Genotype-location interaction defines the SSTR-PET phenotype in paragangliomas and pheochromocytomas.

Abstract

To determine how the somatostatin receptor (SSTR)-PET imaging phenotype in paragangliomas and pheochromocytomas (PPGLs) is driven by tumour location and underlying genetic background. This retrospective study included 174 PPGLs from 91 patients who underwent [⁶⁸Ga]DOTATOC PET/CT. Lesions were classified according to their anatomical location (head-and-neck, thoracic, abdominal, adrenal) and analyzed across patients' genetic status (i.e., presence of germline SDHx pathologic variant). PET-derived parameters (SUVmean, SUVmax, TLSRE), and tumour maximum diameter were extracted using standardized automatic segmentation. Two-way ANOVA was used to assess the effects of genetic status and lesion location, including interaction terms. ANCOVA evaluated the influence of tumour size. Discriminant analysis was performed to assess classification performance. Significant genotype × location interactions were observed for SUVmean (p = 0.0006), SUVmax (p = 0.0003), TLSRE (p = 0.0025), and tumour diameter (p < 0.001). Sporadic paragangliomas exhibited a marked cranio-caudal gradient of uptake, with the highest activity in head-and-neck tumours and substantially lower uptake in abdominal lesions. In contrast, SDHx-related tumours showed more uniform uptake values across locations. Tumour size was slightly associated with uptake (p = 0.021) but did not alter the interaction pattern. Among head-and-neck paragangliomas, TLSRE discriminated between sporadic and SDHx tumours with an overall classification accuracy of 87.3%. Metastatic lesions showed lower [⁶⁸Ga]DOTATOC uptake than non-metastatic tumours, most pronounced in sporadic head and neck paragangliomas. The present study identifies tumour genotype and anatomical location as interdependent determinants of the SSTR-PET phenotype in PPGLs. The distinct topographic uptake patterns observed across genetic groups further indicate that tracer uptake should be interpreted within a combined genotype-location framework.