Preoperative Detection of Liver Involvement by Right-Sided Adrenocortical Carcinoma Using CT and MRI

Simple Summary The major prognosis factor of adrenocortical carcinoma is the completeness of surgery. Focal adrenocortical carcinoma bulge on computed tomography and adrenocortical carcinoma contour disruption on magnetic resonance imaging are highly reproducible signs. These signs are strongly associated with direct liver involvement by right-sided adrenocortical carcinoma on preoperative imaging. These findings may help surgeons plan surgical approach before resection and decrease the complication rate. Abstract The major prognosis factor of adrenocortical carcinoma (ACC) is the completeness of surgery. The aim of our study was to identify preoperative imaging features associated with direct liver involvement (DLI) by right-sided ACC. Two radiologists, blinded to the outcome, independently reviewed preoperative CT and MRI examinations for eight signs of DLI, in patients operated for right-sided ACC and retrospectively included from November 2007 to January 2020. DLI was confirmed using surgical and histopathological findings. Kappa values were calculated. Univariable and multivariable analyses were performed by using a logistic regression model. Receiver operating characteristic (ROC) curves were built for CT and MRI. Twenty-nine patients were included. Seven patients had DLI requiring en bloc resection. At multivariable analysis, focal ACC bulge was the single independent sign associated with DLI on CT (OR: 60.00; 95% CI: 4.60–782.40; p < 0.001), and ACC contour disruption was the single independent sign associated with DLI on MRI (OR: 126.00; 95% CI: 6.82–2328.21; p < 0.001). Both signs were highly reproducible, with respective kappa values of 0.85 and 0.91. The areas under ROC curves of MRI and CT models were not different (p = 0.838). Focal ACC bulge on CT and ACC contour disruption on MRI are independent and highly reproducible signs, strongly associated with DLI by right-sided ACC on preoperative imaging. MRI does not improve the preoperative assessment of DLI by comparison with CT.


Introduction
Adrenocortical carcinoma (ACC) is a rare entity with an estimated incidence of 0.5-2 cases per million per year, and accounts for 0.04-0.2% of all cancer deaths in the Our institutional review board approved this study and informed consent was obtained from all patients. The database of our institution was queried from November 2007 to January 2020 inclusively to identify all consecutive patients who had surgery for ACC. Inclusion criteria were as follows: age > 18 years, history of operated and histologically proven right-sided ACC, and available preoperative imaging (both CT and MRI) within 3 months before surgery. Among 83 patients with operated ACC, 54 were excluded due to incomplete preoperative imaging (14 patients), left-sided ACC (38 patients), or unresected right-sided ACC (2 patients). Figure 1 shows the flow chart of the study. For the 29 patients included, the following data were recorded: age, sex, initial diagnosis date, European Network for the Study of Adrenal Tumors (ENSAT) stage (Supplemental Table S1), surgical and histopathologic findings (Weiss score, Ki67 rate), and overall survival defined as the duration of patient's survival after surgery [12]. Initial characteristics of the 29 included patients are reported in Table 1. There were 8 men and 21 women with a median age of 46 years (q1 = 35; q3 = 59; range: 19-76 years). All patients had en bloc resection for rightsided ACC, according to international recommendations [13], including seven patients who had liver resection (five right hepatectomies and two segmentectomies).    All patients were followed-up at regular intervals with clinical, biological, and imaging examinations, with a minimal follow-up of 6 months.

Procedure and Imaging Protocol
All included patients had both preoperative abdominal CT and MRI performed less than 3 months before surgery. CT examinations were performed before and after intravenous administration of iodinated contrast material, with arterial (30 s) and portal venous (70 s) phases. MRI included at least the following sequences: diffusion-weighted imaging (DWI), T2-weighted half Fourier acquisition single-shot turbo spin-echo (HASTE), fat saturated (FS) T2-weighted BLADE, T1-weighted images (in-and out-of-phase), and dynamic multiphase contrast-enhanced sequences (Supplemental Table S2). Acquisition volume covered the right atrium to exclude right atrial thrombus [14].

