The aim of the study was to determine the characteristics of primary and late latent syphilis cases that were non-reactive on initial screening by rapid plasma reagin (RPR) but reactive by treponemal tests. RPR test results of all primary and late latent syphilis cases in the province of Alberta, Canada during 1980-2001 were compiled. The proportion of syphilis cases that were non-reactive by RPR testing was compared with reactive cases. Multiple logistic regression was used to identify factors associated with non-reactive RPR results. Among primary syphilis cases, 224 (26%) were non-reactive on initial RPR and were compared with 639 reactive primary cases. Independent factors associated with a non-reactive result were male gender (odds ratio, [OR] = 1.99 [1.22-3.26]), Caucasian ethnicity (OR = 1.77 [1.20-2.61]) and diagnosis during the 1980-1986 syphilis outbreak period in Alberta (OR = 3.13 [1.45-6.74]). Of the late latent cases, 512 (39%) were non-reactive by RPR and 791 were reactive. A non-reactive test for late latent cases was associated with male gender (OR = 2.91 [2.15-3.94]), Canadian Aboriginal (OR = 2.91 [1.59-5.32]), Caucasian (OR = 1.87 [1.35-2.58]), black ethnicity (OR = 1.76 [1.28-2.42]) and diagnosis during the 1980-1986 Alberta syphilis outbreak (OR = 1.67 [1.29-2.16]). RPR testing alone is insufficient to diagnose primary and late latent infections under certain conditions and patient characteristics. Case detection was lower for late latent syphilis when compared with primary syphilis based on the properties of the RPR. Further studies are needed to inform the optimal use of the RPR test in laboratory testing algorithms for the diagnosis of syphilis and to determine the cost effectiveness of better screening tests for syphilis.