Background Recent recommendations propose that samples dually reactive by a syphilis enzyme immunoassay (sEIA) and RPR be reported as positive for syphilis without confirmatory testing.
Methods Samples from 1 September 2007 to 19 March 2011 testing reactive by sEIA and tested by INNO-LIA (IL) were extracted from the Alberta Provincial Laboratory for Public Health's DIAL (Data Integration for Alberta Laboratories), a web based application. Syphilis testing history was reviewed for all patients with a reactive sEIA and reactive RPR and negative (NEG) or indeterminate (IND) result by IL. Syphilis infection was defined by a positive confirmatory test (majority IL; a few TPPA and FTA-ABS). The significance of RPR titres in patients NEG/IND by IL with or without evidence of syphilis infection was analysed using χ2 test. Median standard cut-offs (s/co) for the sEIA were compared using the Mann–Whitney U test.
Results 6195 samples from 4695 patients with reactive sEIA were also tested by IL: 15 samples (0.2%) had no reported RPR result, 4232 (48.3%) were non-reactive by RPR, and 1948 (31.4%) samples from 1753 patients were reactive by RPR. 72 (4.1%) of the 1753 patients with reactive RPR had at least one specimen tested NEG/IND by IL (Abstract LBP-1.10 figure 1). 3 of the 72 patients (4.2%) had a serological history of syphilis infection not recognised initially, 15 (20.8%) had no follow-up testing, 23 (31.9%) had a subsequent positive IL, and 31 (43.1%) did not demonstrate syphilis infection on follow-up testing. For the 31 patients with no serological evidence of syphilis infection, 24 remained NEG/IND by IL and seven tested negative by sEIA on follow-up. Overall, 31 patients (1.7%) would have been misclassified as infected based upon an algorithm of dually reactive sEIA and RPR without confirmatory testing. 14.7% of samples that tested negative or IND for IL in confirmed syphilis had RPR titres =1:8 as compared to 20.0% of samples from patients without syphilis on follow-up serology (p=0.5). The median s/co of the screening sEIA for samples from the patients who were infected (4.8, range: 1.0–24.20) differed from the s/co (2.6, range: 1.0–12.4) for those uninfected (p<0.05).