The relationships of INT2 and ERBB2 amplification and of ERBB2 overexpression in primary breast tumors to prognostic factors, recurrence, and survival have generated considerable controversy. The rationale for this study is that long-term, recurrence-free survival is a more direct criterion for testing the validity of a tumor marker than correlation either with prognostic factors or with short-term recurrence and survival. We examined the association of recurrence with INT2 and ERBB2 amplification and ERBB2 expression by comparing primary breast tumors from patients surviving without recurrence for > or = 8.5 years after diagnosis, the LTS group, to tumors from patients recurring within two years, the RR group. The RR (N = 63) and LTS (N = 61) samples were coded and examined for amplification by Southern blotting and for expression by immunohistochemistry. Comparison between the RR and LTS groups demonstrated that INT2 amplification was associated with a significantly (P = 0.018) higher (5.6-fold) risk of recurrence, an association that remained significant after controlling for lymph node (LN), tumor size (TS), and histograde (HG) status. ERBB2 amplification and expression were not associated with a higher recurrence risk. Survival analyses within the RR group, however, demonstrated significantly shorter survival time among cases with than without ERBB2 amplification (P = 0.018, median survival 16 vs 25 months), or ERBB2 expression (P = 0.019, median survival 15 vs 25 months), but not INT2 amplification. Univariate Cox proportional hazards regression models also demonstrated significantly shorter survival among cases with ERBB2 amplification (P = 0.016) or expression (P = 0.049), that remained significant in multivariate analyses (P = 0.022) for ERBB2 amplification. These results indicate a significant positive association between INT2 amplification and risk for tumor recurrence in the RR as compared to the LTS group. The relationship of ERBB2 amplification or overexpression to patient outcome is more complex. ERBB2 amplification and expression have a significant relationship with shorter survival among patients recurrent within two years, but their occurrence in tumors from women surviving without recurrence for > or = 8.5 years suggests that ERBB2 status is not predictive of shorter survival for all breast cancers.