MON-320 Inter-Rater Reliability of T2 MRI Intensity of Somatotroph Adenomas; Endocrinologists vs. Neuroradiologist Pilot Study

Abstract

Background MRI T2 hypointensity of growth hormone (GH) secreting pituitary adenomas (PA) has been associated with better biochemical response to somatostatin receptor ligands and has been suggested to be useful in selecting patients with expected favorable response for pre- and post-surgery medical therapy. However, in most imaging centers, T2 intensity measurement is not part of standard neuroradiologist (NR) reporting. Objective To assess whether endocrinologists (Es) can reliably measure PA T2 signal intensity by calculating inter-rater reliability between Es and NR. Methods Retrospective review of MRI in 20 patients with pituitary somatotroph macroadenoma randomly selected from an IRB-approved PA database who had preoperative MRI available. T2 MRI intensity of the solid portion of the PA was compared to the temporal gray matter (GM) and white matter (WM): hypo- (PA< WM), hyper- (PA> GM), and isointense (WM <PA <GM). Measurements were performed separately by a NR and by two Es trained to take measurements by the same NR. Statistics: SPSS 25; Cohen kappa (κ). Results Patient mean age was 47 ± 20 years, with 12 females; mean largest PA diameter was 22.6 mm (range 11-45 mm). NR measured 12 hyper-, 7 iso- and 1 hypo-intense PA. Agreement was moderate between NR and E#1 (κ 0.72, 95%CI 0.751-1.0, p<0.001) and NR and E#2 (κ 0.638, 95%CI 0.351-0976, p<0.001) and strong between E#1 and E#2 (κ=0.90, 95%CI 0.309-0.903, p=0.001). Hypointense PA (by NR) was read by both Es as isointense. One hyperintense PA (by NR) was read by both Es as isointense. One isointense PA was read by E#2 as hypointense. Overall adenomas were; 9 densely granulated GH, 5 sparsely granulated GH, 3 mixed GH and prolactin, 1 plurihormonal, 1 not classified, and 1 no surgical intervention. Discussion Inter-rater reliability between the 2 Es was strong, however, it was moderate between each E and the NR. Factors that likely contributed to difference in measurement are heterogeneity of the PA, MRI quality, selection bias in choosing “most appropriate” site to measure intensity of adenoma, gray and white matter. Es could be trained to interpret the T2 intensity, although reliability with NR is only moderate. Interestingly, in this sample majority of T2 PA were hyperintense, but densely granulated, suggesting that preoperative identification of densely granulated tumors, which are also predictive of favorable SRL response, might be limited. More studies are needed to assess T2 correlation with pathology. Conclusion As T2 intensity (hyper-, hypo- or iso-) on MRI might be predictive of biochemical response to medical therapy in some patients with PA, we recommend T2 intensity to be part of neuroradiology reporting protocol. Our pilot study showed that endocrinologists could read MRIs after adequate training, but there is only moderate correlation with neuroradiologists.