1

Introduction

Dental caries is still a common disease worldwide and results in tooth substance loss. Inlay or onlay restorations are widely used to treat moderate Class I and II losses. Inlays and onlays can be made of alloy, composite or ceramic material. However, patients often refuse metallic restorations, so practitioners must choose between the 2 esthetic materials: ceramic and composite.

Many studies have compared these materials in vitro . Ceramic inlays and onlays are mainly composed of glass, with some crystals added to increase strength . Composite inlays and onlays are made of a resinous matrix and fillers of different types . Ceramic materials are resistant to compressive forces but are susceptible to tensile stresses and more prone to fracture than are composite materials . However, ceramics are harder than composites and more wear-resistant but can induce more wear than usual with the opposing tooth’s surface . Furthermore, adhesive cement interfaces are made of composite material, so the wear of the interface and restoration material should be closer for composites and marginal integrity could be better . Another disadvantage of composites is their resinous matrix and the possible monomer release if it is incompletely polymerized .

Although in vitro studies generally do not predict in vivo results , few clinical studies have verified these in vitro findings concerning the performance of composite and ceramic for inlays and onlays.

Several clinical studies have evaluated ceramic and composite materials separately . Yet, we can draw no definite conclusions on the best material from these studies. In 1997, Roulet et al. concluded that “very little” was known about the longevity of esthetic inlays. The authors mentioned the difficulty in obtaining a strong bond between the restoration’s surface and the tooth for composite inlays , but this problem has since been solved . In 2001, Hickel and Manhart reported an annual failure rate of 2.3% (range: 0–11.8%) for composite inlays and onlays, as compared with 1.3% (range: 0–7.5%) for ceramic inlays and onlays evaluated in longitudinal studies . In 2004, the same authors reported an annual failure rate of 2.9% (SD 2.6) for composite inlays, as compared with 1.9% (SD 1.8) for ceramic restorations . Note that we have no evidence for linear failure rates of inlays and onlays, so the latter results can be discussed.

Finally, material knowledge has evolved, new materials have been developed, and no systematic review has answered the question posed by practitioners: Is the clinical efficacy of composite or ceramic better for inlay and onlay manufacturing?

We aimed to perform a systematic review of published reports of randomized controlled trials comparing the efficacy of composite and ceramic inlays and onlays for restoring posterior teeth of adults.

2

Methods

The review is registered, and the protocol can be accessed on the Prospero website .

Eligible studies were randomized controlled trials comparing the clinical efficacy of composite to ceramic inlays or onlays. Any composite or ceramic material was eligible. Patients had to be adults, treated in any dental care center or practice. The follow-up had to be 6 months or more. We searched MEDLINE, Embase and the Cochrane Central Register of Controlled Trials (Central) without any restriction on date or language. The last search was on December 24, 2012. The full electronic search strategy is in Appendix 1. We searched ClinicalTrials.gov to identify ongoing trials. The references of all eligible trials were checked for other relevant studies.

Two authors independently and in duplicate screened the titles of records retrieved by the search, then screened the selected abstracts to identify studies that potentially met the eligibility criteria. Any disagreements were resolved by discussion and the reasons for exclusion were recorded. The full text of potentially eligible studies was retrieved and assessed for eligibility by the same authors. Again, the reasons for exclusion were recorded.

A standardized, pre-piloted form was used to extract data from the included studies for assessment of study quality and synthesis of evidence. Data were extracted on (1) general trial information (publication year, number of patients treated, number of inlays or onlays manufactured, and number of operators); (2) intervention characteristics (rubber dam isolation, base or liner application, inlay/onlay materials and manufacturing technique(s), adhesive and cement), and (3) outcome data (clinical scores [United States Public Health Services (USPHS) or Californian Dental Association (CDA) or Fédération dentaire internationale (FDI) criteria] and restoration failures). The clinical scores are usually ordinal variables that measure the quality of the restoration according to a variety of dimensions. When published, in 1971, the USPHS scale comprised 3 grades (Alfa, Bravo and Charlie: Charlie being pejorative and Alfa being favorable) and 5 items (color match, cavosurface marginal discoloration, anatomic form, marginal adaptation, and caries). A fourth grade (Delta) was later added. The CDA scale is similar (except the grades are Romeo, Sierra, Tango and Victor) . The consensus FDI score, published in 2007, comprises 5 grades (from 1 to 5, 1 being ideal and 5 corresponding to a restoration that needs replacement) and 18 items that comprise the 5 initial USPHS items . Any event that required replacement of the inlay or onlay was defined as a failure. Failure types were also extracted and classified according to failure cause: inlay fracture, tooth fracture, recurrent caries, persisting hypersensibility and endodontic treatment.

The risk of bias for each study was assessed by the Cochrane Collaboration Risk of Bias tool, which includes the following domains: methods for sequence generation and maintaining allocation concealment, blinding, and incomplete outcome data . Each domain was rated as low, high, or unclear risk of bias. Within-study selective outcome reporting was not assessed because we did not have access to study protocols. Blinding of dental staff (operators and clinical evaluators) is not possible because both materials can easily be recognized by an experienced eye, so this domain was not assessed and the tool was slightly modified to take into account intervention standardization and the risk of bias during clinical outcome assessment. The standardization risk was high if the cavity preparation or the adhesion procedure differed between treatment groups; it was low otherwise and unclear if these aspects were insufficiently or not reported. The clinical outcome assessment risk was low if 2 independent evaluators or 1 evaluator distinct from the operator had completed the clinical follow-up examinations; it was high otherwise and unclear if these aspects were insufficiently or not reported. The source of funding was also extracted.

