Introduction

Adrenocortical carcinomas (ACCs) are rare tumors, with an estimated incidence less than 2 cases per million people per year [1, 2]. In the pediatric population, this incidence is usually even lower (0.3 cases per million per year), with the notable exception of children in southern Brazil (3.4 – 4.2 cases per million per year) [3, 4]. Despite sharing tumor nomenclature, adult and pediatric ACC may be considered different entities. These tumors do not share many common pathogenic events [5,6,7] and their clinical presentation and outcome are seemingly different [8,9,10]. Furthermore, pathological criteria used for diagnosis and prognostication in adult ACC are not adequate for characterizing the pediatric tumors [11,12,13].

The definitive diagnosis of adrenocortical tumors (ACTs) in the adult population requires the morphological assessment of the whole adrenal mass. In other words, diagnosis is usually only achievable by pathological analysis of surgically-resected specimens; core- and fine-needle biopsies are seldom conclusive for diagnosis and are usually reserved for ruling out metastases from other tissue sites to the adrenal gland [14, 15]. Following surgical resection, adult ACTs are usually classified according to the Weiss histological score in adrenocortical adenomas (ACA, Weiss score < 3 criteria) or ACC (Weiss score ≥ 3 criteria) [16, 17]. Pathological analyses of adult ACCs also include the evaluation of Ki67 labeling index (LI), which along with the tumor’s clinical stage (ENSAT), guide therapeutic decisions [2, 14]. Patients with ACC depicting Ki67 LI ≥ 10% have higher recurrence rates and, for that reason, often receive adjuvant treatment [14, 18].

In the pediatric population, there are no pathological classifications or biomarkers fully consolidated for diagnosis and/or prognostication of ACT [19, 20]. The Weiss score is not adequate for stratifying these tumors [11, 12], and the Wieneke score, which was specifically proposed for assessing biological behavior of pediatric ACT back in 2002 [21], is not widely accepted [6]. Hence, clinical follow-up remains the best indicator of biological behavior of these tumors. Nevertheless, a multi-institutional French cohort recently suggested that pathological analyses of pediatric ACT that include criteria from the Wieneke score and assessment of Ki67 LI may be useful for predicting biological behavior of these tumors. However, the cut-off value for Ki67 should be 15%, which is higher than the 10% value used for prognostication in adult ACC [22]. Similarly, Das et al.had also recently revisited the importance of the Wieneke score and had suggested a role for Ki67 LI for stratifying pediatric tumors [23].

Here, based on a large cohort with long-term clinical follow-up, we reinforce the importance of pathological analysis for risk stratification in adult and pediatric ACT. In tumors from adults, we validate the role of Ki67 LI for patient prognostication but propose a lower cut-off value for defining patients at high risk of recurrence and poor outcome. Among pediatric tumors, we strengthen the evidence that the Wieneke score and Ki67 LI may be useful for predicting biological behavior; we also suggest a potential role for a high Weiss score (≥ 5 criteria) in predicting outcome in pediatric tumors. Finally, we compare clinical and pathological features amongst adult and pediatric ACT and re-emphasize the biological differences between these tumors. Using the same criteria for quantifying Ki67 LI in adult and pediatric ACT, we confirm previous reports that pediatric tumors have higher proliferative indexes compared to their adult counterparts. We expect our results to help further consolidate pathological analyses and Ki67 evaluation in adult and especially, pediatric ACT.

Methods

Cohort

We included 146 adult and 44 pediatric (< 15 y/o) patients with ACT submitted to surgery at Hospital das Clinicas da Faculdade de Medicina da Universidade de Sao Paulo (HC-FMUSP), Brazil. Patients in the adult cohort were described in a previous publication [24], except for 10 cases whose paraffin blocks were no longer available. Follow-up information from all patients and the method of Ki67 evaluation were updated (see digital analysis below). ACT in adults were classified as adenomas (< 3 criteria) or carcinomas (≥ 3 criteria) according to the Weiss criteria [16]. ACT from pediatric patients were classified as clinically benign (CB) or clinically malignant (CM) according to clinical follow-up (i.e. local or distant recurrence, or death). The median time of follow-up in the pediatric population was 129 months (149 months for patients that did not die). The germline status of codons R337H (exon 14) and R227H (exon 8) of TP53 was retrospectively collected from the medical records of 30/44 (68%) pediatric patients.

