OBJECTIVES: Recent randomized controlled trials demonstrated the efficacy of adjuvant chemotherapy for certain patients with early-stage lung cancer (Winton et al. NEJM 2005) while the efficacy of adjuvant radiation therapy hasn't been determined in clinical trials for these patients. Both treatments have been used in practice for years and it isn't clear whether patient survival benefits are available by expanding treatment rates. OBJECTIVES: estimate the effectiveness of these treatments in practice, assess whether survival benefits are available from expanding treatment rates. It has been suggested that instrumental variable estimates on observational data provide such an interpretation (McClellan et al. JAMA, 1994). METHODS: Apply instrumental variable (IV) methods to 7614 early-stage lung cancer patients that received surgery from the SEER-Medicare databases (1991-1999) to assess the effectiveness of each adjuvant treatment. Instruments included area-level Herfindahl indices of patient concentration across providers and area-level treatment rates across time. Covariates included stage, grade, tumor site, age, gender, and race. RESULTS: Using IV methods, we found that our instruments explained a statistically significant portion of the variation in adjuvant treatment use. Using the treatment variation stemming from the instruments, we found the adjuvant chemotherapy had a positive effect on three-year patient survival (p<0.05) while adjuvant radiation therapy didn't. Using standard risk-adjustment models leaving residual confounding in spite of controlling for measured covariates, both adjuvant chemotherapy and adjuvant radiation therapy had statistically significant negative relationships with three-year patient survival (p<0.05). CONCLUSION: Using IV methods provides estimates of effects of treatments in practice, suggesting survival rates would increase if adjuvant chemotherapy rates increased for early-stage lung cancer patients. No survival gains appear available from expanding adjuvant radiation therapy rates. The IV results contrast the negative “treatment effects” in standard risk-adjustment models, suggesting that providers recommended adjuvant treatments to patients with higher unmeasured severity levels.