Real World Data Estimates of Contralateral Nodal Involvement Based on Clinical and Automatically Extracted Treatment Plan Features

Abstract

To evaluate the association of contralateral nodal involvement (CNI) with automatically defined midline proximity in patients (pts) with head and neck cancer based on treatment planning (TP) real world data (RWD).All available TPs for head and neck cancer pts between 2010 and 2017 were analyzed in an automated data analysis pipeline where the pt's midline was determined objectively using their external contour and discounting distortion due to neck lymphadenopathy via linear modeling. Primary gross tumor volumes (GTVp) and nodal volumes (GTVn) were contoured by treating radiation oncologists. Objective tumor laterality (OTL) was calculated as the difference between the GTVp center of mass and the midline in the R/L direction. OTL accuracy compared to the clinically assigned laterality (when available) was assessed. Midline proximity (MP) was the difference between the medial edge of the GTVp bounding box to the patient midline with positive values representing progressively lateralized tumors and negative values representing tumors crossing midline. Nodal involvement (ipsilateral and contralateral levels 1-5 and retropharyngeal) was determined using the TP nodal target contours labeled by neck level as is standard practice at our institution during treatment planning. The primary endpoint, CNI, was defined based on these TP contours. Volumes of GTVp were calculated from the TP. Clinical variables including HPV status, age, smoking, disease site and subsite were used for model development. The factors predictive of CNI were assessed using logistic regression (LR).Of 2458 pts available for analysis, 2031 had a clinically assigned laterality. The accuracy of automatically assigned laterality was 0.91 when compared with the clinically assigned laterality. Using the automatically assigned laterality (n = 2458), the overall rate of CNI was 45% and varied by disease site and nodal level involved: oropharynx (55.1%, n = 629/1142), larynx (17.6%, n = 124/703), nasopharynx (79.7%, n = 231/290). Multivariable logistic regression (LR) including subsite, HPV status, GTVp, GTVp center of mass relative to midline, GTVp midline proximity, and ipsilateral nodal involvement was performed. In the oropharynx, multivariable LR demonstrated GTVp volume (P < 0.01) and ipsilateral nodes (P < 0.01) were associated with CNI. Neither HPV status or subsite (e.g., base of tongue) was significant on multivariable analysis. In the nasopharynx, GTVp and ipsilateral nodes were significantly associated with CNI on LR. In the larynx, GTVp, ipsilateral nodes, and midline position were all significantly associated with CLI. Nomograms estimating CLI were generated for each disease site based on significant features from the LRs.Automated TP analysis can be used to generate RWD-based predictions of CNI. CNI risk remains a critical radiotherapy endpoint and RWD may help determine CLI risk using combinations of anatomic and clinical features.