Purpose: Cyclooxygenase-2 (COX-2)- specific inhibition suppresses carcinogenesis in preclinical models and is a promising strategy for preventing oral cancer. In this pilot randomized phase II study, we evaluated the efficacy and safety of the COX-2 inhibitor celecoxib in patients with oral premalignant lesions (OPL). Experimental Design: Patients were randomly assigned to placebo (n = 18), celecoxib 100 mg twice daily (n = 17), or celecoxib 200 mg twice daily (n = 15) for 12 weeks. Six additional patients received celecoxib (400 mg twice daily) in an unblinded extension of the study. Biopsies were obtained at baseline and week 12. All patients entering the study were required to have at least one histologically confirmed early (atypical hyperplasia, atypical hyperkeratosis, or mild dysplasia) or advanced (moderate to severe dysplasia) OPL. Results: Forty-nine patients (46 of 50 randomized and 3 of 6 open label) were evaluable for efficacy analyses. There were no statistically significant differences between the response rates of the randomly assigned arms: placebo, 33.3% (6 of 18); celecoxib 100 mg twice daily, 41.2% (7 of 17); and celecoxib 200 mg twice daily, 20.0% (3 of 15). Two patients responded on celecoxib 400 mg twice daily. Celecoxib was generally well tolerated. Patients with higher baseline COX-2 mRNA levels had an increased risk of disease progression within 3 months. Conclusions: Celecoxib at 100 or 200 mg twice daily was ineffective in controlling OPLs in this randomized controlled trial. This result and cardiovascular toxicity results of other (large scale) randomized controlled trials of selective COX-2 inhibitors have discouraged the continued investigation of these agents in oral cancer chemoprevention. Better methods for identifying high-risk patients and more active interventions are needed for future oral cancer chemoprevention trials.