The efficacy of powered toothbrushes: A systematic review and network meta‐analysis

Abstract Aim This systematic review and network meta‐analysis synthesizes the available clinical evidence concerning efficacy with respect to plaque scores following a brushing action with oscillating‐rotating (OR) or high‐frequency sonic (HFS) powered toothbrushes (PTB) compared with a manual toothbrush (MTB) as control. Material and methods Databases were searched up to 1 August 2021, for clinical trials that evaluated the efficacy of a PTB with OR or HFS technology compared with an MTB on plaque removal after a single‐brushing action and conducted with healthy adult patients. Meta‐analysis (MA) and a network meta‐analysis (NMA) were performed. Results Twenty‐eight eligible publications, including 56 relevant comparisons, were retrieved. The overall NMA results for the mean post‐brushing score showed a statistically significant difference for the comparison between an OR PTB and an MTB (SMD = −0.43; 95% CI [−0.696;−0.171]). The change in plaque score data showed a significant effect of a PTB over an MTB and OR over HFS. Based on ranking, the OR PTB was highest, followed by the HFS PTB and the MTB. Conclusion Within the limitations of the present study design, based on the outcome following a single‐brushing action, it can be concluded that for dental plaque removal, there is a high certainty for a small effect of a PTB over an MTB. This supports the recommendation to use a powered toothbrush for daily plaque removal. There is moderate certainty for a very small benefit for the use of a powered toothbrush with an OR over an HFS mode of action.

manual toothbrush (MTB) is a relatively simple device that is widely accepted and affordable to most people. 5 Powered toothbrushes (PTB) were first introduced in the 1940s, starting with devices with a circular brush head and a straight brush head. The first generation of electric toothbrushes was essentially mechanized versions of manual toothbrushes, with the bristles moving back and forth in an imitation of how people brush by hand. Over the years, design changes have increased the efficacy of powered toothbrushes in plaque removal, including improved brush head and filament arrangement architecture, 9,10 increased motion 11 and compliance-enhancing features. 12 Modern innovations such as mobile apps give people easy access to oral health knowledge, further improving levels of oral hygiene. 13 These developments over the years have resulted in various types of PTBs with different brush head configurations and modes of action. Currently, the PTBs with oscillating-rotating (OR) and high-frequency sonic (HFS) technology are the most common commercially available products on the market globally. 14 Various systematic reviews have evaluated the efficacy of MTBs and PTBs. In general, they conclude that PTBs are more effective than MTBs in reducing dental plaque, gingivitis and bleeding. 15 Previously, it has been shown that following brushing with an MTB, an average 42% plaque score reduction can be expected. 16 A similar review 14 that evaluated PTBs found an average plaque score reduction of 46%. A more recent review comparing the PTB and the MTB concluded that there is moderate certainty that a PTB is more effective than an MTB for plaque removal. 17 The most recent systematic review evaluated OR and HFS PTBs in particular and concluded that there is moderate certainty of a significant but very small beneficial effect in favour of OR. 18 These studies all evaluate the effect following a single-brushing action and do not consider the benefits of gingival health. They are, however, appropriate for estimating the potential plaque removal, 19 as they facilitate the control of confounding variables such as patient compliance. 20 A Cochrane systematic review (SR) from almost a decade ago involved a direct comparison between PTBs with different modes of action on plaque score reduction. 21 At that time, no definitive conclusions could be drawn regarding the superiority of one particular type of PTB over another. However, some evidence showed that the OR PTBs reduce plaque more than HFS PTBs in the short term. 21 A recently updated SR using the same methodology concluded that the evidence does not suggest the superiority of either OR or HFS PTBs for reducing plaque or gingivitis scores. 22 Based on studies conducted over the last decade, a recent SR compared the efficacy of OR and other PTBs and concluded that there is evidence to suggest that OR is more efficient in plaque removal and reduction in the number of bleeding sites than other PTBs, including HFS. 23 It is the dental care professional (DCP)'s role to provide oral hygiene advice to their patients based on the best available evidence. 24 Thus, evidence-based findings concerning the mechanical plaque removal of a toothbrush have to be established and made readily available. In this respect, an NMA combines both indirect and direct evidence providing the most precise estimate of treatment effects to support decision-making. [25][26][27] The simultaneous comparison of all interventions of interest in the same analysis enables the estimation of their relative ranking for a given outcome. Recently, an NMA 28 was published that only evaluated studies obtained from the database of a PTB manufacturer covering the period 2007-2017. It solely addressed RCTs with a duration up to three months. As the studies were retrieved from a non-public archive, with in addition a non-transparent search strategy and extraction of data, this review had considerable limitations. Therefore, the purpose of the present study is to systematically evaluate the available clinical evidence concerning efficacy with respect to plaque scores following a brushing action with OR or HFS PTBs compared with an MTB as control and synthesize this with an NMA.

