it is likely that our current sample size was not sufficiently powered to detect risk predictors that are only modestly associated with malignancy risk. The small sample size of subgroups and the low frequency of the outcome in these subgroups meant we were unable to predict malignancy in patients with CD or adrenal CS, nor could we estimate the risk of specific malignancies. Moreover, we could not account for certain factors that may influence the risk of malignancy, such as family history of malignancy, duration of exposure to elevated cortisol levels, and the presence of genetic syndromes that predispose individuals to both CS and certain malignancies (e.g., multiple endocrine neoplasia type 1) (Hernández-Ramírez & Stratakis 2018). Finally, the lack of systematic prospective assessment of comorbidities is an important limitation and should raise the standards for future clinical care of these patients and collecting data in new registries. While patients receiving treatment for a particular comorbidity were successfully identified, those without treatment were not systemically recorded, which may have led to underreporting. Such is the case with osteoporosis: only patients who received treatment or whose treating physician decided to send them for a bone density scan were diagnosed, while others without such evaluations were assumed to be free of osteoporosis.
In conclusion, this large nationwide retrospective matched-cohort study found that the risk of cancer was consistently higher in patients with CS, regardless of age or sex, and on average, cancer development occurred 5 years earlier in patients with CS than in controls. The multivariate regression model we developed identified age ≥60 years at CS diagnosis, male gender, and CS of adrenal-origin as independent predictors of malignancy during the 10 years following CS diagnosis. Importantly, we found no association between UFC levels at CS diagnosis and cancer development rates. However, patients with ≥4 CS-associated comorbidities at CS diagnosis were more likely to develop cancer, after adjusting for age and sex. Given previous studies that identified overall cancer as a CS-related comorbidity and as one of the leading causes of death in this population, the results of the current study will help identify patients at high risk of malignancy, emphasize the importance of timely screening tests, in accordance with guidelines for the general population, and highlight the need for larger international cohorts to establish specific cancer screening recommendations for patients with CS.

Supplementary materials

This is linked to the online version of the paper at https://doi.org/10.1530/ERC-25-0059.

Declaration of interest

Yaron Rudman, Genady Drozdinsky, Hiba Masr-Iraqi, Tzippy Shochat, and Shiri Kushnir do not have any financial or personal relationships with other people or organizations to disclose. Maria Fleseriu has been a PI with research funding to the university from Crinetics and Sparrow and has received occasional scientific fees for scientific consulting and advisory boards from Crinetics, Recordati, Sparrow and Xeris. Ilan Shimon has been an investigator for Xeris Biopharma and has received occasional scientific fees for scientific consulting and advisory boards from Medison, CTS pharma, and Neopharm. Amit Akirov has received occasional scientific fees for scientific consulting and advisory boards from Medison, CTS pharma, and Neopharm.

Funding

This work did not receive any specific grant from any funding agency in the public, commercial or not-for-profit sector.

Data availability

The data that support the findings of this study are available from Clalit Health Services. Restrictions apply to the availability of these data, which were used under license for this study. Deidentified individual participant-level data sharing will be considered by the corresponding author of this study, with the permission of Clalit Health Services. All applicants will be asked to sign a data access agreement. All requests will be assessed as to whether data sharing is appropriate, based on the scientific rigor of the proposal.

References

Filed under: Cancer, Clinical trials, Cushing's, Treatments | Tagged: , , , , , , | Leave a comment »

Can We Predict the Risk of Venous Thromboembolism in Patients With Cushing’s Syndrome

Posted on January 22, 2025 by MaryO

Purpose

Patients with Cushing’s syndrome (CS) have an increased venous thromboembolism (VTE) risk with most studies focusing on the perioperative period. The purpose of this study was to assess the 5-year VTE risk and identify predictors of VTE at CS diagnosis.

Methods

A comparative nationwide retrospective cohort study of 609 patients (mean age 48.1 ± 17.2 years, 65.0% women) with CS, and 3018 age-, sex-, body mass index-, and socioeconomic status-individually matched controls. Ectopic CS and adrenal cancer were excluded. The time-to-event of pulmonary embolism (PE) or deep vein thrombosis (DVT) within 5 years of CS diagnosis was examined. VTE risk was calculated with death as competing event.

Results

VTE occurred in 16 cases (2.6%), compared to 17 (0.56%) controls (hazard ratio [HR] 4.71, 95% CI, 2.38–9.33). The 5-year HRs for PE and DVT were 7.47 (95% CI, 2.66–20.98) and 3.32 (95% CI, 1.36–8.12), respectively. After excluding patients and controls with current or prior malignancy the risk for VTE was 7.57 (95% CI, 2.98–19.20). Patients with CS ≥ 60 years at diagnosis (HR, 3.49; 95% CI, 1.30–9.35), with hypertension (HR, 5.53; 95% CI, 1.26–24.27), ischemic heart disease (HR, 3.60; 95% CI, 1.25–10.36), kidney disease (HR, 4.85; 95% CI, 1.39–16.90), or VTE event prior to CS diagnosis (HR, 33.65; 95% CI, 10.07–112.42) had an increased risk of VTE within five years.

Conclusions

In this large cohort of patients with CS, the 5-year VTE risk was 5 times higher compared with matched controls. Key baseline predictors included age ≥ 60, hypertension, heart/kidney disease, and prior VTE.

From https://link.springer.com/article/10.1007/s11102-024-01482-0

 This is a preview of subscription content, log in via an institution  to check access.

Filed under: Cushing's, symptoms | Tagged: , , | Leave a comment »