Table 1: Relapse crude cumulative incidence regression model based on pseudo-values. For each covariate the exponent of regression coefficient is estimate of the ratio between crude cumulative incidences. N0 and N1 indicates absence or presence of axillary lymph node metastases respectively. T is pathological tumour size (in cm). PgR +,- indicates > 25 or <= 25 femtomoles progesterone receptors per milligram of cytosolic protein respectively, ER +,- indicates > 10 fmoli or <= 10 of estrogen receptors femtomoles per milligram of cytosolic protein respectively and Time, time’ and time’’ represents spline basis for time.

	Estimate	Standard Error	Wald statistic	p-value
(Intercept)	-5.074	0.387	172.11	< 0.001
time	0.07	0.008	81.643	< 0.001
time'	-0.284	0.041	47.427	< 0.001
time''	0.48	0.074	42.294	< 0.001
PGR (+ vs. -)	0.083	0.142	0.339	0.561
ER (+ vs. -)	0.313	0.172	3.317	0.069
T (1-2 vs ≤ 1)	0.628	0.187	11.323	0.001
T (> 2 vs. ≤ 1)	0.888	0.203	19.207	< 0.001
N (1 vs. 0)	0.163	0.112	2.117	0.146
age (41-50 vs. ≤ 40)	-0.245	0.162	2.272	0.132
age (51-60 vs. ≤ 40)	-0.229	0.168	1.861	0.172
age (> 60 vs. ≤ 40)	-0.63	0.21	9.044	0.003