| MATERIAL S AND ME THODS
This review is prepared and reported in accordance with the Cochrane Handbook for Systematic Reviews of Interventions. 29 In addition, the guidelines of Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 30 were followed and the consequent extensions for abstracts 31,32 and Network Meta-Analyses (NMA). 33 The protocol that details the review method was developed a priori after an initial discussion among the members of the research team. The review is registered under number CRD42020192418 with the International Prospective Register of Systematic Reviews (PROSPERO). 34 The study was also approved by the Medical Ethical Committee of the Academic Center for Dentistry Amsterdam (ACTA-ETC), the Netherlands, under number: 2021-1835.

| Focussed PICOS question
The PICOS question 35 for this review is: Based on (randomized) controlled clinical trials, what is the efficacy of a PTB with an OR or a HFS technology as compared to an MTB on dental plaque removal following a single-brushing action in healthy participants?

| Search strategy
A structured and comprehensive search strategy was designed to retrieve all relevant publications that satisfied the study purpose with a direct comparison between: • MTB and OR • MTB and HFS

• HFS and OR
Electronic databases were searched for relevant papers.

| Screening and selection
Titles and abstracts of the studies obtained from the searches were screened in detail for suitability by two reviewers (TMJAT, DES) using the Rayyan 36 web application. The reviewers worked independently and were blinded for each other's results during the screening process. Possible duplicates were identified and checked to eliminate those that were identical. Disagreements in the screening and selection process were resolved by consensus or, if disagreement persisted, by arbitration through a third reviewer (GAW) until consensus was reached. The papers that fulfilled all of the inclusion criteria were processed for data extraction.
Studies were deemed eligible for inclusion if they conformed to the following criteria: • Publications written in the English language  • Navy plaque index 40 or Rustogi modified Navy plaque index (RMNPI) 41 • Plaque score data available as mean and standard deviation (SD) for pre-and post-brushing and/or incremental plaque score reduction.

| Heterogeneity assessment
Across the studies, the factors used to evaluate the clinical and methodological heterogeneity of the characteristics of the different studies were as follows: study design and evaluation period, subject characteristics, brushing regimen, technology of mode of action, instructions given and plaque indices or their modifications.
As part of the NMA, heterogeneity was statistically tested by the I 2 statistic, which describes the percentage of variation across studies due to heterogeneity rather than chance. As a rough guide, I 2 was interpreted as follows: an I 2 of 0%-40% may indicate unimportant levels of heterogeneity; an I 2 of 30%-60% may represent moderate heterogeneity; an I 2 of 50%-90% may represent substantial heterogeneity; and an I 2 greater than 75% may indicate considerable heterogeneity. 42

