Highlights
- •
Comparator choice in restorative trials is assessed via network analysis.
- •
Comparisons against certain established standards are common.
- •
Many trials compare within, not between material classes.
- •
It remains unclear if such network geometry indicates a biased research agenda.
Abstract
Objectives
The choice of trial comparators might impact on the validity of the available evidence. We aimed at evaluating dental restorative trial networks and the underlying comparisons made, hypothesizing that certain comparators are disproportionally preferred or avoided.
Methods
A systematic review was performed via Medline, CENTRAL and EMBASE. Randomized controlled trials on dental restoration or adhesive materials published 2005–2015 were included. Social network analysis techniques were used to assess trial networks.
Results
114 studies on 15 321 restorations placed in 5232 patients were included. 57 and 53 trials investigated restoration of cervical and load-bearing cavities, respectively. Four trials on non-cervical, non-load-bearing cavities did not form a network and were not evaluated. The most frequently assessed material combination was hybrid composites placed using 2-step etch-and-rinse adhesives. In cervical cavities, the majority of trials compared adhesives, not restorative materials. In load-bearing cavities, testing other restorative materials (ormocers, compomers) was common, too. In both networks, comparisons within material classes were frequent. There was significant homophily ( p < 0.001), i.e. certain material classes were preferred as comparators, while this preference seemed to change with time. Only very few comparisons yielded significant differences between materials.
Significance
The disproportional use of certain material classes as comparator might be due to their perceived role as gold standard. Compared with other scientific disciplines, dental restorative trial networks seem less prone for bias by comparator choice. Factors underlying the network geometry should be assessed to understand drivers of the research agenda.
1
Introduction
There is increasing interest into the internal and external validity of randomized controlled trials, based on the assessment of, for example, risk of bias , quality of reporting , appropriateness of statistical evaluation, or effects of sponsorship on trial outcomes . One factor which potentially impacts on the totality of available evidence and its robustness is the choice of trial comparators: Certain classes of procedures or products might serve as comparator more often than others, and some comparators might be avoided at all. Using placebo or less effective options for comparisons might distort the impression about the true effectiveness about treatments, and repeated chain-linked comparisons against less-than-optimal standards might in fact heavily bias the totality of evidence . To examine the underlying agenda of clinical trials in a specific field, new methods are required.
To investigate treatment comparisons, network analysis might be used. Such analyses have been performed to demonstrate clear preferences in comparator choice in trials on mycosis and myeloma , but have not been employed in dentistry so far. Network analysis can be used to not only graphically display the undertaken comparisons but also to statistically assess the properties of both the constructed network and therein included comparators .
One of the most prolific fields in dental research is restorative material science, with restorations still being the most frequently performed and overall most expensive treatment provided by dentists . Evaluating restorative trial networks could help to identify and reduce bias associated with the choice of the comparator. Within the present study, we aimed at evaluating dental restorative trial networks and the underlying comparisons made. We hypothesized that certain comparators are disproportionally preferred or avoided in recent randomized controlled trials.
2
Methods
2.1
Study design
With the advent of adhesive dental restorations, an increasing number of restorative materials can be placed using different adhesive strategies. We evaluated networks formed by trials comparing different adhesive and restorative materials and their combinations. First, a systematic review was performed to identify and appraise available trials. Second, data were analyzed using network analysis, with trial networks being assessed graphically and statistically.
2.2
Selection criteria
Randomized, controlled clinical trials (RCT) comparing the survival of two or more different restorative and/or adhesive materials were included. RCTs were excluded, if they compared different treatment techniques (e.g. conventional versus atraumatic restorative treatment), not materials, or placed restorative materials or adhesives as a sealant or for orthodontic bonding, not a restoration.
2.3
Search
The Cochrane Central Register of Controlled Trials, MEDLINE (via Pubmed) and EMBASE (via OVID) were searched at March 02, 2015 for relevant publications. The search strategy and screening procedure are shown in Fig. 1 . As our focus was the assessment of trials investigating current adhesive and restorative materials, our search was limited to studies published from 2005 onwards. Further articles were identified by cross-referencing from retrieved full-text studies. Two calibrated reviewers (FS, GG) independently screened titles and abstracts of the identified studies for eligibility. Inclusion of studies was independently decided by both reviewers. Consensus was obtained by discussion.
2.4
Data extraction
Duplicative data extraction was performed independently by two reviewers (FS, GG) using a piloted spreadsheet. If data of one clinical trial were published at different follow-up times, only data from the most recent publication (longest follow-up) were extracted. Data were recorded according to guidelines outlined by the Cochrane Collaboration . The following data were extracted:
- •
Restorative and adhesive material (class, name, manufacturer).
