1 Summary
Context: 18FFDG PET/CT improves adrenal tumour characterization. However, there is still no 4 consensus regarding the optimal imaging biomarkers of malignancy.
Objectives: To assess the performance of Tumour SUVmax :Liver SUVmax for malignancy-risk and 6 to build and evaluate a prediction model.
Design/Methods. The cohort consisted of consecutive patients with adrenal masses evaluated by 8 18FFDG PET/CT. The gold standard for malignancy was based on histology or a 9 multidisciplinary consensus in non-operated cases. The performance of the previously reported 10 cutoff for Tumour SUVmax :Liver SUVmax (>1.5) was evaluated in this independent cohort. 11 Additionally, a predictive model of malignancy was built from the training cohort (previous study) 12 and evaluated in the validation cohort (current study).
Results. Sixty-four patients were evaluated; 28% of them had a Cushing’s syndrome. Fifty-four 14 adrenal masses were classified as benign and 10 as malignant (including 7 adrenocortical 15 carcinomas). Compared to benign masses, malignant lesions were larger in size, had higher 16 unenhanced densities and higher 18FFDG uptake. CT-derived anthropometric parameters did not 17 differ between benign and malignant masses. A tumour SUVmax :Liver SUVmax >1.5 showed a 18 good diagnostic performance: Se=90.0%/Sp=92.6%/PPV=69.2%/NPV=98.0% and 19 accuracy=92.2%. A predictive model based on tumour size and tumour-to-liver uptake SUVmax 20 ratio for malignancy-risk was validated and provides a complementary approach to the ratio.
Conclusions. Tumour SUVmax :Liver SUVmax uptake ratio is a useful biomarker for diagnosis of 22 adrenal masses. Another tactic would be to calculate with the model an individual risk of 23 malignancy and integrate this information into a shared decision-making process.
Keywords: adrenal; 18FFDG, computed tomography; incidentaloma; adrenocortical carcinoma
Introduction
The adrenal glands can be affected by a variety of pathologic conditions, including 4 hyperplasia, hemorrhage, malignant (primary or secondary) and benign tumours from cortical or 5 medullary origin. Tumours can be hyperfunctioning, when producing an increased amount of6 hormones, or non-hyperfunctioning, which are characterized by normal blood hormone levels. 7 Therefore, in some cases, these lesions can be clinically suspected of malignancy, but in most8 cases adrenal masses are discovered incidentally. Incidentalomas are typically found in 9 approximately 5% of abdominal computed tomographies (CT) and are mainly benign in nature. 10 Regardless of the mode of discovery, clinical history and physical examination, the first step relies 11 on appropriate biochemical evaluation. Characterization of an adrenal mass by appropriate 12 imaging investigations follows.
In this setting, adrenal CT represents the first-level imaging 13 modality for the evaluation of adrenal lesions. Several criteria have been defined on both imaging 14 studies. High specificity for adrenocortical adenomas diagnosis was achieved using an 15 unenhanced CT density cutoff of equal to or less than 10 Housfield units (HU) 1. Moreover, 16 adrenocortical adenomas often exihibit a typical wash-out pattern, with an absolute enhancement 17 washout of ≥ 60% and/or relative enhancement washout of ≥ 40% on contrast enhanced CT 2,3 or 18 demonstrate signal loss in opposed-phased magnetic resonance imaging (MRI) 4. MRI is not 19 superior to CT and among adrenal mass with unenhanced attenuation CT density >30 HU, 66.6% 20 remain indeterminate after chemical shift MRI 5. Beyond these parameters, clinicians should be 21 aware that tumour size is the best predictor of malignancy 6,7. In frequent situations, the masses 22 remain indeterminate on radiologic imaging or are too large for accurately ruling out malignancy 23 risk, except for typical myelolipoma, cyst or hematoma.
Several studies have shown that the assessment of metabolic activity by 18F-fluorodeoxyglucose 25 positron emission tomography-computed tomography (18FFDG-PET/CT) can help to 26 characterize large and/or indeterminate masses. The use of tumour-to-liver uptake maximum 27 standardized uptake values (SUVmax) ratio was found to more accurate than visual analysis in the 28 distinction between benign and malignant tumours, with an optimal threshold varying across 29 studies 8,9. In a prospective study, we have previously shown that a ratio >1.5 (from a ROC curve) 30 was associated with malignancy with sensitivity, specificity, positive predictive value, negative 31 predictive value, and accuracy of 86.7%, 86.1%, 56.5%, 96.9%, and 86.2%, respectively 10 .
