Advanced Statistical Analyses to Reduce Inconsistency of Bond Strength Data

Abstract

This study was designed to clarify the interrelationship of factors that affect the value of microtensile bond strength (µTBS), focusing on nondestructive testing by which information of the specimens can be stored and quantified. µTBS test specimens were prepared from 10 noncarious human molars. Six factors of µTBS test specimens were evaluated: presence of voids at the interface, X-ray absorption coefficient of resin, X-ray absorption coefficient of dentin, length of dentin part, size of adhesion area, and individual differences of teeth. All specimens were observed nondestructively by optical coherence tomography and micro-computed tomography before µTBS testing. After µTBS testing, the effect of these factors on µTBS data was analyzed by the general linear model, linear mixed effects regression model, and nonlinear regression model with 95% confidence intervals. By the general linear model, a significant difference in individual differences of teeth was observed ( P < 0.001). A significantly positive correlation was shown between µTBS and length of dentin part ( P < 0.001); however, there was no significant nonlinearity ( P = 0.157). Moreover, a significantly negative correlation was observed between µTBS and size of adhesion area ( P = 0.001), with significant nonlinearity ( P = 0.014). No correlation was observed between µTBS and X-ray absorption coefficient of resin ( P = 0.147), and there was no significant nonlinearity ( P = 0.089). Additionally, a significantly positive correlation was observed between µTBS and X-ray absorption coefficient of dentin ( P = 0.022), with significant nonlinearity ( P = 0.036). A significant difference was also observed between the presence and absence of voids by linear mixed effects regression analysis. Our results showed correlations between various parameters of tooth specimens and µTBS data. To evaluate the performance of the adhesive more precisely, the effect of tooth variability and a method to reduce variation in bond strength values should also be considered.

Multiparameter evaluation of acrylamide HEMA alternative monomers in 2-step adhesives

“…The µTBS study design, involving the three variables 'Adhesive' (with the four levels DEAA, HEAA, HEMA-and HEMA+), 'Bonding mode' (with the two levels E&R and SE), and 'Aging' (with the two levels 1w and 6m water storage) required an LME statistical model, taking into account the split-tooth design (E&R versus SE) and enabling comparison of inter-tooth and intra-teeth differences, thus to predict proper fitted LME means [40]. Other statistical analytic methods (e.g.…”

Long‐term microtensile bond strength of self‐etch adhesives and influence of 7‐s phosphoric acid etching on adhesion of a 3‐step etch‐and‐rinse adhesive to the dentine of primary teeth

“…Additionally, the thickness of the diamond saw blade used for sectioning had an impact on the beam size. It has to be taken into consideration that there is an inverse relationship between the μTBS and the bonded surface area 18,26‐30 . In general, smaller adhesion areas have a lower probability of having flaws causing failure during μTBS testing compared with beams with larger cross‐sectional areas 30 .…”

“…It has to be taken into consideration that there is an inverse relationship between the μTBS and the bonded surface area. 18,[26][27][28][29][30] In general, smaller adhesion areas have a lower probability of having flaws causing failure during μTBS testing compared with beams with larger cross-sectional areas. 30 Accompanying therewith, the failure mode of small crosssectional areas is mainly adhesive, 18 which was also favorable for the evaluation.…”

Quick bonding using a universal adhesive

“…For statistical analysis, this extensive study set-up required an LME model that also considered the split-tooth design with the UAs applied on each tooth in both E&R and SE bonding mode, hereby taking also inter-tooth differences into account. [33] More conventional statistical analysis methods (e.g. ANOVA, general linear model) would have decreased the statistical power, since the two adhesive modes then could not have statistically been compared on the tooth level.…”