Immunohistochemistry

Immunohistochemistry was performed in the most highly mitotic or most cellular blocks/slides from each tumor: 4 µm slides were cut, deparaffinized and rehydrated. Antigen retrieval was achieved using 10 mM pH 6.0 citrate buffer solution for 40 min in a steamer at 95o C. Following peroxidase block, primary antibody (mouse monoclonal antibody MIB1, 1:100 dilution, DAKO, Denmark) was incubated overnight at 4o C. Signal amplification was performed using Novolink Polymer Detection System (Vision Biosystems™, UK), followed by DAB-solution revelation (3,3′-diaminobenzidine tetrahydrochloride and dimethyl sulfoxide – Sigma, USA). Slides were counter-stained with hematoxylin, dehydrated, and mounted with Entellan (Merck, Germany). Staining of a lymph node was used as an external positive control. In addition, endothelial cells and reactive lymphocytes within the tumors were used as internal positive controls.

Digital Analysis

The stained slides were scanned by the Scanner of histological slides (Zeiss Laser Scanning, Germany) through the Pannoramic Viewer 1:15 software (3DHISTECH, Hungary). Multiple high-resolution images from medium-power fields (200x) were obtained for morphometric analysis. All the images were processed with the Image Pro Plus 4.5 software (MediaCybernetics, USA), and had the positive and total nuclei automatically counted. The images were then evaluated by a pathologist and manual corrections were performed when appropriate: individual non-tumoral cells or areas enriched for non-tumoral cells (e.g., lymphoid aggregates) were manually de-selected from the total count. To normalize the number of cells assessed per patient, only the images with the highest Ki67 LI were considered, until a minimum of 2,000 cells were evaluated.

Statistical Analysis

Categorical variables were summarized as frequency (percentages) and continuous/ordinal variables were summarized as mean (with standard deviation [SD]) and/or median (with range or interquartile range [IQR]). Differences between variables among the cohorts were calculated using the Mann–Whitney test for continuous/ordinal variables, and the Fisher’s exact test for categorical variables. Receiver Operating Characteristic (ROC) curve analyses were used to identify cut-offs (breakpoints) in the continuous variables. Standardized differences were calculated to compare patient groups dichotomized according to a specific variable, as previously described [25]. Higher values for standardized difference suggest bigger differences between patient groups for the variable being analyzed. Survival analyses were performed using the Wald test within the Cox proportional hazard models. Continuous and ordinal variables were also dichotomized, and specific cut-off values were analyzed based on their effect on overall survival (OS) and disease-free survival (DFS). Cut-offs for the survival analyses were derived from values described in previous publications or defined by the results of ROC curve analyses in our cohort. All statistical analysis was performed in R studio with the statistical package or SPSS 22 software (SPSS Inc, Chicago, USA). In all situations, a p < 0.05 was considered significant.

Results

Description of the Adult Cohort

We included 146 surgical samples from ACT in adults (≥ 15 y/o): 76 (52%) tumors were histologically classified as adenoma (Weiss < 3) and 70 (48%) were classified as carcinomas (Weiss ≥ 3). Clinical, imaging, and pathological information from these patients are summarized in Table 1.

Table 1 Description of the adult cohort
Full size table

Comparisons between clinical and pathological features in adults with adenoma and carcinoma revealed that Cushing syndrome was common among patients with adenoma (64%), but less frequent in patients with carcinoma (19%, p < 0.001). Conversely, virilization syndrome, either as an isolated clinical finding (p = 0.046) or in combination with Cushing syndrome (p < 0.001), was enriched in patients with carcinoma. In fact, only two (3%) patients with adenoma presented with virilization syndrome and none presented with combined Cushing-virilization syndrome.

