Validation of a genome-based model for adjusting radiotherapy dose (GARD) in patients with locally advanced rectal cancer

Neoadjuvant radiotherapy is the standard care of locally advanced rectal cancer. Although a majority of patients received the same dose, the curative efficacy varies among individuals. In recent years, cancer treatment has entered the era of precise medical care, and how to identify patients for proper treatment by molecular signature is an important path of individualized therapy. This study aimed to establish and validate a genome-based model for adjusting radiation dose (GARD) for Chinese locally advanced rectal cancer through gene expression microarrays, and to evaluate the response of the GARD model in predicting the efficacy of neoadjuvant radiotherapy. Fresh-frozen primary tumor from 64 patients with locally advanced rectal cancer undergoing neoadjuvant radiotherapy from 2015 to 2018 were included. The gene expression profile was analyzed using Affymetrix 3000Dx gene-chip scanner. The radiosensitivity index (RSI) and GARD were calculated using the pGRT™ algorithm. Neoadjuvant rectal cancer score (NAR) was selected as efficacy evaluation indicators. Patients were divided into high and low NAR scoring groups, and two-sample t-test was used to analyze the differences in GARD values between different NAR subgroups. ROC curves were used to calculate the cut-off values and the area under the curve (AUC) for assessing the validity of the GARD models. The personalized radiation dose ( pGRT dose )can be computed using the formula nd = GARD / (α + βd). Among patients, 1.5% T2, 46.3% T3, and 52.2% T4. Wherein pCR (n = 10; 15.6%) and no pCR (n = 54; 84.4%). The median NAR is 8.43 (rang from 0 to 50.34, IQR 3.75-14.98). NAR > 8.43 (n = 27; 42.2%) and NAR ≤ 8.43 (n = 37; 57.8%), suggesting that there are significant individual differences in clinical efficacy of patients with similar tumor stages and under the same treatment conditions. The median RSI is 0.48 (rang from 0.22 to 0.92, IQR 0.41-0.55). Median GARD was 18.40 rang from (rang from 2.26 to 37.52, IQR 14.94-22.28) within tumor tissue, suggesting individual differences in the efficacy of radiation therapy. The RSI value was significantly lower in the NAR low group (NAR ≤ 8.43) than in NAR high group (NAR > 8.43) (0.44 vs. 0.54, p = 0.0003). The GARD value was significantly higher in the NAR low group (NAR ≤ 8.43) than in NAR high group (NAR > 8.43) (21.01 vs. 15.88, p = 0.0004). Using the Receiver Operating Characteristic (ROC) curve analysis, a GARD threshold of 17 was identified as optimal, covering 37.5% of the 64-patient sample, with an area under the curve (AUC) of 0.75. In the external validation cohort, the high GARD score group demonstrated superior DFS compared to the low GARD score group(p < 0.001). Only 17% of patients had pGRT dose within the guideline recommended dose (45-50 Gy). The differences in NAR values among LARC patients receiving standard neoadjuvant radiotherapy suggest significant individual differences in clinical outcomes among patients with similar tumor stage and the same treatment conditions. Patients with a GARD value exceeding 17 exhibit a more favorable prognosis. Our results suggest that the gene expression-based pGRT™ algorithm has good efficacy prediction performance in preoperative concurrent radiotherapy for locally advanced rectal cancer, suggesting the potential clinical application of this method to guide the designation of individualized radiotherapy doses.