Although this cutoff value is useful, it is well known that some adrenocortical cancer and even 2 large retroperitoneal sarcomas that may mimic adrenal tumours can exhibit lower uptake values. 3 Considering that tumour size and tumour-to-liver uptake SUVmax ratio represent predictors 4 of malignancy, the primary goal of the study was to evaluate the performance of the previously 5 reported tumour-to-liver uptake SUVmax ratio in a new independant cohort of 64 patients. The 6 secondary goals were to built a probability model for predicting malignancy based on a previously 7 reported cohort (87 patients) and validate the model in this new cohort that served as validation 8 cohort. Finally, we also evaluated whether additional parameters such as CT-derived 9 anthropometric parameters or contralateral adrenal gland morphology could add information.
Material and Methods
Study design
The inclusion criteria for validation cohort were (all criteria): 118FFDG-PET/CT 6 performed between January 2017 and December 2019 at La Timone university hospital; 218F7 18FFDG-PET/CT performed for characterization of an adrenal mass of maximal diameter ≥ 30 8 mm on axial CT or ≥ 20 mm and atypical features on adrenal CT (spontaneous density ≥10 HU 9 and slow contrast washout (absolute washout <60% and/or relative washout <40%); 4absence of 10 previous history of any type of cancer, except remission>5 yrs; 5Available complete hormonal 11 work-up. Cases with elevated metanephrines were excluded.
The study was approved by the local ethical committee of Aix-Marseille University. All 13 patients gave informed consent for the use of anonymous personal data extracted from their 14 medical records for research purposes.
Patients and tumour secretory status
Depending on their secretory status, patients were divided in four distinct categories: (1) 19 Cushing syndrome when patients exhibited an absence of cortisol suppression (> 50 nmol/l) after a 20 low dose dexamethasone suppression test, associated with suppressed ACTH secretion and the 21 existence of comorbidities usually associated with hypercortisolism, (2) Subclinical 22 hypercortisolism was defined following the ESE/ENSAT guidelines, as being likely when the 23 cortisol after 1 mg-overnight dexamethasone suppression test was between 51 and 138nmol/L 24 (1.9–5.0µg/dL) and certain when superior to 138 nmol/L, associated with suppressed ACTH 25 secretion and the absence of comorbidities usually associated with hypercortisolism, (3) Other 26 secretion defined by an increased of aldosterone/renin ratio or testosterone levels in women, Non secreting, defined by normal urine and/or plasma metanephrines, normal aldosterone/renin 28 ratio, normal mean of 2 measurements of 24-hour urinary free cortisol levels and cortisol level < 29 50 nmol/l following a 1 mg-overnight dexamethasone suppression test, normal testosterone in women. 18FFDG-PET/CT 18FFDG-PET/CT was performed on a GE Healthcare Discovery PET/CT 710 (General Electric Healthcare) with the three-dimensionnal Time-Of-Flight mode. All patients fasted for 6 4 hours prior to scanning. 18FFDG (3MBq/kg) was intravenously injected. After tracer injection, 5 patients remained at rest and 18FFDG-PET/CT was acquired at approximately one hour post6 injection. A whole-body imaging was perfomed from skull base to mid-thigh, corresponding to 6 7 to 8 steps of 2 minutes each. Slice thickness of the helical CT was 2.5 mm. The attenuation and 8 impulsional response corrected PET was reconstructed with 3D iterative process (with 24 subsets 9 and 2 iterations), using a CT attenuation map. 18FFDG-PET/CT scans were analyzed by a nuclear physician investigator blinded to the results 11 of the other imaging and biochemical studies. The scan being measured was loaded into the Volume Viewer to enable the L3-L4 disc space to 22 be identified using sagittal views. The single cross sectional CT image (L3-L4 disc space) was saved as a DICOM and imported into 24 CoreSlicer (https://coreslicer.com) 11. Attenuation range of HU was set to provide tissue areas 25 (muscle, adipose tissue). Range of HU to select pixels was unknown by users because of the 26 proposed workflow. On the interface, the tissue area is directly selected (muscle, subcutaneous or 27 visceral adipose tissue, bone) and not range of HU. Automated segmentation algorithms are 28 referenced and used in this study to provide body composition. Manual corrections was performed 29 to delete aberrant pixel or add others in the body segmentation. Analysis of fat and lean tissue at 30 this area are higly correlated to corresponding whole body composition and are an important 31 predictor of the metabolic syndrome 12-16. Total body fat mass and lean body mass have been 32 defined using equations published by Mourtzakis et al. 13 . Histology: 1for adrenocortical tumours, malignant tumours