This is a longitudinal, pragmatic, cluster randomized, non-inferiority clinical trial consisting of SDF combined with fluoride varnish versus therapeutic sealants with fluoride varnish given biannually to each study participant in primary schools. This trial is reported following the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) guidelines and has received approval from the New York University (NYU) School of Medicine Institutional Review Board (IRB) (#i17–00578). Any changes to the study protocol will be communicated to the IRB and funder in quarterly reports, and investigators will cooperate with any independent audit on behalf of the IRB or funding organization.
Prior to the study, schools meeting the inclusion criteria will be solicited for participation and randomly assigned to receive fluoride varnish/SDF or fluoride varnish/sealants in 6-month intervals (± 1 month). Informed consent for all children in participating schools will be distributed to parents at the beginning of each school year. At each observational period, study participants with completed consent will receive a comprehensive oral examination provided by a licensed dental hygienist and complete a brief psychological assessment for oral health-related quality of life (Fig. 1) [36, 37]. Following the oral evaluation, participants will receive the assigned treatments (Additional file 1).
The primary aim is to evaluate the non-inferiority of SDF versus sealants in the arrest and prevention of dental caries. Secondary study aims include assessing differences in provider type, quality of life, academic performance, and school attendance. To determine provider effects, treatments in the simple prevention group will be given by either dental hygienists or registered dental nurses and the rates of arrest and prevention will be compared across provider. Data for educational outcomes will be solicited from the New York City (NYC) Department of Education (DOE). Any participant presenting with a medical emergency will be referred to school nurses for follow-up care. The anticipated schedule of enrollment, interventions, and assessments is shown in Fig. 2.
Clinical evaluation and diagnosis
At each observation for each participant, examining dentists will dry tooth surfaces with gauze squares and perform full-mouth oral examinations, including examination of all teeth and tooth surfaces to determine if sound, decayed, missing, filled, or having pulpal involvement. The exam will also include an assessment of pain, swelling, infection, and abscess presence.
Diagnosis of cavitated carious lesions will be made based on a visual-tactile oral examination and follow criteria for cavitated lesions as specified by the Diagnostic Criteria and Procedures of the Oral Health Surveys of the National Institute of Dental Research [38].
Examiners will be calibrated by examining ten students independently at baseline and discussing whether caries are present or not. Following this review, examiners will be calibrated by examining another ten students independently and comparing results. To standardized delivery of care, examiners will be trained to use all interventions prior to participating in the study. Standardization of diagnostic criteria will consist of reviewing diagnoses at the mid-point of each study year and discussing with examiners. Both calibration and standardization will be conducted every year of the study.
Treatments in the complex arm will be provided by dental hygienists with the support of dental assistants. Treatments in the simple arm will be provided by registered nurses or dental hygienists with the support of dental assistants, both overseen by a nurse practitioner.
Treatment description and regimen
Simple prevention
One drop (0.05 ml) of SDF (Advantage Arrest™) solution at 38% concentration (2.24 F-ion mg/dose) will be dispensed per child. Posterior tooth surfaces to be treated will be dried, after which the SDF will be applied with a microbrush to all asymptomatic cavitated lesions and to all pits and fissures on bicuspids and molar teeth for 30 s. Fluoride varnish (5% NaF, Colgate PreviDent™) will then be applied to all teeth. Simple prevention will be provided by either dental hygienists or registered nurses.
Complex prevention
All primary and permanent teeth will be dried prior to application. Pits and fissures on all bicuspids and molar teeth will be sealed with glass ionomer sealants (GC Fuji IX). Interim therapeutic restorations will be placed on all frank asymptomatic cavitated lesions. Cavitated lesions will be cleaned using a tooth brush prior to placement of ITR. Fluoride varnish (5% NaF, Colgate PreviDent™) will then be applied to all teeth. Complex prevention will be provided by dental hygienists.
Across both arms, non-cavitated lesions will be treated with fluoride varnish. Both arms will also receive toothbrushes, fluoride toothpaste, and oral hygiene instruction. Clinical care will be provided in a dedicated room in each school using mobile equipment and disposable supplies. Any child with pulpal involvement or in need of an extraction will be referred to local dentists for follow-up care.
