Tumour cell density quantified by artificial intelligence is associated with differential benefit from irinotecan-based chemo-radiotherapy in locally advanced rectal cancer: a post-hoc study of the phase 3 ARISTOTLE trial.

内容摘要

Tumour cells and tumour-associated stroma are key components of the tumour microenvironment, and their interaction impacts disease progression and treatment resistance in rectal cancer. This study introduces a computational approach to quantify tumour cell density (TCD) within epithelial and stromal regions and assess whether treatment response differs according to TCD status in patients with locally advanced rectal cancer (LARC) undergoing neoadjuvant chemoradiotherapy (nCRT). The data of 414 ARISTOTLE trial (ISRCTN09351447) participants with available digitised pre-treatment biopsies were analysed in this study. We defined TCD as the proportion of tumour cells within the tumour and stroma tissues and quantified TCD using an AI framework applied to digitised haematoxylin and eosin-stained whole-slide images. The patients were stratified as TCD-high/TCD-low using a cut-off value of 0.5 (50% of tumour cells). The TCD status was combined with treatment arms [CRT (capecitabine + radiotherapy) and IrCRT (experimental capecitabine + irinotecan + radiotherapy)] to stratify the disease-free survival (DFS), overall survival (OS) and pathological complete response (pCR) rates. Among the patients analysed, 188 (45%) of 414 patients were classified as TCD-high and 226 (55%) as TCD-low. A significant treatment-TCD interaction was observed for both DFS (χ2 = 6.88, p = 0.009) and OS (χ2 = 10.61, p = 0.001). In the TCD-high subgroup, the use of IrCRT was associated with significantly longer DFS (HR = 0.57, 95% CI: 0.36-0.90, p = 0.014) and OS (HR = 0.50, 95% CI: 0.30-0.84, p = 0.008), along with a higher pCR rate (20 [23%] vs. 8 [11%], OR = 2.46, 95% CI: 1.01-5.98; p = 0.042 [p = 0.13 after adjustment for multiple testing]), compared to CRT. However, in the TCD-low subgroup, no significant difference in DFS was observed between treatment arms (HR = 1.29, 95% CI: 0.85-1.95, p = 0.22), while patients who received IrCRT exhibited a trend towards worse OS (HR = 1.55, 95% CI: 0.96-2.51, p = 0.07). No significant difference in pCR rates was observed (12 [14%] vs. 23 [22%], OR = 0.60, 95% CI: 0.28-1.29; p = 0.19). In this post-hoc, hypothesis-generating analysis of the ARISTOTLE trial, higher TCD was associated with differential outcomes after irinotecan-intensified neoadjuvant chemoradiotherapy compared with standard chemoradiotherapy. These findings support further evaluation of AI-derived TCD as a candidate predictive biomarker in independent retrospective and prospective cohorts. Cancer Research UK Radiation Research Network - Project Seed Funding (RRNPSF-Jan21/100001), Cancer Research UK ARISTOTLE sample collection grant (A18745), UK Research and Innovation Future Leadership Fellowship (No. MR/T040785/1) and the Radiation Research Unit at the Cancer Research UK City of London Centre Award (C7893/A2899).