Types of studies

All randomised controlled clinical trials comparing the use of adhesive restorations ton non‐ adhesive restorations in NCCLs in permanent teeth, with a minimum follow up of one week. We will include both parallel and split‐mouth designs.

Types of participants

Adults aged 18 years and over with permanent teeth and at least one NCCL in a vital tooth with a prescribed adhesive restoration, non‐adhesive restoration or not treatment

Types of interventions

Any type of adhesive restoration placed on a vital tooth with an NCCL, including but not limited to resin‐based restorations, flowable composite, compomers glass ionomers and resin glass ionomers.

We will include studies where the comparison groups received any non‐adhesive intervention, such as amalgam, or any other intervention (e.g. periodontal therapy or chemical therapy, including but not limited to dentifrices and any other desensitiser agents), or no treatment.

Types of outcome measures

Electronic searches

We will search the following databases:

• the Cochrane Oral Health Group Trials Register (to present);

• the Cochrane Central Register of Controlled Trials (CENTRAL; latest issue);

• MEDLINE via OVID (1946 to present);

• EMBASE via OVID (1980 to present);

• LILACs via BIREME Virtual Health Library (1980 to present).

We will place no restrictions on language or date of publication.

Searching other resources

We will search the following trial registries for ongoing studies:

• the US National Institutes of Health Clinical Trials Database (http://clinicaltrials.gov);

• the World Health Organization International Clinical Trials Registry Platform (http://apps.who.int/trialsearch/).

We will also search the reference lists of related relevant articles for additional trials. We will attempt to search for unpublished studies if available and will contact recognised authorities in the field in an attempt to collect all possible data.

We will also search abstracts from the scientific meetings and conferences of the American Association for Dental Research and the International Association for Dental Research for appropriate studies.

We will handsearch relevant journals for relevant articles. We will obtain translations of studies written in a language other than English from the authors of those studies and their collaborators, if needed.

Selection of studies

Two review authors (AVK and SS) will independently screen titles and abstracts identified by the electronic searches in order to identify potentially eligible studies for further evaluation to determine whether they meet the inclusion criteria for this review. A third review author (HE) will moderate any disagreements. We will obtain full‐text copies of all eligible and potentially eligible studies, which will be further evaluated in detail by two review authors (AVK and SS) to identify those studies that actually meet all the inclusion criteria. A third review author (HE) will moderate any disagreements. We will record those studies not meeting the inclusion criteria in the excluded studies section of the review and will note the reasons for exclusion in a 'Characteristics of excluded studies' table. A PRISMA study flow chart will be created to summarise this process.

Data extraction and management

We will design a data extraction form specifically for use in this review, which two review authors (AVK and SS) will pilot independently. We will enter details of all included studies in a 'Characteristics of included studies' table in Review Manager (Revman) version 5.1.0. AVK will enter the final data. We will resolve any disagreements by discussion and, if necessary, by consulting a third review author (HE).

The following details will be recorded for each study:

• publication details (e.g. year of publication, language);

• trial methods;

• participants;

• intervention;

• control (comparison);

• outcomes;

• duration of follow‐up;

• sample size calculation;

• funding details.

Assessment of risk of bias in included studies

We will assess the risk of bias in all included studies using the criteria suggested in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

We will grade each study for risk of bias in the seven specific key domains suggested in the Cochrane Handbook for Systematic Reviews of Intervention:

1. sequence generation (selection bias);

2. allocation concealment (selection bias);

3. blinding of participants and personnel (performance bias);

4. blinding of outcome assessor (detection bias);

5. completeness of outcome data (attrition bias);

6. risk of selective data reporting (reporting bias);

7. risk of other bias.

We will make an overall judgement of a low risk of bias for a study when we consider that any plausible bias across all seven domains was unlikely to have altered the results. We will make an overall judgement of unclear risk of bias for a study when any plausible bias across one or more of the key domains raises some doubt that it may have altered the results. We will make an overall judgement of high risk of bias for a study when any plausible bias across one or more of the key domains seriously weakens our confidence in the results reported in that study.

Two review authors (AVK and SS) will conduct the assessment of risk of bias independently and in duplicate. A third review author (HE) will moderate any disagreement. A 'Risk of bias' table, as described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), will be presented for each included study. Our judgement of the risk of bias in studies will be presented graphically.

Measures of treatment effect

For dichotomous outcomes, we will express estimates of treatment effect as risk ratios (RRs) (e.g. the number of failed restorations in each group) together with 95% confidence intervals (CIs).

For continuous outcomes (such as mean VAS scores), we will use mean differences and 95% CIs.

We will use the methods for estimating missing standard deviations as described in Section 7.7.3 of the Cochrane Handbook for Systematic Reviews of Interventions, where appropriate (Higgins 2011).

Unit of analysis issues

It is anticipated that the trials included in this review may randomise either participants or teeth. In order to avoid unit of analysis errors, we will follow the guidance in Chapter 9 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Our final analysis will take into account the level at which randomisation occurred.

The number of observations in the analysis should match the number of 'units' that were randomised.

We will consider in each study:

• participants who were individually randomised to one of the interventions (e.g. single parallel‐group clinical trials);

• participants undergoing more than one intervention (e.g. split‐mouth designs).

Dealing with missing data

We will contact the authors of included trials to request any missing data, unclear data or for clarification of the trial methodology. We will exclude missing data from our analyses until further clarification is provided. We will consider and discuss the potential impact of missing data in the overall conclusions of the review. The analysis will include only the available data other than imputing missing standard deviations, as necessary. The methods used for this will be those outlined in Section 7.7.3 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) if appropriate.

Assessment of heterogeneity

A minimum of two review authors will assess the extent of clinical heterogeneity by examining the type of interventions and outcomes in each study.

We will also assess statistical heterogeneity by inspection of the point estimates and CIs on forest plots. The lack of overlap of CIs may indicate heterogeneity.

We will assess the variation in treatment effects by means of Cochran's Q test for heterogeneity and the I² statistic. We will consider heterogeneity to be statistically significant if the P value is < 0.1. A rough guide to the interpretation of the I² statistic given in the Cochrane Handbook for Systematic Reviews of Interventions is: 0 to 40% might not be important, 30 to 60% may represent moderate heterogeneity, 50 to 90% may represent substantial heterogeneity, 75 to 100% considerable heterogeneity (Higgins 2011).

When possible, we will use independent variables to subgroup outcome data in order to investigate possible the effects of heterogeneity.

Assessment of reporting biases

If there are more than 10 studies included in a meta‐analysis, we will consider using a funnel plot to explore potential publication bias.

Data synthesis

Using RevMan 5.3 software (RevMan 2014), we will conduct a meta‐analysis only if there are studies involving similar comparisons reporting the same outcome measures. We will combine RRs for dichotomous data and mean differences for continuous data using the random‐effects model. If there are three studies or fewer, we will use a fixed‐effect model.

If there are insufficient data, we will present a narrative synthesis.

Subgroup analysis and investigation of heterogeneity

If possible, we will analyse the data by:

• age of participants;

• adhesive systems/etching technique used;

• tooth preparation and surface treatment;

• duration of follow‐up;

• extent of lesion.

If sufficient data are available, we will attempt to classify participants' baseline sensitivity into degrees of severity so that the outcome can be assessed in relation to baseline severity. If there are insufficient data, we will pool the results together as one category.

Sensitivity analysis

Provided that a sufficient number of trials is included in the review, we will conduct sensitivity analyses to evaluate the impact of the following factors:

• studies at high or unclear risk of bias

• unpublished studies;

• studies with imputed missing data;

• use of fixed‐ or random‐effect model of analysis;

• impact of funding.