Among the pathological variables, tumor size (p < 0.001) and weight (p < 0.001) were significantly different in adenoma compared to carcinoma samples. The standardized difference (see methods) between benign and malignant tumors for tumor size of 10.5 cm was 55.3, higher than the standardized differences for the sizes of 8 cm (standardized difference = 6.3) and 15 cm (standardized difference = 44.9), suggesting that 10.5 cm, among the values tested, was the best size to distinguish carcinomas from adenomas. Indeed, there was no adenoma ≥ 10.5 cm, whereas less than half of all carcinomas were < 10.5 cm (31/70, 44%).

The median Ki67 LI of adult adenomas and carcinomas was 0.5% (IQR: 0.2% – 0.9%) and 4.9% (IQR: 0.8% – 19.5%), respectively. Ki67 LI = 3% showed higher standardized difference (99.1) in separating carcinomas from adenomas compared to Ki67 LI = 10% (standardized difference = 78.3) and Ki67 LI = 20% (standardized difference = 37.2). ROC curve analysis correlating the Ki67 LI with the pathological diagnosis of carcinoma (Area Under the Curve [AUC] = 0.821, p < 0.001) validated these results by depicting a specificity of 99% for the diagnosis of adult carcinomas in tumors with Ki67 LI ≥ 3%. In fact, only 1/70 (1.4%) adult adenomas presented with Ki67 LI ≥ 3%, and it came from a female patient with a clinical diagnosis of Cushing syndrome. Gross examination depicted a 3.5 cm tumor and histology showed a Weiss score = 1. Disease relapse was not observed during the patients’ clinical follow-up (401 months). Despite the high specificity, the sensitivity of the Ki67 LI ≥ 3% for the diagnosis of carcinoma is low (sensitivity = 57%). In other words, multiple tumors classified as carcinomas according to the Weiss criteria had a Ki67 LI < 3% (n = 30/70, 43%).

Survival Analyses in Adult Adrenocortical Carcinoma

The median follow-up for patients diagnosed with carcinoma in the adult cohort was 37 months (85 months for patients who did not die). Overall- (OS) and disease-free-survival (DFS) for these patients are described in Table 2. Patients with metastasis at diagnosis were excluded from the DFS analysis.

Table 2 Survival analysis of adrenocortical carcinoma in adults
Full size table

Among clinical features, advanced ENSAT stage (III + IV) showed impact in both OS (p < 0.001) and DFS (p = 0.007), and combined Cushing-virilization syndrome showed impact only in OS (p = 0.027; DFS: p = 0.41). Among pathological features, increase in the Weiss score, tumor size and tumor weight showed impact in both OS and DFS. We next dichotomized the samples based on specific cut-offs for each pathological variable and found that a Weiss score ≥ 5 (OS and DFS: p < 0.001) and tumor size ≥ 10.5 cm (OS: p = 0.003; DFS: p = 0.004) showed the highest impact in OS and DFS amongst the values tested for these variables.

Finally, an increase of 1% in Ki67 LI showed a significant impact in OS and DFS (p < 0.001). Furthermore, Ki67 LI ≥ 10%, among the cut-offs tested, showed the highest hazard ratio (HR) in predicting recurrence, in accordance to the results reported by Beuschlein et al. [18]. However, in our cohort, Ki67 LI ≥ 3% showed the highest hazard ratio in predicting overall survival. Moreover, the mortality rates of patients with Ki67 LI ranging from 3% – 9.9.% (5/10, 50%) were equal to the rates of patients with Ki67 ranging from 10% – 19.9% (7/14, 50%; Fisher exact test calculated by events: p = 1). Mortality rates in patients with Ki67 LI < 3% were lower (4/30, 13%) than patients with Ki67 LI ≥ 3% (25/40, 63%; p < 0.001) and, as expected, rates in patients with Ki67 LI ≥ 20% were higher (13/16, 81%) than in those with Ki67 LI < 20% (16/54, 30%; p < 0.001).

Description of the Pediatric Cohort

The clinical, imaging, and pathological information from the 44 patients in the pediatric cohort (< 15 y/o) are summarized in Table 3. Most patients were female (27/44, 61%), younger than 3 y/o (35/44, 80%), and were diagnosed at early clinical stages (International Pediatric Adrenocortical Tumor Registry [IPACTR] stages I + II, 40/44, 91%). Virilization syndrome, as an isolated finding (27/44, 61%) or combined with Cushing syndrome (16/44, 36%) was common (overall: 97%).

