Automated imaging-based tumor burden and pre-treatment circulating tumor DNA in HPV-associated oropharynx cancer
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
Artificial intelligence (AI)-based imaging analysis has applications for the diagnosis of head and neck malignancies, and serum circulating tumor-associated DNA (ctDNA) is an emerging biomarker being evaluated for response assessment and risk stratification in human papilloma virus (HPV)-associated oropharynx squamous cell carcinoma (HPV-OPSCC). The relationship between automated imaging biomarkers and ctDNA has not yet been explored.
Objective
To test the association between ctDNA and AI-derived measures of tumor burden among patients with HPV-OPSCC.
Design, Setting, and Participants
This cross-sectional study included patients who were treated with curative intent for HPV-OPSCC between 2020-2023, prospectively enrolled on a blood collection protocol (Clinical trials.gov identifier: NCT04965792 ).
Exposures
Clinical factors including demographics, AJCC 8 th edition clinical staging, and HPV genotype.
Main Outcomes and Measures
Pre-treatment serum measurement of circulating tumor-tissue modified viral (TTMV) HPV-DNA using a commercially available test, measured as a continuous value (fragments/mL). Primary tumor and nodal volumes, total tumor volume, and cystic/necrotic nodal volume were generated on pre-treatment diagnostic or radiation CT- planning scans using a prospectively validated AI auto-segmentation algorithm. Assessments of model fit: Akaike information criterion (AIC) and Bayesian information criterion (BIC).
Results
170 patients with HPV-OPSCC were included in the study. On univariable regression, primary tumor volume (coeff=39.43, p<0.001), nodal volume (coeff=39.54, p<0.001), AJCC 8 th edition Tumor (T) stage (coeff = 1031.09, p=0.009), Nodal (N) stage (coeff=1840, p=0.018), HPV subtype 16 (coeff=3072.40, p=0.006), and CCI (coeff=-596.60, p=0.038) were associated with ctDNA. Cystic nodal volume was not associated with ctDNA (coeff=0.31, p=0.11). On multivariable analysis, primary tumor and nodal volumes were associated with ctDNA (coeff=34.79, p=0.001 and coeff=24.68, p=0.022, respectively), but T and N stage were not (coeff=-439.28, p=0.37 and coeff=238.19, p=0.29, respectively). Including automated tumor and nodal volumes improved model fit compared to T and N stage alone (3420.96 vs 3435.88 AIC, 3449.18 vs 3457.83 BIC).
Conclusions and Relevance
AI-automated volumetrics on pretreatment imaging are independently associated with ctDNA, controlling for clinical stage. The association is stronger than staging and improved predictive capacity of regression models. AI-automated volumetrics may provide a practical correlate to ctDNA levels and help risk stratify patients.
