Most cancer vaccine trials are based on efficacy studies against transplanted mouse tumors that poorly reflect the clinical situation. We constructed adenoviruses expressing interleukin-7 and B7.1 and tested their therapeutic efficacy after transfer into established transplanted and nontransplanted 3-methylcholanthrene-induced tumors. The adenoviruses efficiently induced rejection of transplanted tumors, leaving behind systemic immunity. Against nontransplanted tumors of similar size, there were almost no therapeutic effects. This result was not due to the site of tumor development, tumor type, general immune suppression, or differences in transduction efficacy. Adenoviral expression of beta-galactosidase as a surrogate antigen in nontransplanted tumors induced cytotoxic T cells that were unable to quantitatively reach the tumor site. Based on rigorous mouse models and an effective in situ immunization procedure, it is suggested that cancer vaccines can be effective, if at all, against "minimal residual disease"; additional experimental procedures must be found against established nontransplanted tumors.