Effectiveness of remineralising agents in prevention and treatment of orthodontically induced white spot lesions: a protocol for a systematic review incorporating network meta-analysis

White spot lesions (WSLs), defined as ‘subsurface enamel porosity from carious demineralisation’ that manifest as ‘milky white opacities located around the bracketed gingival area in the tooth surface’, are one of the most common adverse effects in fixed orthodontic treatment [1‐3].

Demineralisation occurs when a low pH persists for a longer period of time and large amounts of calcium ions are released from the enamel of the tooth (mainly through hydroxyapatite; HAP) [4]. In addition, bonding attachments hinder conventional oral hygiene and limit natural oral self-cleaning mechanisms, which leads to accumulation of plaque and a lower pH [5, 6]. WSLs cause cosmetic problems in anterior teeth and influence patients’ satisfaction with their smile [7‐9]. The reported prevalence of WSLs was relatively high, ranging from 23.4 to 49.6% after debonding, depending on the measurement method or criteria [7, 10]. Along with the importance of the prevention and treatment of WSLs with remineralising agents, the safety of the remineralising agents is equally important, since an intake of excess fluoride may lead to fluoride toxicity. As such, proper techniques and safety measures must be adapted when applying remineralising agents.

Remineralising agents are the most commonly used interventions in the prevention of WSLs and the first-line treatment for post-orthodontic WSLs. Topical fluoride use is the most reported method of prevention and treatment for early enamel demineralisation, including fluoride materials of different concentrations (in the form of bonding materials, sealants, gels, mouth rinses, toothpastes, or varnishes) [11‐13]. There are alternatives to these, such as casein phosphopeptide-amorphous calcium phosphate (CPP-ACP) creams, casein phosphopeptide-amorphous calcium phosphate with fluoride (CPP-ACFP), and bioactive glass toothpaste, among others [14‐16].

Many interventions are self-applied agents, such as mouth rinses and toothpastes containing fluoride. Some interventions are performed by dentists, such as a fluoride varnish (Duraphat; Colgate-Palmolive, New York, NY) or a fluoride film (Sheer; CAO Group, West Jordan, UT) with minimal clinical chair time [15, 17]. Almost all of these interventions are combined with tooth-brushing guidance and oral hygiene education [14, 15].

Almost all preventive measures and remineralising interventions aim to enhance enamel resistance to acid and prompt the process of remineralisation to reverse the caries process. Prior studies have shown that during the remineralisation process, enamel lesions preferentially adsorb fluoride ions onto partial enamel lesions with demineralised HAP crystals or redeposit fluorohydroxyapatite (FHAP) [17]. Therefore, increased fluoride can enhance remineralisation and form a low-solubility veneer, which is an acid-resistant mineral on the remineralising crystals [18].

CPP-ACP has a beneficial sub-surface effect, whereby its milk-protein-based formulation can promote the natural salivary healing process. In addition, its nanoclusters of ACP are small enough to access demineralised areas through an existing remineralised surface zone [19‐21].

Bioactive glass is a ceramic material consisting of amorphous sodium calcium-phosphosilicate [22]. It can release sodium ions for exchange with hydrogen cations to release calcium and phosphate ions, whereby the topical pH increase would precipitate the extra calcium and phosphate ions to form a calcium phosphate layer. This layer crystallises into hydroxycarbonate apatite (HCA) as these reactions continue [23]^.

Previous studies have compared the effectiveness of many remineralising agents in WSLs prevention and treatment [13‐15]. However, many of these studies are not well-designed and have limited methodology. In addition, there are differences in purpose, design, and treatment among current studies regarding WSLs remineralisation agents [18, 24‐26]. Furthermore, previous studies have not clearly explored the safety of remineralising agents.

Several systematic reviews have been published to investigate the comparative effectiveness of remineralising agents for WSLs prevention and treatment [27‐29]. However, all these systematic reviews have only found pairwise evidence from head-to-head comparisons and have thus failed to evaluate the comparative effectiveness of all available remineralising agents. In addition, the effectiveness rankings of the remineralising agents are still unclear. Network meta-analysis (NMA) can integrate the evidence from direct and indirect comparisons. It enables inference about every possible comparison between a pair of interventions in the network, even when some comparisons have never been evaluated in a trial. Through the comparison of multiple interventions, we can rank all evaluated interventions with minimal adverse effects [30‐32].

Therefore, this study will involve a systematic review and frequentist framework network meta-analysis to compare different remineralising agents for the prevention and treatment of WSLs containing different concentrations of fluoride, CPP-ACP, CPP-ACFP, bioactive glass, and the safety of the current remineralising agents (Fig. 1).

In this review, we will compare the effectiveness and safety of remineralising agents used in the prevention and treatment of WSLs in patients treated with fixed orthodontic appliances in any type of orthodontic setting.

