Immune infiltration marks who gains most from pertuzumab

Design

secondary biomarker analysis of a randomised phase 3 trial, median follow-up 74.1 months

Population

4,262 tumour samples from 4,805 patients with early HER2-positive breast cancer in APHINITY

Primary outcome

invasive disease-free survival by stromal TIL level and scoring method

Effect

higher TILs HR 0.41–0.93; pertuzumab benefit at higher TILs HR 0.36–0.48; 6-year absolute gain 12.1 percentage points (SD 2.8) in node-positive, sTIL ≥70%

APHINITY randomised 4,805 patients with early HER2-positive breast cancer to chemotherapy plus trastuzumab with either pertuzumab or placebo. This secondary analysis scored stromal tumour-infiltrating lymphocytes on 4,262 haematoxylin and eosin images by four methods — standardised manual assessment, automated digital scoring, an AI lymphocyte quantification, and two AI spatial features — with median follow-up 74.1 months.

Manual scoring was reproducible: intraclass correlation 0.84 (95% CI 0.79 to 0.88) across five independent pathologists on 262 samples. Concordance between manual and automated methods was only modest, which matters more than it sounds — AI-based scoring reclassified 120 of 1,035 node-positive tumours (11.6%) from immune-low to immune-high, and that reclassified group showed greater separation between pertuzumab and placebo curves than tumours both methods called immune-low.

Higher lymphocyte levels were associated with better invasive disease-free survival by every method (hazard ratios 0.41 to 0.93), and pertuzumab benefit was greater at higher levels (0.36 to 0.48). The largest absolute gain at six years was in node-positive tumours with manual scores of 70% or more: 12.1 percentage points (SD 2.8). AI spatial hotspot scores added consistent information to any density measure (all p < 0.010).

So a cheap stain, read to a standard protocol, separates who gains a lot from dual blockade from who gains little — and the automated methods are not yet interchangeable with the manual read they are meant to replace.

• Report stromal TILs using the standardised manual method on node-positive HER2-positive tumours
• Do not treat digital or AI scores as interchangeable with manual scoring — concordance was only modest
• Use TIL level to inform the strength of the pertuzumab conversation, not to withhold it
• Record which method produced a reported score, since the methods reclassify patients

Why it matters

Which scoring method is used changes which patients are called immune-low, and therefore who appears to benefit.

Don't overread it

A secondary biomarker analysis — it has not been validated in an independent cohort and should not yet select who receives pertuzumab.

The statistics, in plain English

An intraclass correlation of 0.84 means pathologists agree well — this is a measure that travels between readers, which most histological scores do not. The hazard ratio ranges of 0.41 to 0.93 span the different scoring methods, so the prognostic effect is consistent in direction but varies in size with how it is measured. A 12.1 percentage point absolute gain at six years is roughly one extra person free of invasive disease for every eight treated in that subgroup. This is a secondary analysis with subgroups, so it generates rather than confirms a stratification.

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.