Types of studies

Randomised controlled trials (RCTs), including those that use a split‐mouth design, were included in this review.

Types of participants

Any participant with full arch fixed orthodontic appliance(s) with molar tubes, bonded to first or second permanent molars, was included. Participants who had previously undergone fixed appliance treatment have been excluded. Participants with cleft lip or palate have been excluded due to the higher prevalence of molar crossbite in this group which has been associated with a greater incidence of molar attachment failure (Hodges 2001). Those with other craniofacial syndromes, or where orthognathic surgery was required, have also been excluded.

Types of interventions

Studies which compared any type of adhesive used to bond molar tubes (stainless steel or titanium) with any other adhesive, were included. Trials were also included where:

1. a tube was bonded to a molar tooth on one side of an arch and a band cemented to the same tooth type on the opposite side of the same arch;

2. molar tubes had been allocated to one tooth type in one patient group and molar bands to the same tooth type in another patient group.

Studies have been excluded that:

1. compared adhesives from the same group, that used the same curing mechanism;

2. varied etching times;

3. bonded a molar tube to a surface other than intact buccal human enamel (e.g. to a gold, porcelain or amalgam substrate or to hypoplastic enamel);

4. used tubes of different metals or of different bonding base size or configuration on the same tooth type on opposite sides of the same arch;

5. used lip bumpers to bonded molar teeth;

6. bonded tubes to primary molars or premolars or to different molar teeth on opposite sides of the mouth;

7. did not follow participants to the end of treatment.

Types of outcome measures

The primary outcome for the review was the success of each adhesive i.e. first time bond (or band as appropriate) failure. Dichotomous data on whether the molar tube stayed cemented to the tooth or not would be recorded.

Dichotomous data on the presence or absence of decay (decalcification) associated with or around the tubes would also be recorded. If data existed on size/area of decalcifications, these would also be included.

Data on adverse events (i.e. illness, allergy, bad taste, mucosal trauma), damage to teeth on tube removal, length of treatment, treatment cost and time to replace tubes with an adhesive were recorded also.

Electronic searches

The following electronic databases were searched:

• Cochrane Oral Health's Trials Register (whole database, to 15 February 2017) (Appendix 1);

• the Cochrane Central Register of Controlled Trials (CENTRAL; 2017, Issue 1) in the Cochrane Library (searched 15 February 2017) (Appendix 2);

• MEDLINE Ovid (1946 to 15 February 2017) (Appendix 3);

• Embase Ovid (1980 to 15 February 2017) (Appendix 4).

No restrictions were placed on the language or date of publication when searching the electronic databases.

Searching other resources

We searched the following trial registries for ongoing studies (see Appendix 5 for details of the search strategy):

• ClinicalTrials.gov (whole database, to 15 February 2017);

• The World Health Organization International Clinical Trials Registry Platform (whole database, to 15 February 2017).

All the references lists of the included studies were checked manually to identify any additional studies.

All the first authors of trial reports were contacted in an attempt to identify any unpublished studies and clarify information about the published trials (including missing data, method of randomisation, blinding and withdrawals).

Manufacturers were contacted to confirm the cement/adhesive type and were asked about their knowledge of any unpublished or ongoing clinical trials.

Cochrane Oral Health's Trials Register and CENTRAL already contain the handsearching results for the European Journal of Orthodontics (1979 to 2007), the American Journal of Orthodontics and Dentofacial Orthopedics (1970 to 2007), the British Journal of Orthodontics (became Journal of Orthodontics in 2000) (1973 to 2008) and the Angle Orthodontist (1979 to 2007).

No additional handsearching of journals was undertaken.

Selection of studies

The titles and abstracts (when available) of all reports identified through the searches were scanned by two review authors. Full reports were obtained for trials appearing to meet the inclusion criteria or for which there was insufficient information in the title and abstract to make a clear decision. All full reports were assessed for eligibility independently and in duplicate without blinding to the authors, adhesives used or results obtained. All disagreements were resolved by discussion.
A statistician was to be consulted with regard to data analysis and where doubt existed about inclusion.

Data extraction and management

Data extraction was carried out independently and in duplicate. The following data were entered on a customised data collection form.

• Date that the study was conducted.

• Year of publication.

• Treatments including details of type of adhesive used to cement molar tubes (and bands where appropriate) and type of fixed appliance used.

• Sample size by study group.

• Age of subjects.

• Number of male subjects and female subjects per study group.

• Details of withdrawals by study group.

• Outcome measures.

The primary outcome measures were first time bond failure only (relative numbers/proportion of failures per group) and decalcification in each group.

