Table 2 Prognostic independence of relevant clinical and histological parameters in the analyzed the Cancer Genome Atlas (TCGA) patient cohorts

From: Multi-resolution deep learning characterizes tertiary lymphoid structures and their prognostic relevance in solid tumors

 

HR

95% CI

p-value

LUSC (manual TLS)

   

STAGE(III+IV)

1.448147

[0.9927, 2.113]

0.0546

Manual TLS

0.989205

[0.9730, 1.006]

0.1984

LUSC (predicted TLS)

   

STAGE(III+IV)

1.439338

[0.9876, 2.0977]

0.0581

Predicted TLS

0.984524

[0.9696, 0.9997]

0.0457

KIRC (manual TLS)

   

STAGE(III+IV)

3.326768

[2.1327, 5.189]

1.16e-07

Age

1.021708

[1.0025, 1.041]

0.0267

Manual TLS

1.057970

[0.9769, 1.146]

0.1657

KIRC (predicted TLS)

   

STAGE(III+IV)

3.404224

[2.1901, 5.291]

5.22e-08

Age

1.024197

[1.0050, 1.044]

0.0135

Predicted TLS

1.048095

[0.9659, 1.137]

0.2596

  1. Clinical, pathological and histological parameters found to have a significant association with overall survival in the whole TCGA cohorts (training + validation + test sets) by univariate Cox regression were tested for their independence by multivariate Cox regression. Stage (I+II versus III+IV) was used as a categorical variable and all other parameters were used as continuous variables. Tertiary lymphoid structure (TLS) and germinal center (GC) density was determined as the number of manually annotated or predicted regions per cm2 of the analyzed tissue. Hazard ratio (HR), its 95% confidence interval (CI) and p value for each parameter are reported.
Back to article page