Table 2 Cox regression model estimates of the significance of predictive factors for overall survival

(a) Univariate Cox regression model

  Non-papillary gross finding

0.062

1.76 (0.97–3.18)

  Pathological T category ≥ 2

0.086

1.94 (0.91–4.12)

  Pathological T category ≥ 3

0.018

2.03 (1.13–3.63)

  High grade tumor

0.87

1.17 (0.16–8.54)

  Lymphovascular invasion

 < 0.0001

3.27 (1.84–5.80)

  Concomitant subtype histology

0.097

1.68 (0.91–3.09)

  Concomitant carcinoma in situ

0.23

0.61 (0.28–1.37)

  Pathological lymph node metastasis^a

0.043

2.34 (0.99–5.54)

  Positive surgical margin

0.020

2.39 (1.15–4.96)

  High budding

 < 0.0001

3.82 (2.19–6.67)

  High TAICs (score 2–3)

0.51

1.20 (0.69–2.09)

  High periostin expression

 < 0.0001

4.92 (2.68–9.01)

(b) Multivariate Cox regression model

  Pathological T category ≥ 3

0.42

0.74 (0.35–1.55)

  Lymphovascular invasion

0.074

1.97 (0.94–4.13)

  Pathological lymph node metastasis^a

0.73

0.82 (0.27–2.47)

  Positive surgical margin

0.31

1.59 (0.65–3.89)

  High budding

0.43

1.38 (0.62–3.04)

  High periostin expression

0.00072

3.62 (1.72–7.64)