Impact of Dentine Pretreatment with Matrix Metalloproteinase Inhibitors on Bond Strength of Coronal Composite Restorations: A Systematic Review and Meta-analysis of In Vitro Studies

Abstract Matrix metalloproteinase (MMP) enzymes participate in collagen matrix degradation, including in dentine, potentially compromising bond strength. Therefore, MMP inhibitors have been hypothesized to improve restoration bond strength and stability. This systematic review aimed to evaluate the influence of different MMP inhibitors applied as dentine surface pretreatments on the immediate (24 hours) and longer term (months) bond strength of direct coronal composite restorations. This systematic literature review followed the Preferred Reporting Items for Systematic Review and Meta-analyses (PRISMA) statement. A systematic literature search of three databases (Ovid MEDLINE, Ovid Embase, and Google Scholar) was conducted independently by two reviewers from inception to April 2022. An adapted quality assessment tool was independently applied by two reviewers for risk of bias assessment. RevMan v5.4 software was used for meta-analyses. A randomeffectsmodel was used to generatemean differences with 95% confidence intervals for treatment and control comparisons. The Q-test and I2-test were used to test for heterogeneity. The proportion of total variance across studies attributable to heterogeneity rather than chance was calculated. Overall effects were tested using the Z-test, while subgroup differences were tested using Chi-squared tests. Of 934 studies, 64 studies were included in the systematic review and 42 in the meta-analysis. Thirty-one MMP inhibitors were reported, three of which were included in the meta-analysis: 2% chlorhexidine (CHX), 0.3M carbodiimide (EDC), and 0.1% riboflavin (RIBO). Pretreatment with 2% CHX for 30 and 60 seconds did not significantly improve bond strength compared with controls either immediately or after long-termageing. However, pretreatment with 0.3MEDC and 0.1% RIBO (but not CHX) significantly improved bond strength compared with control groups both immediately and over time. Most studies showed a medium risk of bias. These in vitro findings pave the way for rationale clinical trialing of dentine surface pretreatment with MMP inhibitors to improve clinical outcomes.


Introduction
Since their introduction around six decades ago, restorative adhesives have undergone numerous improvements. 1,2Despite these advances, adhesive restorations often lose their bond strength over time, leading to their failure. 3,4Adhesive restorations critically rely on their bond with the tooth structures for strength, with the interface-the hybrid layer -crucial in determining the bond's longevity and stability. 5,6he collagen fibrils in dentine (mainly type 1 collagen) are key to establishing a strong bond, and their deterioration is thought to be the main reason underlying bond failure to dentine. 7ecent studies have examined the role of endogenous enzymes present within the dentine extracellular matrix (ECM) and their effect on bond stability.Among these enzymes, matrix metalloproteinases (MMPs) represent a group of calcium-and zinc-dependent host-derived enzymes. 8MMPs are divided into six subgroups: collagenases (MMP-1 and MMP-8), stromelysins (MMP-3, MMP-10, MMP-11, and MMP-20), gelatinases or type-IV collagenases (MMP-2 and MMP-9), matrilysin (MMP-7), metalloelastase (MMP-12), and membrane-type metalloproteinases (MMP-14, MMP-15, MMP-16, and MMP-17). 9Of these, four MMPs have been identified within the dentine extracellular matrix: MMPs-2, -8, -9, and -20, with MMP-2 and -9 as the most abundant. 10,11These enzymes are secreted by odontoblasts during odontogenesis and remain silenced and inactive within the dentine ECM. 12 However, these MMPs are activated either by biological acids produced by cariogenic bacteria 13 or acids introduced during acid etching. 14,15hen activated, they start to degrade the exposed collagen fibrils within the dentine. 16Therefore, inhibiting MMPs could help to preserve the hybrid layer and, therefore, bond stability.
Several types of MMP inhibitor (synthetic and natural) have been described including benzalkonium chloride, 17,18 chlorhexidine, [18][19][20][21] galardin, 22 green tea extract, 23,24 and zinc. 25MMP inhibitors can be administered either as dentine surface pretreatments or those incorporated into the adhesive.The current systematic review and meta-analysis aimed to collect and analyze the available in vitro evidence on the influence of different MMP inhibitors applied as dentine surface pretreatments on the immediate and long-term bonding strength of coronal composite restorations.The null hypothesis was that there would be no difference in bond strength after MMP inhibitor use compared with controls.

