Abstract
Objectives
The risk of ‘caries adjacent to restorations’ (CAR) might depend on the used restorative materials. In situ studies are often used to compare the risk of caries adjacent to different materials. We aimed to review in situ studies to evaluate how different materials contribute to risk of CAR.
Data sources
We included in situ controlled trials comparing directly placed restorative materials, reporting on caries (mineral loss, measured via radiography or micro-hardness) adjacent to these materials. Medline, Embase and Cochrane CENTRAL were systematically searched. Screening and data extraction was performed independently by two authors. Materials were classified according to the used adhesive and restorative materials. Fixed-effects pairwise and frequentistic network meta-analyses were performed
Study selection
Nine studies (132 patients, 8 materials) were included, yielding inconsistent results. We could not identify underlying reasons, as confounders were only limitedly reported. The resulting material rankings come with great uncertainty, and raise doubts as to the validity and transferability of in situ studies as well as the applicability of their findings.
Conclusions
The current body of evidence of in situ studies is insufficient for firm conclusions as to the caries risk adjacent to different materials. The validity and applicability of included studies remain uncertain.
Clinical significance
While single in situ studies seem to convey consistent and applicable information, the overall body of evidence is inconsistent, limiting the conclusions which can be drawn from it.
1
Introduction
Secondary caries or ‘caries adjacent to restoration’ (CAR) is often stated as the main complication of dental restorations, leading to their replacements later in life and limiting their lifetime . While there is some debate around these lesions being truly “secondary”, i.e. somewhat associated with the restoration, or merely primary caries next to a restoration, it is understood that many factors seem to contribute to the risk of developing such lesions, like the patient’s general caries risk, the operator’s skills during restoration placement, as well as the specific local conditions, like cervical restoration extent . One factor which could plausibly determine the risk for CAR would be the used restoration material, as the material determines the bacterial colonization, growth and activity on the tooth-restoration interface via its surface roughness and possible antibacterial effects, but also the mineral loss resulting from any cariogenic bacterial activity via its buffering or remineralization capacities . Antibacterial restoration materials, containing for example 12-methacryloyloxydodecylpyridinium bromide (MDPB) have been postulated to reduce the risk for CAR . The same is claimed for materials releasing fluoride, like glass ionomer cements (GI) or other minerals, like calcium phosphate .
Comparing materials for their effect on CAR risk is thus relevant. In vitro studies allow a standardized assessment of materials, and can measure the resulting mineral loss adjacent to the restoration. However, they have limited external validity, employing either chemical or simplified biofilm caries models . Clinical studies are much more valid, but are usually limited by their follow-up periods and the inherent lack of standardization of lesions, material placements, and assessments . Moreover, technical and ethical constraints do usually not permit to assess the true outcome of the disease (mineral loss), which is why surrogate outcomes with unclear validity (like margin discoloration or gap formation) are used instead . In situ studies therefore seem to be an ideal study design for studies comparing material effects on CAR: While these study employ a variety of designs, differing in the used appliances, the specimens (usually bovine or human enamel or dentin ), or the sequence of groups or unit of randomization (parallel, cross-over, split-mouth studies) , they allow standardization of study conditions, the assessment of mineral loss, and can be performed with greatly reduced efforts compared with clinical studies.
A number of reviews assessed how the risk of CAR impacts on the clinical longevity or in vitro performance of restorations , with more recent reviews focusing on the effects of antibacterial adhesives . To date, no systematic appraisal of in situ studies investigating the risk of caries adjacent to different materials has been performed. The present review systematically assessed in situ studies comparing the effects of different materials on CAR. We further compared and ranked materials using network meta-analysis. This should allow to derive conclusions for clinical practice, but also regarding the robustness and validity of the findings of such studies.
2
Methods
2.1
Eligibility criteria
This systematic review included trials that
- •
were in situ controlled trials on directly placed restorative materials
- •
compared minimum two restorative material or material combinations
- •
reported caries adjacent to these materials
Trials needed to have compared materials in conditions which can be considered clinically acceptable; studies which had only placed materials adjacent to gaps or placed inappropriately (composites without bonding, purposive lack of moisture control etc.) were excluded. Moreover, materials needed to be available on the market (i.e. outdated materials with limited applicability for today’s practitioners were excluded).