The eligibility criteria for our systematic review have been developed according to the PICOS acronym. These criteria are presented here and details are summarised in Table 1 [35].

No other limitations will be imposed on unpublished studies, the language of publication, study settings, studies of all durations, and those conducted during all points in time are eligible for inclusion.

We assume that patients who fulfil the inclusion criteria are equally eligible to be randomised to any of the interventions we plan to compare.

Two reviewers will extract data from each included study using Microsoft Excel 2010 (Microsoft, Redmond, WA) with a specifically developed data extraction form, independently and in duplicate. Data extraction forms will be piloted initially on a small number of included studies. All disagreements will be resolved by discussion or by involving a third assessor. We will record sources of funding if stated. We will record the study details in the ‘Characteristics of included studies’ table (Additional file 4). We will collate multiple reports of the same study. We will extract estimates of 2 × 2 tables (dichotomous data), means and standard deviations from effect estimates, confidence intervals, and other forms of data. From each trial, the following data/information will be collected [35]:

The primary outcomes are the severity of WSLs at pre-intervention and at last available follow-up assessment measured by objective methods as follows:

Secondary outcomes will be subjective measures of WSLs as follows:

Most outcomes are observed among multiple time points because remineralising agents require long-term use. While combining outcomes that were measured at different follow-up times might not be appropriate, to date, there is no consensus on the appropriate duration of a remineralising treatment cycle. In order to synthesise the data as much as possible, we will divide the time points into short-term (< 3 months) and long-term (> 3 months) and will extract the outcome data from the longest follow-up (closest to 3 months or longest after 3 months, respectively).

We will conduct both fixed-effects and random-effects model NMA to synthesise all evidence for each outcome. The choice between models will be based on the expectation of whether the intervention effects are truly identical or the funnel plot asymmetry or other conditions [43]. We will obtain a comprehensive ranking of all treatments (ranking probabilities, the surface under the cumulative ranking curve (SUCRA), and mean ranks) [32]. The estimated relative ranking of interventions will be generated according to primary outcomes. If the NMA and SUCRA cannot be performed on primary outcomes, we will refer to the NMA and SUCRA results of secondary outcomes.

We will explicitly describe the process leading to node grouping [44, 45]. The network of treatments will be judged based on the characteristics of the available studies and presented and evaluated graphically. We will evaluate the following: (1) if the network is disconnected, (2) if there is a sufficient number of comparisons in the network with available direct data, (3) if there is a high number of comparisons based on a single study, and (4) if any key treatments are missing. Next, the feasibility of the network meta-analysis will be assessed checking the following: (1) transitivity (i.e., the comparable distribution of effect modifiers across comparisons), which will be examined using boxplots or percentages to visually inspect potential effect modifiers of treatment effect [46]; (2) consistency between direct and indirect estimates of the effects, which will be examined using the node-splitting method [47], and globally (i.e., evaluating the network as a whole) using the design-by-treatment interaction model [48]; and (3) the amount of variability, which we will quantify, that can be attributed to heterogeneity and inconsistency rather than sampling error, by calculating the I² statistic [49].

We will include both two-arm trials and multi-arm trials, and those comparing different active ingredients or dosages of the same active ingredients will be analysed separately for a global analysis of all outcomes. Different concentrations of the same active ingredient at the same dosage will be merged together for a global analysis of all outcomes (e.g., 900 ppm NaF toothpaste and 1100 ppm NaF toothpaste), with the exception of huge concentration differences (e.g., 900 ppm NaF toothpaste and 5000 ppm NaF toothpaste). Moreover, we will not distinguish interventions according to the duration and administration frequencies. Both the placebo and treatment-free groups will be merged together as the control group for global analysis of all outcomes.

Two review authors will independently assess the risk of bias for each study using the criteria outlined in section 8.2 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019) and in section 3.5 of Chaimani’s study (2017) [34, 50]. All disagreements will be resolved by discussion or by involving a third assessor. We will assess the following domains as ‘low’, ‘unclear’, or ‘high’ risk of bias:

We will report these assessments in a ‘Risk of bias’ table for each included study and we will provide supporting judgments for each assessment.

We will provide summary assessments of the risk of bias for each important outcome (across domains) within and across studies following Table 8.7.a in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) [50]:

We will perform our NMA model with contrast-level data by multivariate meta-analysis (commands network meta and mvmeta) in STATA (version 14, Stata Corp, College Station, TX, USA) within a frequentist framework. The restricted maximum likelihood method will be used to estimate the between-study variance in the NMA [51, 52]. We will perform both traditional pairwise analyses and NMA. If NMA is not appropriate due to high global inconsistency or other conditions, a pairwise meta-analysis only will be considered. If a pairwise meta-analysis is also not possible, studies will be summarised narratively.

