Although kits to detect cryptococcal antigen are used widely to diagnose cryptococcal infection, the comparative performance of commercially available assays has not been evaluated in the past decade. Therefore, we compared the sensitives and specificities of five commercially available kits for detecting cryptococcal antigen (four latex agglutination test kits--Calas [Meridian Diagnostics])--Crypto-LA [International Biological Labs], Myco-Immune [MicroScan], and Immy [Immunomycologics]--and an enzyme immunoassay kit, Premier [Meridian Diagnostics]) with culture for the diagnosis of cryptococcal meningitis and fungemia. Of 182 cerebrospinal fluid (CSF) and 90 serum samples submitted for cryptococcal antigen and fungal culture, 49 (19 and 30 samples, respectively) from 20 patients had a culture positive for Cryptococcus neoformans. For CSF specimens, the sensitivities and specificities of all kits were comparable (sensitivity, 93 to 100%; specificity, 93 to 98%). There was a significant difference in sensitivities of the kits when serum samples were tested with the International Biological Labs and MicroScan kits, which do not pretreat serum with pronase. These kits were less sensitive (sensitivity, 83%) than the Immy and Meridian latex kits (sensitivity, 97%), which do pretreat with pronase. The sensitivity of the Meridian enzyme immunoassay kit was comparable to that of the pronase-containing latex kits. These kits were of equivalent specificities (93 to 100%) when testing serum. Some of the currently available kits have limitations that need to be recognized for proper interpretation of results. Specifically, the use of pronase on serum samples reduces the number of false-positive results, and a titer of < or = 1:4 can be a false-positive result when CSF samples are being tested.