Types of studies

The review is confined to randomised controlled trials comparing manual and powered toothbrushes. It excludes trials only comparing different kinds of powered brushes or those comparing different kinds of manual brushes.

In the current update an agreement was made that cross‐over trials were eligible for inclusion if the wash‐out period length was more than two weeks. This was particularly important to diminish any carry‐over effects of the different toothbrushes on clinical gingivitis. Split‐mouth trials were excluded, as these were not considered representative of 'everyday use'.

Studies were included irrespective of publication status or language.

Types of participants

We included individuals of any age with no reported disability that might affect toothbrushing. We also included individuals wearing orthodontic appliances.

Types of interventions

The toothbrushes included in the review were all forms of manual brushes and all forms of powered brushes. Trials instituting combined interventions, e.g. brushing combined with the use of mouthrinse or irrigation, were excluded. However, trials where participants were permitted to continue with their usual adjuncts to oral hygiene, such as flossing, were included.

Trials were excluded where the brushing intervention was carried out or was supervised by a professional less than 28 days before a follow‐up assessment.

Trials of 28 days and over were eligible and a subgroup analysis was carried out on the duration of trials for the different outcome measures.

Powered toothbrushes were divided into seven groups according to their mode of action.

1. Side to side action, indicates a brush head action that moves laterally from side to side.

2. Counter oscillation, indicates a brush action in which adjacent tufts of bristles (usually six to 10 in number) rotate in one direction and then the other, independently. Each tuft rotating in the opposite direction to that adjacent to it.

3. Rotation oscillation, indicates a brush action in which the brush head rotates in one direction and then the other.

4. Circular, indicates a brush action in which the brush head rotates in one direction.

5. Ultrasonic, indicates a brush action where the bristles vibrate at ultrasonic frequencies (> 20 kHz).

6. Ionic, indicates a brush that aims to impart an electrical charge to the tooth surface with the intent of disrupting the attachment of dental plaque.

7. Unknown, indicates a brush action that the review authors have been unable to establish based on the trial report or confirm with the manufacturers.

An additional group was added in a parallel review of the effectiveness of different powered brushes (Deacon 2010). This 'multidimensional group' included brushes with two of the above action types. Due to the limited number of trials conducted using this brush type, they were considered as part of the rotation oscillation group in this update.

It was agreed from the earlier reviews that analysis of filament arrangement, orientation, size, shape and flexibility, brush head size and shape along with presence or absence and characteristics of a timer would prove difficult to define across time and brush types.

Types of outcome measures

Electronic searches

We searched the following electronic databases:

• the Cochrane Oral Health Group's Trials Register (to 23 January 2014) (Appendix 1);

• the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2014, Issue 1) (Appendix 2);

• MEDLINE via OVID (1946 to 23 January 2014) (Appendix 3);

• EMBASE via OVID (1980 to 23 January 2014) (Appendix 4);

• CINAHL via EBSCO (1980 to 23 January 2014) (Appendix 5).

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

Searching other resources

We searched the following databases for ongoing trials, seeAppendix 6 for details of the search strategy:

• US National Institutes of Health Trials Register (http://clinicaltrials.gov) (to 23 January 2014);

• the WHO Clinical Trials Registry Platform (http://apps.who.int/trialsearch/default.aspx) (to 23 January 2014).

Only handsearching done as part of the Cochrane Worldwide Handsearching Programme and uploaded to CENTRAL was included (see the Cochrane Masterlist for details of journal issues searched to date).

All references cited in the included trials were checked for additional studies. Identified manufacturers were contacted and additional published or unpublished trial reports requested.

Selection of studies

Two review authors independently reviewed the titles and abstracts identified in the search. If in the opinion of both authors an article clearly did not fulfil the defined inclusion criteria it was considered ineligible. We obtained full reports of all trials of possible relevance for assessment. On receipt of the full article, two review authors assessed each study independently using specifically designed data extraction forms. Disagreements were resolved by discussion with the review team.