| Risk of bias and (in)directness
Two reviewers (TMJAT and DES) individually scored the methodological qualities of the included studies according to the method described by Van der Weijden et al., 43 and in greater detail by Keukenmeester et al. 44 In summary, the study was classified as having an estimated 'low risk of bias' when random allocation, defined eligibility criteria, masking of examiners, masking of patients, balanced experimental groups, identical treatment between groups (except for the intervention) and reporting of follow-up were present. The study was considered to have an estimated 'moderate risk of bias' when one of these seven criteria was missing. When two or more of these criteria were missing, the study was estimated to have a 'high risk of bias'. The potential risk of bias was estimated, and the acquired evidence was graded.
For the assessment of indirectness in the context of the NMA two components were considered: the similarity of the studies in the analysis to the target PICO-question 35 (i.e. the extent to which the evidence relates to the population, intervention(s), comparisons and outcomes of interest); and the transitivity assumption, which is the comparison between two treatments via a third one. 45 For the present review, risk of bias and the assessment of indirectness were checked for each included study by two reviewers (TMJAT and DES). If disagreements in the quality assessments were found, this was resolved by consensus after discussion. TA B L E 1 Search terms used for MEDLINE-PubMed and Cochrane-CENTRAL. The search strategy was customized according to the database being searched. The following strategy was used in the search {(<intervention AND outcome>)} {<[(MeSH terms) Toothbrushing OR (text words) toothbrush OR toothbrushing OR toothbrush*> AND <(MeSH terms) dental plaque OR dental plaque index OR dental deposits OR [text words] plaque OR dental plaque OR plaque removal OR plaque index OR dental plaque removal OR dental deposit* OR dental deposits* OR dental deposit OR dental deposits>} Note: The asterisk (*) was used as a truncation symbol.

| Data extraction
The data from the publications that met the selection criteria were extracted and processed for further analysis. Custom-designed data extraction forms were used by two independent reviewers (TMJAT and DES) for mean pre-and post-brushing and incremental plaque score data and SD. If studies provided a standard error (SE) of the mean, these values were converted to SD based on the sample size (SE = SD/√N).
In all cases, to ensure an accurate estimate, any data approximation in figures was avoided. In case of missing data or undetermined information, attempts were made to contact the first or corresponding author of the included publications for clarification or to retrieve additional data. For studies with multiple treatment arms, and for those in which data from the control group were compared with more than one other group, the number of participants (N) in the control group was divided by the number of comparisons. 46 Disagreements in the data extraction were resolved by discussion and consensus.

Network meta-analysis
From the selected papers, the mean plaque scores, the standard deviations and the number of participants per group for the consequent plaque scores were used for the NMA. The NMA was performed using MetaInsight 47 (with either the fixed-or random-effects model, as appropriate). MetaInsight 47 is an interactive web-based tool for analysing, interrogating, and visualizing network meta-analyses using R-shiny and netmeta (see: https://crsu.shiny apps.io/MetaI nsigh t/). 48 Irrespective of which plaque index score the data were related to, the overall effect size analysis was calculated as the standardized mean difference (SMD). The difference of means (DiffM) was used for the sub-analysis per plaque index score. The 95% confidence interval (CI) is presented for both the SMD and the DiffM. The NMA was performed on overall data as well as for the direct and indirect comparison. In addition, the 95% prediction interval (PI) was calculated.
Treatments were ranked based on the NMA and ranking was performed by P-scores. The P-scores are based solely on the point estimates and standard errors of the network estimates. They measure the extent of certainty that one treatment is better than another, averaged over all the competing treatments. 49 This interpretation is comparable to that of the surface under the cumulative ranking curve (SUCRA), which is the rank of a treatment within the range of treatments, measured on a scale from 0 (worst) to 1 (best). 50

| Evidence profile
The Grading of Recommendations Assessment, Development and Evaluation (GRADE) system, as proposed by the GRADE working group, 51 was used to rank the body of evidence emerging from this review. Two reviewers (TMJAT and DES) used the GRADE 51 approach and the Confidence in Network Meta-Analysis Software (CINeMA) 52,53 to evaluate the strength of evidence for results at the end of treatment from the NMA. CINeMA is a web application that simplifies the evaluation of confidence in the findings from an NMA and is based on the framework developed by Salanti et al. 25 and refined by Nikolakopoulou et al. 53 It is a single page application that communicates to an R back-end server. In particular, the package's meta and netmeta are used. 54 The CINeMA platform provides a transparent framework to evaluate evidence from systematic reviews with multiple interventions. 25 Six domains that affect the level of confidence in the NMA results are considered: (i) within-study bias, (ii) across-studies bias, (iii) indirectness, (iv) imprecision, (v) heterogeneity and (vi) incoherence. Any disagreement between the two reviewers was resolved after additional discussion with the third reviewer (GAW).

