Prognostic features of serum galactomannan (GM) remain poorly defined in patients with GM-positive invasive aspergillosis (GPA). We identified 93 patients with proven or probable invasive aspergillosis (IA) and GM values of >or=0.50 from January 2005 to March 2009. We used Cox modeling of time to 6- and 12-week mortality for the GM level at the time of diagnosis (GM(0)), GM decay in the week following diagnosis in 72 patients with >or=2 GM values, other predictors of mortality, and antifungal use during the week following diagnosis. Six-week mortality was 55% in the whole cohort and 43% in patients with >or=2 GM determinations. The hazard ratio (HR) of GM(0) per unit increase and 1-week GM decay per unit decline per week were 1.25 (95% confidence interval [CI], 1.01 to 1.54; P = 0.04) and 0.78 (95% CI, 0.63 to 0.96; P = 0.02), respectively, adjusting for other predictors of IA mortality; these values remained stable after adjusting for antifungal use and were predictive of all-cause mortality at 12 weeks with similar adjusted HR values. We conclude that the combination of GM(0) and 1-week GM decay is predictive of all-cause mortality in patients with GPA, independent of other traditional risk factors for mortality and antifungal exposure, supporting GM decay as a potential surrogate endpoint for future antifungal therapeutic trials.