Data extraction and management

For this update, piloting of data extraction was performed independently by two authors on eight pilot articles. However, all authors reported back on the design of the data extraction forms and their interpretation of the inclusion and exclusion criteria along with their understanding of the outcome measures and new risk of bias (ROB) assessment. On the basis of this feedback the data extraction forms were altered and the inclusion, exclusion, outcome measures and ROB assessment were redefined to avoid misinterpretation. All data extraction for the included studies was then undertaken independently and in duplicate.

The final data extraction protocol considered the following information.

1. Bibliographic details of the study.

2. Funding source for the trial. A trial was considered to have been funded by a brush manufacturer if it was reported that any material sponsorship from the manufacturer occurred, including the donation of brushes. It was considered unclear, if there was no statement on funding. A trial was only considered to be unsponsored by a manufacturer if it clearly stated so.

3. Inclusion eligibility.

4. Baseline characteristics of the participants in the study, including age, number of participants in the study and gender. Also, specific groups, such as dental students or orthodontic patients were noted, where mentioned.

5. Intervention characteristics including type of brush and its mode of action, duration of use and delivery of instructions.

6. Outcomes including plaque and gingivitis indices.

7. Additional information on a priori calculation of sample size, duration of study, reliability and validity of outcomes measures and monitoring of compliance.

Trials were considered as 'short term' or 'long term'. 'Short‐term' data included follow‐up between 28 days and three months. 'Long‐term' data included follow‐up beyond three months. Within each category of long term and short term, where a trial reported multiple end points, only the latest data were extracted.

Data from trials that reported follow‐up before and after three months were included in the short‐ and long‐term meta‐analyses. Likewise, data from trials that reported both plaque and gingivitis would be included in meta‐analyses for both outcomes. These were the only circumstances when data from the same trial were considered more than once.

Many different indices of plaque and gingivitis were used across trials and some trials reported multiple indices. A frequencies table was prepared of the indices used and they were ranked based on common usage and simplicity. For plaque we extracted, where possible, data reported as the Turesky modification of the Quigley‐Hein plaque index (Quigley 1962; Turesky 1970). For gingival inflammation we extracted where possible data reported as the gingival index of Löe and Silness (Löe 1963) or, if unavailable, bleeding on probing (Ainamo 1975). Data for Russell's periodontal index were excluded because this index fails to distinguish between gingivitis and periodontitis (Russell 1967).

Where available, data were extracted for whole as opposed to part‐mouth scores. Where only part‐mouth scores were reported in a study, they were extracted and a sensitivity analysis carried out to consider their impact on the results of the review. Part‐mouth scoring was said to have occurred if plaque or gingivitis or both were not recorded around all erupted teeth, except third molars.

Completed data extraction forms were compared. Where there was disagreement between review authors with regard to any part of the extraction details it was resolved by discussion between the authors and a note made on the data collection forms. Any disagreement, unresolved between the two authors, was settled by majority vote of the entire panel of review authors. Authors were contacted for clarification where necessary.

Assessment of risk of bias in included studies

We conducted this assessment using the recommended approach for assessing risk of bias in included studies for Cochrane reviews (Higgins 2011). All included studies were assessed independently and in duplicate by two review authors as part of the data extraction process. The risk of bias tool evaluates six specific domains.

• Sequence generation (selection bias).

• Allocation concealment (selection bias).

• Blinding of outcome assessment (detection bias).

• Incomplete outcome data (attrition bias).

• Selective outcome reporting (reporting bias).

• Other sources of bias; comparability of groups at baseline.

Risk of bias assessment.

• A trial was considered to have adequately generated a random sequence of allocation, if it fully reported the type of allocation generation and it satisfied the CONSORT guidelines as true randomisation (http://www.consort‐statement.org/).

• A trial was considered to have adequate blinding, if the report indicated that the method of outcome assessment did not allow the recording clinician to know to which group the participants had been allocated, with no other contradicting statement.