The outcomes of continuous and dichotomous variables will be presented as MDs and RRs with 95% CIs.

A network plot will be created to show what treatments can be compared as many as possible (Fig. 1). The effectiveness of each treatment among all treatments will be ranked, and we will generate plots of the treatment rank probabilities to rank the various treatments for each outcome.

The participants will be the unit of analysis. Where cluster-randomised trials are included, we will undertake data analysis at the same level as the randomisation or at the individual level accounting for the clustering. In doing so, we will follow the advice provided in section 23.1.3 of the Cochrane Handbook for Systematic Reviews of Interventions [53].

In studies where data are unclear or missing, we will contact the principal investigators. If missing data are unavailable, we will follow the advice given in Section 10.12.3 of the Cochrane Handbook for Systematic Reviews of Interventions [43]. If results are only reported graphically, we will graphically obtain the values, whenever applicable.

We will assess the assumptions of transitivity and similarity based on clinical and methodological characteristics. We will assume that intervention effects are transitive in this network meta-analysis, and we will investigate similarity based on clinical characteristics, such as the same mechanism of treatment, the similar period of treatment, and the severity of WSLs at baseline. All of these effect modifiers will be judged and reported before the network meta-analysis is conducted.

We will assess evidence for consistency in three ways: loop, design-by-treatment interaction, and node/side-splitting [47, 48, 54‐57]. First, when there is a loop connecting three or more treatments, it is possible to evaluate the consistency between direct and indirect evidence. Second, we will use the design-by-treatment interaction model that provides a single inference, using the chi-squared test, about the plausibility of assuming consistency throughout the entire network. Third, we will use the node-splitting method to calculate the inconsistency of the model, which separates evidence on a particular treatment contrast into direct and indirect evidence.

If there are 10 or more studies in a meta-analysis, we will investigate reporting biases (such as publication bias) using funnel plots. We will assess the asymmetry of funnel plot by visual evaluation and statistical tests. For continuous outcomes, we will use the test proposed by Egger (1997), and for dichotomous outcomes, we will use the test proposed by Harbord [58, 59]. If asymmetry is detected in any of these tests or is suggested by visual assessment, we will perform exploratory analyses to investigate it.

We will carry out sensitivity analyses to assess the robustness of our review results. Sensitivity analysis will be conducted by repeating meta-analysis with studies with an unclear or low risk for bias. More issues suitable for sensitivity analysis will be identified during the review process. We will report sensitivity analyses by producing a summary table [43].

We will create a ‘Summary of findings’ table for the comparisons of the primary and secondary outcomes. We will also describe the use of additional summary measures assessed, such as treatment rankings and SUCRA values, as well as modified approaches used to present summary findings from meta-analyses.

We will use the GRADE system (GRADE 2004) and the GRADEpro GDT software 2015 to create the table [60]. We will consult the latest literature on GRADE for network meta-analyses quality evaluation [61‐64]. Four steps will be used to assess the quality of treatment effect estimates from NMA: (a) present direct and indirect treatment estimates for each comparison of the evidence network, (b) rate the quality of each direct and indirect effect estimate, (c) present the NMA estimate for each comparison of the evidence network, and (d) rate the quality of each NMA effect estimate [61, 62].

All differences between the protocol and the final review will be reported with the rationale for these changes. We will also report the influences of these modifications.

This protocol uses direct and indirect comparison methods, which helps in determining the best agents before and after the occurrence of WSLs. By integrating currently available clinical data, the network meta-analysis will provide a high-quality conclusion for patients and clinicians about the prevention and treatment of the WSLs. Potential limitations and challenges include but are not limited to study level (the possibility of clinical heterogeneity, poor-quality reporting in the included trials, bias like patient intervention self-administration without the study assessing compliance with the instructions, analyses disregarding the clustering of measurements) and review level (a lack of available treatment comparisons to build robust nodes, the network unable to cover all the interventions and combinations). Summary and implications including the conclusion of recommended or ineffectual interventions/combinations will be discussed in the final manuscript.

This network meta-analysis will compare the effectiveness of remineralising agents in the prevention and treatment of orthodontically induced WSLs. Previous systematic reviews have only done pairwise comparisons. One study made a preliminary attempt on network plotting, but it was still a traditional meta-analysis [28]. For a better understanding of the effectiveness of various conservative WSL treatments, we will conduct a network meta-analysis that allows us to estimate the relative effectiveness of all available treatments. Simultaneously, it will improve the efficiency in comparative effectiveness research and in the quality of decision-making. We believe that the results of this review will be beneficial for clinical decision-making and will also advance future clinical studies.

Any amendments (such as nodes combination, results reporting) made to the current protocol will be published as a supplement in the final manuscript, any amendments made to this protocol when conducting the study including the date of and rationale for the amendments, with the final manuscript.