- •
Setting (number of patients; number of teeth; number of lesions; follow-up and drop-out, outcome measure).
- •
Included teeth (cavity location, i.e. cervical, load-bearing [occlusal and/or proximal], non-load bearing non-cervical; indication for treatment, dentition).
As this review was part of a larger investigation, the following data were also extracted, but not reported on in the present study:
- •
Methodological issues (trial registration, unit of randomization, method of analysis).
- •
Failures (number of failures per group).
- •
Risk of bias, assessed using the Cochrane Risk of Bias tool .
Restorative and adhesive materials were categorized as follows:
- –
Restorative materials: (1) hybrid composites (including nano- and micro-filled composites); (2) ormocer composites; (3) bulk fill composites (both flowable and packable bulk-fills), (4) siloranes; (5) compomers; (6) amalgam; (7) glass ionomer cements or resin-modified glass ionomer cements (only RMGICs were investigated within included studies).
- –
Adhesive materials: (1) 4- or 3-step etch-and-rinse adhesives; (2) 2-step etch-and-rinse adhesives; (3) 2-step self-etch adhesives; (4) one-step self-etch adhesives; (5) no adhesive used.
2.5
Analysis
Our primary analysis focused on the combinations of adhesive and restorative materials, as the choice of such combination would usually be clinical reality (i.e. most materials would be placed using an adhesive). In our secondary analyses, networks of compared adhesive and restorative materials were assessed too, while such comparisons are largely artificial (placement of an adhesive would be followed clinically by placement of a restoration).
Network analysis was performed using techniques developed for social network analysis. The latter describes links between different actors in a network and has been used as framework to assess scientific interactions and information exchange . It can be applied to evaluate the distribution of comparisons made in clinical controlled trials . In our networks, network nodes are formed by different treatment (adhesive and restorative materials and their combinations), which are connected by ties (trial comparisons) indicating a direct comparison within an RCT.
Network analysis allows to measure the properties of both the entire treatment network and each node (treatment). We mainly assessed network level properties, and graphically displayed networks to evaluate the network structure, commonly and less commonly used treatments, and preferred or avoided comparisons. The following network properties were evaluated via Node XL 2014 ( www.srmfoundation.org ):
- (1)
Density, which indicates the average or maximum proportion of the performed per all possible comparisons within a network. Values <25% and >50% indicate low and high density, respectively. Networks with low density are more prone for changes in the cohesion of the network if further trials are added .
- (2)
Degree, which represents the number of comparisons each node has been subject to, i.e. corresponds to density by indicating how well connected treatments are with each other . We assessed average and maximum degree.
- (3)
Betweenness, which indicates the ability of a treatment to link to other treatments . We assessed average and maximum betweenness.
- (4)
The clustering coefficient, which indicates how well connected treatments are, with values of 1 representing a network where all possible connections are made, while values of 0 indicate that only the minimum number of connections for forming the network are available .
- (5)
Homophily, which is a measure for treatment of the same class being compared with each other more often than expected . For evaluating homophily, classes were defined by the used restorative materials. Testing for homophily draws on simulated networks produced by permutations using the same nodes and constant network density . p -Values are derived by comparing the observed and the simulated (expected) pattern of comparisons . Homophily was calculated using UCINET 6 (Analytical Technologies, Lexington, USA).
In addition, networks were plotted graphically, with node diameter indicating the number of trial arms involving the specific treatment, and the width of lines (ties) representing the number of performed direct comparisons. Both graphical and statistical evaluations were performed separately for differently located (cervical, load-bearing) cavities, as the location affects choice of materials.
Last, we assessed if the number of trial arms involving certain adhesive or restorative materials changed over time, i.e. if there was a temporal evolution of comparator preference, using Excel 2013 (Microsoft, Redwood, USA).
2
Methods
2.1
Study design
With the advent of adhesive dental restorations, an increasing number of restorative materials can be placed using different adhesive strategies. We evaluated networks formed by trials comparing different adhesive and restorative materials and their combinations. First, a systematic review was performed to identify and appraise available trials. Second, data were analyzed using network analysis, with trial networks being assessed graphically and statistically.
2.2
Selection criteria
Randomized, controlled clinical trials (RCT) comparing the survival of two or more different restorative and/or adhesive materials were included. RCTs were excluded, if they compared different treatment techniques (e.g. conventional versus atraumatic restorative treatment), not materials, or placed restorative materials or adhesives as a sealant or for orthodontic bonding, not a restoration.
