LT Comparative Effectiveness of 2 Caries Treatments

Study Design: The original design was a cluster randomized clinical trial (RCT) in 33 publicly subsidized housing facilities/sites in NE Ohio and housing sites randomized into 2 arms in a roughly 1:1 ratio (SDF vs. ART + FV) and is a continuation of the original study "Reducing Oral Health Disparities of Older Adults: Comparative Effectiveness of 2 Treatments". All participants will be followed for one year and will receive intervention when new decay is present. The study hypothesis for the LTFU study is noninferiority of SDF to ART + FV for each of the primary outcomes. Primary outcomes are assessed through clinical and questionnaire evaluations. We will maintain blinding of participants to the study arm. Other clinical and study staff cannot be blinded due to the nature of the interventions.

Study Population and Setting: For the Intervention Arm, a total of 287 participants from 18 facilities will receive biannual 38% SDF, and for the Control Arm, a total of 193 participants from 15 facilities will receive ART + biannual FV application for new decay. Housing sites are located in three counties in Northeast Ohio (Cuyahoga, Summit, and Lorain). IRB approval was given allowing study staff to contact participants from the original study for future studies. Informed consent for LTFU will be obtained in-person at scheduled "Dental Days" at the housing sites. These will be scheduled in coordination with the housing site Service Coordinators, who will be briefed well in advance of LTFU activities.

A total of 480 participants (out of 568 from the original study) are expected to be retained into the LTFU study. Among the 480 participants to be consented and enrolled in the long-term follow-up, 26% (n-124), 36% (n=173), 17% (n=82), 21% (101) will contribute data at 3, 4, 5, 6 years post-randomization respectively. Each participant will be followed for one year (total of 3 visits -baseline (LTFU #1), 26 weeks (LTFU #2), 52 weeks (LTFU #3).The study population are predominantly low socioeconomic status (SES) and are adults ≥ 62 years of age residing in 33 participating HUD Section 202 and other publicly subsidized housing facilities. The population is diverse: African American/Black, Caucasian/White including Hispanic/Latino, and Asian. Medicaid participation among the tenants is 76% among facilities. The characteristics of the 480 in this long-term follow-up vs. the 568 participants that completed the cRCT are similar as follows: mean age 69.5±6.6 vs. 69.6±6.8; female 52% vs. 51%; Hispanic 2.9% vs. 2.7%; black 66% vs. 66%, white 29.6% vs. 29.3%; high school education or less 57% vs. 57.2%; living alone 93.7% vs. 93.3%; annual income $12, 490 or less 53.4% vs. 53.1%. Bi-variate analysis of the 480 LTFU participants vs. 88 non-participants indicate no significant differences in any of the characteristics indicating representativeness of the LTFU study population. These participants are also a diverse representation of the overall population in NE Ohio.

Analytic Plan: Analyses will follow the intent-to-treat principle; namely, the SDF and ART groups to be compared will be defined according to the randomization of sites (and original site residency of individuals). Preliminary analyses will involve the construction of one-sided 95% confidence interval based on a t-test (or other appropriate test) for the difference in mean responses (ART minus SDF) at selected long-term follow-up times. To account for clustering, a modified approach for these t-tests to allow for a possible non-zero within-facility correlation will be used. If this confidence interval lies below (for an outcome where higher is better) or above (where lower is better) the tolerance value then we may make the preliminary conclusion of 'non-inferiority' of SDF relative to ART with respect to that outcome. Confidence intervals will also be used to determine superiority of one intervention relative to the other.

To corroborate initial results, we will conduct generalized estimating equations (GEE) analyses. Specifically, for each outcome at the targeted long-term follow-up time, we will fit a GEE (marginal) model including, as covariates, an indicator variable for treatment and a selected set of baseline variables (including sociodemographic - e.g., race, gender and age, medical conditions, and oral health behavior variables) that are prognostic for the given outcome. Appropriate link functions will be used: a logit link for binary or binomial outcomes (including the arrest rate), a log link for count outcomes (number of newly missing teeth and number of newly decayed teeth), and an identity link for continuous outcomes (tooth pain score and GOHRQoL). An exchangeable working correlation matrix will be used to account for correlations among subjects within site. Inferiority and superiority (based on appropriate - possibly scale-transformed - tolerance values) will be tested using robust t tests with correction for a small number of clusters (or via corresponding 95% confidence intervals).

Secondarily, we will apply, to the repeated measures of each outcome (including initial study as well as long-term follow-up measurements) extended GEE models that include (in addition to the above mentioned covariates) time and possibly a time by treatment interaction, as well as an appropriate within-site working correlation structure (e.g., including an autocorrelation structure for the repeated measures over time). For each outcome model, we will estimate and test, via a robust t-test, the time by intervention interaction term to compare trends of the mean outcome over time for the two interventions; we will also separately assess intervention effects at multiple times in the long-term follow-up as well as consider more complex trends, e.g., including a quadratic time term. The same approach as above will be conducted for the secondary outcome, gingival inflammation. In that case, being a continuous outcome, an identity link will be used in the GEE model.

Sensitivity Analyses: We will check sensitivity of conclusions to model assumptions as well as methodological decisions such as outcome definition. For the GEE analyses, sensitivity will be assessed in part by using alternative working covariance structures and (for repeated measures analyses) alternative functions of time. For some outcomes, alternative distributions will be considered; for example, for count outcomes such as number of newly missing teeth, we will consider Poisson, negative binomial, zero-inflated Poisson (ZIP), and zero-inflated negative binomial (ZINB) distributions. Sensitivity to assumptions regarding missing data will also be investigated as described further below.

Missing Data: (a) Methods to prevent and monitor missing data: All questionnaires (using tablets) and dental assessment forms (using paper), will be managed in an electronic database (REDCap) by study staff. Weekly quality control checks will be run for outliers, entry errors, missing data, and potential data anomalies. Statistical analyses, summary and missing data reports will be generated by the study biostatistician monthly during the study. (b) Statistical Methods to Handle Missing Data and Account for Statistical Uncertainty Due to Missingness: In the preliminary approach, we will utilize complete cases for included baseline variables and will use all available measurements of each outcome for each individual; this approach assumes missing completely at random (MCAR) data, that is, missingness related only to baseline variables. In additional, a multiple imputation approach will be conducted as described below. (d) Plans to Record and Report Dropout and Missing Data: The trial data will be managed using REDCap electronic data capture platform as in the original study. REDCap is a secure web-based application providing an intuitive interface for validated data entry, audit trails for tracking data manipulation and export procedures, and automated export procedures for seamless data downloads to common statistical packages. Missing data reports will be generated weekly from REDCap for timely resolution and reporting purposes. (e) Plans to Examine Sensitivity of Inferences to Missing Data Methods: The MCAR assumption seems reasonable for missingness due to variation in measurement times (which depend on when the subjects entered the initial study), but more questionable for missingness due to death or other causes of dropout. We will test for a relationship between missingness and available (earlier) outcomes to evaluate the plausibility of the MCAR assumption. As a sensitivity analysis, we will conduct multiple imputations (relaxing MCAR and making the missing at random, MAR, assumption), for example using the multiple imputation by chained equations (MICE) algorithm with an appropriate imputation model, e.g., using logistic regression or predictive mean matching) depending on the type of outcome.

Reporting Plan: To allow for assessment of study's internal and external validity, we will adhere to standard reporting guidelines for randomized clinical trials as specified by: EQUATOR network; SPIRIT 2013 statement; and CONSORT 2010 statement.