Instrumental variables (IV) techniques were used to estimate outcome differences between treatment options among marginal patients with early stage prostate cancer diagnosed during 1986-1993. Access to care, area socioeconomic and healthcare market characteristics affected choice of treatment and these associations varied among age groups. Unbiased estimates of treatment effects also varied among these age groups. Results suggest that an increase in aggressive treatment for patients with presumed localized early stage prostate cancer would have resulted in increased survival rates but these effects decreased with patient age. Further, an increase in surgery among those patients aggressively treated would have also increased survival and these effects increased with patient age. health policy should be tailored to these differences between age subgroups of older men. Though results suggested that increased aggressive treatment in general, and surgery in particular, would have been cost-effective, conclusions about cost-effectiveness are tenuous because Medicare costs were available only from 1991-93.