was based on a Weiss score for ACCs ≥ 3; 2for oncocytomas, malignant was based on Bisceglia scoring system: the 7 existence of at least one major criterion defines a malignant oncocytoma. Oncocytomas 8 with uncertain malignant potential (borderline) defined by one to four minor criteria were 9 classified as benign. For non operated adrenal masses: lesions were classified as benign by a multidisciplinary staff based on imaging features on CT and CT follow-up (≥ 6 months). Stable disease was 12 assumed when the mass remained stable or had minimal increase in size (<15%) of the 13 tumour diameter on the last CT. Statistical analysis Statistical analysis was performed using IBM SPSS Statistics version 20 (IBM SPSS Inc., 17 Chicago, IL, USA). Continuous variables are expressed as means SD or medians with range (min, max), and categorical variables are reported as count and percentages. All the tests were 19 two-sided. Statistical significance was defined as p<0.05. Two different populations were defined: all the cases and the subgroup of adrenocortical 21 tumours. For each population, comparisons of imaging findings between two groups, benign and 22 malignant masses, were performed using student t-test or Mann-Whitney U for continuous 23 variables and Chi-Square test (or Fisher’s exact test, as appropriate) for qualitative variables. First, a tissue-based biomarker prediction model was built from a prospective cohort (training cohort) using a 25 logistic regression model including 2 parameters: the tumour SUVmax: liver SUVmax uptake ratio 26 and the tumour diameter. To quantify the discrimination performance of the model, the area under 27 the receiver operating characteristic (AUC) curve was measured. Calibration plots were used to 28 assess the calibration of this model, accompanied with a Hosmer–Lemeshow.
The prospective training cohort consisted of 87 patients from 8 French university hospitals and 30 has been previously described 10. In the training cohort, among the 87 masses, 72 were classified 31 as benign and 15 as malignant. Briefly, histology was obtained in 64 patients and identified 15 32 malignant tumours (11 ACCs, one metastasis from lung carcinoma, two leiomyosarcomas, one liposarcoma); 47 benign tumours including 34 ACAs and two oncocytomas with uncertain 2 malignant potential. The remaining 23 cases remained stable on 12-months follow-up CT and 3 were therefore considered as benign lesions.
Second, the performance of the model was tested on an independent cohort (validation 5 cohort). The logistic regression formula from the initial cohort was applied to the validation cohort 6 and the probability for each patient was calculated. To quantify the discrimination performance of 7 the model, the area under the receiver operating characteristic (AUC) curve was measured. 8 Calibration plots were used to assess the calibration of this model, accompanied with a Hosmer– 9 Lemeshow chi-square test. The cutoff points were calculated from the ROC curves that 10 maximized both sensitivity (Se) and specificity (Sp). Negative predictive value (NPV), positive 11 predictive value (PPV), and accuracy were provided with their 95% confidence intervals.
Results
Patients and tumours
During the inclusion period, 75 consecutive patients with adrenal masses were referred from 6 endocrinologists and endocrine surgeons of our institution for 18FFDG-PET/CT. Eleven patients 7 were excluded from the analysis: 5 cases due to the presence of an active cancer, 2 due to lack of 8 information regarding their secretory status, 4 patients due to elevated metanephrines (i.e., 9 pheochromocytoma). The study population consisted of 64 patients (31 women, 33 men; mean age 10 of 58.3 years): 35 had masses ≥ 40 mm including 18 with atypical feature on CT; 29 had masses 11 <40 mm including 17 with atypical feature on CT. Seventeen patients had cortisol hypersecretion 12 alone (15 overt Cushing's syndrome, 2 subclinical Cushing's syndrome), 1 mixed secretions 13 (hypercortisolism and hyperaldosteronism) and 3 had isolated hyperaldosteronism. Management Management and therapeutic decisions were made with the knowledge of the PET/CT 18 findings. Thirty-three Biomass allocation patients (51%) underwent adrenalectomy.
Final diagnosis
In the validation cohort, according to the gold standard, 54 masses were classified as benign 23 and 10 as malignant. Histology (following surgery or biopsy for two cases) was obtained in 35 24 patients and identified 10 malignant tumours (7 adrenocortical carcinomas-ACC, 1 metastasis 25 from hepatocarcinoma, 1 from a renal cell carcinoma (RCC) (>5 years remission) and 1 26 undifferentiated carcinoma from an unknown origin), and 25 benign tumours (16 adenomas-ACA, 27 2 hematomas, 3 myelolipomas, 1 benign oncocytoma, 1 borderline oncocytoma of uncertain 28 malignant potential, 2 cysts).
