Background. A number of factors concerning the prognosis of cancer pat
ients have been reported, but the evaluation of individual factors is
not always consistent. The author considered that the discrepancies ma
y lie in part in that factors may have a changing influence on relapse
with time. The present study was intended to reveal the phenomenon an
d its clinical significance in patients with breast cancer. Methods. E
leven pathological prognostic factors in 462 patients with primary bre
ast cancer were analysed with special reference to the influence on re
lapse and its changes with time. The Cox regression analysis was perfo
rmed for whole and three stratified non-overlapping observation period
(within 2, 2-5 and 5 years more), using factors including extent of n
odal involvement (N), nuclear grade (NG), tumor infiltrating lymphocyt
es (TIL), tumor size, fat infiltration (F) and others. Results. Cox re
gression analysis for whole observation period showed N followed by NG
, TIL, F and others as the most important factors on relapse. But the
analysis performed for 3 stratified intervals revealed that the influe
nce on relapse decreased with time for N and NG, but tended to increas
e with time for F. The suppressive effect of TIL on relapse was not ev
ident up to 2 years but was significantly emphasized after 2 years pos
toperatively. Conclusions. In conclusion, a distinct time-dependency w
ith varied patterns was found in the influence of some prognostic fact
ors on relapse, which implies the assessment of prognostic factors may
vary according to the length of the observation period.