| Search and selection results
A search of the MEDLINE-PubMed and Cochrane-CENTRAL databases identified 4467 unique papers (for details, see Figure 1).
Screening the titles and abstracts resulted in 100 papers, which were obtained in full text. After careful, extensive and detailed reading, 72 papers were excluded (for details, see online Appendix S1).
This resulted in 28 9,28,55-78 papers for inclusion in this review, describing in total 56 comparisons. Of these, 34 comparisons used the Q&HPI 37 or a modification, 38,39 and 22 used the RMNPI. 40,41 In total, 25 compared the MTB to OR PTBs, nine compared the MTB to HFS PTBs, and 22 compared the HFS to OR PTBs.

| Study characteristics
All but one 70 of the 28 selected studies were RCTs. The characteristics of each study are displayed in online appendix Table S2, and the extracted data are presented by means and SD separately per plaque index of interest (online appendix, Table S3a-b). The number of participants varied from 12 to 216 per group, and the age ranged provided visual instructions to their participants and one study 70 did not give any instruction (for details, see online appendix Table S2).

| Risk of bias and (in)directness
A summary of the evaluation of the risk of bias in the individual studies is shown in Table S3a-b in the online appendix, along with the indirectness assessment. Of the 28 selected studies, 24 were estimated to have a low risk of bias, three a moderate risk and one a high risk (online appendix, Table S4). The indirectness was scored low for 17 and moderate for 11 comparisons.

| Meta-analysis
At pre-brushing, no statistically significant difference between the comparisons was found (OR vs. MTB, SMD = 0.10; 95% CI (0.00;0.19); HFS vs. MTB, SMD = 0.08; 95% CI (−0.04;0.19)) (for details, see online appendix Tables S5, S6 and S7). Tables 2 and 3 present an overview of the outcomes of the performed NMA on post-brushing plaque scores, and Tables 4 and 5 show the incremental reduction in plaque scores. An overall analysis and a sub-analysis per plaque index score are presented. The corresponding forest plots are displayed in the online appendices (S8 through S13).
Only for the post-brushing scores of the comparison between OR and MTB did the overall NMA show a statistically significant difference (SMD = −0.43; 95% CI (−0.696;−0.171)) ( Table 2). The sub-analysis (Table 3)  Analysis of the outcomes on incremental plaque score data (Table 4) showed a significant reduction for all three comparisons of interest. In detail, the two types of PTBs present both a significant difference compared to an MTB, and OR shows a significant difference in comparison with HFS. The sub-analysis per plaque index score (Table 5) shows that only when (M)Q&HPI 37-39 was used could no difference be found between the PTB technologies (DiffM = 0.04; 95% CI (−0.068;0.141)).

| Heterogeneity assessment
The studies included in the NMA showed considerable heterogeneity, 29 as I 2 statistic values range from 76.3% and 94.7% (Tables 2 and   4 (Tables 3 and 5).

| Network meta-analysis graph
The graphs of the NMA 53 for post-brushing scores are shown in The different nodes represent a device, risk of bias and sample size.
The width of the edge represents the number of included comparisons and the indirectness (for details, see Appendices S14 and S15).

| Confidence in network meta-analysis
The six domains that affect the level of confidence in the NMA were estimated with CINeMA 53 (see Appendices S16-S21). Based on the main concerns regarding heterogeneity and incoherence, a moderate confidence rating of the overall NMA for both post-brushing and incremental reduction in plaque index scores was present.