Methods Eligibility Criteria
The systematic review was developed according to the PICO scheme (►Table 1) 26 and was conducted according to the Preferred Reporting Items for Systematic Review and Metaanalyses (PRISMA) guidelines 27 : Population: all studies examining extracted human teeth, caries-free dentine, healthy dentine, sound dentine, cariousaffected dentine, or affected dentin.
Interventions: all studies examining MMP inhibitors as dentine surface pretreatments prior to direct coronal composite restoration placement.Therefore, studies that used luting cements and glass ionomer cements were excluded.
Comparator(s)/control(s): teeth without intervention (i.e., without the addition of MMP inhibitor).Studies that included no comparator were excluded.
Outcome: the main outcome was bond strength or bond stability at the microscale (by microtensile and microshear testing).Studies that tested bond strength at the macroscale were excluded.Included studies needed to have aged the samples for at least 24 hours in water or artificial saliva.Thus, studies with ageing up to 24 hours only and/or studies that used ageing solutions other than water or artificial saliva were excluded.

Types of Searched Studies
The search included published, peer-reviewed in vitro studies presenting the results (means and standard deviations [SDs]) quantitively and numerically in the English language.Thus, studies that reported the results in graphs or figures only were excluded.Non-peer reviewed studies, conference posters, letters, theses, reviews, and editorials were excluded.

Period of Reviews (Timing) and Databases
A systematic literature search was conducted in three databases: Ovid MEDLINE (1946-April 2022), Embase (1974-April 2022; see ►Table 1), and Google Scholar (up to April 2022).
With respect to the search strategy for Google Scholar, the following terms were used: "Extracted human teeth" OR "human teeth" OR "Sound dentine" OR "healthy dentine" OR "affected dentine" OR "Carious affected dentine" OR "Caries affected dentine" OR "Dentine" AND "Matrix metalloproteinase inhibitors" OR "MMP inhibitors" AND "Bond strength" OR "Bond stability."controls either immediately or after long-termageing.However, pretreatment with 0.3MEDC and 0.1% RIBO (but not CHX) significantly improved bond strength compared with control groups both immediately and over time.Most studies showed a medium risk of bias.These in vitro findings pave the way for rationale clinical trialing of dentine surface pretreatment with MMP inhibitors to improve clinical outcomes.
Impact of Dentine Pretreatment with MMP Inhibitors on Bond Strength Jamal et al. 975

Data Selection and Collection Processes
Full texts of all eligible studies were uploaded to reference management software (EndNote X9.3.1) and duplicate publications were removed automatically.Two authors (H.J. and R.Y.) screened the titles and abstracts, and the full text of studies meeting the inclusion criteria was read.Two evaluators (H.J. and R.Y.) independently screened each full-text paper based on the eligibility criteria.In case of discrepancies about study eligibility between the two reviewers, a further evaluator was involved (H.A. or P.A.).A data extraction form included the following: authors' names, year of publication, type of MMP inhibitor used, duration of MMP inhibitor used as dentine pretreatment, substrate condition, type of bonding agent, type of ageing solution, period of ageing, type of bond strength test, and bond strength means.Two reviewers (H.J. and R.Y.) were independently involved in data collection.An experienced third reviewer (P.A.) independently extracted data from 10% of studies to check process consistency.Conflicts of opinion were resolved through consensus by consulting a further reviewer (H.A. or A.Y.).

