OBJECTIVES: To assess the performance of routine syphilis screening during 5 year follow up of Ethiopian factory workers, participating in a cohort study on HIV/AIDS. METHODS: Syphilis serology test results of factory workers, who each donated at least six blood samples were evaluated. Screening in 1997-8 had been performed by the Treponema pallidum particle agglutination (TPPA) assay and in 1999-2001 by the rapid plasma reagin (RPR) test. TPPA had been followed by RPR or RPR by TPPA, in case of a positive screening result. Samples of study subjects showing inconsistent sequential TPPA and/or RPR results were retested independently by three laboratory technicians. RESULTS: A total of 540 cohort participants (8.3% HIV positive at enrollment) donated 4,376 blood samples (mean 8.3 per subject). From 93 of the 176 participants with at least one positive TPPA result during follow up, 152 samples were retested by RPR and/or TPPA. Based on the revised syphilis test results, the 540 cohort participants were classified as having no (70.5%), past (20.6%), prevalent (6.9%), or incident (2.0%) syphilis. The RPR screening test was difficult to interpret and yielded 8.2% biological false positive (BFP) RPR results, or 3.2% if weak positive results were excluded. There was no correlation between HIV infection and BFP RPR reactions. Sample mix-ups were detected in 1.2%. CONCLUSION: Evaluation of routine syphilis screening as performed in a long term cohort study on HIV/AIDS in Ethiopia showed difficulties encountered in syphilis screening programmes such as a high percentage of BFP RPR, inconsistencies in interpretation of the RPR test, and sample mix ups. The findings stress the need to develop a syphilis screening assay that is easy to perform and interpret and to implement quality assurance programmes.