The unit of analysis was the tooth. For clinical score outcomes, we reported only percentages of restorations assessed with the best grade (i.e., 1 for FDI criteria, A for USPHS criteria, R for CDA criteria). For each item, we estimated risk ratios for the restoration to be assessed with the best grade (considered the event). To take into account patients with missing outcome data, we assumed that the proportion of patients with the best grade was the same in complete cases and in patients with missing outcome data. For the dichotomous failure outcome, the measure of treatment effect was the risk ratio. To allow for an intention-to-treat analysis, we imputed missing outcome data as success.

When a study included multiple composite or multiple ceramic groups, all composites were combined into a single composite group and/or all ceramics were combined into a single ceramic group . We synthesized trials comparing at least one composite and one ceramic with the same outcome (item score or failure) at a given follow-up time. The decision whether or not to combine the results of individual studies depended on the assessment of heterogeneity in forest plots and by I ² coefficients. Combined estimates and associated 95% confidence intervals were calculated by use of Mantel–Haenszel fixed-effect methods.

2

Methods

The review is registered, and the protocol can be accessed on the Prospero website .

Eligible studies were randomized controlled trials comparing the clinical efficacy of composite to ceramic inlays or onlays. Any composite or ceramic material was eligible. Patients had to be adults, treated in any dental care center or practice. The follow-up had to be 6 months or more. We searched MEDLINE, Embase and the Cochrane Central Register of Controlled Trials (Central) without any restriction on date or language. The last search was on December 24, 2012. The full electronic search strategy is in Appendix 1. We searched ClinicalTrials.gov to identify ongoing trials. The references of all eligible trials were checked for other relevant studies.

Two authors independently and in duplicate screened the titles of records retrieved by the search, then screened the selected abstracts to identify studies that potentially met the eligibility criteria. Any disagreements were resolved by discussion and the reasons for exclusion were recorded. The full text of potentially eligible studies was retrieved and assessed for eligibility by the same authors. Again, the reasons for exclusion were recorded.

A standardized, pre-piloted form was used to extract data from the included studies for assessment of study quality and synthesis of evidence. Data were extracted on (1) general trial information (publication year, number of patients treated, number of inlays or onlays manufactured, and number of operators); (2) intervention characteristics (rubber dam isolation, base or liner application, inlay/onlay materials and manufacturing technique(s), adhesive and cement), and (3) outcome data (clinical scores [United States Public Health Services (USPHS) or Californian Dental Association (CDA) or Fédération dentaire internationale (FDI) criteria] and restoration failures). The clinical scores are usually ordinal variables that measure the quality of the restoration according to a variety of dimensions. When published, in 1971, the USPHS scale comprised 3 grades (Alfa, Bravo and Charlie: Charlie being pejorative and Alfa being favorable) and 5 items (color match, cavosurface marginal discoloration, anatomic form, marginal adaptation, and caries). A fourth grade (Delta) was later added. The CDA scale is similar (except the grades are Romeo, Sierra, Tango and Victor) . The consensus FDI score, published in 2007, comprises 5 grades (from 1 to 5, 1 being ideal and 5 corresponding to a restoration that needs replacement) and 18 items that comprise the 5 initial USPHS items . Any event that required replacement of the inlay or onlay was defined as a failure. Failure types were also extracted and classified according to failure cause: inlay fracture, tooth fracture, recurrent caries, persisting hypersensibility and endodontic treatment.

The risk of bias for each study was assessed by the Cochrane Collaboration Risk of Bias tool, which includes the following domains: methods for sequence generation and maintaining allocation concealment, blinding, and incomplete outcome data . Each domain was rated as low, high, or unclear risk of bias. Within-study selective outcome reporting was not assessed because we did not have access to study protocols. Blinding of dental staff (operators and clinical evaluators) is not possible because both materials can easily be recognized by an experienced eye, so this domain was not assessed and the tool was slightly modified to take into account intervention standardization and the risk of bias during clinical outcome assessment. The standardization risk was high if the cavity preparation or the adhesion procedure differed between treatment groups; it was low otherwise and unclear if these aspects were insufficiently or not reported. The clinical outcome assessment risk was low if 2 independent evaluators or 1 evaluator distinct from the operator had completed the clinical follow-up examinations; it was high otherwise and unclear if these aspects were insufficiently or not reported. The source of funding was also extracted.

The unit of analysis was the tooth. For clinical score outcomes, we reported only percentages of restorations assessed with the best grade (i.e., 1 for FDI criteria, A for USPHS criteria, R for CDA criteria). For each item, we estimated risk ratios for the restoration to be assessed with the best grade (considered the event). To take into account patients with missing outcome data, we assumed that the proportion of patients with the best grade was the same in complete cases and in patients with missing outcome data. For the dichotomous failure outcome, the measure of treatment effect was the risk ratio. To allow for an intention-to-treat analysis, we imputed missing outcome data as success.

When a study included multiple composite or multiple ceramic groups, all composites were combined into a single composite group and/or all ceramics were combined into a single ceramic group . We synthesized trials comparing at least one composite and one ceramic with the same outcome (item score or failure) at a given follow-up time. The decision whether or not to combine the results of individual studies depended on the assessment of heterogeneity in forest plots and by I ² coefficients. Combined estimates and associated 95% confidence intervals were calculated by use of Mantel–Haenszel fixed-effect methods.

3

Results