2.1.1
Outcomes
The primary outcome was mineral loss of the carious lesion, assessed either via microradiography or micro-CT, or via hardness measurements (surface micro-hardness, cross-sectional micro-hardness).
The secondary outcome was adverse events, i.e. allergy, discomfort, taste affection etc. caused by the restorative material. We also counted any discomfort caused by the appliance as adverse event, while it should be noted that this is unrelated to the tested materials.
2.2
Information sources
2.2.1
Electronic searches
We have searched Medline via PubMed, Embase via Ovid and Cochrane CENTRAL at July 27th 2016. Grey literature was not searched, as we expected the information content to be insufficient to extract the required details. Reference lists of identified full-texts were screened and cross-referenced. We contacted study authors if required to obtain full-texts. Neither authors nor journals were blinded to reviewers. No language restriction was set.
2.2.2
Search strategy
A three-pronged approach was taken for screening, combining ((Secondary caries) OR caries adjacent to restoration) OR recurrent caries AND (in situ) AND Search ((((((mineral loss) OR micro-hardness) OR hardness) OR demineralisation) OR demineralization) OR TMR) OR TWIM ( Fig. 1 ). Details on the search strategy for each database can be found in the Appendix (Table S1).
2
3 Study records
2.3.1
Selection process
Two authors (HA, FS) independently screened titles and abstracts to decide consideration of full-texts. In case of disagreement, a full-text was obtained and assessed. Full texts were evaluated independently after de-duplication. Studies were included after agreement with consensus in cases of disagreement being reached through discussion.
2.3.2
Data collection process
A piloted spreadsheet was used for data extraction and management. Data extraction was performed independently by two reviewers (HA, FS). Disagreements were resolved through discussion.
2.4
Data items
The following items were collected: study, year and country of study conductance; study design (cross-over, parallel) and duration; specimens substrate (bovine or human enamel or dentin); compared restorative materials; other interventions; cavity dimensions; participants (number, age, caries risk, dentition); appliance design (upper or lower jaw), number of specimens per appliance, position of the specimens in the appliance (palatal, buccal); method of mineral loss measurement (micro-hardness, microradiography, μ-CT).
The used materials were then classified for syntheses. To account for certain materials (composites) being placed in combination with different adhesives, material class combinations were used in this case. The following classification was used: (1) Amalgam (AM), (2) Glass ionomer cement (GI), (3) Resin-modified GI (RMGI), (4) conventional composites placed using etch-and-rinse adhesives (ER/CC), (5) conventional composites placed using conventional self-etch adhesives (SE/CC), (6) conventional composites placed using antibacterial, MDPB-containing self-etch adhesives (MB/CC), (7) silorane composite placed using the respective adhesive (SI), (8) compomer (CMP), which had been placed only with an etch-and-rinse adhesive.
2
5 Data synthesis
2.5.1
Meta-analysis
Fixed-effects pairwise meta-analysis was performed using STATA. The effect size estimate was the Standardized Mean Difference, i.e. the intervention effect in each study relative to the variability observed in that study . SMD was used, as included studies used different scales to measure mineral loss. A negative SMD indicates a smaller mineral loss in the test than the comparator group. Heterogeneity was assessed using I 2 -statistics . Note that we did not separately assess splitmouth or cross-over studies, as this would have reduced the number of included studies per analysis.
As most studies reported mineral loss in different interfacial depths or distances from the restoration, we decided to only evaluate the deepest reported mineral losses in each study (as we assumed the largest differences between groups to occur here) as well as only those measurements closest to the material (and the interface), again assuming largest differences between groups in these locations.
