OBJECTIVE The objective of this randomized controlled trial was to evaluate the clinical performance of a 2-hydroxyethyl methacrylate (HEMA)-free one-step adhesive. METHODS Two hundred sixty-seven cervical lesions in 52 patients were restored with the composite Gradia Direct (GC), bonded with either the one-step self-etch adhesive G-Bond (GC) or the three-step etch-and-rinse adhesive Optibond FL (Kerr) in a random order. The restorations were evaluated for retention, marginal integrity, marginal discoloration, and caries occurrence after 5 years. Specific statistics were used to account for the clustered data (multiple restorations per patient). RESULTS The clinical success rate for G-Bond (87.4 %) was not significantly different from that of Optibond FL (90.9 %). Both adhesives showed progressive marginal deterioration, but G-Bond exhibited more small enamel defects and marginal discoloration, and unlike previous recalls, several restorations failed because of deep microleakage. Large sclerotic lesions were a significant risk factor for retention loss with G-Bond. Irrespective of the adhesive, almost all restorations with retention loss were located in the lower jaw. CONCLUSIONS After 5 years of clinical service, restorations bonded with the HEMA-free one-step adhesive did not need repair or replacement more often than those with the three-step etch-and-rinse adhesive, and both adhesives had a high retention rate (>90 %). There were indications that G-Bond did not (self-)etch enough in some clinical situations, as G-Bond exhibited more incisal defects and marginal discolorations, and sclerotic lesions were at higher risk of retention loss. CONCLUSIONS The clinical performance of the HEMA-free one-step adhesive was clinically acceptable after 5 years.