• Attrition was considered to have been adequately reported if there was a clear indication of how many withdrawals occurred in each group during the trial and an attempt made to give reasons why the withdrawals occurred.

The first part of the entry involved authors' describing what was reported in the study. The second part involved the authors' judgements of the adequacy of the study, that is, whether they are at low, high or unclear risk of bias. Numerical data extracted from the included trials were checked by a third author for accuracy and entered into Review Manager (RevMan) software (RevMan 2012).

Two risk of bias figures were generated to illustrate the findings of the assessment. A 'Risk of bias graph' illustrated the proportion of studies across the domain with each of the judgements ('low risk', 'high risk', 'unclear risk'). A 'Risk of bias summary' summarised all of the judgements for a study entry. We assumed that the risk of bias of outcomes was equally important both within and across studies. They were assessed as follows.

Measures of treatment effect

The estimate of effect used was the mean difference (MD) and corresponding 95% confidence intervals (CI). However, different indices for plaque measure the same concept on different scales, with high correlation between the different indices. The same is true for gingivitis. As it is not possible to combine the results from different indices, the effects were expressed as standardised values, which have no units, before combining. The standardised mean difference (SMD) was therefore calculated along with the appropriate 95% CI and was used as the effect measure for each meta‐analysis where results were available for more than one index (Deeks 2001). Where only one index was presented in a comparison, the treatment effect was measured as the MD with 95% CI.

Unit of analysis issues

No units of analysis issues were anticipated other than cross‐over studies which were included using the generic inverse variance (GIV) approach (Elbourne 2002; Higgins 2011).

Dealing with missing data

Trial authors were contacted to retrieve missing data where necessary. Data remain excluded until further clarification becomes available. Standard deviations were imputed as in section 7.7.3 of the Cochrane Handbook for Systematic Reviews of interventions (Higgins 2011).

Assessment of heterogeneity

We assessed heterogeneity by inspection of a graphical display of the estimated treatment effects from the trials along with their 95% CI and by Cochran's test for heterogeneity undertaken before each meta‐analysis as described in the Cochrane Handbook for Systematic Reviews of interventions (Higgins 2011). The heterogeneity was quantified using the I² statistic, where a guide for interpretation in the Cochrane Handbook for Systematic Reviews of interventions is (Higgins 2011):

• 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.

Assessment of reporting biases

A funnel plot (plots of effect estimates versus the inverse of their standard errors) was drawn. 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 performed using the method proposed by Egger et al (Egger 1997). This was carried out using Stata version 12.0 (Stata Corporation, USA) using the program Metabias.

Data synthesis

Statistical values such as SMD have no inherent clinical meaning. Therefore we back‐translated key effect scores using the clinical indices from a study where the difference was similar to the SMD. Such examples are given in the Discussion. Random‐effects models were performed where four or more studies were to be combined, otherwise fixed‐effect models were used.

Data from cross‐over trials were included with that of similar parallel group trials, using the techniques described by Elbourne and colleagues (Elbourne 2002). This was done using the generic inverse variance method within RevMan (Higgins 2011).

Subgroup analysis and investigation of heterogeneity

Subgroup analyses were undertaken for assessments based on full mouth recording versus those based on a partial recording and to examine the effects of concealed allocation, randomisation generation and blinded outcome assessment on the overall estimates of effect for important outcomes.

Additional subgroup analyses were undertaken to explore heterogeneity. Evidence of variability in any subgroup was further explored by examining funnel plots.

Sensitivity analysis

Sensitivity analyses were conducted to test whether the assumptions involved in the design of this review affected the findings. These analyses were undertaken by repeating the meta‐analyses in the following cases: where a full mouth index had been used, where adequate concealment of randomisation occurred, where there was adequate generation of randomisation sequence, where there was blinding of the outcome assessor, if the trial was funded by a manufacturer, with adequate information about attrition and for trials that were not restricted to participants only wearing fixed orthodontic appliances.