Should statistical analysis of bond-strength data include or exclude cohesive failures?

Abstract

The aim was to investigate shear bond strengths and failure modes of four self-etch bonding agents to bovine dentin and enamel and to compare evaluation of data sets with or without exclusion of cohesive failure specimens. Composite-cylinders were bonded perpendicularly to bovine dentin and enamel surfaces. Shear-strengths were measured 24h post-bonding of: Scotchbond Universal® (SBU, 3M), OptiBond™ XTR (OBXTR, Kerr), OptiBond™ universal (OBU, KaVo-Kerr) and Prime & Bond active® (PBA, Dentsply-Sirona). Analysis of overall data was made via a linear mixed-model. This was repeated after exclusion of specimens associated with cohesive failures. When both adhesive and cohesive failures were considered, OBU and OBXTR showed comparable dentin and enamel bond strengths, whereas lower strengths were found on enamel for SBU (p<0.001) and PBA (p=0.015). For OBXTR higher shear strengths were measured for specimens associated with cohesive failures. When cohesive failures were excluded, the majority of shear bond strengths of adhesive failure specimens were only slightly different from overall results. However, uniquely with OBXTR dramatically lower shear bond strengths were found for dentin substrate. After exclusion of cases with cohesive failures OBXTR adhesive fell behind other materials in the sequence of average shear strengths. This did not reflect the actual performance of the material. Therefore, in statistical analysis we do not recommend exclusion of data based on a specific fracture mode.