Direct tumour-fat contact at the invasion front predicted worse colorectal cancer survival

Design

Meta-analysis of retrospective cohorts, adjusted hazard ratios

Population

Eight cohorts; 5,932 patients for cancer-specific survival

Primary outcome

Cancer-specific and overall survival

Effect

CSS HR 1.77 (1.55–2.01); OS HR 1.53 (1.39–1.68)

This meta-analysis, published in July 2026, pooled multivariable-adjusted hazard ratios from retrospective colorectal cancer cohorts in which the invasion front showed direct tumour-adipocyte contact without intervening desmoplastic stroma or inflammation — a feature first termed SARIFA. Seven cohorts (5,932 patients) reported cancer-specific survival and five (4,545) overall survival.

Tumour-adipocyte contact was associated with worse cancer-specific survival (HR 1.77, 95% CI 1.55–2.01) and overall survival (HR 1.53, 1.39–1.68), with no meaningful heterogeneity. All studies were of good methodological quality but retrospective.

The feature needs nothing beyond the routine H&E, and the effect size is comparable with established adverse features. It is not yet in staging systems or required datasets, so it should not change management on its own. But recording it costs nothing and builds the evidence for prospective validation.

• Examine the deepest invasion front for tumour glands or clusters directly touching fat cells
• Require no intervening desmoplastic stroma or inflammatory cells before calling it
• Consider recording its presence or absence as an additional comment in colorectal resection reports
• Do not use it to change staging or treatment recommendations until prospectively validated

Why it matters

A free, reproducible H&E feature could add prognostic information to every colorectal resection.

Don't overread it

All evidence is retrospective; the feature is not part of current staging or reporting datasets.

The statistics, in plain English

A hazard ratio of 1.77 means patients with the feature died of colorectal cancer at about 1.8 times the rate of those without, after adjustment for other prognostic factors. Low heterogeneity means the cohorts agreed, but all were retrospective, so reporting bias cannot be excluded.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.