Development of a Magnetic Resonance Grading System to Predict Cavernous Sinus Compartmental Invasion of Pituitary Adenomas

Objective: Cavernous sinus (CS) invasion is of paramount importance in the surgical planning of pituitary adenomas. The most widely used classification (Knosp grading) has several limitations, such as its evaluation of only superior and inferior compartments, and variability in the position of the lines, leading to variable incidences of true invasion and low interobserver agreement. Therefore, we set out to develop a new classification to address these shortcomings.

Methods: A classification was developed based on the evaluation of each CS compartment (superior, inferior, lateral, and posterior) and clinoidal space. For each compartment, a line bisecting the carotid artery, following anatomical planes, was used to grade the lateral extension of the adenoma. Tumors medial to these lines were graded as 0, and those lateral to it were graded as 1. The classification was used to evaluate MRIs of 255 patients (510 CS) that underwent endoscopic endonasal approach (EEA), either with transcavernous extension (139 CS) or without it (371 CS). Invasion of each CS compartment was evaluated by intraoperative assessment in transcavernous cases, and by postoperative MRI in the rest. Results were compared with the Knosp classification, stratified by the presence of apoplexy, carotid tortuosity, and previous surgery, and interobserver agreement (IOA) was calculated.

Results: Cavernous sinus invasion was present in 19.4% of CS, with superior compartment invasion in 13.3%, posterior compartment in 12%, inferior compartment in 11.2%, lateral compartment in 4.5%, and clinoidal space in 10.2%. Our classification yielded an overall sensitivity of 90.8% (95% CI: 86.6–94%), specificity of 98.4% (95% CI: 97.8–98.9%), PPV of 86.5% (95% CI: 82.3–89.8%), NPV of 99% (95% CI: 98.5–99.3%), and accuracy of 97.6% (95% CI: 96.9–98.2%), with comparable results in all compartments. Previous surgery increased the rate of invasion of grade 0 in superior and posterior compartments, as well as in Knosp grade 2 cases. IOA was rated as almost perfect for superior and inferior compartments, and substantial for lateral, posterior, and clinoidal space. Knosp classification achieved only moderate agreement.

Conclusion: Our classification provided a reliable and accurate tool to evaluate CS compartment invasion, outperforming Knosp classification.

Fig. 1 Classification of superior (A, B) and inferior (C, D) compartments in coronal MR. A vertical line bisects the horizontal carotid artery, with adenomas staying medial to this line graded as 0, and those with extension lateral to this line as 1.

Fig. 2 Classification of lateral compartment in coronal MR. The line bisects the carotid on the horizontal plane. Adenomas crossing this line on the lateral aspect of the carotid, either from the superior (A, B) or inferior (C, D) compartment are graded as 1, while adenomas not crossing this line are graded as 0.

Fig. 3 (A, B) Posterior compartment classification: The line bisects the short vertical segment of the carotid artery parallel to the midline. Adenomas staying medial are graded as 0, and those lateral as 1. (C, D) Clinoidal space classification. The line bisects the clinoid segment of the carotid artery. Adenomas staying medial, behind the clinoid carotid, are graded as 0, and those lateral as 1.

Publikationsverlauf

Artikel online veröffentlicht:
07. Februar 2025