Table 1 Multivariable Cox proportional hazard models estimating the contribution of several prognostic variables on MFS on CANTO

From: Deep learning assessment of metastatic relapse risk from digitized breast cancer histological slides

CANTO N =  1169

Variable

Cox model without RlapsRisk BC

Cox model with RlapsRisk BC

 

Unit HR (95% CI)

p

Unit HR (95% CI)

p

 

Histological grade 1–2

1 (ref)

N.A.

1 (ref)

N.A.

Multivariable Cox model

Histological grade 3

1.79 (0.94–3.41)

0.08

1.32 (0.69–2.54)

0.41

Age

1.18 (0.90–1.55)

0.24

1.10 (0.84–1.45)

0.49

Lymph node invasion

1.08 (1.03–1.14)

<0.005

1.08 (1.02–1.13)

<0.005

Tumor size

1.02 (1.01–1.04)

<0.005

1.02 (1.00–1.03)

0.03

Ki67 expression

1.28 (1.05–1.55)

0.01

1.23 (1.01–1.50)

0.04

RlapsRisk BC score

N.A.

N.A.

1.67 (1.31–2.13)

<0.005

  1. p Values correspond to two-sided Wald tests evaluating the null hypothesis that the hazard ratio (HR) for each variable is equal to 1. RlapsRisk BC scores are derived from inferences made on the CANTO external validation WSIs, while the model has been trained exclusively on the discovery cohort, remaining blind to the external validation data.
Back to article page