Risks of Bias and Quality Assessment
A quality assessment tool adapted from a previous study 28 was independently used by two reviewers (H.J. and R.Y.).The tool evaluated bias in terms of sample randomization, substrate condition, duration of dentine pretreatment, the use of materials according to the manufacturer's instructions, storage medium, interface surface area, restorative and bond tests performed by a single operator, sample size calculation (power analysis), and blinding of the operator during bond strength testing.Minor modifications were added to the risk of bias evaluation tool, which are "dentine pretreatment duration" and "storage medium".For each component of the tool, the letter "Y (yes)" was assigned if the author reported the item and "N (no)" if it was not reported.The grading judgement of "low," "medium," or "high" for the study was based on the total number of "Ys" as follows: one to five (high), six, or seven (medium), and eight or nine (low).

Data Synthesis
Findings were summarized narratively using text and tables.For example, findings were summarized according to type of MMP inhibitor used, duration of dentine pretreatment, substrate condition (caries-free or caries-affected), type/mode of bonding agent, type of ageing solution, period of ageing, type of bond strength test, and mean bond strength.

Meta-analysis
Review Manager (RevMan) version 5.4 software from the Cochrane Collaboration was used for meta-analyses using the following information: the average difference in outcome measures between the intervention and control groups, the number of teeth in each treatment group, and the standard deviations.These data were categorized into three time periods: 24 hours, 6 months, and 12 months, where applicable, and further divided into the type of MMP inhibitor, the adhesive application method used (self-etching or etch and rinse), and the pretreatment duration.Only MMP inhibitors applied for 30 and 60 seconds were included as they contained enough data for the meta-analysis.
The mean differences (MDs) and their 95% confidence intervals (CIs) were calculated.Findings from all comparisons were generally pooled according to the three time periods (24 hours, 6 months, and 12 months).After establishing the pooled MDs according to time, additional pooling was carried out depending on the various parameters indicated.A positive MD supports the experimental group, whereas a negative MD favors the control group.A random-effects model was used to generate MDs with 95% CIs for treatment and control comparisons.
The Q-test and I 2 -test were used to test for heterogeneity.The I 2 statistics was interpreted according to the Cochrane guidelines, with 0 to 29% as being low, 30 to 50% as moderate, and 50 to 90% as considerable heterogeneity. 29he proportion of total variance across studies attributable to heterogeneity rather than chance was calculated.Finally, the overall effects were tested using the Z-test, while subgroup differences were tested using Chi-squared tests.
The following analyses were carried out:

Study Selection
A flowchart summarizing the selection process according to the PRISMA statement is shown in ►Fig. 1. 27 During the initial search, 934 potentially eligible studies were retrieved.After removal of duplicates, 763 studies remained of which 193 remained after reviewing the titles and 163 after reviewing the abstracts.Following reading the full texts, 64 studies were included in the study and 42 were included in the meta-analysis.

Study Characteristics
The data obtained from the included publications are listed in ►Table 2. The 64 included in vitro studies were published between 2009 and 2022.Thirty-one different types of MMP inhibitors were used, 14 synthetically derived and 17 naturally derived.The microtensile bond strength test was used in all included studies except for five studies that used microshear bond strength testing.Most studies (n ¼ 53) used caries-free dentine substrate, 13 used caries-affected dentine, two studies used eroded dentine, and one study used dentine without mentioning its condition.All studies used permanent teeth except for one study that used primary teeth.
With respect to storage medium, the majority of studies used distilled water (40 studies) and 22 used artificial saliva.Two studies used both distilled and deionized water.The majority of the studies applied MMP inhibitor for 60 s (n ¼ 47), six studies applied it for 30 seconds, four for 120 seconds, three for 5 seconds, two for 15 seconds, and one each for 20 and 180 seconds.One study did not report the application duration.Only MMP inhibitors applied for 30 and 60 seconds were included in the meta-analysis, as they contained enough data.
Ageing periods ranged from 24 hours to 5 years, and various thermocycling ageing protocols were also used.The majority of studies (n ¼ 62) aged samples for 24 hours as an immediate ageing period.With respect to long-term ageing, 31 studies aged the samples for 6 months, 19 aged them for 12 months, five aged them for 3 months, three for 2 years, three for 18 months, two for 9 months, and one study each for 3 days, 1 week, 15 days, 15 months, and 5 years.Eleven studies used thermocycling for ageing: four used 1,000 cycles, two used 5,000 cycles, and one study each used 2,500, 3,000, 10,000, and 25,000 cycles.

