Introduction: The influence of non-obstructive coronary artery disease (CAD) on long-term mortality, and therefore optimal intensity of treatment, has not been well defined. Hypothesis: Increasing grades of coronary stenosis at invasive angiography will predict long-term all-cause mortality, even among patients with non-obstructive disease. Methods: We identified 31,121 consecutive adult patients without acute coronary syndrome or prior revascularization undergoing first invasive coronary angiography at Duke Hospital to evaluate CAD between 1986 and 2007. We evaluated all-cause death using Cox proportional hazards models including CAD severity (as defined by a Modified Duke CAD Index - see figure legend), baseline characteristics, and risk factors. Complete follow-up was obtained in 99% of patients by survey, Social Security Master Death File search and National Death Index search. Results: Median follow-up time was 9.3 years (IQR 5.0-15.5 years) with 14,588 deaths. Unadjusted hazard ratio of death increased with CAD grade, even with non-obstructive CAD (Figure). In multivariable analysis, Modified Duke CAD Index remained the strongest predictor of mortality after age, CHF class, and ejection fraction (Wald X 2 = 411.0, HR 1.10 per grade, p<0.0001). Adjusted mortality increased with increasing CAD grade at 5 years (10%, 15%, 18%, 19%, 22%, 27%, 31%) and 10 years (19%, 29%, 34%, 36%, 41%, 48%, 53%). Conclusions: The relationship between CAD severity and mortality is continuous, including non-obstructive CAD. Studies to determine optimal treatment strategies for patients with non-obstructive CAD may be indicated.