Table 3 Description of the pediatric cohort
Full size table

Tumors in patients with combined Cushing and virilization syndromes showed a trend towards malignant behavior (p = 0.053). Malignant tumors also showed increased tumor size (p = 0.01) and higher Wieneke scores (p < 0.001) compared to benign tumors. ROC curve analysis confirmed the high accuracy of the Wieneke score in predicting malignant biological behavior (AUC = 0.89, confidence interval [CI] = 0.77, 1; p < 0.001). Furthermore, the recommended Wieneke score of 3 showed high sensitivity (89%) and fair specificity (76%) for the diagnosis of ACC in our cohort. In fact, only one patient with Wieneke score < 3 experienced disease-relapse (1/27,4%); after the resection of the relapse nodule, this patient did not experience additional recurrence until the end of the clinical follow-up (249 months). Conversely, half of the patients (8/16) with a Wieneke score ≥ 3 experienced disease-relapse, from which 6/8 (75%) later passed away. Interestingly, the Weiss score was also significantly higher in CM compared to CB tumors; a score ≥ 5 was able to significantly predict biological behavior in the pediatric cohort (p = 0.016).

The median Ki67 LI of CB and CM pediatric tumors was 7% (IQR: 1.2% – 13.4%) and 26.3% (IQR: 21.9% – 46.3%), respectively (p < 0.001). Like the Wieneke score, Ki67 LI also showed high accuracy in predicting biological behavior (AUC = 0.9, CI: 0.8, 0.99, p < 0.001). Ki67 LI of 10% showed high sensitivity (100%), but low specificity (51%) for the diagnosis of ACC. Conversely, the Ki67 LI of 20% showed lower sensitivity (78%) but high specificity (89%), and the Ki67 LI of 15% showed intermediate values (sensitivity = 89%; specificity = 77%). In other words, Ki67 LI < 10% ruled out malignant biological behavior in the pediatric cohort, but a high number of benign tumors also showed a LI ≥ 10% (17/35, 49%). Conversely, few cases with LI ≥ 20% had benign biological behavior (4/35, 11%), but some malignant cases are missed when using this cut-off value for risk prediction (2/9, 22%) (Fig. 1). Of note, 3/4 tumors with Ki67 LI ≥ 20% that had benign clinical course showed high Weiss (≥ 5) or Wieneke (≥ 3) scores. Thus, it is also possible that these cases were biologically malignant, but cure was achieved through complete surgical resection.

Fig. 1
figure 1

Adrenocortical tumors in the pediatric population with low Ki67 LI (< 10%) show benign clinical behavior (A). In contrast, most tumors with high Ki67 LI (> 20%) show higher recurrence and mortality rates (B)

Full size image

Survival Analyses in Pediatric Adrenocortical Tumors

Survival analysis in the pediatric population is described in the Table 4. Age ≥ 3 correlated with worse DFS (p = 0.01) and showed a trend towards poor OS (p = 0.07). Virilization syndrome, either isolated or combined with Cushing syndrome, correlated with OS, and showed a trend towards poor DFS. Increase in the Wieneke score strongly correlated with both OS (p = 0.002) and DFS (p = 0.001), reinforcing the role of pathological criteria for risk stratification in pediatric ACT. From the values tested, the Wieneke score ≥ 3 showed the highest hazard for DFS (p = 0.008), while Wieneke score ≥ 5 showed the highest hazard for OS (p = 0.001). Finally, an increase in the Ki67 LI strongly correlated with poor OS (p = 0.004) and DFS (p < 0.001). The Ki67 LI ≥ 15% showed the highest hazard for both OS (p = 0.04) and DFS (p = 0.007).

Table 4 Survival analysis of adrenocortical carcinoma in children
Full size table

Comparisons Between Adult and Pediatric ACT

We next looked for differences between ACT in the pediatric and adult populations (Table 5). The predilection for female gender was common in adult and pediatric ACT, but the prevalence of female patients was significantly higher in adults (p = 0.003). Combined Cushing and virilization syndromes were more common in pediatric than in adult patients (36% vs. 21%, p = 0.047). Also, Cushing syndrome without virilization was much more common in adults (43% vs. 5%, p < 0.001), whereas virilization without Cushing syndrome was much more frequent in children (61% vs. 7%, p < 0.001).

