Background: Quantitative cytomegalovirus (CMV) DNA-specific polymerase chain reaction (PCR) analysis is widely used as a surveillance method for hematopoietic stem cell transplant (HCT) recipients. However, no CMV DNA threshold exists in bronchoalveolar lavage (BAL) to differentiate pneumonia from pulmonary shedding. Methods: We tested archived BAL fluid samples from 132 HCT recipients with CMV pneumonia and 139 controls (100 patients with non-CMV pneumonia, 18 with idiopathic pneumonia syndrome [IPS], and 21 who were asymptomatic) by quantitative CMV and beta-globin DNA-specific PCR. Results: Patients with CMV pneumonia had higher median viral loads (3.9 log10 IU/mL; interquartile range [IQR], 2.6-6.0 log10 IU/mL) than controls (0 log10 IU/mL [IQR, 0-1.6 log10 IU/mL] for patients with non-CMV pneumonia, 0 log10 IU/mL [IQR, 0-1.6 log10 IU/mL] for patients with IPS, and 1.63 log10 IU/mL [IQR, 0-2.5 log10 IU/mL] for patients who were asymptomatic; P < .001 for all comparisons to patients with CMV pneumonia). Receiver operating characteristic curve analyses and predictive models identified a cutoff CMV DNA level of 500 IU/mL to differentiate between CMV pneumonia and pulmonary shedding, using current CMV pneumonia prevalence figures. However, different levels may be appropriate in settings of very high or low CMV pneumonia prevalence. The presence of pulmonary copathogens, radiographic presentation, or pulmonary hemorrhage did not alter predictive values. Conclusion: CMV DNA load in BAL can be used to differentiate CMV pneumonia from pulmonary shedding.