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.

A question frequently asked about repair of restorations is whether the repair material bonds adequately to the existing restorations. When the clinician places composite restorations in increments, he or she relies on the oxygen-inhibited layer to make the bonding of subsequent increments possible.^7,8 Bonding new composite material to an existing restoration presents a different challenge; while there is no oxygen-inhibited layer, some unreacted double bonds remain. To ensure bonding between the repair composite and the existing composite restoration, the clinician may consider some sort of mechanical roughening of the prepared restoration surface. The use of adhesion promoters, such as enamel/dentin bonding agents and silane coupling agents, may improve the bonding between the two composites.^7–18 For example, in one study, air abrasion of the surface often yielded significantly higher bond strength than did other roughening procedures; however, the effect of surface roughness in the range of 0.1 to 0.5 micrometer was found insignificant.¹² The use of bonding agents often improves the repair strength values.^7,13 Silane coupling agent has shown the greatest magnitude of shear bond strength regardless of material variation,¹⁴ even though in another study, minimal or no improvement of the transverse strength was found with abraded surface or surface etched with 9.6 percent hydrofluoric acid at the interface.¹⁵

When the clinician decides to repair a composite restoration, he or she may not have complete information about the nature of the restorations to be repaired. Despite the fact that in vitro studies consistently have concluded that the strength values of repaired composites are adequate, this lack of information may compromise the bonding between the repair composite and the existing restoration. The clinician may use additional mechanical features, such as undercuts, to enhance bonding. An undercut preparation, therefore, may be included as part of the procedure during the exploratory cavity preparation into an old restoration.

We undertook an in vitro study to test the hypothesis that the presence of undercuts on the proposed surface of the existing restorations could improve the strength of the bond between the old and the new composite materials using a three-point bending test.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

 
We prepared three 6-millimeter cylinders (each 16 mm in diameter) by increment from each of the three repair substrate materials tested (Table 1) in a mold with a 16-mm inner diameter. Each increment, about 1.5 mm in thickness, was light cured for 60 seconds. We ground flat one end of each cylinder with 320-grit silicon carbide abrasive paper, taking care to make the flattened surface perpendicular to the axis of the cylinder. To prepare specimens with undercuts, we used two cylinders from each material group. We made a grid of 2.25 x 2.25–mm squares (Figure) on the nonflattened end of the specimen with a diamond wheel saw (Model 650, South Bay Technology, San Clemente, Calif.). The grid aided in indexing the location of undercuts on the flattened surface.

View larger version (22K): [in this window] [in a new window]   Figure. Preparation of specimens with mechanical undercut.

We prepared the remaining cylinders as described in the figure as well, except that we did not index and prepare fissurelike undercuts. These cylinders are referred to as "etched-only cylinders." We prepared one additional 12-mm cylinder from each repair substrate material shown in Table 1 for the purpose of assessing their cohesive strength values. We also embedded these cylinders in gypsum and sliced them with the diamond saw to a dimension of 1 x 1 x 12 mm. We made the specimens that were to receive undercuts with larger cross-sections so that they could accommodate the undercuts.

We stored all specimens in water for seven days before testing. Using a three-point bending test using an Instron material testing machine (Model 5500R; Instron, Canton, Mass.), we determined the load of failure with a crosshead speed of 0.5 mm/minute. Tables 2 and 3 show the number of specimens tested. We applied the load at the repair interface. We calculated the flexure strength from the following relationship:

where FS is the flexure strength, P is the load at fracture, l is the distance from the lower span support to the point of loading, w is the width of the specimen cross-section and h is the height of the cross-section. We used the Wilcoxon rank sum test; general linear model, or GLM (an analysis of variance, or ANOVA, for unbalanced data); and Tukey’s honestly significant difference, or HSD, test and to analyze the flexure strength values. We observed fracture surfaces with a x 30 magnification stereomicroscope to determine the mode of failure.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

 
Table 2 shows the cohesive flexure strength values of nonrepaired intact specimens and Tukey’s HSD grouping of the materials used in the study. Wilcoxon rank sum test showed that the flexure strength values were statistically significantly higher for nonrepaired intact bars than their respective repaired bars (P < .0001) (Tables 2 and 3), except for the Durafill (Ivoclar Vivadent, AG, Schaan, Liechtenstein) bar, in which we observed no difference (P = .7979).