Table 5 Comparison of ACT in the adult and pediatric populations
Full size table
Table 6 Comparison of adrenocortical carcinoma in the adult and pediatric populations
Full size table

Among pathological features, mean tumor weight (p = 0.38) and size (p = 0.64) did not differ between pediatric and adult tumors in this surgical series. Interestingly, the Weiss score of pediatric ACTs was consistently higher than in adult tumors. At this time, we could not assess the Wieneke score in many adult ACTs, and comparisons for this variable are not available. Finally, the Ki67 LI index was higher in pediatric compared to adult tumors (p < 0.001), but the range of this index was similar in both cohorts (0 to 57% in pediatric and 0 to 58% in adult ACT).

Significant differences in clinical presentation and pathological features were not observed when comparing only the malignant cases from the adult and pediatric populations (Table 6), which could suggest that ACC from these cohorts are not as different as anticipated. However, the low number of malignant tumors in the pediatric cohort could have compromised the sensitivity to detect some differences. Even so, the Ki67 LI of pediatric ACC was significantly higher than the index in adults (26% vs. 5%, p < 0.001). Similar findings were observed in benign tumors (pediatric ACA: 7% vs. adult ACA: 0.5%, p < 0.001; Fig. 2).

Fig. 2
figure 2

Distribution of the Ki67 LI in adult and pediatric ACT

Full size image

Discussion

We performed a comprehensive evaluation of pathological features in a large series of adult and pediatric ACT from a single academic institution. With long-term clinical follow-up, we were able to further validate the importance of clinical features, pathological scores, and the proliferative index by Ki67 immunohistochemistry, in predicting the biological behavior of ACT in both age cohorts. We also compared adult to pediatric ACT and found more corroborating evidence for these age-separated tumors having clinicopathological differences. These findings provide further support for the notion that pediatric and adult ACT are indeed different entities. Of note, by scoring tumors with the same quantification method, we confirmed that Ki67 LI in pediatric ACT (median: 10.27% IQR: 2.27%—23.54%) is much higher than in adult ACT (median: 0.83%; IQR: 0.24%—4.68%).

Clinical management of adult patients with ACC takes into consideration the Ki67 LI of the tumor, which is based on strong evidence that an increased proliferative index is associated with worse clinical outcome [18, 26]. Current guidelines suggest the cut-off value of 10% for recommending adjuvant therapy following surgical resection of ACC in adults [14]. These results are mostly supported by the robust analyses performed by Beuschlein et al.that included multiple cases of ACC from different European Centers [18]. Of note, in that study, even a Ki67 LI ≥ 5% correlated with outcome in ACC. Similarly, our results also suggest that a lower Ki67 LI could predict outcome. By performing digital analyses, which arguably improves Ki67 quantification compared to visual estimation, we showed that Ki67 LI ≥ 3% correlated with worse outcome. Furthermore, most negative events (local or distant recurrence and death) in patients with ACC having Ki67 LI < 10% happened in tumors with Ki67 LI ranging from 3% to 9.9%; mortality in Ki67 LI ranging from 3% to 9.9% also did not differ from the 10% – 19.9% LI range. Altogether, these results suggest that patients with ACC with Ki67 LI ≥ 3% might already benefit from adjuvant therapy following surgical resection, which is usually reserved for ACC with Ki67 ≥ 10%.

Stojadinovic et al. reported that in adults, Ki67 LI > 5% is more frequent in ACC as compared to ACA [27]. In accordance with these results, we also found a significantly higher Ki67 LI in carcinomas as compared to adenomas in the adult population. These results may be used for ACT characterization in clinical situations for which histological diagnosis is not straightforward such as during the evaluation of underrepresented (or consult cases) cases. Again, we suggest Ki67 LI ≥ 3% to indicate patients at higher risk of recurrence and in need of closer clinical follow-up. Moreover, a Ki67 cut-off value of 3% also showed a high specificity (albeit low sensitivity) for the pathological diagnosis of ACC. In other words, most of the ACT with Ki67 LI ≥ 3% would likely have a Weiss score ≥ 3 and would be pathologically classified as carcinomas.

