A multicenter study was conducted to assess the inter- and intralaboratory reproducibility of the Etest for susceptibility testing of the rapidly growing mycobacteria. The accuracy also was evaluated by comparing Etest results to those obtained by broth microdilution. Ten isolates (four of the Mycobacterium fortuitum group, three of Mycobacterium abscessus, and three of Mycobacterium chelonae) were tested against amikacin, cefoxitin, ciprofloxacin, clarithromycin, doxycycline, imipenem, and trimethoprim-sulfamethoxazole in each of four laboratories. At each site, isolates were tested three times on each of three separate days (nine testing events per isolate) using common lots of media and Etest strips. Interlaboratory agreement among MICs (i.e., mode +/- 1 twofold dilution) varied for the different drug-isolate combinations and overall was best for trimethoprim-sulfamethoxazole (75% for one isolate and 100% for all others), followed by doxycycline and ciprofloxacin. Interlaboratory agreement based on interpretive category also varied and overall was best for doxycycline (100% for all isolates), followed by trimethoprim-sulfamethoxazole and ciprofloxacin. Interlaboratory reproducibility among MICs was most variable for imipenem, and agreement by interpretive category was lowest for imipenem and amikacin. Modal Etest MICs agreed with those by broth microdilution only for doxycycline and the sulfonamides. For all other drugs, the modal MICs by the two methods differed by more than +/- 1 twofold dilution for one or more isolates. In all cases, the Etest MIC was higher and would have caused reports of false resistance. In summary, the Etest in this evaluation did not perform as well as broth microdilution for susceptibility testing of the rapidly growing mycobacteria. It was problematic for most species and drugs, primarily because of a trailing endpoint and/or high MICs compared to broth. Its use will necessitate further investigation, including determination of the optimal medium and incubation conditions and clarification of endpoint interpretation.