Same Rectal Cancer Treatment, but Who Truly Benefits? Pathology AI Finds a Clue in Tumor Cell Density

When treating locally advanced rectal cancer, adding another chemotherapy drug may not benefit every patient and could result in more severe side effects. A UK research team has now used artificial intelligence to reexamine pathology slides from a large clinical trial, identifying a group of patients who had been obscured by the overall results: those with higher tumor cell density appeared more likely to benefit from combining irinotecan with standard chemoradiotherapy.

This post hoc analysis drew on the phase 3 ARISTOTLE trial. Conducted across 75 UK hospitals, the trial was originally designed to assess whether adding irinotecan to capecitabine-based radiotherapy could improve outcomes for patients with locally advanced rectal cancer; the full trial did not show a general benefit across all participants. The researchers therefore turned to pretreatment biopsies to ask whether the tumors themselves contained identifiable differences.

The team applied a pathology AI trained on large open datasets to 414 biopsy images, enabling the system to identify tumor regions, distinguish cancer cells from other cells, and calculate tumor cell density. The analysis classified 188 samples into the high-density group and 226 into the low-density group. This type of work, which requires identifying large numbers of cells individually, is time-consuming and makes it difficult to maintain consistent standards in large studies when performed entirely by hand.

The results showed that, in the high-density group, adding irinotecan was associated with an approximately 43% lower risk of recurrence and an approximately 50% lower risk of death; the estimate for the risk of death was based on five years of follow-up. The low-density group did not show the same benefit. The findings offer a possible explanation: a treatment with limited effectiveness in the overall population may still have value for patients with specific histological characteristics.

If this stratification approach proves valid, its clinical significance would lie not only in identifying patients who should receive intensified treatment, but also in sparing those unlikely to benefit from additional toxicity. Irinotecan may increase burdens such as diarrhea and reduced white blood cell counts, so treatment escalation is not without cost when efficacy is uncertain. The research team has developed Octopath, an online tool that allows biopsy images to be uploaded for analysis, but the tool's availability does not mean that evidence supporting its use in medical decision-making has been established.

The most important limitation at present is that this remains a post hoc subgroup analysis of existing trial data, rather than a prospective trial in which treatment was assigned in advance according to AI results. The research team also stated clearly that the findings must first be reproduced in an independent patient cohort, and that their robustness across different hospitals, scanning equipment, and slide-processing workflows must be further confirmed before classification thresholds and clinical uses can be defined. Until these validations are completed, tumor cell density is better regarded as a predictive biomarker worth testing than as a basis for directly deciding whether to add irinotecan.