Risks and adverse events
Each intervention used in this trial is currently used in clinical practice as a standard of care procedure. The potential risks for study participants are minimal and identical to the risk for children obtaining care in a dental office. The greatest risk is an allergic reaction to fluoride varnish, SDF, or glass ionomer. All adverse events occurring during the study period will be recorded: at each contact with the study participant, investigators will seek information on adverse events by specific questions and an oral examination. Evidence of adverse events will be recorded on electronic health records and appropriate case report forms. The clinical course of each event will be followed until resolution, stabilization, or until it has been determined that participation in the study was not the cause. Serious adverse events ongoing at study end will be followed to determine the final outcome. Adverse event reports will be reported to the IRB within five working days from the time that investigators become aware of the event.
Definition of outcome measures
Primary outcome measures
Primary outcomes include clinically evaluated caries arrest and the prevention of new caries. Caries arrest will be evaluated after 2 years and the prevalence of new caries will be evaluated after 4 years.
Secondary outcome measures
Secondary outcomes include oral health-related quality of life measured using the Child Oral Health Impact Profile – Short Form (COHIP-SF) [37], academic performance measured using mathematics and English language standardized assessments given to all NYC school children between grades 3 and 8, and the number of school days missed for each participant over each academic year.
Recruitment and eligibility
All schools meeting the inclusion criteria were solicited by the NYC DOE to participate in the program. School principals were mailed letters describing the study protocol and interventions, and any interested principal opted into the study. Prior to the beginning of each school year, electronic rosters for each participating school will be provided to study investigators from the DOE, which will include a unique student identifier, contact information, demographic and socioeconomic variables, and Medicaid identification (if available). School rosters will be used to electronically create personalized informed consent for every student in the school, which will then be combined with a letter from the principal explaining the study and distributed to parents of children in each school. Completed informed consent will be collected at the school by NYU investigators. Schools will be recruited and enrolled over the first 2 years of the study. However, children within schools will be consented and enrolled in each year of the program to accommodate newly registered students each academic year. Recruitment for this study is pending.
Inclusion criteria
Any primary school in NYC with a student Hispanic/Latino population greater than 50% and a low-income population (defined as a student receiving free or reduced-price lunch) of at least 80% is eligible to participate. Within participating schools, all children are eligible to participate in the study regardless of age, sex, race/ethnicity, insurance status, or other sociodemographic variable. Those with informed consent and assent will receive care.
Exclusion criteria
The only schools that are ineligible to participate are those that already have a pre-existing school-based dental health program. Within participating schools, exclusion criteria for children include those without informed consent or those with consent but without assent.
Randomization
Participating schools will be block randomized at the school level to receive either the simple or complex treatment using a random-number generator. First, schools will be ordered by the total student population. Blocks of two schools will then be selected sequentially from the list. Schools in each block will then be randomly assigned to either simple or complex arms using a random-number sequence generated from a computer program. Schools will enter into the study in three phases: in phase 1, 16 schools will be randomized. Six months later (phase 2), an additional 14 schools with be randomized, followed by 30 more schools 6 months after that (phase 3).
Block randomization will be conducted at the school level as the NYC Departments of Education and Health and Mental Hygiene requested that interventions be identical within schools. Additionally, interventions in the simple arm are provided by nurses and hygienists, while treatments in the complex arm are provided by hygienists only. Thus, study coordination is more feasible if provided at the school level.
Blinding
Due to the nature of the treatments, nurses and dental hygienists providing care will not be independent from study protocols and, therefore, are not blinded. Additionally, all participants will be provided with a letter describing the care that they received. Assignment to treatments will follow a predetermined randomization list at the school level, and all students with consent in participating schools will receive the assigned treatment. However, all data for caries arrest will be masked prior to analysis such that which schools were assigned to which treatment, and the subsequent treatments given to each participants (e.g., SDF or ITR), will not be able to be determined during analysis. Only data on the sound or decayed/filled status of individual teeth will be provided per participant. Following analysis of caries arrest, this can no longer be guaranteed.
Data collection, transmission, and storage
Data collected from each participant will be recorded on a password-protected tablet computer using a propriety software system that is pre-populated with the demographic information of the participant. Recorded data will also include responses to the short form version of the Child Oral Health Impact Profile (COHIP-SF). COHIP questions will be posed to each participant and examiners will record responses.
