2026 Poster Presentations
P503: QUANTIFYING SURGICAL COMPLEXITY IN KOOS GRADE IV VESTIBULAR SCHWANNOMAS: EXTENDED ANALYSIS ON THE RELIABILITY AND VALIDITY OF A NOVEL GRADING SYSTEM
Yunjia Ni, MD1; Shawn M Stevens, MD, FACS2; Kaith K Almefty, MD3; Randall W Porter, MD3; Julio Juarez Alvarez, BFA4; Marie E Clarkson, BFA4; Jonathan Tangsrivimol, MD5; Kivanc Yangi, MD5; Takuma Maeda, MD6; Hirotaka Sato, MD6; Anthony M Asher, MD3; Anna Huguenard, MD3; Jonathan Lam, BS3; Mark C Preul, MD5; Richard D Dortch, PhD1; Michael T Lawton, MD3; 1Barrow Neuroimaging Innovation Center, Department of Translational Neuroscience, Barrow Neurological Institute, St Joseph's Hospital and Medical Center, Phoenix, AZ; 2Department of Otolaryngology and Skull Base Surgery, Barrow Neurological Institute, St Joseph's Hospital and Medical Center, Phoenix, AZ; 3Department of Neurosurgery, Barrow Neurological Institute, St Joseph's Hospital and Medical Center, Phoenix, AZ; 4Neuroscience Publications, Barrow Neurological Institute, St Joseph's Hospital and Medical Center, Phoenix, AZ; 5The Loyal & Edith Davis Neurosurgical Research Laboratory, Department of Neurosurgery, Barrow Neurological Institute, St Joseph's Hospital and Medical Center, Phoenix, AZ; 6Barrow Aneurysm & AVM Research Center, Department of Translational Neuroscience, Barrow Neurological Institute, St Joseph's Hospital and Medical Center, Phoenix, AZ
In very large (Koos Grade IV) vestibular schwannomas (VS), surgical complexity varies between cases and within specific components of an individual resection due to inter and intra-tumoral heterogeneity. Tumor vascularity, adhesion, texture, and facial nerve shape all may impact the difficulty of a given surgery. Quantitative pre-operative prediction of these features could significantly impact surgical planning and patient counseling. To this end, MRI biomarkers have shown promise for their potential to predict these features. However, establishing the predictive value of MR biomarkers requires a validated grading system of VS surgical complexity, which does not currently exist. To address this gap, we sought to develop and investigate the reliability and validity of a novel grading system (VSGS) that quantifies the local surgical complexity across distinct surgical domains of VS microsurgery.
108 surgical video clips (≤ 5?min each) from seven Koos Grade IV VS operated by the senior neurosurgeon were comprehensively sampled across three domains (tumor-nerve interface, tumor-brain interface, tumor-specific domain). After training on use of the VSGS, ten surgeons of varying skull base experience (one neurotologist and nine neurosurgeons, PGY5–37, further classified as young and senior neurosurgeons) independently rated the clips in a randomized and single-blind fashion. Inter-rater reliability was primarily assessed by Gwet’sAC2 with quadratic weights alongside 95% CIs (item-bootstrap). Criterion validity was measured by quadratically weighted Cohen’s?kappa against the senior surgeon’s retrospective unblinded ratings. To compare group rating differences, cumulative link mixed model was used to assess subgroup rating differences. Subgroup comparisons of Gwet’s AC2 and weighted kappa were also reported. Landis–Koch scale was used to benchmark agreement metrics.
Within-clip rater variability was modest (overall median SD = 0.52, IQR 0.47–0.65), with only four clips breaching SD ≥ 1.0. VSGS demonstrated almost-perfect (≥0.83) reliability for all domain-features except for tumor-nerve shape (substantial, AC2 = 0.841; 95% CI 0.794–0.880) and tumor-nerve adhesion (substantial, AC2 = 0.828; 95% CI 0.776–0.879). In terms of validity, we observed fair to moderate agreement (0.2-0.6) in vascularity features across all domains. Validity for adhesion mostly ranged from slight to moderate (0.11-0.45), with more agreement clustered in tumor-brain interface. Tumor-nerve shape demonstrated slight to moderate (0.10-0.45) validity, while tumor-specific texture feature showed lowest validity. Raters’ training level impacted adhesion ratings at the tumor–brain interface (OR?1.63,?p?=?0.03). Subgroup analysis results paralleled the overall analysis.
In this internal validation study, VSGS reliably quantified local vascularity, adhesion, and CN VII shape features in Koos IV VS. In contrast, validity of the rubrics is fair to moderate, with the tumor texture having the lowest validity. Ongoing work will include more detailed analysis of grading notes and refining the rubrics based on the preliminary data. In addition, involvement of larger sample size and validating the rubrics on different operating surgeons would be necessary. Aggregation of local scales to offer a global surgical complexity score for a single case is also a viable research avenue. Ultimately, a fully validated VSGS would be used to correlate with preoperative imaging (predictive capacity to anticipate intraoperative findings).
.png)
