Within our large multi-center cohort, we found a nearly 100% rate of PPBS loss in germinoma and NGGCT of the hypothalamo-neurohypophyseal axis. This finding is important in pediatric imaging, because it serves as a rather reliable diagnostic criterion for early tumor detection in malignant GCTs and has the potential to reduce delays in diagnosis.

GCTs are the most prevalent suprasellar pathology in pediatric patients with DI, particularly in the Asian population [22]. In the general population, an absence of the PPBS is observed in 4.1% of individuals without endocrinopathies, with a correlation to increasing age [23]. In a small cohort of 17 GCTs, absence of the posterior pituitary bright sport was reported in 82% [10]. We observed a loss of the PPBS in 96.1% in the total cohort and in the cases with available clinical information regarding an existing DI, each MRI revealed a PPBS loss.

The occurrence of a loss of the PPBS in other sellar tumors, such as adenomas, is uncommon, accounting for a mere of 20% [24]. Furthermore, adenoma exhibit an expansion of the adenohypophysis, whereas microadenomas are distinguished by sparse areas of contrast enhancement. The distinction between small GCTs and Langerhans cell histiocytosis may pose a greater challenge on MRI. However, the evaluation of alpha-1-fetoprotein and beta human chorionic gonadotropin can serve as an additional diagnostic tool to ascertain NGGCTs and in non-secreting tumors skeletal evaluation may provide the diagnosis of a sellar lesion. Idiopathic central diabetes determined by inflammatory/autoimmune causes is noted as one important differential diagnosis especially in cases with pituitary stalk thickening [25]. Since pituitary stalk thickening in inflammatory diseases should regress within the first six months, close clinical follow-up is necessary for a differentiation from small GCTs [26, 27]. However, it must be noted that pure germinomas are often associated with immune cell infiltration, and therefore, treatment with corticosteroids could lead to an apparent reduction in tumor size in the short-term follow-up [28].

Histologically, intracranial GCT do not differ from their extracranial counterparts. Furthermore, extra- and intracranial GCT most often occur within the midline. It is hypothesized that GCTs develop from non-eliminated mismigrated primordial germ cells that migrate along the sympathetic trunk can reach the midline of the brain [3]. Teilum hypothesized that extragonadal GCT, germinoma and NGGCT, originate from stray primordial germ cells [29]. But germinoma are more sensitive to radiotherapy and chemotherapy which exhibits a better prognosis compared to NGGCT. This may be explained by their rapidly progressive, undifferentiated cells leading to the assumption that NGGCT differentiate from different cells of origin. Therefore, it is postulated that the primordial germ cells theory rather explains the evolution of germinomas, but for NGGCT the pluripotent stem cell hypothesis is more plausible [30]. Recently, it was suggested that the transformation from primordial germ cells to transformed embryonic stem cells is the most logical mechanism for all intracranial GCTs [31]. Both theories do not explain the midline location sufficiently, the authors assumed lateral ventricle location of GCT because neural stem cells are present in the subependymal zone. We only evaluated midline location of GCT and each tumor in our study involved the neurohypophysis, where hypothalamic neurons terminate. However, this does not explain the location as well. The fact that small tumors were almost exclusively located inferior to the optic chiasm may be a further hint of tumor origin. In these cases, the diagnostic value of the loss of PBBS is profound. Already Fujisawa et al. [32] postulated on a basis of seven patients that germinoma of the hypothalamo-neurohypophyseal axis arise from the neurohypophysis and a theory by Tan [4] was that anterior midline structures are highly active in hormone production and concentration in factors like GnRH, potentially playing a role in tumorigenesis. Ultimately, radiology cannot provide a definitive explanation for the origin of GCTS; however, we aspire that our findings will aid in the advancement of theories. Recently a novel MRI classification for intracranial GCT, compromising bifocal and basal ganglia manifestations, was introduced [33]. Within this report no GCT was localized only intrasellar. Therefore, they hypothesized that suprasellar GCT arise from the tuber cinereum and median eminence and infiltrate the stalk. We found 13 GCTs in the neurohypophysis not involving the third ventricle floor and therefore, we cannot support the results. One explanation for these different results may be that Esfahani et al. [33] evaluated only post-contrast T1WI.

