Summary of main results

There is insufficient evidence of a difference from two trials, judged to be at a low risk of bias, on the effectiveness of fluoride varnish applied, by a professional, every six weeks in reducing new demineralised lesions (DLs) during fixed orthodontic treatment (summary of findings Table for the main comparison). Evidence from one trial judged to be at unclear risk of bias, provides a low level of certainty that fluoride foam (12,300 parts per million (ppm) fluoride), professionally applied every two months, reduces the incidence of new DLs after fixed orthodontic treatment (summary of findings Table 2). Similarly, one trial, assessed as at unclear risk of bias, provides a low level of certainty that use of a high‐concentration fluoride toothpaste by patients throughout their fixed orthodontic treatment reduces the incidence of new DLs, compared with a conventional concentration of fluoride toothpaste (summary of findings Table 3).

Evidence is insufficient to show whether the use of a professionally‐applied amine fluoride gel reduces new DLs (Additional Table 1), or patient use of a sodium fluoride/amine fluoride mouthrinse (summary of findings Table 4) or amine fluoride and stannous fluoride toothpaste/mouthrinse combination (Additional Table 2) is more or less effective than a non‐fluoride mouthrinse or a sodium fluoride toothpaste/mouthrinse combination respectively. Finally, there is no evidence that either a fluoride‐containing material to bond brackets (summary of findings Table 5) or an intraoral fluoride‐releasing glass bead device (Additional Table 3) reduces the incidence of new DLs.

Overall completeness and applicability of evidence

As with the last version, the update of this review has included only parallel‐group trials, in which the individual participant is the unit of randomisation. This was decided upon because of the possibility of cross‐contamination between experimental and control teeth in the same mouth, either between upper and lower arches or between sides of the mouth, which might lead to under‐estimation of the effectiveness of any fluoride products.

Interventions that rely on the patient for delivery, including fluoride mouthrinse and toothpaste, will work only if they are used regularly. They rely greatly on patient compliance to succeed; however, evidence suggests that compliance with mouthrinsing is poor among orthodontic patients. One study (Geiger 1992) found that only 42% of participants rinsed with a sodium fluoride mouthrinse at least every other day. Results also showed that those who complied least with fluoride rinsing regimens tended to have more DLs. It is important to consider the acceptability of interventions to both adolescents and adults with a view toward increasing compliance with recommended dental hygiene practices.

Interventions that are professionally applied and deliver fluoride 'passively', such as fluoride varnish, fluoride‐releasing bracket cements and fluoride‐releasing elastics, avoid the need for patient compliance. In addition, these materials deliver fluoride close to the bracket, where it is most needed. Many fluoridated materials release large amounts of fluoride initially, but the level drops rapidly and might not be sufficient to prevent decay over the whole course of orthodontic treatment. Reapplication of fluoride varnish and frequent replacement of fluoride‐releasing elastics are likely to be required. In the parallel‐group trial of a fluoride varnish intervention included in this review, varnish was reapplied every six weeks at each orthodontic check‐up appointment. We found no parallel‐group trials of fluoride‐releasing cements or elastics that met the inclusion criteria for this review.

An interesting addition, since the initial review was carried out, is the further development of materials that produce a slow and sustained release of fluoride (Luther 2005). This trial was small and at high risk of bias, and evidence was insufficient to reveal whether these devices are more or less effective than a mouthrinse in reducing the development of DLs. It is possible, that with further refinement, this technique could potentially be effective. Intraoral fluoride‐releasing devices should be evaluated by parallel‐group randomised controlled trials with appropriate masking of participants, clinicians/operators and assessors.

When examining the effectiveness of a fluoride product in preventing dental decay, one should consider two aspects: first, whether the fluoride product reduces the number of DLs appearing during treatment, and second, whether it reduces the severity of DLs in terms of the size or area of the tooth surface affected, the amount of mineral lost or the depth of decay. Banks et al (Banks 2000) developed the Enamel Decalcification Index, which is an ordinal index that includes an assessment of the area covered. Assessment of the size of the lesion is a useful outcome measure, but none of the studies included in this review reported this outcome.

Ideally the appearance of the tooth should be recorded before and after orthodontic treatment, so that the change in appearance of the tooth is measured (incidence), not just its appearance at the end (prevalence). There are many different causes of coloured marks on the teeth, many of which occur during their development. It is important that these development lesions, as well as decay that has occurred before the brace is fitted, are excluded from the analysis, hence the need for the clinical photographs or fluorescent images taken before treatment. Measurement of both incidence and severity will depend on the method used to record DLs. Two main methods may be used: visual inspection and clinical images. Both methods are associated with problems. One problem with visual inspection is that the examiner or examiners will require calibration at the start and regular recalibration throughout the experimental period to ensure consistency of measurement. The duration of the experiment, including the recruitment and data collection will be long because, as discussed later, the product should be tested over the entire length of orthodontic treatment. This can take between 18 and 30 months ‐ sometimes longer. Another problem with visual recording clinically involves masking of the assessor to the allocated intervention. To reduce bias, the examiner should not know whether the participant has received a fluoride product and this will complicate the way the experiment is run.

Images have the advantage of providing a permanent record of the appearance of the tooth. Assessments can be carried out by several people independently or in groups, whereby a consensus is achieved. The images can be placed in a random order and the judges masked to group allocation. In addition, because the assessment can be performed over a short period of time the problem of examiner drift, whereby an assessor might subtly change his or her assessment over time, is reduced. The challenge of using clinical photographs consists of achieving consistency in lighting and reducing reflections that can mask or mimic DLs. When a careful photographic technique is applied, however, the advantages of photographs outweigh their potential disadvantages. Several optical and fluorescent methods are available for measuring lesions on the teeth (Angmar‐Mansson 1996). These methods require specialised equipment, which would add considerably to the cost of a clinical study, but they provide an objective measurement of the amount of decay in terms of mineral loss or lesion depth or both.