| Ranking of interventions
When the toothbrushes are ranked based on the P-scores as result of the NMA by the program MetaInsight 47 , either based on postbrushing (Table 6) or incremental change (Table 7) in plaque scores, the OR ranks first and MTB last. The exception is the sub-analysis for the post-brushing scores when (M)Q&HPI 37-39 is used, where HFS ranks first. For details on ranking data, see online appendices S8-S13. Table 8 shows the evidence profile based on a summary of the various factors used to rate the quality of evidence and the level of certainty. It accumulates into an estimation of the strength and direction of the recommendation according to GRADE. 49 With respect to removal of dental plaque, there is a high certainty for a small effect, which supports the recommendation to advise using a PTB rather than an MTB. In addition, there is moderate certainty for a very small beneficial effect for the use of an OR mode of action PTB over a HFS PTB.

| Evidence profile
TA B L E 2 Meta-analysis for the standardized mean difference (SMD) evaluating efficacy of a manual toothbrush (MTB), an oscillating-rotating powered toothbrush (OR) and a high-frequency sonic powered toothbrush (HFS) using the MQ&HPI and the RMNPI. Overall results, independent of the plaque indices used. Post-brushing data

| Network meta-analysis
The aim of this study was to use an NMA to systematically compare and rank the effect of two different PTBs (OR and HFS technology) compared with an MTB with respect to plaque removal as evaluated following a single-brushing action research model. An NMA is a novel approach that takes the assumptions of a MA one step further. 84 An NMA 29 incorporates direct and indirect comparisons based on the principle of transitivity, which relies on the fact that combined studies have a common comparator. 85 The addition of indirect comparisons by incorporating evidence from other sources makes the results more robust. 86 When the network is well connected and provides both direct and indirect comparisons, these can be pooled together into 'mixed evidence', which increases statistical power and the precision of the estimates. 84 The use of added information also allows for more robust recommendations compared with conventional pairwise metaanalyses. 87

| Plaque indices
From previous reviews on toothbrushing efficacy that have evalu- Conversely, with an OR mode of action, most evaluations are based on the Q&HPI. [37][38][39] This phenomenon may contribute to reporting bias. In our earlier work, 14, 16 we have shown that the outcomes with TA B L E 3 Meta-analysis for the difference of means (DiffM) evaluating efficacy of a manual toothbrush (MTB), an oscillating-rotating power toothbrush (OR) and a high-frequency sonic powered toothbrush (HFS) using the MQ&HPI and the RMNPI. Sub-analysis per index. Post-brushing data    the Q&HPI 37-39 result in a smaller plaque score reduction compared with the RMNPI. 40,41 This effectively contributes to a wider CI, which is evident from the data presented in Tables 2-5.

| Oral hygiene instruction
If optimal results are to be achieved with a toothbrush, professional instruction and reinforcement are needed. 90

| Study design
The study design that evaluates a single-brushing action provides an assessment under ideal conditions in which all participants comply with the use of the device to which they are randomly assigned. 16 Although the design is clearly restricted to an instant evaluation of one-time brushing action under controlled circumstances, when data indicate that a specific toothbrush shows a greater potential in reducing plaque scores, it can be supposed that it offers improved plaque control over time. Consequently, it may also have long-term benefits for gingival health. 16 The American Dental Association (ADA), in their acceptance programme requirements for an ADA-seal, requests a minimum study duration of 30-days to show improved reduction in plaque and gingivitis scores. 92 Ideally, study should involve a combination of single-brushing designs and short-and long-term studies that establish robust evidence for a particular toothbrush. The clinical importance of the findings also deserves attention. In this respect, phase IV studies are needed to confirm the long-term clinical importance of PTB use and for safety surveillance. Studies extending over several years not only provide data related to the prevention of periodontal diseases but also to the prevention of caries. 93,94 The practical difficulty is that PTB manufacturers frequently change toothbrush design or technology, adjust brush heads and Therefore, it seems that PTB usage in the long run helps maintain the number of teeth in the oral cavity and reduces the progression of periodontal disease burden. 95

