Development of a Predictive Model for the Risk of Metastatic Disease in PPGLs, a Retrospective Cohort Study

Phaeochromocytomas and paragangliomas (PPGLs) are tumours of the adrenal medulla and extra-adrenal sympathetic nervous system, respectively, that often secrete catecholamines(1). The tumours are derived from the neural crest and approximately 50% are caused by a germline variant. Because of the strong heritability, genetic testing is recommended in all patients with PPGLs (2-5). Most PPGLs are non-malignant (defined as non-metastatic) tumours and respond well to surgical treatment. Approximately 15% of PPGLs are metastatic and incurable by currently available therapies(1). The latest WHO Blue Book classifies all PPGLs as possibly malignant due to their unpredictable behaviour(6).

PPGLs are very heterogeneous tumours predisposed by more than 20 distinct genes(1). In two thirds of the tumours a germline or somatic variant can be identified and thus PPGLs have been divided into distinct biological subgroups on the basis of these variants(1). A consensus statement from 2017 stated the necessity for a targeted gene panel (Next-Generation Sequencing, NGS) in all PPGL patients and testing of tumour DNA if possible to identify potential future targets for drug therapies(2). Furthermore, epigenetic changes in PPGLs and other endocrine tumours have been shown to be associated with the risk of metastatic disease(7-11) and future research should be focused on understanding both epigenetic and genetic changes in PPGLs as this will open opportunities for targeted molecular therapies and personalized medicine(11,12). It has been demonstrated that DNA methylation-based tumour classification is a valuable asset for clinical decision making(13).

Many histopathological predictive risk factors for poor prognosis have been proposed and scoring systems have been developed to identify those PPGLs that have a metastatic potential. The PASS (Pheochromocytoma of the Adrenal Gland Scaled Score) and GAPP (Grading system for Adrenal Phaeochromocytoma and Paraganglioma) scores are the most well-known (1,14-18). In a recent meta-analysis the algorithms were evaluated and it was concluded that both algorithms had a low positive predictive value but a high negative predictive value for PPGLs which indicated that the models were relevant for ruling out metastatic potential for both tumour types rather than identifying cases with an increased risk of disseminated disease. This could most probably be attributed to subjective assessment of criteria involved in these scoring systems. The authors concluded that the inclusion of NGS data and an overall molecular approach is needed to accurately pinpoint cases at risk for future metastases(19). Furthermore, there is a need for an algorithm based on objective measurements to ensure objectivity in the pathological diagnosis of PPGLs, similar to algorithms for adrenocortical carcinomas like the Helsinki Score and the Reticulin Algorithm(19-21). To this end computer-assisted techniques have shown promising results(11,20).

Correctly identifying PPGLs with a future risk of metastatic disease is a clinical challenge which impacts not only the physicians but also the patients and their family members. Today's clinical guidelines are based on 'rule all in' methods which have a costly impact on the healthcare resources and the quality of life of the patients. An algorithm based on histopathological, genetic, epigenetic, and clinical variables may be helpful in stratifying patients into risk categories with different needs for clinical follow-up.

Pilot Study: Patients followed at Copenhagen University Hospital, Copenhagen have been part of a pilot study on the effect of genetic screening of family members with SDHX variants on the clinical presentation of cases. Variants of the SDHX (SDHA, -AF2, -B, -C, -D) genes are a frequent cause of familial PPGLs. The investigators found distinct differences in the clinical and histopathological characteristics between genetic variants in SDHB. Family screening for SDHB variants resulted in earlier detection of tumours in two families. Patients with SDHA, SDHC and SDHD variants also had severe phaenotypes, underlining the necessity for a broad genetic screening of the proband. The study corroborated previous findings of poor prognostic markers and found that the genetic variants and clinical phenotype are linked and therefore useful in the decision of clinical follow-up. This study has been presented as master thesis at the University of Copenhagen 2019 (Grade A) and the manuscript has been published(22). The study corroborates the feasibility of the current project.

Process of the study

Part 1: Description of the genotype-phaenotype association in a nation-wide cohort of approximately 400 patients with PPGLs or germ-line genetic variants predisposing to PPGLs using clinical data, biochemical variables, tumour characteristics and imaging results. This national cohort will provide characterization of patients making it possible to identify a metastatic and non-metastatic clinical course, respectively.

Part 2: Identification of novel prognostic biomarkers using formalin-fixed paraffin-embedded tumor tissue samples for genetic testing (NGS for somatic variants), DNA methylation, PASS scoring, and immunohistochemical markers. Twenty patients with a poor prognosis (proof of metastases) and 20 patients with a good prognosis (non-metastatic) will be included from the Capital Region of Denmark. The results will be used to develop a new algorithm for clinical prognosis.

Part 3: Validation of the new algorithm in a second subcohort of patients with tumor tissue specimens selected at random from the national cohort. Tumor specimens will be subjected to investigation of the same novel prognostic biomarkers identified in part 2. Results will be entered into the new algorithm to test whether the clinical course can be accurately predicted.

Dissemination Positive, negative or inconclusive results will be published in peer-reviewed international journals and presented at scientific meetings and in relevant patient fora. The study group members are affiliated to ENDO-ERN (European Reference Network on Rare Endocrine Conditions), a European collaboration of highly specialized centres treating rare endocrine conditions through which results can be disseminated (https://endo-ern.eu/).

To ensure completeness and transparency in the reporting of results the investigators will use the STROBE reporting guidelines relevant to cohort studies.

Patient involvement:

Patients with lived experiences contribute additional expertise and give valuable, novel insights and thus improve the effectiveness of the study. Furthermore, the patient perspective is essential in the presentation of the results. Relevant patients with the disease in question from our outpatient clinic will be invited to participate in the Steering Committee.

Safety and ethical considerations Treatment of personal data: General Data Protection Regulation will be strictly adhered to. Approval by the Ethics Committee for the Capital Region has been granted (H-20065699, 19th of February 2021).

Database: The database will be established under RedCap, which is accessible nation-wide and managed centrally by the data protection unit of The Capital Region.

Statistics: Based on previous studies which have developed similar algorithms (20,21) the investigators will need approximately 200 patients to achieve statistical power in relation to validation of the algorithm in the larger cohort. Study population characteristics will be presented using descriptive statistics. Group differences will be tested primarily using non-parametric tests (Mann-Whitney, Wilcoxon) or t-test. Associations between the genotype and phaenotype will be analysed by ANOVA (analysis of variance) and regression analysis controlling for relevant confounders, i.e. age and sex, where appropriate. The sensitivity and specificity of the algorithm to predict poor or good outcome will be determined using multiple regression analyses, Chi2 and ROC (Receiver Operating Characteristics). The investigators will seek help from professional statisticians at Copenhagen University at Statistical Advisory Service for PhD-students and students at the Faculty of Health Sciences.