Table 3 shows for the repaired specimens the mean (standard deviation, or SD) repair strength values and corresponding percentage of the cohesive flexure strength, as well as a summary of the surface analysis. Examination of the fracture surfaces showed that the mode of failure was different among the materials and between the surface preparations. For the etched-only specimens, only one specimen each of Filtek Z250 Universal Restorative (3M/ESPE) and Beautifil (Shofu, Kyoto, Japan) did not show fracture initiated at the interface. More than one-half of the Durafill specimens fractured within the Durafill but near the interface (Table 3). A Wilcoxon rank sum test showed there was no statistical difference between the mean repair strength values of the two modes of failure of Durafill (P = .0605); we used all values for further statistical analysis.

For the specimens with undercuts, we noted that a substantial number of undercuts were not filled completely (Table 3) despite the effort to fill them with the help of an explorer. The mean flexure strength values of filled specimens were slightly higher than those of partially filled specimens; however, Wilcoxon rank sum tests showed no statistical differences between the mean strength values of filled and partially filled specimens (P = .6496 for Beautifil, P = .0539 for Durafill and P = .0605 for Filtek Z250). In addition, about one-half of the repaired Filtek Z250 specimens fractured at the interface of increments of Filtek Z250. The mean (SD) flexure strength of all the specimens fractured at the increment interface was 45.0 (9.2) megapascals (n = 23), and the mean flexure strength of the specimens fractured at the repair interface was 56.9 (14.3) MPa (n = 21). A Wilcoxon rank sum test showed a statistical difference between the two groups (P = .0022). We excluded the values of the specimens fractured at the increment interface from further statistical analysis.

Two-way GLM showed that material (P < .0001) and surface treatment (P = .0309) significantly influenced the flexure strength values of the repaired specimens and that there was interaction between the two variables (P < .0001). Further analysis on the effect of surface preparation on the repair strength values for each material revealed that compared with etched-only specimens, undercuts yielded a higher mean flexure strength value with Beautifil (P = .0004), a lower value for Filtek Z250 (P = .0122) and no effect with Durafill (P = .1745).

Strength values of repaired restorations are lower than those of the cohesive flexure strength of the repair substrates.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

 
Adhesive bond strength values between two materials reported in dental literature often are given in terms of shear bond strength^12,14,19; however, flexure strength,¹⁷ tensile bond strength²⁰ and diametral tensile strength²¹ values also are reported. The shear test is used more often than any other test method, judging from the literature, but it also has been criticized for the test arrangement that produces high stress concentration at the point of contact.^22,23 The arrangement of the tensile strength test specimen, on the other hand, allows a uniform stress distribution within the interface and is considered a better approach to evaluating interfacial strength.²⁴ Recently, tensile tests using specimens with cross-sections of 1 x 1 mm² or smaller, also known as "microtensile tests," have been used widely in investigating bonding between dentin and various composite materials²⁵ and repair of porcelain.²⁶ The method should have been applicable for our study. However, a preliminary study showed that the cohesive strength values of the composites selected for this study were so high that specimens kept separating from the testing device without fracture. Therefore, we obtained the flexure strength values using a three-point bending test.

The strengths of this study of the repaired specimens in this study were in the range of 31 to 98 percent in comparison with the nonrepaired specimens. The range was consistent with findings in the literature.^7,10,17 The results confirmed that strength values of repaired restorations are lower than those of the cohesive flexure strength of the repair substrates. For Filtek Z250 and Beautifil specimens without undercuts, their level of strength in repairing is 37 to 40 percent of their respective cohesive flexure strengths. This range is about the same as the value of 40 percent reported in a study in which a bonding agent was used before the application of a new composite material.¹⁷

On the basis of the shear bond strength values obtained from materials of various elastic module, it has been suggested that the shear bond strength value should closely follow the elastic modulus of the material because of favorable stress distribution associated with high elastic modulus.²⁷ That conclusion seems contradictory to our results, which showed that Beautifil with the highest elastic modulus (14.0 gigapascals) exhibited the lowest strength value. Since the stress distribution within a bent specimen in a flexure test is different from that in shear test, one can expect that different results may arise. In a specimen under bending stress, the portion with higher elastic modulus will bear more stress than the portion with a lower elastic modulus as the specimen is being flexed. It means that the bending moment induced at the interface by the high elastic modulus portion of the specimen is likely to be higher than the failure load indicates.