Following each day of the study, electronic records will be uploaded to a secure server and stored at the Boston University Data Coordinating Center (DCC) and evaluated for quality assurance. The DCC will also maintain sociodemographic information provided by the NYC DOE for each study participant. When the DCC transmits data to investigators, no identifying information will be provided with the exception of a unique student identifier. This data will be kept at the NYU on a secure, password-protected server.
Data for secondary school outcomes (academic performance and attendance) will be transmitted to the NYU investigators by the NYC DOE at the end of each school year.
Sample size calculation
The presented study is powered for the primary outcomes of caries arrest and prevention. Approximately 14,100 students are expected to be enrolled across 60 schools over the duration of the study (n = 235 per school cluster). Based on a pilot study of school-based caries prevention conducted in NYC from 2016 to 2017, the baseline prevalence of caries of the primary and permanent dentition at the child level was 40.7%. The prevalence of untreated decay on primary dentition only was 36.4%, and the prevalence of untreated decay on permanent dentition only was 35.1%. The average dmft was 1.40 (SD = 2.04) and the average DMFT was 0.34 (SD = 0.84). Power for the primary outcome of caries arrest was calculated using a two-arm non-inferiority trial design. The per-person proportion of carious teeth at baseline that were treated with complex or simple prevention and stayed arrested over a 2-year period will be analyzed at the child level. Calculations assume an equal proportion of success, π, of teeth with caries being arrested in complex and simple arms. The non-inferiority margin (δ) was set at 10%. Additional parameters of a two-sided type-I error rate of 5% and statistical power of 0.80 yielded a total sample size per group of 198 (ntot = 396) [39]. To adjust for any potential school-level clustering based on inclusion criteria of schools consisting of high-need children, an intraclass correlation coefficient (ICC) of 0.10 was assumed. The design effect associated with this ICC increased the effective sample size by a factor of 24.5, resulting in a total required sample of 9702.
For caries prevention, power for the two-arm, longitudinal, cluster randomized design was estimated using the method of Diggle et al. (2002) for generalized estimating equations [40]. Estimates assume an average number of visits per child of six, statistical power of 0.80, and a two-sided type I error rate of 5%. A repeated measures correlation of 0.5 and a per-visit attrition rate of 20% were also assumed. For a given minimally detectable effect size (standardized effect size difference) of 0.25, a cluster-adjusted (ICC = 0.10) sample size of 12,874 is required. Based on the expected participant enrollment, the study is powered for these conservative assumptions for caries prevention.
For secondary outcomes, the NYC citywide average absenteeism rate is 14.2%. The average performances for reading and math examinations (grades 3–5) are 298 (SD = 17) and 299 (SD = 21), respectively. Sample sizes were calculated using a simple two-group cluster randomized comparison of means (standardized test performance) and proportions (school absences). Power calculations assume an intraclass correlation of 10%, statistical power of 0.80, and a two-sided type I error rate of 5%. Based on these assumptions, the anticipated sample size is sufficient to detect a 9% decrease in absenteeism (to 5%) and a standardized test difference of 6.44 (reading) and 7.95 (math).
Quality of life will be measured using the short form version of the Child Oral Health Impact Profile (COHIP-SF) [37]. In a validation study of the COHIP-SF, a pediatric sample of children (aged 7–17 years) had average scores of 56.2 (SD = 9.3). For a simple cluster randomized, two-group mean comparison, with a two-sided type I error rate of 5%, statistical power of 0.80, and ICC of 0.10, the study is powered to detect a difference of 2.2 on the COHIP-SF scale.