| Clinical relevance
As with an MA, an SMD as a summary of multiple plaque indices can be calculated with an NMA. Furthermore, a DiffM can be calculated for a specific plaque index. The results of the present NMA show larger differences between the indirect and direct evidence analysis when comparing the outcomes for the SMD with the DiffM. This is presumably due to the variation within and between indices and the subsequent SDs. Hence, it appears legitimate to perform the analysis separately per plaque index score and synthesize the data according to the DiffM. This also allows a direct interpretation of outcomes relative to its original scale, which is crucial for estimating and judging the clinical relevance of the observed difference. This also helps clinicians to interpret the scientific findings in their daily practice.
The contribution of indirect data to the overall NMA appears to be more pronounced for the post-brushing data than the data of the incremental change (see Tables 2-5). This may be because the incre-  First, the cost of a PTB is substantially more than an MTB and also comes with a variety of models and prices. Second, it needs to be ascertained how much the expected improvement in plaque removal and subsequent preventive effect of improved gingival health and reduced caries risk will cost. As periodontitis accounts for a considerable proportion of edentulism and masticatory dysfunction, it has a negative impact on general health and results in significant dental care costs. 101,102 Thus, indirectly, adequate plaque removal reduces the need for treatment and, consequently, toothbrushing presumably reduces oral healthcare costs in both the short and long term.
Based on this complex series of considerations and consequent calculations, VBOHC can be used to gain more insights into which oral health outcome can be obtained for a specific person by using a PTB.
This insight will answer the question raised earlier of whether the financial expense for a PTB is realistic and beneficial for everyone. it has been optimized, and this also applies to brush head design.

| Limitations and recommendations
-The new development of digital software to optimize patients' oral hygiene performance-such as timers, pressure sensors, apps, and artificial-intelligence brushing recognition and guidance-is not considered in this review.

| CON CLUS IONS
Within the limitations of the present study design, based on the outcome of single acts of brushing, it can be concluded that for dental plaque removal, there is a high certainty for a small effect of PTB Note: Strength and direction of the recommendation: With respect to removal of dental plaque, there is high certainty for a small effect of a PTB over an MTB which is in support of a recommendation to advice a PTB over an MTB. There is moderate certainty for a very small benefit for the use of an OR mode of action PTB over a HFS PTB. efficiency compared with an MTB. This supports the recommendation to advise using a PTB rather than an MTB. There is moderate certainty for a very small benefit in using an OR mode of action PTB rather than a HFS PTB.

| Scientific rationale for the study
Toothbrushing is considered the most efficient way to remove dental plaque and prevent periodontal diseases. At present, no network meta-analysis (NMA) of the available literature has been performed concerning the efficacy of different powered toothbrush technologies on plaque removal.

| Principal findings
The NMA demonstrated that an oscillating-rotating (OR) or a highfrequency sonic (HFS) powered toothbrush (PTB) is more effective than a manual toothbrush. When comparing the two PTB technologies, OR ranks higher than HFS.

| Practical implications
When recommending a toothbrush to a patient, a PTB is more effective than a manual toothbrush and should be considered the first choice. The clinical relevance of the very small but significant difference in favour of OR over HFS technology needs further appraisal based on long-term studies.

ACK N OWLED G EM ENTS
The authors gratefully acknowledge the support of Eveline van der Sluijs, Therese Elkerbout, Martijn Rosema and Nienke Hennequin-Hoenderdos, the Periodontal Prevention and Therapy research team at ACTA, in preparing this publication. Additionally, for their advice and consultation, we thank Dr. Cinzia del Giovane regarding the use of CINeMA and Prof. Dr. Alex Sutton for MetaInsight. We are also grateful to Joost Bouwman, the head librarian at ACTA, who helped to retrieve the full-text papers.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflicts of interest.

AUTH O R S CO NTR I B UTI O N S
All authors approved the final version of this manuscript before submission and agreed to be accountable for all aspects of the work ensuring that questions related to the accuracy or integrity of any part of the work were appropriately addressed and resolved. TMJAT contributed to design, search and selection, analysis and interpretation and drafted the manuscript, and DES contributed to conception and design, search and selection, analysis and interpretation and critically revised the manuscript. GAW contributed to conception and design, analysis and interpretation and critically revised the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data are derived from public domain resources. The data that support the findings (the included studies) of this study are available from search databases PubMed/Medline and Cochrane-CENTRAL.
These data were derived from resources available in original papers that are published in the public domain.