Few data exist regarding cement failure load and marginal leakage of castings cemented to implant-supported abutments subjected to load and thermal cycling, especially with newer cements. This study evaluated the cement failure load and marginal leakage of castings cemented to Steri-Oss abutment/analog assemblies with one of seven luting protocols (n = 5). Protocols consisted of a zinc phosphate control, zinc phosphate with petroleum jelly, TempBond, TempBond NE, ImProv, ImProv with petroleum jelly and Provilink. Cemented castings were subjected to cyclic axial compression of 75 N for 100 000 cycles, 1000 thermal cycles (5-55°C), and immersed 24 h in 0.5% basic fuchsin. The castings were then loaded in tension until cement failure. Cement failure load was analysed using anova and Fischer's PLSD (α = 0.05). Marginal leakage on the intaglio of the casting was scored with the unaided eye on a scale of 0-2, and analysed with Duncan's multiple range (α = 0.05). Similar groups are noted by superscript. Cement failure load (from lowest to highest): ImProv with petroleum jellya, TempBonda, ImProva, Tempbond NEa, Provilinka, zinc phosphate with petroleum jelly b, zinc phosphatec (P < 0.0001). Marginal leakage scores (from lowest to highest): ImProva, Provilinka, ImProv with petroleum jellya, zinc phosphateab, zinc phosphate with petroleum jellyab, TempBondbc, and TempBond NEc (P = 0.01). Addition of petroleum jelly to zinc phosphate lowered cement failure load without affecting marginal leakage. ImProv and Provilink demonstrated relatively low cement failure load, as well as lowest marginal leakage. © 2007 Blackwell Publishing Ltd.