Overall, 29 non operated masses were classified as benign by a multidisciplinary staff. Median 30 and mean follow-up CT were 15 months (range 6-39 months) and 18.3 months, respectively. 31 20/29 patients had stable disease at follow-up CT (≥ 12 months). 9/29 had stable disease at follow32 up CT (6 months ≤ CT < 12 months). Among this 9 patients 4 had previous history of stable disease for more than 6 months prior PET study (>6 months to 5 years).
18FFDG-PET/CT findings
Compared to benign lesions, malignant lesions were larger in size, had a higher 6 spontaneaous density and higher 18FFDG uptake values. These results were observed in the 7 entire population (N=64), as well as in the group of histologically proven adrenocortical tumours (N=25) (Table 1).
One malignant tumour exihited a tumour SUVmax: liver SUVmax ratio <1.5 and corresponded 10 to 1 adrenal metastasis from a RCC (ratio 1.2). None of the patients with malignant tumour 11 exihibit a contralateral adrenal nodule. Benign tumours with the highest tumour SUVmax: liver SUVmax ratio corresponded to a 13 borderline oncocytoma (ratio=13.2) and 2 adenomas with an oxyphile cells component of 50% 14 and 10% (ratio=4.7 and 2.8, respectively). Among the 29 non-operated benign masses: 25 had tumour SUVmax: liver SUVmax ratio <1, two 16 had a ratio=1 and the remaining 2 cases had a ratio>1 (1.4 and 1.2) but a stable disease on CT follow-up (Figure 1).
The performance of the previously reported cutoff values (based on the prospective study) 19 for tumour SUVmax: liver SUVmax (>1.5 in the entire cohort and >1.6 adrenocortical tumours) 20 applied to validation cohort were :
In the entire population (n=64) using a ratio >1.5: Se=90.0% (59.6-98.2); Sp=92.6% (82.522 97.1), PPV= 69.2% (42.4-87.3), NPV= 98.0% (89.7-99.7) and accuracy= 92.2% (83.0-96.6).
In adrenocortical tumours (n=25) using a ratio >1.6: Se=100% (64.6-100); Sp=77.8% (54.824 91.0), PPV= 63.6% (35.4-84.8), NPV= 100% (78.5-100) and accuracy= 84% (65.4-93.6).
Predictors of malignancy
Two models for calculating probabilities of malignancy that takes into account both tumour 29 size (diameter) and tumour SUVmax: liver SUVmax were built from a prospective cohort of patients 30 (See materials and methods) and expressed as follows:
The AUC values for model 1 and 2 applied to validation cohort were 0.88 (95% CI: 0.78 to 0.95, 23 p <0.0001), and 0.91 (95% CI: 0.72 to 0.99, p <0.0001), respectively. The cutoff values of risk 24 probability of malignancy in the model 1 was 16.1% with a sensitivity of 90% (55.5-99.7) and 25 specificity of 74.1% (60.3-85.0). For the model 2, the cutoff was 25.8% with a sensitivity of 26 100% (59-100) and specificity of 83.3% (58.6-96.4). A graphical representation of the probabilities of malignancy according prediction model 1 is 28 shown in Figure 2. A calculator configured to calculate individual malignancy risks with the two 29 models is provided as supplemental file. Analysis of CT-derived anthropometric parameters As shown in Table 2 and supplemental Table 1, in women and men, there were no statistical 4 differences between benign and malignant adrenal masses for V:S ratio, V:TA ratio and total body lean mass. Discussion Characterisation of adrenal masses is a challenging clinical scenario since a delay in 4 malignancy diagnosis may affect the prognosis and excessive adrenalectomies (Adx) for 5 undeterminated masses can lead to excessive resection of benign tumours with potential 6 postoperative and endocrine morbidity. The present study aimed to evaluate the value of 7 previously reported cut-off ratios for tumour SUVmax: liver SUVmax in an independent cohort and 8 to build a predictive test for malignancy with two independent populations (training and validation 9 cohorts). We also evaluated for the first time the potential role of CT-derived anthropometric 10 parameters in this clinical setting. The principal conclusions that can be drawn from this study include: firstly; the previously 12 reported performances of the cutoff values for tumour SUVmax: liver SUVmax ratio in 13 indeterminate and/or large adrenal masses are confirmed; secondly; a reliable predictive model has 14 been generated; and finally, the measurement of CT-derived anthropometric parameters did not 15 add information. In non-oncologic patients or after complete remission of cancer, most of the malignant 17 masses are represented by adrenocortical carcinomas (ACC): 73% in our previous study and 70% 18 in the present series. One of the most difficulties for determining a “universal” cutoff value for 19 tumour SUVmax: liver SUVmax ratio mainly relies on the heterogenous nature of the ACC. In our 20 previous prospective cohort, two malignant tumours exihited a tumour SUVmax: liver SUVmax ratio 21 <1.5 and corresponded to a liposarcoma (ratio= 0.8) and an ACC (ratio= 1.4) (Sensitivity of the 22 1.5 cutoff= 86.7%). Additionally, benign oncocytomas which are characterized by impairment of 23 oxidative phosphorylation processes and a compensatory excessive mitochondria biogenesis, 24 usually exihibit highly elevated uptake ratio values and represent a potential false positive finding. 