Imaging Analysis
Two radiologists blinded to clinical outcomes (A.K. and M.B. with, respectively, 3 and 7 years of experience in abdominal imaging) independently reviewed all imaging examinations after anonymization on a picture-archiving and communication system (PACS) viewing station (DirectView ® , 11.4.0.1253 sp1 version, Carestream Health, Rochester, NY, USA). For each examination, adrenal mass characteristics including dimensions on CT, and eight candidates imaging criteria of DLI were noted: (i) disappearance of fat border between ACC and liver; (ii) periadrenal fat infiltration; (iii) ACC contour disruption; (iv) macroscopic mass effect on inferior vena cava; (v) macroscopic mass effect on right hepatic vein; (vi) focal ACC bulge; (vii) periadrenal hepatic parenchyma enhancement; and (viii) ACC inclusion by hepatic parenchyma >180 • (Figures 2-4). Some of these signs have already been studied for locoregional invasion of ACC. Contrariwise, other signs have not been specifically analyzed for the assessment of DLI by ACC, but they have already been studied in other malignancies. Since the aim of our work was to identify new imaging signs associated with DLI by right-sided ACC, we tried to transpose these signs and their definitions for the study of locoregional extension of ACC. The fat border between ACC and liver was considered to be non-measurable and absent when <1 mm [10]. Periadrenal fat densification was defined on CT by a difference in attenuation values >10 Hounsfield units (HU) between normal retroperitoneal fat and peritumoral fat [15]. On MRI, it was characterized by hyperintense areas in the periadrenal fat on T2-weighted BLADE sequences. ACC contour disruption was defined on both CT and MRI by an adrenal capsular defect, which did not show any enhancement, contrary to a thin marginal enhancement of the lesion suggesting an intact adrenal capsule [11]. Macroscopic mass effect on inferior vena cava or right hepatic vein was defined by an anterior displacement of the vessel in association with direct contact with the tumor, with or without changes in caliber [16]. Focal ACC bulge was defined as focal protrusion of tumor into hepatic parenchyma [17]. Periadrenal hepatic parenchyma was considered enhanced by comparison with distant parenchyma if there was a measurable attenuation difference on portal phase images, with a threshold of 20 HU [18]. Inclusion by hepatic parenchyma was considered on axial images when the liver parenchyma surrounded the tumor over its half-circumference (Table 2).

Standard of Reference
At surgery, DLI by ACC was considered in the presence of macroscopic tumoral capsular rupture associated with invasion of the Glisson capsule, macroscopic invasion of the fat border between the tumor and the liver, or disappearance of the anatomical space between the tumor and the liver, making a free-margin resection impossible to obtain [7].
After resection, all ACCs were graded using Weiss histopathologic criteria of malignancy, based on the evaluation of nine features: nuclear grade, mitotic rate, atypical mitotic figures, cytoplasm, diffuse architecture, necrosis, venous invasion, sinusoid invasion, and invasion of tumor capsule [19,20]. Histopathologic diagnosis of ACC was made if there were at least three of the nine malignancy criteria.
The definite diagnosis of DLI was made using surgical and histopathological findings. Patients who had no liver resection were considered as having no DLI by ACC.  (B) A 34-year-old man with a right-sided cortisol-secreting ACC (Weiss score = 9, Ki67 = 70%), with DLI. T2-weighted BLADE fat saturated (FS) image in the transverse plane shows periadrenal fat infiltration (white arrow). (C) A 57-year-old woman with a right-sided cortisol-secreting ACC (Weiss score = 9, Ki67 = 60%), associated with DLI. T1-weighted 3D volumetric interpolated breath-hold gradient-echo (VIBE) image in the transverse plane obtained during the venous phase following intravenous administration of gadoterate meglumine shows ACC contour disruption (arrow). (D) A 23-year-old woman with a right-sided cortisol-secreting ACC (Weiss score = 8, Ki67 = 50%), without DLI. T1-weighted VIBE image in the transverse plane obtained during the venous phase after intravenous administration of gadoterate meglumine shows macroscopic mass effect on inferior vena cava (arrow). (E) A 45-year-old woman with a right-sided ACC (Weiss score = 9, Ki67 = 16%), with DLI. T1-weighted 3D VIBE image in the transverse plane obtained during the venous phase after intravenous injection of gadoterate meglumine shows macroscopic mass effect on right hepatic vein (white arrow). (F) A 36-year-old woman with a right-sided noncortisol-secreting ACC (Weiss score = 7, Ki67 = 9%) associated with DLI. CT image in the transverse plane obtained during the venous phase following intravenous administration of iodinated contrast material shows focal ACC bulge (arrow). (G) A 55-year-old woman with a right-sided cortisol-secreting ACC (Weiss score = 6, Ki67 = 7 %), with DLI. CT image in the transverse plane obtained during the venous phase after injection of iodine based intravenous contrast agent shows periadrenal hepatic parenchyma enhancement (arrow). (H) A 36-year-old woman with a right-sided noncortisol-secreting ACC (Weiss score = 7, Ki67 = 9%) associated with DLI. T1-weighted VIBE image in the transverse plane obtained during the venous phase after intravenous administration of gadoterate meglumine shows ACC inclusion by hepatic parenchyma >180 • (arrow).

