BACKGROUND: The authors compared the clinical performance of complex amalgam restorations, replacing at least one cusp--retained either mechanically with self-threading pins or bonded--with a filled, 4-methacryloxyethyl trimellitate anhydride, or 4-META-based resin designed for amalgam bonding. METHODS: The authors placed 60 amalgam restorations (28 pin-retained and 32 bonded), each restoration replacing at least one cusp. They used self-threading stainless steel pins in the pin-retained group and a filled, 4-META-based bonding resin in the bonded group. For both groups, the authors left in place any retention form remaining after removal of an old restoration but did not enhance it. RESULTS: At four years, six restorations had failed. At five years, of the 40 restorations available for evaluation, three had failed, for a total of nine failed restorations; seven of those were pin-retained and two were bonded. Using the Fisher exact test to compare the groups at five years, the authors found no significant difference in failure rate, marginal adaptation, marginal discoloration, secondary caries, tooth sensitivity or tooth vitality. CONCLUSIONS: At five years, there was no difference in the performance of pin-retained amalgam restorations and bonded amalgam restorations. This study will be continued for at least a sixth year. CLINICAL IMPLICATIONS: Bonding with a filled, 4-META-based bonding resin appears to be a satisfactory method of retaining large amalgam restorations replacing cusps.