Serum samples from 85 patients with proven typhoid fever, 11 patients with p-typhoidal fever, 101 patients with febrile non-typhoidal, and 130 healthy subjects were tested for immunoglobulin G (IgG), IgA, and IgM antilipopolysaccharide (LPS) of Salmonella typhi antibodies by enzyme-linked immunosorbent assay (ELISA) and Widal test. The levels of all three classes of immunoglobulin anti-LPS of S. typhi were higher in typhoid patients than in healthy or febrile nontyphoidal groups; we selected various combinations between the three classes of immunoglobulin to obtain the best combination of sensitivity and specificity. The sum of the absorbance values obtained from the ELISA assay for IgG+IgA+IgM (sigma lgs) was the best choice for diagnostic utility for typhoid fever. We selected a positive test at a decision level of sigma lgs > or = 1.2 with a sensitivity of 94% and a specificity of 92% with a frequency of false negative of 5.9%. The frequency of false positives for healthy controls was 7.7% and, for the febrile nontyphoidal group, it was 7.9%. We also compared receiver (or relative) operating characteristic (ROC) curves for the diagnostic usefulness of the ELISA with that of the Widal test, whose merits and limitations, especially in endemic regions, are discussed. The ELISA assay was much more sensitive and specific than any combination of the Widal test, and hence it could be a useful tool for the serologic diagnosis of typhoidal fever with a single blood sample.