Statistical Analysis
Statistical analysis was performed by using SPSS 23.0 software (SPSS, Chicago, IL, USA) and MedCalc, version 11.3.0 (MedCalc Software Ltd., Ostend, Belgium). Continuous variables were expressed as medians, interquartile ranges, and ranges. Qualitative variables were expressed as raw numbers, proportions, and percentages along with their 95% confidence intervals (CIs). Normality of distributions was assessed by using histograms and Shapiro-Wilk test. Continuous and categorical variables were compared by using Mann-Whitney and Fisher exact tests, respectively. Overall survival (OS) was estimated by using the Kaplan-Meier method. The log-rank test was used to compare survival curves. Overall survival was calculated from the date of surgery until death. Interobserver agreement for categorical variables was assessed using the weighted kappa (K) test, and K values were reported with their 95% CI [21].
Univariable analyses were conducted by logistic regression model to identify candidate features associated with DLI and to estimate odds ratios (ORs) and their 95% CIs. To consider confounders of DLI, a multivariable analysis was performed by using a logistic regression model with backward stepwise selection of covariates and with entering and removing limits of p < 0.10 and p > 0.05. Correlations between all variables were examined. In case of a strong correlation between two variables, one or another variable was included in the multivariable model. A multivariable model was built for CT features and a second for MRI features. ROC curve analysis was performed for these two models, and the area under the ROC curves (AUROC) was compared using the De long test to evaluate the added value of MRI. Significance was set at p < 0.05.

Univariable and Multivariable Analysis
At univariate logistic regression analysis, three qualitative features and ENSAT stage were associated with DLI: disappearance of fat border between ACC and liver (Figures 2A  and 3), ACC contour disruption ( Figures 2C and 4), and focal ACC bulge ( Figure 2F) (Tables 1 and 6). Table 6. Evaluation of the association between independent imaging findings and the actual status of DLI using logistic regression analysis in 29 patients with operated right ACC.
The diagnostic capabilities of both CT and MRI were assessed using subgroup analysis restricted to patients presenting with at least focal ACC bulge and ACC contour disruption on CT or MRI. The same diagnostic capabilities were found for both focal ACC bulge and ACC contour disruption. Each sign, visible indifferently on CT or on MRI, had 100% sensitivity (

Discussion
Preoperative prediction of DLI by right-sided ACC is mandatory in order to ass the benefit to risk balance of the procedure and tailor perioperative care. In the pres study, we found that ACC contour disruption and focal ACC bulge on preoperative aging were associated with DLI with high ORs and accuracies. Interestingly, we fou that relatively small ACC may cause DLI, as the smallest lesion with DLI in our coh measured 63 × 39 × 45 mm 3 , and there were no differences in terms of size between b groups.
Increased peritumoral hepatic parenchyma enhancement was identified as a sign DLI in other studies. According to Tseng at al., hyperintensity of the liver parenchy adjacent to gallbladder carcinoma is an indirect sign of DLI [18]. Perfusion hepatic dis ders adjacent to tumor can be found in a variety of tumors, particularly when they aff a focal part of the hepatic parenchyma [22]. Some hypotheses have been suggested to plain this phenomenon. First, correlation with histopathological examination reveals t hepatic perfusion abnormalities indicate direct locoregional involvement of hepatic renchyma by tumor or inflammatory reaction secondary to peritumoral sinusoidal dila tion and edema [23]. Finally, compressed or blocked portal blood vessels within the li due to an extrinsic mass-effect can lead to a blood compensation via the hepatic arte possibly producing a focal higher parenchyma enhancement [24]. However, in our stu we did not find any significant association between hyperenhancement of peritumo hepatic parenchyma and DLI.
In our study, patients who had liver resections had decreased OS compared to th who did not. This suggests that our study population is representative and similar to th of previous series. Indeed, ENSAT stage III ACCs whatever the organ invaded are kno to have worst prognosis than lower stage ACCs and more specifically have poorer pr nosis compared to those without hepatic involvement [5]. For selected patients, an aggr sive surgical approach for ACC liver metastasis (ENSAT stage IV) is associated with lo

