Protocol and registration

The review protocol was devised a priori and registered in the International Prospective Register of Systematic Reviews (PROSPERO; CRD42019137627).

Eligibility criteria

The PICO framework was:

Participants: Orthodontic patients undergoing fixed or removable orthodontic treatment, without age or gender restrictions.

Interventions:

Comparison: Either placebo or another treatment.

Outcomes: The primary outcomes were incidence of WSLs and their inhibition (arrest and/or reversal), as measured using either ICDAS or DMFT. The total number of carious lesions at the end compared to the start of the study was utilised to obtain a risk ratio.

The secondary outcomes included differences in lesion size between control and intervention groups. This was measured at the end of the study, using quantitative light induced fluorescence (QLF). Self-reports of adverse events, side-effects and tolerability were also recorded.

For binary outcomes (absence or presence of WSLs), the relative risk (RR) was computed at 95% confidence intervals (CI). For continuous data (change in mean scores of WSLs from start to end of the study), the mean difference was calculated at 95% CI.

Study design

English language RCTs in humans from 1970 to June 2019, with a duration of at least six months, were included. Exclusion criteria were non-randomised designs, laboratory-based (in vitro, in vivo and ex vivo), retrospective, split mouth studies and trials that did not record WSLs at the onset and at the end of treatment. Post-orthodontic WSL inhibition studies were also excluded.

Information sources and literature search

A systematic search of Embase, MEDLINE and Cochrane Central Register of Controlled Trials (CENTRAL) was performed. The search strategy is provided in Supplementary Table I. Unpublished studies were sought by searching ClinicalTrials.gov. The reference lists of included papers were hand searched to screen for relevant studies.

Supplementary Table I.

MEDLINE OVID search strategy.

Study selection

Following the removal of duplicates, the titles and abstracts were screened for inclusion. Relevant articles had their full texts assessed to determine suitability for inclusion or exclusion. The study selection was done independently by two authors (JM and MS) (Figure 1). A third author (DM) was contacted to confirm the randomisation process in older studies.

Figure 1.

PRISMA flowchart presenting the study selection and identification process.

Data collection and data items

Data extraction was performed by the first author (JM) using Excel (Microsoft, 2019, Chicago), and audited by MS. Disagreements were resolved by discussion with DM. Data items included sample size, age of participants, intervention, comparison, frequency of intervention, follow-up period, outcome measure, baseline and outcome data. When required, authors were contacted via email correspondence for clarification of missing data.

Risk of bias in individual trials

The Cochrane RoB 2 tool was used to evaluate bias in the included RCTs.¹⁹ This tool evaluates bias across five domains: the randomisation process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result.

Summary measures and approach to data synthesis

Statistical heterogeneity was assessed by the I² statistic, with significance set at p < 0.05. These statistical analyses and meta-analysis were performed using Jamovi (an open statistical software) with a random-effects model, due to expected heterogeneity.²⁰

Additional analyses

Following NHMRC recommendations, GRADE was used to determine the quality of evidence, to facilitate the development of a clinical guideline for the prevention of WSLs during orthodontic treatment.²¹ Publication bias was planned for assessment using funnel plots, if more than 10 studies were included in the quantitative analysis. Subgroup analyses for age, gender, preventive measure, outcome measure and type of appliance were planned.

Study selection

A total of 2494 citations were identified through the searches. Following duplicate removal, 1430 articles were available for title and abstract screening. This led to the exclusion of 1368 studies. Subsequently, 62 articles had their full texts read. Based on the predetermined inclusion/exclusion criteria, 44 articles were excluded (Supplementary Table II), and 18 RCTs were included in the qualitative analysis. Following the risk of bias assessment, three RCTs were pooled for quantitative synthesis. The study selection process is illustrated in Figure 1.

Supplementary Table II.

Reasons for exclusion following full-text assessment.

Study characteristics

The characteristics of the 18 RCTs included in the qualitative analysis are summarised in Supplementary Table III. Ten trials investigated F, three assessed CHX, two CPP-ACP, two oral hygiene, and one assessed laser intervention.^22-39

Supplementary Table III.

Data from studies included in the qualitative synthesis.

Following the RoB assessment, three RCTs were included in the quantitative analysis, since the rest were heterogeneous and at high RoB.^25,27,31 Included studies had a total of 433 patients (10–20 years of age), of which 216 participants were allocated to the intervention group.

Risk of bias within studies

The eighteen RCTs were assessed for RoB (Table I). Ten studies were determined to have a high RoB and were excluded from the quantitative synthesis.^{22,23,26,28,30,33,34,35,38,39} This led to the consideration of eight RCTs for meta-analysis.^{24,25,27,29,31,32,36,37}

Table I.

Risk of bias assessment of RCTs included in the qualitative analysis.

