Table 1 Published predictive models for mortality in breast cancer patients.

Jimenez-Lee et al.⁸

BC stages 0–III

389

5

T, N and hormone receptors

Stage IV was excluded

The score for each variable was not justified and the definition of the risk groups was not justified

No

Chang et al.⁹

BC stages 0–III and follow-up >1 year

818

5

Age, diagnostic methods, tumor grade, N, hormone receptors and chemotherapy

Assessed all-cause mortality, stage IV was excluded, diagnostic methods are not useful, the model was not explained

They used a logistic regression model with censored data and the events-per-variable was less than 10.

AUC = 0.894; p-value (H-L) = 0.945

Fan et al.¹⁰

Invasive BC with mastectomy

1016

2 and 5

Age, T, N, M and estrogen receptors

Assessed all-cause mortality, a specific population and the model is laborious

No comment

C-statistic = 0.80; Calibration plot

Faneyte et al.¹¹

BC without distant metastasis, <54 years, ≥4 positive lymph nodes, no previous other malignancies, treated with surgery and adjuvant therapy

739

2

Grade, estrogen receptors and N

Assessed all-cause mortality, a very specific population, the follow-up was very short and the model was not explained

They used a logistic regression model with censored data and it was not internally validated

No

Fontein et al.¹²

Postmenopausal, endocrine-sensitive and early BC

2602

5

Age, hormone receptors, grade, T, N, HER2, treatment and locoregional recurrence

Assessed all-cause mortality, a very specific population, early BC was not defined, the model is laborious to use and the web tool does not work*

Internal validation was not assessed with bootstrapping

C-index = 0.70–0.79; HSF = 0.995

The Nottingham Prognostic Index^13,14,15.

Invasive primary operable BC with no other malignancies

387

5

T, N and grade

Assessed all-cause mortality and a specific population

The choice of cut-points was not explained, the risk groups changed in each publication and no bootstrapping

Graphically