Design: Retrospective cohort with whole-exome and whole-transcriptome sequencing
Population: 90 intrahepatic cholangiocarcinomas, 25 FGFR2/3-altered
Primary outcome: Clinicopathological and molecular features; survival
Effect: Small-duct type 100% vs 32%; univariate DSS HR 0.29 (95% CI 0.09 to 0.95), not significant on multivariate

A retrospective study sequenced 90 intrahepatic cholangiocarcinomas with combined whole-exome and whole-transcriptome sequencing. Twenty-five (28%) carried FGFR2 or FGFR3 alterations, mostly FGFR2 fusions with a wide range of partners.

Compared with wild-type tumours, the FGFR2/3-altered group was younger (median 61 vs 68 years) and had larger tumours (9.6 vs 6.4 cm). All were small-duct type (100% vs 32%), they were mostly non-mucin-producing (84% vs 3%) and more often well differentiated. BAP1 alterations were more common and TP53 less common. Disease-specific survival was better on univariate analysis (HR 0.29, 95% CI 0.09 to 0.95) but not after adjustment.

For the reporting pathologist, the morphology is the useful part: a small-duct, non-mucinous intrahepatic cholangiocarcinoma is the tumour most likely to carry an FGFR2 fusion, which has targeted treatment options. That is a reason to make sure such tumours reach molecular testing.

The survival advantage did not survive multivariable analysis and the series is small, so it is not a prognostic marker on its own.

Record small-duct versus large-duct type in intrahepatic cholangiocarcinoma reports.
Note mucin production; non-mucinous small-duct tumours were enriched for FGFR2/3 alterations.
Flag such tumours for comprehensive genomic profiling, including fusion detection.
RNA-based testing helps catch the wide range of FGFR2 fusion partners.

Why it matters

Morphology can point the oncologist towards a targetable alteration.

The statistics, in plain English

A univariate hazard ratio of 0.29 suggested better survival, but after adjusting for other factors the estimate became 2.52 with a huge interval (0.14 to 45), meaning the data cannot tell either way.