Results of individual studies and data synthesis

Of the eight trials considered for quantitative synthesis, five evaluated the same intervention – topical F application.^{24,25,27,29,31} The five RCTs identified a reduction in WSL development. The other trials investigated chlorhexidine, oral hygiene, and CPP-ACP.^32,36,37 A slight reduction in WSL incidence was obtained by these interventions, which was reported as non-significant. The studies could not be pooled for meta-analysis, since they were heterogeneous interventions.

Of the five F intervention RCTs, three used placebo controls.^25,27,31 The other two RCTs used F in their intervention and comparison arms.^24,29 One RCT compared a toothpaste combining SnF₂ and AmF to a NaF toothpaste with the same F concentration.²⁴ While another compared two toothpastes of different F concentration.²⁹ Since these studies prevented the estimation of effect size, it was decided to synthesise data from the three trials that compared the effects of F to a placebo.^25,27,31

Two studies compared the effect of professionally applied F.^25,27 They investigated the effects of a six-weekly application of 0.9% difluorsilane (1000 ppm F) in a polyurethane varnish base and a two-monthly application of 12,300 ppm APF foam.^25,27 Conversely, the effect of daily at-home use of a 250 ppm NaF rinse was investigated using quantitative light induced fluorescence (QLF) by another trial.³¹ Compliance with the use of a mouthrinse was not recorded. The F interventions were supplementary to standard oral hygiene practice of twice daily brushing with a fluoridated toothpaste.

In the three RCTs, a clinical assessment was performed under artificial white lighting following plaque removal.^25,27,31 Patients with developmental defects of enamel (DDE) such as hypoplasia and fluorosis were excluded in one study.²⁷ The other two RCTs did not consider DDE.^25,31 The diagnosis and assessment of WSLs differed between the studies. The teeth were dried with sterile gauze prior to clinical examination in one RCT.²⁷ Clinical photographs were taken after drying the teeth in two RCTs.^25,31 The bonding materials used to attach appliances varied between the RCTs. One study did not provide this information.³¹ Resin composite was used in one trial, while the other study used a resin applied as a pre-coated bracket (Victory Twin APC II bracket, 3M Unitek, CA, USA).^25,27 All three RCTs had some concerns related to RoB (Figure 2).

Figure 2.

Risk of bias summary of studies included in the meta-analysis.

The pooled effect size from the random-effects model showed a risk difference of -0.23 (95% CI: -0.35, -0.11, p < 0.001; Figure 3). A 23% reduction in WSL incidence was obtained with F interventions. Similar results were obtained using a fixed-effects model (Supplementary Figure 1). Statistical heterogeneity was moderate at 49.3% (Figure 4). Only one study investigated the effect of F on lesion size.³¹ Since there were no WSLs at the start of the study for these participants, the effect on lesion inhibition could not be ascertained.

Figure 3.

Meta-analysis of WSL risk reduction with fluoride interventions, using a random-effects model. The forest plot shows the pooled effect of fluoride on WSL incidence.

Figure 4.

Heterogeneity assessment.

Additional analyses

The publication and reporting biases were not analysed since fewer than ten studies were included in the meta-analysis. Subgroup analyses were not conducted due to insufficient data. A pooled adverse effect profile for F could not be constructed, since only one study reported adverse events, of which there were none.²⁷

Clinical guideline

Quality of evidence (GRADE)

The findings from the quantitative analysis of the three included RCTs were used for grading F interventions.^25,27,31 The quality of evidence for CHX, CPP-ACP, oral hygiene program and laser interventions were graded from the initial full-texts analysed. This included the RCTs and observational studies.

According to the GRADE evaluation, a moderate quality of evidence for the use of F during orthodontic care was obtained (Figure 5). Quality was downgraded to moderate due to statistical heterogeneity and a small number of RCTs. Very low-quality evidence was obtained for CHX, CPP-ACP, oral hygiene program and laser therapy, due to non-randomised studies that were sponsor funded and measured surrogate outcomes (Figure 5).

Figure 5.

GRADE evidence profile: caries prevention during orthodontic treatment.

Resource use and economic evaluation (GRADE)

The primary resource to evaluate is the intervention, both in-chair and at-home. The cost of F interventions may vary within and across countries, and could be shared by governments, private insurers and patients. A cost-benefit analysis regarding F for caries prevention has not been conducted previously. Therefore, an economic evaluation could not be performed in the present review. Since these caries preventive resources are not labour intensive, a resource use evaluation may not be important in developing clinical recommendations.

Clinical recommendations (GRADE)

For policy makers in governments, institutions and health insurers, the use of F during orthodontic treatment is recommended for adoption as a policy. For most healthy patients undergoing orthodontic treatment, the use of F as a caries preventive measure is advocated. For patients who prefer to avoid F, appropriate advice from the clinician is recommended. Clinicians are advised to routinely use topical F during orthodontic care.