With repair substrate of lower elastic modulus, such as Filtek Z250 (11.3 GPa), a higher portion of load is taken up by the flowable portion of the specimen. A higher load will be needed to attain the same level of bending moment at the interface induced by the Filtek Z250 portion of the specimen. Since Durafill has an elastic modulus (4.4 GPa) close to that of the flowable repair material (4.0 GPa), the materials are likely to bear the load equally. A much higher load to fracture should be expected, but should not exceed the cohesive strength values of Durafill. Fracture surface analysis confirmed that majority (58 percent) of failure was within the Durafill restorations. It also is possible that Durafill, being a microfilled composite material, has more resin on the repair surface that facilitates more resin-resin bonding.¹²

The undercut preparations in this study did not yield the anticipated improvement of the repaired flexure strength for all three composite repair substrates. Only Beautifil exhibited significant improvement in the repair strength (approximately 20 percent). The high elastic modulus of Beautifil probably made crack propagation more difficult than it was with Durafill and Filtek Z250. Examination under x 30 magnification stereomicroscopy showed that a high percentage of the undercuts were only partially filled. One may conclude that the clinician must make a special effort in placing the material or the design of undercut to ensure complete filling. However, the mean strength values appear to indicate that there is no need to fill the undercut completely, since there was no statistically significant difference between the completely and partially filled specimen groups. On the contrary, it is an indication that a shallower preparation (much less than 1 mm depth), which is more likely to be filled completely, may be sufficient to realize the improvement that may exist.

Stereomicroscopic observation also indicated that all fractures were initiated at the repaired interface and propagated into the repair substrates. This mode of fracture implies that the presence of the repair material filled in the undercut may have retarded the crack propagation as intended, but the undercut preparation itself may have acted as a flaw facilitating the crack propagation. Drilling with a frac14; round bur on repair substrate results in a cavity with a near 90-degree shoulder and a rounded bottom. The 90-degree shoulder presents a site of stress concentration and should be eliminated. Rounding the shoulder or widening the opening of the undercut should suffice. The preceding discussion indicates that the undercut should be deep enough only to facilitate adequate filling of the repair material. Surface modification by preparing a shallow and wide open undercut essentially results in an increased area for bonding. This increased surface area should be more effective in increasing overall bonding of the repair material to the repair substrate than are the mechanical undercuts that often are used to retain restorations.

The role of an oxygen-inhibited layer in providing bonding between increments is well-established.^7,8 To build a 16-mm diameter cylinder with a 1.5-mm increment requires a great deal of manipulation of the composite material and time to spread it to a relatively flat surface before curing. The observation of adhesive failure between increments is likely due to the contamination of the oxygen-inhibited layer during the manipulation that reduced the bonding between the increments.

	CONCLUSIONS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

 
The mean repaired flexure strength values were lower than the cohesive flexure strength values of the materials being repaired. The level of strength of the repaired specimens with respect to the cohesive flexure strength was consistent with published data.

Preparing undercuts on the composite surface to improve repair strength between a flowable repair material and an unknown repair substrate is a novel idea. However, this in vitro study showed that undercuts of 1 mm in depth resulted in improvement of one repair substrate and reduction of the two other substrates. The dimension of undercuts used in this study appeared to be too deep to allow complete filling of the undercut. In addition, the sharp shoulders might have created an unnecessary area of stress concentration. If an undercut is regarded as necessary, it should be a shallow one without a sharp shoulder, aiming at increasing the area for bonding without undue stress concentration.

View larger version (122K): [in this window] [in a new window]   Dr. Shen is an associate professor, Department of Dental Biomaterials, College of Dentistry, University of Florida, P.O. Box 100446, Gainesville, Fla. 32610-0446, e-mail "cshen{at}dental.ufl.edu". Address reprint requests to Dr. Shen.

View larger version (105K): [in this window] [in a new window]   Mr. Mondragon is a laboratory technician, Department of Operative Dentistry, College of Dentistry, University of Florida, Gainesville.

View larger version (130K): [in this window] [in a new window]   Dr. Gordan is an associate professor, Department of Operative Dentistry, College of Dentistry, University of Florida, Gainesville.

View larger version (119K): [in this window] [in a new window]   Dr. Mjör is a professor, Department of Operative Dentistry, College of Dentistry, University of Florida, Gainesville.