Statistical analysis
For the non-inferiority of caries arrest, the per-patient proportion of carious lesions at baseline treated with simple versus complex prevention that stayed arrested throughout the first 2 years of study observation will be determined. Any deciduous teeth with treated carious lesions that are lost due to exfoliation during the course of the study will be considered as arrested throughout the observed lifetime of the tooth. Data will record the last observed status of exfoliated teeth to be used in subsequent analyses. Thus, tooth-level indicators are able to be present for both primary and permanent dentitions at the same time. With this approach, each carious tooth treated with either simple or comprehensive prevention is a single trial with outcomes either of arrested (1) or failed to arrest (0). The percentage of arrested caries (at the child level) will be modeled using multilevel binomial regression with a logit link:
$$ {Y}_j\sim B\left({\pi}_{\mathrm{j}}\right),E\left({Y}_{\mathrm{j}}\right)={\pi}_{\mathrm{j},} $$
where πj is the probability of success (as defined above). Analysis will be conducted for all teeth, permanent teeth only, and primary teeth only. As described, the non-inferiority margin, δ, is set at 10%. While there is no “gold standard” criterion for the selection of this margin, the margin was set based on collaborative discussion with clinicians to determine what is considered clinically unimportant. The null hypothesis is that the experimental treatment (simple prevention) is inferior to the standard treatment (complex prevention) by at least δ: πsimple − πcomplex ≥ δ. The alternative hypothesis is that πsimple − πcomplex < δ.
Based on results from multilevel binomial models, differences in effect sizes estimated by confidence intervals will be used to determine clinical non-inferiority of the two prevention methods [41]. Confidence intervals will be calculated for the difference between the two interventions, with the width of this interval signifying the extent of non-inferiority. If the difference between the two interventions lies to the right of δ, then non-inferiority will be concluded. Though this is method is preferred by reporting guidelines, p values will also be reported, in keeping with other recommendations [41].
For the prevention of new caries, longitudinal data will be analyzed using generalized estimating equations (GEE) and mixed-effects multilevel regression models (ME-MLM) with the appropriate error distribution for the prevalence and incidence of untreated caries. The number of teeth at risk for each child during each follow-up interval will be identified and the number of those teeth in which new caries is observed at the examination that ends that interval will be determined. Primary teeth lost in each interval and new permanent teeth will not contribute to data for that interval. Data from baseline visits will be omitted from analyses and used as an indicator of any untreated decay at baseline. Analysis will be conducted for all teeth, permanent teeth only, and primary teeth only.
To explore non-linear trends in untreated decay between simple and complex prevention, longitudinal data will be analyzed using generalized additive models (GAMs) with non-parametric smoothers, linking the known proportion:
$$ {p}_{it}=\mathrm{E}\left({\mathrm{y}}_{\mathrm{it}}=1|{\mathrm{x}}_{\mathrm{ijt}},{\mathrm{z}}_{\mathrm{it}}\right) $$
to a non-linear non-parametric predictor using the link function:
$$ {n}_{it}=g\left({u}_{it}\right)=\ln \left({u}_{it}/1-{u}_{it}\right)={\sum}_{j=1}^p{s}_j\left({x}_{jit}\right)+{z}_{it}^T{u}_i, $$
where sj are smooth non-parametric functions and ui are random effects assumed to be iid ~ N(0, D(ψ)) [42]. Heterogeneity and correlation among subjects will be accounted for through random effects.
Longitudinal effects of simple and complex prevention on academic outcomes, compared to untreated children, will be analyzed using propensity score matching and multilevel modeling. First, propensity scores will be estimated for each participant at baseline for the probability of treatment assignment conditional on observed covariates (e.g., prior academic performance). Propensity scores will be used to match treatment students to students not receiving treatment, considering multiple forms of matching such as nearest neighbor and caliper. Potential comparator students not receiving either simple or complex treatments will be drawn from “peer-schools,” schools identified by the NYC DOE as similar to treated schools based on socioeconomic, academic performance, and teacher-quality indicators. This data is anonymized and can be used without consent as a secondary data source. Treated students and matched comparators will then be analyzed using ME-MLM (for academic achievement) and Poisson regression (for school absences). If a different mechanism drives initial versus continued absences, school absences will be analyzed using zero-inflated negative binomial multilevel modeling.
Finally, change in OHRQoL between groups over time will be analyzed using ME-MLM. Baseline quality of life will be included as a covariate and models will include predictors for time, treatment, and the treatment-time interaction. Hypothesized a priori confounders, including sociodemographic variables at the child and school level, will be included.
Missing data will be adjusted for using multiple imputation and inverse probability weighting (IPW). Statistical analysis will be performed following intention-to-treat and analyzed using Stata v14.0 (StataCorp LLC, College Station, TX, USA) and R v3.1.1.