Acellular pertussis vaccines are being evaluated in multiple clinical studies, and human immunogenicity data will likely be pivotal in the appraisal of vaccine responses between populations and the responses to different vaccine combinations. Antibody response to pertussis antigens is also used in the diagnosis of pertussis. An international study was designed to assess the comparability of data generated in different laboratories by enzyme-linked immunosorbent assays (ELISAs). Thirty-three participating laboratories were asked to quantitate specific antibody to pertussis toxin (PT), filamentous hemagglutinin (FHA), pertactin (PRN), or fimbrial proteins (FIM) in 21 samples. Samples were to be assayed in triplicate in five independent assays by each ELISA routinely performed in the laboratory to assess intra-assay, interassay, and population variability. The mean sample values were used to compare quantitative results among the laboratories. Thirteen of the 32 laboratories which submitted evaluable data for an assay to measure antibodies to PT, 12 of 30 laboratories with assays for FHA, 10 of 17 laboratories with assays for PRN, and 6 of 13 laboratories with assays for FIM maintained a coefficient of variation below 20% for 75% of the samples tested. Assays that measure antibodies to FIM appear to be less precise than the other assays. Precision varied among laboratories that used similar methods. The relative values of intra- and interassay variabilities were not consistent for a given assay within a laboratory, indicating that the sources of these variability components may be unrelated. Precision and agreement appeared less reliable for samples with low antibody levels. Ranking and regression analyses suggest that some laboratories generated comparable quantitative results, although direct comparison or combination of results from different laboratories remains difficult to support. Calibration to the U.S. Reference Pertussis Antisera appears to have been successful at standardizing the results in some laboratories. Statistical analyses are affected by assay precision and are not necessarily reliable sole predictors of biologically relevant differences in quantitative results. If results from different laboratories must be compared, appropriate studies of precision and quantitative agreement should be conducted to support the specific comparisons.