Risk of Bias Evaluation
►Table 3 shows the evaluated risk of bias of the included studies.Overall, almost half of included studies showed a medium risk of bias (33 of 64), 17 of 64 studies showed a high risk of bias, and 14 studies were classified as a low risk of bias.

Meta-Analysis
Of the 64 studies, data from 42 studies were subjected to further evaluation in meta-analyses (►Figs.2-6).In the first analysis (2% CHX vs. control in the baseline, immediate bond strength values), 16 etch-and-rinse studies were included, representing 28 datasets considered.There was no statistically significant difference between groups (Z-test ¼ 1.26, p ¼ 0.21), and there was considerable heterogeneity (I 2 ¼ 54%).Eight self-etching studies were included, with 11 datasets considered.There was no significant difference between groups (Z-test ¼ 0.76, p ¼ 0.45), and there was moderate heterogeneity (I 2 ¼ 35%).Overall (self-etching and etchand-rinse), there was no statistically significant difference       The second analysis (2% CHX vs. control at 6 months of ageing) included 14 etch-and-rinse studies, representing 25 datasets.There was overall a higher bond strength for the experimental group compared with controls, but this was not statistically significant (Z-test 1.81, p ¼ 0.07) and heterogeneity was considerable (I 2 ¼ 88%).Six self-etching studies were included, with nine datasets considered.There was no statistically significant difference between groups (Z-test ¼ 0.86, p ¼ 0.39), and again there was considerable heterogeneity (I 2 ¼ 73%).Tests for overall effect showed significantly higher bond strength in the experimental group compared with controls (Z-test ¼ 2.33, p ¼ 0.02), with considerable heterogeneity between subgroups (I 2 ¼ 86%; ►Fig.2B).
The third analysis (2% CHX vs. control at 12 months of ageing) included five etch-and-rinse studies with seven datasets.There were overall higher bond strength values in the experimental group compared with the control group.but this was not statistically significant (Z-test ¼ 1.09, p ¼ 0.28) and heterogeneity was considerable (I 2 ¼ 91%).For self-etching, three studies were included with four datasets considered, and there was no statistically significant difference between groups (Z-test ¼ 0.18, p ¼ 0.86) but with considerable heterogeneity (I 2 ¼ 84%).Tests for overall effect favored the experimental group over the control group but without statistical significance (Z-test ¼ 1.66, p ¼ 0.10) and with considerable heterogeneity between subgroups (I 2 ¼ 90%; ►Fig.2C).
For the fourth analysis (0.3 EDC vs. control at baseline), only etch-and-rinse studies met the inclusion criteria.Three studies were included, representing six datasets.Overall, the effect was not statistically significant (Z-test ¼ 0.33, p ¼ 0.74).Heterogeneity between groups was low (I 2 ¼ 0%; ►Fig.3A).
For the fifth analysis (0.3 EDC vs. control at 12 months), again, three etch-and-rinse studies representing six datasets were included.Overall, there were significantly higher bond strength values in the experimental group compared with the control group (Z-test ¼ 2.58, p ¼ 0.01) but with considerable heterogenicity (I 2 ¼ 66%; ►Fig.3B).
For the sixth analysis (0.1% RIBO vs. control at baseline), only two etch-and-rinse studies met the criteria, representing three datasets.There was overall a significant difference favoring the experimental group over the control group (Ztest ¼ 3.12, p ¼ 0.002), with considerable heterogeneity (I 2 ¼ 99%; ►Fig.4A).
For the seventh analysis (0.1% RIBO vs. control at 6 months), two studies representing three datasets showed significantly higher bond strengths in the experimental group than the control group (Z-test ¼ 5.78, p < 0.00001) but with considerable heterogeneity I 2 ¼ 98% (►Fig.4B).
For the eighth analysis of pretreatment for 30 seconds (2% CHX vs. control at baseline), only four etch-and-rinse studies were included, representing seven datasets.There was overall a statistically significant difference favoring the control   group over the experimental group (Z ¼ 2.42, p ¼ 0.02), and heterogeneity was low (I 2 ¼ 0%; ►Fig.5A).
For the ninth analysis of pretreatment for 30 seconds (2% CHX vs. control at 6 months), only three etch-and-rinse studies met the criteria, representing six datasets.There was overall no statistically significant difference between groups (Z ¼ 0.28, p ¼ 0.78), and heterogeneity was considerable (I 2 ¼ 55%; ►Fig.5B).
For the 11th and final analysis of pretreatment for 60 seconds (2% CHX vs. control at 6 months), 11 etch-and-rinse studies were included, representing 16 datasets.Overall, the experimental group was slightly, but not significantly, favored over the control group (Z-test ¼ 1.73, p ¼ 0.08), with considerable heterogeneity (I 2 ¼ 91%).Five self-etching studies were included representing five datasets.Overall, the experimental group was slightly, but not significantly,  favored over the control group (Z-test ¼ 1.22, p ¼ 0.22), with considerable heterogeneity (I 2 ¼ 83%).The tests for overall effect favored the experimental group but this was not statistically significant (Z-test ¼ 2.35, p ¼ 0.73).Heterogeneity between subgroups was considerable (I 2 ¼ 90%; ►Fig.6B).