Summary of evidence

In the present SR, the effect of interventions in the prevention of WSLs during orthodontic treatment was evaluated. Previous SRs have several limitations (Supplementary Table IV), including the inclusion of high-risk studies. These include in vitro, in vivo, ex vivo, retrospective, non-RCT and split mouth design studies (Supplementary Table V). It is critical to conduct sensitivity analyses to evaluate whether studies associated with a high risk of bias influence the pooled estimate. This aspect has been lacking in previous published SRs. Additionally, several SRs included studies that considered WSL inhibition post-orthodontic treatment. Since complete reversal is unachievable with current interventions, it is imperative to prevent WSLs.^40,41 This counters the undesirable aesthetic and biologic consequences of WSL development.

Supplementary Table IV.

Previous systematic reviews and meta-analyses.

Supplementary Table V.

Reasons for exclusion of studies included by previous SRs.

The present review addressed these limitations through a rigorous approach. An a priori registration on the PROSPERO database, extensive search on multiple databases, and an independent analysis approach provided robustness in this review. The use of an updated Cochrane RoB 2 was also advantageous, since it identified an overall risk within the studies, and was less subjective compared to previous versions.

Eight studies were excluded due to their short duration of less than six months. A longer time frame was a key eligibility criterion because a short follow-up period is likely to be insufficient for clinically detectable signs of dental caries to develop in many individuals. Even when carious lesions are detected early, prior to cavitation, the demineralisation process may have been present for months or years.² The lack of adequate follow-up could lead to a type I error due to under-reporting of WSLs. Furthermore, several studies were excluded because they failed to include the presence of carious lesions (WSLs) as an outcome, and instead, resorted to report changes in microbiome, plaque and saliva. While these studies provided valuable insights into the mechanisms by which preventive measures reduce the risk of developing carious lesions, they cannot be considered equivalent to measuring the actual outcome of identifying WSLs. A reduction in certain cariogenic pathogens may not have an effect on WSL incidence, since the remineralisation-demineralisation balance is influenced by other factors.² These surrogate measures cannot yet be utilised as accurate markers of clinical disease.

The three RCTs included in the quantitative analysis investigated different F concentrations, and none were considered a low RoB. Two studies in the meta-analysis did not have an initially calculated sample size.^25,31 One RCT calculated the sample size required as 132 participants per group.²⁵ Following attrition, this became 132 and 125 per group. Similarly, another RCT required 47 participants per group following their power calculation.³¹ Subsequent to dropouts, the groups were 36 and 45. These discrepancies could affect the statistical power of the studies, due to the risk of a type II error. One study had no pre-specified analysis plan, which created a potential for bias in the analysis post hoc.²⁷ Although the outcome measures differed between the three RCTs as the analysis was based on the proportion of WSL incidence, this did not affect the synthesis. The heterogeneity obtained could be explained by the differing concentration and delivery vehicles of F in these trials. Since the number of studies assessed was low, heterogeneity analysis might be misleading due to decreased sensitivity.

The GRADE evaluation considers the number and design of studies in addition to effect sizes, which highlighted the lack of RCTs on the topic. The recommendation to use F during orthodontic treatment supports the current widespread usage by dental professionals.⁴² F therapy could include professional delivery as part of clinical care, or at-home use of F rinses and toothpastes. However, reported poor compliance with mouth rinsing by orthodontic patients reduces the clinical effectiveness of at-home therapies.⁴³ Since F interventions alone provide a 23% risk reduction of WSL development during orthodontic care, the combination of other preventive measures could reduce the risk further. Proper oral hygiene and non-cariogenic dietary practices in combination with F interventions may be of greater importance to reduce the risk of WSL incidence. The present review was unable to evaluate the cost-effectiveness of interventions, as no appropriate studies were identified, an important consideration that is likely to favour at-home therapy. No RCTs were conducted to investigate the preventive effects of self-assembling peptides (SAP), RMGIC coatings and other calcium products such as TCPs. The meta-analysis was planned to be performed using Stata, but Jamovi statistical software was used, due to its open and free access.

Limitations

There are several limitations associated with this review. The World Health Organization International Clinical Trials Registry Platform was not sought for unpublished studies, which could have provided additional relevant papers. The exclusion of non-English papers may have omitted well-conducted relevant studies. Since only three RCTs were pooled for meta-analysis, the estimates of risk difference may differ from the results obtained. Nevertheless, both random and fixed effects models provided similar outcomes. The results related to the inhibition of WSL progression and patient reported outcomes were not investigated in the included studies.

Figure 6.

Pooled risk difference obtained from a fixed-effects model.

Recommendations for future research

Future RCTs should explore the efficacy of interventions such as SAP, RMGIC, CPP-ACP and TCP on the prevention of WSLs during orthodontic treatment. Authors are encouraged to report on caries increment/lesion transition and adverse effects. Prospective RCTs should register their study protocols a priori and have a follow-up period of the duration of care, or otherwise at least six months. Future investigators are encouraged to design trials for patients undergoing removable orthodontic treatment.