Quality of the evidence

Two studies included in this review were judged to be at low risk of bias in all domains; however clarification over some issues was required from some of the authors before these judgements could be confirmed. Both the design and the reporting of trials of fluorides for preventing DLs have improved since the first version of this review and we hope that this may continue. Undoubtedly the increasing use of the CONSORT guidelines for the reporting of randomised controlled trials has been a factor.

Several of the studies were assessed at some risk of bias as a result of the proportion of participants randomised to the intervention not been included in the analysis following withdrawal or dropout. It is inevitable that some patients will move away during a course of orthodontic treatment, but every effort should be made to collect records and data at the appropriate times, even if the treatment is finished early.

Another issue in orthodontic research is that it is often not possible to mask the clinician/operator to the allocated intervention. In these situations the participant can also usually work out to which group they have been allocated and, therefore, triple‐blinding is difficult, if not impossible to achieve. It is, however, not clear how much this will affect the behaviour of the clinician and the participant. It is known that patients respond favourably to simply being part of a clinical trial, irrespective of whether they have received the active intervention or the control (McCarney 2007). This is the so‐called Hawthorne effect, named after a Chicago electrical company works, whose workers were subjected to a series of experiments during the 1920s and 1930s. The investigators noticed that no matter how they changed the working conditions of the workers their productivity improved and concluded it was because they knew they were part of an experiment. We believe that it would be harsh to judge all trials that are not triple‐blind as a potentially high risk of bias when the effects are not known and perhaps this should be an area of research in the future.

Potential biases in the review process

We undertook a sensitive search of several electronic sources, supplemented by searches of references lists. We placed no restriction on language or publication status. The review authors have tried, as far as possible, to identify all possible studies that might meet the inclusion criteria for this review. Study authors have been contacted, and many have replied; however, some were not able to supply the requested information, as their records have been destroyed or lost.

When a product, such as a bonding material, can be applied to single teeth, it is tempting to use an experimental design whereby the material being tested is used in two quadrants of the mouth and the control material is used in the other two quadrants. This is called a split‐mouth design. The main advantage of the split‐mouth design over a conventional parallel‐group study design, in which the two materials are tested in two separate groups of individuals, is that the experimental material is tested in the same mouth, under the same conditions as the control material. In theory, any differences in outcome between the two materials are due only to their properties ‐ not to other factors, such as differences in oral hygiene and diet between participants (with a parallel design) or even differences in oral hygiene and diet over time within the same participants (with a cross‐over design).

Unfortunately, when one is examining the ability of fluoride products to reduce decay, it is highly unlikely that the fluoride released will be confined to only the quadrants/teeth in which the experimental material has been placed, and some contamination of the 'untreated' teeth is inevitable. This contamination will reduce the difference in outcomes between treated and untreated teeth. The previous version of this review included split‐mouth studies, which failed to show any difference between treated and untreated teeth; this may be due to cross‐over contamination between control and experimental sides and may reflect our contention of contamination. For this reason, we have decided to exclude split‐mouth studies from the previous update of our review.

Agreements and disagreements with other studies or reviews

Three other systematic reviews gathering evidence for the most effective means of preventing caries/demineralisation during fixed orthodontic appliance treatment have been reported in the literature. Derks et al (Derks 2004) examined all preventive measures for preventing demineralisation ‐ not just fluoride products. These review authors had to exclude many published studies as well, because of inappropriate research design or poor reporting and were unable to provide firm, evidence‐based recommendations as to the prevention of DLs during fixed orthodontic treatment.

A second systematic review (Chadwick 2005) investigated the effectiveness of topical fluorides used alone in preventing demineralisation during orthodontic treatment. These review authors included seven studies in their review; these studies, however, were excluded from our review, because the outcomes were not appropriate (DMFT/DMFS (decayed, missing, and filled teeth/surfaces)), or the participants were not examined immediately after removal of the fixed appliance(s). Although they suggest that according to their outcome measure (preventive fraction), some evidence shows that the addition of a topical fluoride preparation helps in the prevention of demineralisation during fixed orthodontic treatment, this conclusion must be viewed with caution, because these review authors were not able to calculate confidence intervals. We support their request that researchers design and report their studies using standard outcomes, so that in the future, data may be pooled and overall recommendations on preventive measures may be provided.

Rogers et al (Rogers 2010) included 10 studies in their systematic review investigating the effectiveness of fluoride‐containing bonding adhesives used in orthodontics to prevent demineralisation. Five of these studies were excluded from our review because they were not randomised, and a further three studies were excluded because data in the report were insufficient, and the study authors, when contacted, were unable to provide requested data. Rogers' conclusions are consistent with ours with regard to the design of trials and the quality of reporting and statistical analyses.

Wang 2013 included 20 studies (19 articles) in their review, which is written in Chinese and only includes data from Chinese studies. They included studies with quasi‐randomised designs and it is not clear if they included studies with a within‐person (split‐mouth) design and those not following patients for the full length of orthodontic treatment.

The following reviews were also examined: Höchli 2017; Lapenaite 2016; Lopatiene 2016; Nascimento 2016; Rahimi 2017; Sardana 2019a; Sardana 2019b; Tasios 2019. These reviews either investigated interventions other than fluoride to prevent DLs, such as casein phosphopeptide‐amorphous calcium phosphate (CCP‐ACP), treatment of DLs after the appliance was removed or included quasi‐ and non‐randomised studies, studies with short follow‐up or outcomes that are not particularly relevant to patients and clinicians. No new studies were identified from the references in these articles to include in this review.