Image-based consensus molecular subtypes and colon cancer recurrence: Understanding the impact of lifestyle factors across subtypes of colon cancer.

A nested case-control study was implemented in two prospective cohorts of colon cancer patients. Cases were patients with recurrence (n = 167), and controls (n = 668) were matched using incidence density sampling. Formalin-fixed paraffin-embedded tumour tissue was used to determine image-based CMS (imCMS) from hematoxylin and eosin-stained sections using a deep learning model. Lifestyle factors and dietary intake were assessed at diagnosis using questionnaires to construct the lifestyle inflammation score (LIS) and the dietary inflammation score (DIS). Associations between imCMS, DIS, LIS and recurrence were investigated using multivariable conditional logistic regression analyses.

Individuals with an imCMS4 tumour had the highest risk of a recurrence. In addition, the inflammatory potential of the diet was associated with recurrence only in patients with imCMS4.

Patients with an imCMS4 compared to those with an imCMS2 tumour had a higher risk of recurrence (IRR_{per_1_unit_increase} 2.22, 95%CI 1.28; 3.85). A higher DIS was not associated with recurrence (IRR_{per_1_unit_increase} 1.04, 95%CI 0.96; 1.12), while the risk of recurrence was higher in patients with a higher LIS (IRR_{per_1_unit_increase} 1.21, 95%CI 0.97; 1.52), albeit non-significant. In the subgroup of patients with an imCMS4 tumour, a higher DIS was associated with a 39% higher risk of recurrence (I IRR_{per_1_unit_increase} 1.39, 95%CI 1.08; 1.78). Associations between the LIS and risk of recurrence were not clearly different across imCMS subgroups.

We investigated associations between consensus molecular subtype (CMS) and colon cancer recurrence, and explored how the inflammatory potential of diet and lifestyle relate to recurrence across CMS subgroups.