3062 - Toward Non-Operative Management of Locally Advanced Rectal Cancer Patients Undergoing Short Course Chemoradiotherapy: Early Response Prediction via Longitudinal Apparent Diffusion Coefficient

J. Pham¹, Q. Lyu², Y. Yang³, M. Cao¹, J. B. Weidhaas¹, A. Raldow¹, and X. Qi⁴; ¹Department of Radiation Oncology, University of California, Los Angeles, Los Angeles, CA, ²University of California, San Francisco, San Francisco, CA, ³Department of Radiation Oncology, University of California, Los Angeles, CA, ⁴UCLA, Los Angeles, CA

Purpose/Objective(s): To investigate longitudinal apparent diffusion coefficient (ADC) variation as an early treatment response predictor for locally advanced rectal cancer (LARC) patients undergoing short course radiotherapy followed by chemotherapy (SCRT-TNT). Materials/

Methods: A pilot cohort of nine LARC patients on a prospective clinical trial (NCT04703101) were included in the study. All patients were simulated and treated on a MR-guided linear accelerator, with a prescription of 25 Gy in 5 fractions followed by 16 weeks of mFOLFOX. ADC maps were acquired at 3 timepoints: 1) during simulation (pre-treatment, baseline), 2) 3rd fraction, and 3) 5th fraction of radiotherapy treatment using single-shot echo-planar diffusion-weighted imaging (DWI) sequence (b-values = 0, 300, 500 s/mm²). DWI ADC quantification was estimated and validated using Horos ADC plugin and a NIST QIBA Diffusion Phantom. Longitudinal DWIs were rigidly registered to the planning MRI and ADCs of rectal gross tumor volume (GTV) were measured at each timepoint. Treatment responses were assessed at 8-12 weeks after completion of neoadjuvant therapy using digital rectal examination endoscopy and MRI. Patients were classified as clinical complete response (cCR) or incomplete response. Longitudinal ADC change was correlated with post-treatment response. Prediction performance was evaluated using area-under the curve (AUC), accuracy, sensitivity, and specificity. ADC prediction threshold was determined at the optimal operating point of the ROC curve.

Results: Four out of nine (44%) patients achieved a cCR and were recommended for NOM. AUCs of mean ADC and ADC change for the GTV between baseline and fraction 3 (?ADC₁₂), baseline and fraction 5 (?ADC₁₃), and fraction 3 and 5 (?ADC₂₃) are shown in Table 1. Increasing ADC was associated with cCR. Mean ADC change between baseline and last fraction was shown to be most predictive of treatment response with an AUC of 0.85, where the optimal ?ADC₁₃ of +0.28 mm²/s demonstrated an accuracy, sensitivity, and specificity of 0.89, 0.75, and 1.00.

Conclusion: Longitudinal ADC may be used as a robust predictor for early treatment response in LARC patients undergoing SCRT-TNT. Early prediction of treatment response may help physicians tailor treatments for LARC patients interested in NOM.

Abstract 3062 – Table 1

	Complete Responder (mm2/s)	Incomplete responder (mm2/s)	AUC
ADC1	1.09 ± 0.23	1.09 ± 0.24	0.60
ADC2	1.20 ± 0.20	1.11 ± 0.31	0.65
ADC3	1.45 ± 0.14	1.14 ± 0.22	0.85
?ADC12	0.11 ± 0.25	0.02 ± 0.26	0.55
?ADC13	0.36 ± 0.23	0.05 ± 0.16	0.85
?ADC23	0.25 ± 0.24	0.03 ± 0.25	0.75