In general, each tumor in our cohort involved the posterior pituitary gland or the stalk and most tumors (61.8%) showed an expanded growth with compression of diencephalic structures. While suprasellar and bifocal location did not differ, we found that suprasellar NGGCT were significantly more often (85%) expanded compared to suprasellar germinoma (46.8%). In bifocal GCT the posterior pituitary was significantly less often involved than in the suprasellar cohort, but this may result from the lower incidence of bifocal germinoma. The pituitary stalk was involved in nearly 100% in both location groups and was independent from histology. In the work by Zhang et al., the posterior lobe was less often infiltrated than the stalk in a bifocal GCT cohort [34]. The involvement of the suprasellar region was deemed as a metastatic seeding from a primary pineal GCT and rated as “false” bifocal location. False GCT were differentiated from true GCT by their shape. Interestingly, the tumors rated being “false” bifocal, showed no involvement of the posterior lobe. In four cases we found a present PPBS, but only one in a bifocal metastasized germinoma. Finally, the signal intensity of the PPBS in all four cases was lower than expected or not well delimitable to the clivus and accordingly, consensus reading was necessary (of a total of 11 consensus readings).

It may be challenging to distinguish bone marrow fat from the PPBS on MRI. In GCTs, the cortical layer of the clivus is not a reliable tool for spatial delineation, because, as we observed in suprasellar GCTs, it can be destroyed or infiltrated by the tumor. As a consequence, we propose that each patient with endocrinopathy should be evaluated by a non-contrast sagittal T1WI with fat-suppression for a better discrimination.

Kilday et al. [10] argued that a present PPBS in germinoma indicates residual pituitary function. Unfortunately, we do not have sufficient data on DI and serum sodium values in our patients to support this hypothesis. In our cohort, each patient with DI or polydipsia had no PPBS at diagnosis.

Consensus recommendations set a cut off of the diameter of the inferior pituitary stalk of 3 mm [19, 20]. Our study revealed that 27 out of the 64 (42.2%) measurable inferior stalks had a diameter within that range indicating that this is not a reliable indicator for early tumor detection. Zhang et al. [34] assumed that bifocal intracranial germinoma are “false”, if the inferior pituitary stalk is measuring less than 3.0 mm or the bright sport sign is absent. We observed only one bifocal germinoma patient with a PPBS and an inferior stalk measuring 1.1 mm. In addition, this patient showed true leptomeningeal dissemination and therefore the “false” bifocal theory is quite possible.

There was, however, no noticeable difference in the diameter of the inferior stalk between suprasellar and bifocal GCT in our study. Of note, the diameter was substantially smaller in bifocal germinoma compared to NGGCT (p = .014) and in 40% within the range of the healthy population.

Even if the spectrum of causes for a PPBS loss is wide, each differential diagnosis needs rapid clarification. In pediatric GCT patients, a delay in diagnosis has been reported in 83% with a median time interval between symptom onset and diagnosis of 25 months [16]. It is known that prolonged symptom intervals in GCT patients are associated with higher risk of developing a leptomeningeal dissemination [17]. Furthermore, poorer outcomes have been reported for germinoma patients treated after a delay [15]. The PPBS loss is only systematically reported in suprasellar germinoma [10]. With our work, we are the first to report this sign as common also in NGGCT. Our results led us to conclude that the tumor could be overlooked on the first MRI. For early tumor detection, especially small GCTs require a sufficient marker. The earlier treatment start in the bifocal cohort was expectable, but the maximum interval to treatment of 204 days in bifocal germinoma demonstrates the difficulty in smaller tumors as well. Regarding bifocal GCT, it is important to note the recently published consensus recommendation on GCT in this context. When central diabetes insipidus is present with the absence of the PPBS on MRI, and the pituitary stalk or median eminence is not thickened, patients with pineal tumors and negative serum and cerebrospinal fluid α-fetoprotein and human chorionic gonadotropin markers should be considered as having sufficient evidence of bifocal germinoma [20].

In our cohort, there was a delay in the time interval between MRI and treatment start, which was significantly longer in the suprasellar cohort (median 9 days in the bifocal cohort versus median 24 days in the suprasellar cohort). Moreover, the time interval was more prolonged in suprasellar germinoma compared to NGGCT (median 54 days versus 8 days) and in suprasellar circumscribed tumors when compared to expanded tumors (median 312 versus 9 days). We believe that this is caused by the significantly smaller tumor size that we have observed. A mono-institutional study determined the time interval between symptoms onset and diagnosis by surgery or biopsy exactly and reported a median interval of 25 months in sellar germinoma [16]. Because our study was retrospective with patients from multiple hospitals, there was a lack of clinical information. We were able to determine the time interval between the first MRI and the commencement of treatment or surgery. Hence, the duration of the delay is less than 25 months. Nonetheless, it is imperative to assume that the genuine interval based on clinical symptoms is significantly longer, as evidenced by the individual reports with available clinical data in our suprasellar cohort. Clinical reports of the bifocal group were rare (n = 5) and the symptom interval with up to six months less prolonged compared to the suprasellar cohort.