Correlations between pathological criteria and outcome are not as evident in pediatric as in adult ACT. First, pediatric tumors are rarer, and accruing a significant number of cases for robust analysis is difficult. Furthermore, there is no absolute histological feature or exact breakpoints in pathological scores that distinguish ACC from ACA in the pediatric population. In fact, the original manuscript that proposed the pathological criteria (Wieneke score) for evaluating pediatric ACT described malignant cases with a score as low as 1 [21]. Similarly, one of the patients in our pediatric cohort that progressed with metastatic disease also had a tumor with a Wieneke score of 1. The lack of an absolute threshold to rule out malignancy makes it hard to suggest biological behavior when individual cases are being analyzed. Nevertheless, increase in the pathological score is accompanied by higher recurrence and mortality rates. Furthermore, malignant cases with low scores were the exception in the original publication [21], as they are in a more recent report [23] and our study. Finally, analysis of additional criteria such as the Ki67 LI may help predict the biological behavior in pediatric ACT with low pathological scores. For instance, the malignant tumor in our cohort with Wieneke score = 1 had a Ki67 LI = 40%. It is possible that in this tumor, sustained proliferation may have preceded other cancer hallmarks [28] and phenotypes.

Establishing the ideal Ki67 cut-off value to suggest malignancy in pediatric ACT is also challenging. In the cohort described by Picard et al., no pediatric patient with ACT having a Ki67 LI < 15% had disease recurrence [22]. Although we have used a similar method for Ki67 quantification in our cohort, we failed to validate this finding: a two-year-old patient with ACT that had Ki67 LI = 11% recurred 14 months after the resection of the original tumor; the patient passed away shortly thereafter. Importantly, the Wieneke score of this tumor was 5, again suggesting that this pathological score and the Ki67 LI may complement each other for risk stratification in pediatric ACT. Similarly, but by using a slightly different Ki67 quantification method, Das et al. also reported two pediatric patients with poor outcome that had ACT with Ki67 LI < 15% (10.2% and 12.5%) [23]. In all three cohorts, there was no malignant tumor with Ki 67 LI < 10%. Despite the high sensitivity, in our cohort, a Ki67 LI ≥ 10% showed low specificity for the diagnosis of ACC in the pediatric population, which greatly limits the use of this cut-off to suggest biological behavior or to indicate adjuvant therapy in children. Therefore, although the Ki67 cut-off of 15% may miss few cases of malignant behavior in the pediatric population, it offers increased specificity for the diagnosis of ACC and may help identify patients that would benefit from adjuvant treatment. Moreover, in our cohort, this cut-off showed the highest hazard ratio for predicting outcome, which supports the results reported by Picard et al. [22].

Most of the pediatric patients included in our study were aged 3 years or less (35/44, 80%). Thus, our observations in the pediatric cohort could be biased towards younger patients. Ideally, the prognostic value of the Wieneke score and the Ki67 LI should be assessed in different age groups within the pediatric population. However, in our study, the low number of patients older than 3 years limited our ability to do such an analysis. This will hopefully be addressed in future studies. Similarly, our cohort is enriched for patients from southern Brazil, who show a high prevalence of germline TP53 mutations associated with development of adrenocortical tumors. While our results were comparable to those of other geographic locations [22, 23], it would have been ideal to directly compare tumors from patients with different genetic backgrounds or from different geographic locations. Finally, the current work lacks molecular analyses of somatic events, which could improve patient prognostication beyond clinical and pathological features.

In summary, adult and pediatric ACT are seemingly distinct tumors, as illustrated by differences in clinical presentation, outcome, pathological scores, and proliferative indexes. Nevertheless, risk assessment of ACT in adults and children will benefit from careful pathological analyses, including immunohistochemical evaluation of the Ki67 LI. Indeed, in both age cohorts, increase in pathological scores and in the Ki67 LI strongly associated with diagnosis of carcinoma and poor OS and DFS.

Grant Support

SNMF was supported by the CAPES doctorate grant.