Networks of interventions were constructed by plotting different treatments (as nodes) and comparisons (as edges), and were inspected for geometry and asymmetry . We performed frequentist network meta-analyses using the network command in STATA to compare different treatments . Network meta-analyses synthesized direct and indirect evidence based on the assumption of transitivity. If treatment A is better than treatment B in one trial, and treatment B is better than treatment C in a different trial with a similar patient population to the previous trial. Transitivity means that we could expect that treatment A is better than treatment C if they were compared in a hypothetical trial . When incorporating direct and indirect evidence, NMA model assumes direct and indirect evidence is consistent, and in our network meta-analysis this assumption was evaluated by the loop inconsistency model proposed by Lu and Ades . We reported mean SMD and their 95% confidence intervals (95% CI) for treatment comparisons. Different material combinations were ranked according to their probability of having the lowest mineral loss , and the average rank calculated. The surface under the cumulative ranking (SUCRA) line was plotted and the area under the plot (SUCRA value) calculated. Comparison-adjusted funnel plots were used to assess publication bias . All our statistical analysis was performed using STATA, version 14.
In studies employing different materials (product) from the same material class combination, we randomly selected one of these for the analysis. We performed sensitivity analysis to account for the resulting uncertainty. For studies reporting on different hygiene or sample regimens (fluoride application or not, aged or non-aged restoration), we chose the group most likely to reflect clinical conditions (aged restorations, fluoride supply).
2.5.2
Subgroup and meta-regression analyses
We separately analysed carious lesions in enamel and dentin.
2.6
Confidence in data
Selection bias (sequence generation and allocation concealment for participants and specimens), detection bias (blinding of participants, operator, appliance construction and examiners), attrition bias (loss-to-follow-up and missing values or participants) and reporting bias (selective reporting, unclear withdrawals, missing outcomes) were recorded, assessed and classified according to Cochrane guidelines . Sequence generation and allocation concealment of the specimens as well as the blinding of the operator or the technician who constructed the appliance were additionally assessed. The latter criteria were added to account for specific bias in in situ studies.
2
Methods
2.1
Eligibility criteria
This systematic review included trials that
- •
were in situ controlled trials on directly placed restorative materials
- •
compared minimum two restorative material or material combinations
- •
reported caries adjacent to these materials
Trials needed to have compared materials in conditions which can be considered clinically acceptable; studies which had only placed materials adjacent to gaps or placed inappropriately (composites without bonding, purposive lack of moisture control etc.) were excluded. Moreover, materials needed to be available on the market (i.e. outdated materials with limited applicability for today’s practitioners were excluded).
2.1.1
Outcomes
The primary outcome was mineral loss of the carious lesion, assessed either via microradiography or micro-CT, or via hardness measurements (surface micro-hardness, cross-sectional micro-hardness).
The secondary outcome was adverse events, i.e. allergy, discomfort, taste affection etc. caused by the restorative material. We also counted any discomfort caused by the appliance as adverse event, while it should be noted that this is unrelated to the tested materials.
2.2
Information sources
2.2.1
Electronic searches
We have searched Medline via PubMed, Embase via Ovid and Cochrane CENTRAL at July 27th 2016. Grey literature was not searched, as we expected the information content to be insufficient to extract the required details. Reference lists of identified full-texts were screened and cross-referenced. We contacted study authors if required to obtain full-texts. Neither authors nor journals were blinded to reviewers. No language restriction was set.
2.2.2
Search strategy
A three-pronged approach was taken for screening, combining ((Secondary caries) OR caries adjacent to restoration) OR recurrent caries AND (in situ) AND Search ((((((mineral loss) OR micro-hardness) OR hardness) OR demineralisation) OR demineralization) OR TMR) OR TWIM ( Fig. 1 ). Details on the search strategy for each database can be found in the Appendix (Table S1).
2
3 Study records
2.3.1
Selection process
Two authors (HA, FS) independently screened titles and abstracts to decide consideration of full-texts. In case of disagreement, a full-text was obtained and assessed. Full texts were evaluated independently after de-duplication. Studies were included after agreement with consensus in cases of disagreement being reached through discussion.
2.3.2
Data collection process
A piloted spreadsheet was used for data extraction and management. Data extraction was performed independently by two reviewers (HA, FS). Disagreements were resolved through discussion.