25 In the present cohort, 1 malignant tumour (1 renal cell cancer metastasis) had a tumour 26 SUVmax: liver SUVmax<1.5. Of note, during our longstanding experience (>15 yrs) of adrenal 27 imaging, we have had very few cases with uptake ratio <1.5 (none in this series). Other previous 28 series have also reported various cutoff values for SUV-derived metabolic indices 17-20. The main 29 limitation of all studies on adrenal mass characterization relies on the heterogeneous nature of the 30 population. Therefore, alternative approaches should be developed. Since tumour size represents another powerful predictor of malignancy and may also affect 32 18FFDG quantification (via partial volume effect), we have developped a model that takes into account tumour diameter and tumour SUVmax: liver SUVmax ratio, both parameters being easily 2 measured on PET/CT examinations. The idea would be to provide a tool for clinicians that may 3 help to calculate in a given patient the risk of malignancy and discuss the benefit-risk of Adx vs 4 short-term imaging surveillance in each individual situations. Two models have been described, 5 model 1 for the entire cohort and model 2 if the adrenocortical nature of the mass is known (i.e., steroid hormone secretion, tumour uptake of an adrenocortical tracer, metabolomics analysis, 7 adrenal biospy). For example, using prediction model 1, for a tumour SUVmax: liver SUVmax= 1.3 8 the estimated probabilities of malignancy range from 7% for 3 cm diameter to 22% for 6 cm 9 diameter (Figure 2). As shown in our series, when estimated risk of malignancy is low, the 10 presence of a contralateral adrenal nodule can be considered as an additional reassuring argument. 11 In our study, we have also measured CT-derived anthropometric parameters. Eighteen 12 patients of our cohort had hypercortisolemia. It is well established that hypercortisolemia results in 13 central adiposity which confers insulin resistance, dyslipidemia and increased risk of mortality 14 from cardiovascular disease. In overt Cushing’s syndrome, CT studies demonstrated increased 15 visceral fat 21. In a retrospective cross-sectional analysis, CT-derived fat compartment volumes 16 were analyzed in 125 incidentalomas patients and 9 women with overt Cushing’s syndrome. An 17 increased V:TV was observed between men and women in cases with positive (serum cortisol 18 greater than 1.8 μg/dL) versus negative low-dose dexamethasone suppresion test 22. There was no 19 significant difference in terms of V:TV between the groups with cortisol greater than 1.8 μg/dL 20 (1.8-2.9; 3-5, >5 μg/dL) and those with overt Cushing’s syndrome. In another study, visceral fat 21 measurements were comparable between patients with autonomous cortisol secretion compared to 22 non-secreting masses. However, an increased visceral fat content was observed during follow-up 23 (3 years) in patients harbouring autonomous cortisol secretion 23. Therefore, many factors may 24 influence fat redistribution such as duration of exposure before diagnosis which is not easy to 25 estimate, amount and patterns of secretion (fluctuating or permanent), individual tissue sensitivity 26 to glucocorticoids, altered glucose metabolism. In the present study, we failed to identified any 27 differences between benign and malignant masses in men and women. Therefore, it can be 28 estimated that the analysis of CT-derived anthropometric parameters have very limited value in 29 this clinical setting.
We acknowledge several limitations of the present study: the observational Selleckchem GSK’872 nature of the 31 study, the limited sample size of the validation cohort, and the absence of pathologically proof for 32 all masses with short follow-up CT (<12 months) in some cases. In conclusion, although informative, it remains elusive to find a universal cutoff value for 2 tumour SUVmax: liver SUVmax that allow diagnosis of all malignant tumours. The integration into 3 the decision-making process of a predictive model for risk-malignancy could be an alternative 4 approach that would enable to estimate, in a given situation, the benefit-risk of surgery versus 5surveillance, taking into account that Adx inexperienced centers has low morbidity.