[1171] Practical Clinical Application of a Histologic Risk Assessment Model for Head and Neck Squamous Cell Carcinoma.

Robert J Cabay, Tudor Vladislav, Hui Xie, Iffat Choudhry, Steven A Garzon, Vidya Nagrale, Seung Park, Douglas Washing, ShouJin Wu, Yi Zhou, Andre Kajdacsy-Balla. University of Illinois at Chicago

Background: A very promising histologic risk assessment model for head and neck squamous cell carcinoma (HNSCC) has been proposed in which a risk score is derived from the sum of scores for infiltrative pattern, perineural invasion, and lymphocytic response (Brandwein-Gensler et al. Am J Surg Pathol 2005;29:167-178). Risk assessment model validation included live training sessions for the pathologists who applied the model (Brandwein-Gensler et al. Am J Surg Pathol 2010;34:676-688). We questioned whether interobserver agreement would differ if observers applied the model after reading the articles cited above without the benefit of live training sessions, more closely simulating how the model would most likely be implemented in clinical practice.
Design: 6 pathologists and 4 residents at our institution applied the model to 20 cases of HNSCC from our files after learning the scoring method in the model by reading the articles cited above. No training sessions were given. Interobserver agreement was measured using the scores that the observers assigned.
Results:

Interobserver Agreement (as expressed by Kendall's Coefficients of Concordance)

Observers	Infiltrative Pattern	Perineural Invasion	Lymphocytic Response	Risk Score
Pathologists	0.46	0.52	0.63	0.68
Residents	0.43	0.60	0.63	0.71
Pathologists + Residents	0.38	0.49	0.59	0.66

Observers reached complete agreement on risk classification (low/intermediate risk vs. high risk) in 7 of the 20 cases.
Conclusions: In our study, interobserver agreement showed some variation across the risk score elements in this HNSCC risk assessment model. Pathologist, resident, and overall risk score correlation coefficients obtained without training sessions were similar (0.68, 0.71, and 0.66, respectively). Prior to widespread implementation of the model, evidence of repeated achievements of greater levels of interobserver agreement would be highly desirable. The establishment of an Internet-based, annotated image bank may provide an easily-accessible training resource that may lead to a successful, broad incorporation of the model into clinical practice.
Category: Head & Neck

Tuesday, March 1, 2011 1:00 PM

Poster Session IV # 145, Tuesday Afternoon