Sensitivity and specificity of rapid diagnostic tests for detection of group B streptococcal antigen in bacteremic neonates

Abstract

Latex particle agglutination (LPA) testing for antigen to group B streptococcus (GBS) has been useful in the diagnosis of GBS sepsis in newborns. However, recent reports have demonstrated that the sensitivity of LPA assays may be as low as 27 to 54%. The purposes of the present study were to directly compare the abilities of four urine antigen assays to detect GBS antigen with clinical urine samples from neonates with GBS bacteremia and to evaluate the effect of the urine concentration on the sensitivities and specificities of these assays. Urine samples were collected serially from neonates with blood cultures positive for GBS or on admission from healthy full-term infants. One milliliter of urine was removed, and the remainder was concentrated to a volume of 1 ml. Unconcentrated samples were serially diluted with normal saline and were assayed to determine the highest dilution which would produce a positive test result. The Wellcogen, Bactigen, and Directigen LPA tests and ICON immunoassay were directly compared by using concentrated and unconcentrated urine specimens and urine specimens with known titers. A total of 94 urine specimens, including 61 concentrated and 75 unconcentrated specimens, from bacteremic infants were available for sensitivity testing, and 220 urine specimens from uninfected infants were available for specificity testing. There were significant differences in sensitivity among the four assays when they were performed on concentrated urine specimens, as follows: Directigen, 98%; Bactigen, 92%; ICON, 89%; Wellcogen, 68%. When the assays were performed on unconcentrated urine specimens, the Directigen (84%) and Bactigen (76%) assays were each significantly more sensitive than the ICON (59%) or Wellcogen (43%) assay. All four assays were significantly more sensitive in detecting GBS antigen in concentrated than in unconcentrated urine. The Directigen assay detected antigen in higher dilutions (geometric mean titer, 1:5) than the ICON (1:3), Bactigen (1.2), or Wellcogen (1:1) assay. The specificity was 99.5% for all four assays when concentrated urine was used and for the Bactigen, Directigen, and ICON assays when unconcentrated urine was used; the Wellcogen assay was 100% specific when unconcentrated urine was used. We conclude that there are significant differences in sensitivity but not specificity among the commercially available assays for the detection of GBS antigenuria when concentrated and unconcentrated urine specimens are tested. These differences in sensitivity may affect the abilities of clinicians to accurately diagnose GBS sepsis before culture results are available.