Discussion
This meta-analysis revealed that at least some MMP inhibitors significantly alter bond strength, both immediately and over the longer term.Accordingly, the null hypothesis was rejected.
Of all MMP inhibitors considered for meta-analysis, two MMP inhibitors improved bond strength: 0.3 M EDC and 0.1% RIBO.The 0.3 M EDC did not improve bond strength immediately (24 hours) but showed benefit after ageing for 12 months, while 0.1% RIBO showed statistically significant increases in bond strength both immediately (24 hours) and over the long term (6 months) compared with controls.Conversely, 2% CHX showed a slight but nonsignificant improvement in bond strength after 6 months of ageing but not immediately (24 hours) or after 12 months.The lack of immediate benefit with 2% CHX is consistent with two previous meta-analyses, 28,89 but the long-term results differ, possibly due to the different concentration of CHX used in previous studies.It has been suggested but not consistently proven that MMP inhibition by CHX is dose dependent. 90,913][94][95][96][97] With respect to adhesive systems, a previous systematic review 28 found that both types of adhesive system (self-etching and etch and rinse) benefited from 2% CHX in vitro.This, however, was also not consistent with the current meta-analysis results, since we found no significant difference according to the adhesive system used.
EDC and RIBO have a different mechanism of MMP inhibition to CHX through their cross-linking action.][100] These additional functions could explain their superiority in maintaining adhesive interface integrity.
Pretreatments of 30 and 60 seconds with 2% CHX met the inclusion criteria for meta-analysis.Generally, neither pretreatment protocol significantly improved bond strength either immediately (24 hours) or over the long term (6 months).Indeed, when 2% CHX was applied for 30 seconds, there was a significant negative effect on bond strength over 24 hours.After 6 months of aging, there was a slight improvement in bond strength, still favoring the control group.With pretreatments of 60 seconds, 2% CHX showed no effect on bond strength and was similar to controls and, while slightly improved bond strength was observed with CHX after 6 months, it was nevertheless not statistically significant.
Our results show some inconsistencies with previous systematic reviews which might be due to differences in the inclusion criteria.For example, Montagner et al 28

