A randomized, prospective study of 200 consecutive established hand infections was designed to compare the efficacy of two antibiotics, cefamandole and nafcillin. Bacteriologic data revealed 63.5% of the patients grew multiple organisms (2.3 organisms per culture) and 26% of the patients had anaerobic infections. Complications were noted in 13% of all patients--26% in patients who grew aerobes and anaerobes and 9.8% in patients who grew aerobes alone (p less than 0.05). Despite the fact that 95% of all organisms were sensitive in vitro to cefamandole whereas only 67% of organisms were sensitive to nafcillin (p less than 0.01), complications occurred more frequently in patients treated with cefamandole. We conclude that the empirical selection of a broad-spectrum antibiotic is reasonable based on in vitro sensitivity studies; however, other factors such as treatment delay, initial extent of infection, anatomic location of infection, cause of infection, and extent of surgical debridement are important in the development of complications.