Methods. The authors used three composites to fabricate cylinders as repair substrates. The etched-only group was ground, etched, dried and built up with a flowable composite. For the undercut group, the authors introduced arrays of fissures on the surfaces before preparing the specimens for subsequent buildup in the manner described for the etched-only group. They made nonrepaired cylinders for baseline measures of strength. They sliced all finished cylinders into slender bars with a diamond saw. Flexure strength values were determined by a three-point-bending test.

Results. Nonrepaired bars exhibited statistically significantly higher flexure strength values than did repaired bars, as determined by Wilcoxon rank sum test. Two-way general linear model showed that both material (P < .0001) and undercut (P = .0207) exhibited a statistically significant influence on the repaired flexure strength. Repair substrate with elastic modulus close to that of repair material exhibited a greater percentage of recovery of the respective cohesive strength. Compared with the etched-only group, the undercut group yielded a higher mean flexure strength with one composite but a lower mean flexure strength with the other two. Examination of the fractured surfaces showed that a significant number of undercuts were filled only partially.

Conclusion. Flexure strengths of repaired specimens always were lower than the cohesive flexure strengths of the materials being repaired. Undercuts did not generally improve repair strength.

Clinical Implications. Small undercuts on the surface of composite often are difficult to fill completely, resulting in areas of stress concentration that result in no improvement in the repair strength.