2.3
Search
The Cochrane Central Register of Controlled Trials, MEDLINE (via Pubmed) and EMBASE (via OVID) were searched at March 02, 2015 for relevant publications. The search strategy and screening procedure are shown in Fig. 1 . As our focus was the assessment of trials investigating current adhesive and restorative materials, our search was limited to studies published from 2005 onwards. Further articles were identified by cross-referencing from retrieved full-text studies. Two calibrated reviewers (FS, GG) independently screened titles and abstracts of the identified studies for eligibility. Inclusion of studies was independently decided by both reviewers. Consensus was obtained by discussion.
2.4
Data extraction
Duplicative data extraction was performed independently by two reviewers (FS, GG) using a piloted spreadsheet. If data of one clinical trial were published at different follow-up times, only data from the most recent publication (longest follow-up) were extracted. Data were recorded according to guidelines outlined by the Cochrane Collaboration . The following data were extracted:
- •
Restorative and adhesive material (class, name, manufacturer).
- •
Setting (number of patients; number of teeth; number of lesions; follow-up and drop-out, outcome measure).
- •
Included teeth (cavity location, i.e. cervical, load-bearing [occlusal and/or proximal], non-load bearing non-cervical; indication for treatment, dentition).
As this review was part of a larger investigation, the following data were also extracted, but not reported on in the present study:
- •
Methodological issues (trial registration, unit of randomization, method of analysis).
- •
Failures (number of failures per group).
- •
Risk of bias, assessed using the Cochrane Risk of Bias tool .
Restorative and adhesive materials were categorized as follows:
- –
Restorative materials: (1) hybrid composites (including nano- and micro-filled composites); (2) ormocer composites; (3) bulk fill composites (both flowable and packable bulk-fills), (4) siloranes; (5) compomers; (6) amalgam; (7) glass ionomer cements or resin-modified glass ionomer cements (only RMGICs were investigated within included studies).
- –
Adhesive materials: (1) 4- or 3-step etch-and-rinse adhesives; (2) 2-step etch-and-rinse adhesives; (3) 2-step self-etch adhesives; (4) one-step self-etch adhesives; (5) no adhesive used.
2.5
Analysis
Our primary analysis focused on the combinations of adhesive and restorative materials, as the choice of such combination would usually be clinical reality (i.e. most materials would be placed using an adhesive). In our secondary analyses, networks of compared adhesive and restorative materials were assessed too, while such comparisons are largely artificial (placement of an adhesive would be followed clinically by placement of a restoration).
Network analysis was performed using techniques developed for social network analysis. The latter describes links between different actors in a network and has been used as framework to assess scientific interactions and information exchange . It can be applied to evaluate the distribution of comparisons made in clinical controlled trials . In our networks, network nodes are formed by different treatment (adhesive and restorative materials and their combinations), which are connected by ties (trial comparisons) indicating a direct comparison within an RCT.
Network analysis allows to measure the properties of both the entire treatment network and each node (treatment). We mainly assessed network level properties, and graphically displayed networks to evaluate the network structure, commonly and less commonly used treatments, and preferred or avoided comparisons. The following network properties were evaluated via Node XL 2014 ( www.srmfoundation.org ):
- (1)
Density, which indicates the average or maximum proportion of the performed per all possible comparisons within a network. Values <25% and >50% indicate low and high density, respectively. Networks with low density are more prone for changes in the cohesion of the network if further trials are added .
- (2)
Degree, which represents the number of comparisons each node has been subject to, i.e. corresponds to density by indicating how well connected treatments are with each other . We assessed average and maximum degree.
- (3)
Betweenness, which indicates the ability of a treatment to link to other treatments . We assessed average and maximum betweenness.
- (4)
The clustering coefficient, which indicates how well connected treatments are, with values of 1 representing a network where all possible connections are made, while values of 0 indicate that only the minimum number of connections for forming the network are available .
- (5)
Homophily, which is a measure for treatment of the same class being compared with each other more often than expected . For evaluating homophily, classes were defined by the used restorative materials. Testing for homophily draws on simulated networks produced by permutations using the same nodes and constant network density . p -Values are derived by comparing the observed and the simulated (expected) pattern of comparisons . Homophily was calculated using UCINET 6 (Analytical Technologies, Lexington, USA).
In addition, networks were plotted graphically, with node diameter indicating the number of trial arms involving the specific treatment, and the width of lines (ties) representing the number of performed direct comparisons. Both graphical and statistical evaluations were performed separately for differently located (cervical, load-bearing) cavities, as the location affects choice of materials.
Last, we assessed if the number of trial arms involving certain adhesive or restorative materials changed over time, i.e. if there was a temporal evolution of comparator preference, using Excel 2013 (Microsoft, Redwood, USA).
Stay updated, free dental videos. Join our Telegram channel
VIDEdental - Online dental courses