Assessment of risk of bias in included studies

Eligible trials were assessed according to the following criteria:

• generation of random sequence;

• concealed allocation of treatment;

• blinding of participants/caregivers (where feasible);

• incomplete outcome data;

• selective reporting.

A description of the domains was tabulated for each included trial, along with a judgement of low, high or unclear risk of bias as described in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 (Higgins 2011).

A summary assessment of the risk of bias for the primary outcome (across domains) across studies was undertaken (Higgins 2011). Within a study, a summary assessment of low risk of bias is given when there is a low risk of bias for all key domains, unclear risk of bias when there is an unclear risk of bias for one or more key domains, and high risk of bias when there is a high risk of bias for one or more key domains. Across studies, a summary assessment is rated as low risk of bias when most information is from studies at low risk of bias, unclear risk of bias when most information is from studies at low or unclear risk of bias, and high risk of bias when the proportion of information is from studies at high risk of bias sufficient to affect the interpretation of the results.

Measures of treatment effect

For dichotomous outcomes, the estimates of effect of an intervention were expressed as risk ratios (or hazard ratio) together with 95% confidence intervals. For continuous outcomes, mean differences together with 95% confidence intervals were used.

Unit of analysis issues

This would be the participant or sites within the participant in split‐mouth studies. The statistical analysis was considered inappropriate if:

1. a split‐mouth design did not take the clustering of the teeth or 'pairing' into account;

2. all failures were included without taking into account multiple failures on the same tooth.

Dealing with missing data

Authors were contacted for further data where required.

Assessment of heterogeneity

Clinical heterogeneity was to be assessed by examining the participants, interventions and outcome measures included in the trials. Statistical heterogeneity was to be assessed by inspection of a graphical display of the estimated treatment effects from the trials along with their 95% confidence intervals. The significance of any discrepancies in the estimates of the treatment effects from the different trials was to be assessed by means of Cochran's test for heterogeneity and quantified by the I² statistic.

Assessment of reporting biases

A funnel plot was to be drawn if sufficient trials (10 or more) were identified. Asymmetry of the funnel plot may indicate publication bias and other biases related to sample size, though it may also represent a true relationship between trial size and effect size. A formal investigation of the degree of asymmetry was to be undertaken following the recommendations presented in Chapter 10 of the Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 (Higgins 2011).

If evidence of small‐study effects had been identified, a sensitivity analysis exploring issues such as publication status and language of publication would have been undertaken.

Data synthesis

Comparisons were to be made firstly between any of the five main types of adhesive. If possible, comparisons were to be made within groups and, where appropriate, between chemical‐ and light‐cured adhesives as follows:

1. chemically‐cured composite (CC) ‐ variables on composite matrix and primer;

2. light‐cured composite (LC) ‐ variables on composite matrix and primer (including self‐etching primer);

3. conventional glass ionomer (GIC) ‐ variables on powder and liquid (product is not light‐cured);

4. poly‐acid modified composite (Compomer) ‐ variables on composite matrix and glass ionomer particles;

5. resin‐modified glass ionomer (RMGIC) ‐ variables on type of acid, resin and polymerisation mechanism.

Within group comparisons assessing products of different brand names to see if any adhesive of the same type performs better than another of the same type, were also to be undertaken if data allowed. For each adhesive group and between adhesive groups, comparisons were also to be undertaken between the orthognathic cases and non‐orthognathic cases, if data allowed.

Meta‐analyses were to be undertaken only on studies of similar comparisons reporting the same outcome measures. Risk ratios along with 95% confidence intervals were calculated for dichotomous data, and mean differences and 95% confidence intervals calculated for continuous data. Data were combined using a random‐effects model (fixed‐effect models used if less than three studies in meta‐analysis). The number needed to treat (NNT) was to be calculated to prevent one extra bonded molar tube failing, as appropriate.

Sensitivity analysis

Sensitivity analysis was to be undertaken for aspects of study quality and for potential sources of heterogeneity specified a priori as follows: excluding/including unpublished studies, excluding/including studies of low quality and excluding/including one or more large studies to assess how much they dominate the results. In addition, analysis by angle classification, inclusion of orthognathic surgery and level of participant co‐operation was to be undertaken if data allowed. The association of these factors with estimated effects was to be examined by performing random‐effects metaregression analysis in STATA version 7.0 (STATA Corporation, USA), using the program Metareg. Further potential sources of heterogeneity were to be investigated as determined from the study reports, although these would clearly have been identified as 'post‐hoc' analyses and the results treated with caution.