Discussion
Preoperative prediction of DLI by right-sided ACC is mandatory in order to assess the benefit to risk balance of the procedure and tailor perioperative care. In the present study, we found that ACC contour disruption and focal ACC bulge on preoperative imaging were associated with DLI with high ORs and accuracies. Interestingly, we found that relatively small ACC may cause DLI, as the smallest lesion with DLI in our cohort measured 63 × 39 × 45 mm 3 , and there were no differences in terms of size between both groups.
Increased peritumoral hepatic parenchyma enhancement was identified as a sign of DLI in other studies. According to Tseng at al., hyperintensity of the liver parenchyma adjacent to gallbladder carcinoma is an indirect sign of DLI [18]. Perfusion hepatic disorders adjacent to tumor can be found in a variety of tumors, particularly when they affect a focal part of the hepatic parenchyma [22]. Some hypotheses have been suggested to explain this phenomenon. First, correlation with histopathological examination reveals that hepatic perfusion abnormalities indicate direct locoregional involvement of hepatic parenchyma by tumor or inflammatory reaction secondary to peritumoral sinusoidal dilatation and edema [23]. Finally, compressed or blocked portal blood vessels within the liver due to an extrinsic mass-effect can lead to a blood compensation via the hepatic artery, possibly producing a focal higher parenchyma enhancement [24]. However, in our study we did not find any significant association between hyperenhancement of peritumoral hepatic parenchyma and DLI.
In our study, patients who had liver resections had decreased OS compared to those who did not. This suggests that our study population is representative and similar to those of previous series. Indeed, ENSAT stage III ACCs whatever the organ invaded are known to have worst prognosis than lower stage ACCs and more specifically have poorer prognosis compared to those without hepatic involvement [5]. For selected patients, an aggressive surgical approach for ACC liver metastasis (ENSAT stage IV) is associated with long-term OS (5-year survival rate of 39% for operated metastatic patients) compared to conservative management (5-year survival of 15-20% for non-operated metastatic patients) [25].
In our study, we found that MRI does not improve the preoperative prediction of DLI by comparison with CT. The respective capabilities of these two imaging techniques for the diagnosis of adrenal tumors have already been studied. CT is usually considered as the most useful modality for identification and characterization of adrenal tumors. When lesions cannot be characterized adequately with CT or when CT is contraindicated, MRI remains an alternate option. However, MRI does not convey better spatial resolution, and this might explain why there is no difference for preoperative assessment of DLI between MRI and CT [16]. Nevertheless, there is no clear recommendation for the use of CT versus MRI for the diagnosis and preoperative staging of ACCs [13].
Our results demonstrate a high reproducibility of candidate signs, since the kappa values show nearly complete agreement between two radiologists with different levels of experience in abdominal radiology for the three most important signs. In addition, we used a strong standard of reference for the diagnosis of DLI because it was based on the association of surgical and histopathological criteria.
Our study has some limitations. First, its retrospective and monocentric design may limit generalizability. However, our center is a nationwide tertiary referral center for this rare disease, and the characteristics of our population are similar to those of other series reporting patients with ACCs [26,27]. Then, we included patients from 2007 to 2020, which may lead to interpretation bias toward imaging modalities. In fact, CT and MRI techniques evolved since 2007, especially concerning the quality of image and spatial resolution.
In conclusion, ACC contour disruption and focal ACC bulge are the two most accurate and reproducible signs associated with DLI on preoperative imaging. This may help surgeons plan a surgical approach before resections and decrease the complications rate. However, further prospective validation is needed to confirm these findings.