(P109) FDG-PET Feature and Texture Analysis as Potential Predictors of Pathologic Outcome Following Neoadjuvant Chemoradiotherapy for Locally Advanced Rectal Cancer

Richard Tuli, MD, PhD, Yong Yue, PhD, Jason Naziri, BSc, Mourad Tighiouart, PhD, Hendifar Andrew, MD, Benedick Fraass, PhD, Howard Sandler, MD, Wensha Yang, PhD; Cedars-Sinai Medical Center

Background: Intriguing data suggest that clinical complete response following neoadjuvant chemoradiation (CRT) may obviate the need for definitive resection in patients with locally advanced rectal cancer. However, accurate radiographic assessment of clinical response remains challenging. Whereas prospective data suggest that positron emission tomography/computed tomography (PET/CT) scans do not have adequate value for predicting a pathologic complete response (pCR), minimal data are available on the value of PET/CT in predicting tumor regression grade (TRG)-based pathologic response. Herein, we investigate whether morphologic, textural, and quantitative features of PET/CT can be used to predict pathologic response in rectal cancer.

Methods: With institutional review board (IRB) approval, we reviewed patients with locally advanced rectal cancer treated with neoadjuvant CRT from 2011–2013 who also underwent fluorodeoxyglucose (FDG)-PET/CT scans within 6 weeks pre-CRT and 6–8 weeks post-CRT. PET/CT images were deformably registered to the radiation therapy (RT) planning CT. Tumor volumes of interest were divided into 4.8-mm³ subvolumes, characterized by mean RT dose and comprehensive texture (energy, correlation, variance, sum mean, cluster tendency, inverse variance) and feature (pre/post-CRT maximum standardized uptake value [SUV_max], mean SUV, coefficient of variation [CV], total lesion glycolysis [TLG], tumor size, MTV40 [metabolic tumor volume 40%]) analysis. Mann-Whitney test and Cox modeling were used to identify predictors of complete/near-complete response (TRG 0–1) vs moderate/minimal/no response (TRG 2–3).

Results: Fifteen patients with locally advanced rectal cancer were identified, all of whom received pre-CRT PET, while 9 of 15 also received post-CRT PET. Pathologic downstaging was noted in 60% of patients, and 13% had a pCR (ypT0N0). TRG of 0–1 and 2–3 was noted in 27% and 73% of patients, respectively. Pre-CRT tumor volume (P = .01), TLG (P = .04), and MTV40 (P = .01) were found to be significant predictors of TRG, whereas no correlation was seen with SUV_max (P = .38), mean (P = .29), or pre-/post-PET textural feature analysis between pathologic responders and nonresponders.

Conclusions: Parameters of metabolic response were generally not predictive of pathologic response or TRG in locally advanced rectal cancer, with the exception of pre-CRT tumor volume, TLG, and MTV40. Further research is warranted to further validate these predictors, given the potential clinical significance.