2.4
Data items
The following items were collected: study, year and country of study conductance; study design (cross-over, parallel) and duration; specimens substrate (bovine or human enamel or dentin); compared restorative materials; other interventions; cavity dimensions; participants (number, age, caries risk, dentition); appliance design (upper or lower jaw), number of specimens per appliance, position of the specimens in the appliance (palatal, buccal); method of mineral loss measurement (micro-hardness, microradiography, μ-CT).
The used materials were then classified for syntheses. To account for certain materials (composites) being placed in combination with different adhesives, material class combinations were used in this case. The following classification was used: (1) Amalgam (AM), (2) Glass ionomer cement (GI), (3) Resin-modified GI (RMGI), (4) conventional composites placed using etch-and-rinse adhesives (ER/CC), (5) conventional composites placed using conventional self-etch adhesives (SE/CC), (6) conventional composites placed using antibacterial, MDPB-containing self-etch adhesives (MB/CC), (7) silorane composite placed using the respective adhesive (SI), (8) compomer (CMP), which had been placed only with an etch-and-rinse adhesive.
2
5 Data synthesis
2.5.1
Meta-analysis
Fixed-effects pairwise meta-analysis was performed using STATA. The effect size estimate was the Standardized Mean Difference, i.e. the intervention effect in each study relative to the variability observed in that study . SMD was used, as included studies used different scales to measure mineral loss. A negative SMD indicates a smaller mineral loss in the test than the comparator group. Heterogeneity was assessed using I 2 -statistics . Note that we did not separately assess splitmouth or cross-over studies, as this would have reduced the number of included studies per analysis.
As most studies reported mineral loss in different interfacial depths or distances from the restoration, we decided to only evaluate the deepest reported mineral losses in each study (as we assumed the largest differences between groups to occur here) as well as only those measurements closest to the material (and the interface), again assuming largest differences between groups in these locations.
Networks of interventions were constructed by plotting different treatments (as nodes) and comparisons (as edges), and were inspected for geometry and asymmetry . We performed frequentist network meta-analyses using the network command in STATA to compare different treatments . Network meta-analyses synthesized direct and indirect evidence based on the assumption of transitivity. If treatment A is better than treatment B in one trial, and treatment B is better than treatment C in a different trial with a similar patient population to the previous trial. Transitivity means that we could expect that treatment A is better than treatment C if they were compared in a hypothetical trial . When incorporating direct and indirect evidence, NMA model assumes direct and indirect evidence is consistent, and in our network meta-analysis this assumption was evaluated by the loop inconsistency model proposed by Lu and Ades . We reported mean SMD and their 95% confidence intervals (95% CI) for treatment comparisons. Different material combinations were ranked according to their probability of having the lowest mineral loss , and the average rank calculated. The surface under the cumulative ranking (SUCRA) line was plotted and the area under the plot (SUCRA value) calculated. Comparison-adjusted funnel plots were used to assess publication bias . All our statistical analysis was performed using STATA, version 14.
In studies employing different materials (product) from the same material class combination, we randomly selected one of these for the analysis. We performed sensitivity analysis to account for the resulting uncertainty. For studies reporting on different hygiene or sample regimens (fluoride application or not, aged or non-aged restoration), we chose the group most likely to reflect clinical conditions (aged restorations, fluoride supply).
2.5.2
Subgroup and meta-regression analyses
We separately analysed carious lesions in enamel and dentin.
2.6
Confidence in data
Selection bias (sequence generation and allocation concealment for participants and specimens), detection bias (blinding of participants, operator, appliance construction and examiners), attrition bias (loss-to-follow-up and missing values or participants) and reporting bias (selective reporting, unclear withdrawals, missing outcomes) were recorded, assessed and classified according to Cochrane guidelines . Sequence generation and allocation concealment of the specimens as well as the blinding of the operator or the technician who constructed the appliance were additionally assessed. The latter criteria were added to account for specific bias in in situ studies.
Stay updated, free dental videos. Join our Telegram channel
VIDEdental - Online dental courses