Limitations
There are a few limitations to our study.This review only included in vitro studies since there have been very few in Fig. 5 Forest plots according to pretreatment duration for 30 seconds: pretreatment with 2% CHX vs. control group at 24 hours (A) and 6 months (B).CHX, chlorhexidine; CI, confidence interval; SD, standard deviation.Fig. 6 Forest plots according to pretreatment duration for 60 seconds: pretreatment with 2% CHX vs. control group at 24 hours (A) and 6 months (B).CHX, chlorhexidine; CI, confidence interval; SD, standard deviation.. vivo studies or clinical trials in the literature.More in vivo studies will ultimately be crucial for providing high-quality evidence of the safety, toxicity, and efficacy of a given intervention in a complex model.Furthermore, although strict measures were taken during the search of the articles included for meta-analysis, several data demonstrated high heterogeneity.It is worth mentioning that most of the results with high heterogeneity were observed in the long-term ageing periods, unlike the immediate ageing periods which showed lower heterogeneity.Factors that could influence this may include the different brands of adhesive systems used and the ageing methods utilized.Similar findings were observed in the study by Montagner et al 28 which found that the aging methods were the greater influencing factor in the high heterogeneity.It is also worth noting that there are no standardized protocols for evaluating bond strength which previously shown will inevitably increase the heterogeneity of results 101 .To improve the reliability and quality of future bond strength testing studies, robust and strict guidelines for laboratory testing must be developed and implemented.
Many of the studies carried a risk of bias, and only one study mentioned blinding of the operator testing the bond strength; this parameter will be important to include in future studies to reduce the risk of bias.Moreover, only six studies calculated the sample size and reported a power analysis.
Nevertheless, these in vitro findings pave the way for rationale clinical trialing of dentine surface pretreatment with MMP inhibitors to improve clinical outcomes.

Conclusion
The data suggest that using 2% CHX had no significant positive effect on bond strength either immediately or over the longer term.Pretreatments with 2% CHX for either 30 or 60 seconds do not improve the bond strength.Both 0.3 M EDC and 0.1% RIBO improve bond strength immediately and over time.There was considerable heterogeneity between the different adhesive systems used, limiting our meta-analysis.Given the limited clinical evidence available, more research is required to confirm the beneficial use of MMP inhibitors.
and Kiuru et al 89 included different concentrations of CHX other than 2%, as well as various bond strength tests other than microtensile bond strength testing.

Table 2
Characteristics of the included studies

Table 2 (
Continued) Impact of Dentine Pretreatment with MMP Inhibitors on Bond Strength Jamal et al. 981 European Journal of Dentistry Vol. 17 No.4/2023 © 2022.The Author(s).

Table 2 (
Continued)Impact of Dentine Pretreatment with MMP Inhibitors on Bond Strength Jamal et al.

Table 2 (
Continued) Impact of Dentine Pretreatment with MMP Inhibitors on Bond Strength Jamal et al. 983 European Journal of Dentistry Vol. 17 No.4/2023 © 2022.The Author(s).

Table 2 (
Continued) European Journal of Dentistry Vol. 17 No.4/2023 © 2022.The Author(s).Impact of Dentine Pretreatment with MMP Inhibitors on Bond Strength Jamal et al.

Table 2 (
Continued) Impact of Dentine Pretreatment with MMP Inhibitors on Bond Strength Jamal et al. 985 European Journal of Dentistry Vol. 17 No.4/2023 © 2022.The Author(s).

Table 2 (
Continued)Impact of Dentine Pretreatment with MMP Inhibitors on Bond Strength Jamal et al.

Table 2 (
Continued) Impact of Dentine Pretreatment with MMP Inhibitors on Bond Strength Jamal et al. 987 European Journal of Dentistry Vol. 17 No.4/2023 © 2022.The Author(s).

Table 2 (
Continued)Impact of Dentine Pretreatment with MMP Inhibitors on Bond Strength Jamal et al.

Table 2 (
Continued) Impact of Dentine Pretreatment with MMP Inhibitors on Bond Strength Jamal et al. 989 European Journal of Dentistry Vol. 17 No.4/2023 © 2022.The Author(s).

Table 2 (
Continued)Impact of Dentine Pretreatment with MMP Inhibitors on Bond Strength Jamal et al.

Table 3
Quality assessment and risk of bias