Title: 0577 - Accounting for Missing Subjects in a Practice-Based Caries Management Study
John Featherstone (Presenter)
University of California - San Francisco
Benjamin Chaffee, University of California San Francisco
Peter Rechmann, University of California - San Francisco
Objectives: In a recent practice-based research network (PBRN) clinical trial of caries management by risk assessment (CAMBRA), marked reductions in percentage of patients at high risk and with new clinical caries were observed at 12, 18, 24 months (Rechmann et al., Adv Dental Res, 2018). The objective of the present study was to evaluate how statistically accounting for participant drop-out would affect the results reported from the CAMBRA-PBRN trial.
Methods: Twenty-one dentists participated in a randomized, controlled, parallel-arm, double-blind clinical trial with individual-level assignment of 460 patients to standard-of-care (control) versus active CAMBRA treatment (intervention). Control or active antimicrobial and remineralizing agents were dispensed at baseline, and 6-monthly recall visits according to risk level and treatment arm. Primary outcome was caries risk level at recall. Secondary outcome was clinical disease indicators. Generalized estimating equations were used to fit log-linear models for each outcome while accounting for repeated measurements. Different imputation approaches were used to account for missing observations, including last-observation carried-forward and multiple imputation (MI), with or without baseline treatment category as a possible predictor of the missing outcomes.
Results: At 24 months, 181 patients remained in the study. Of those classified at high caries risk at baseline, 25% in the intervention and 54% in the control group remained at high risk in the observed data (RR=0.42, P=0.003). MI yielded similar estimates when baseline treatment category was used as a predictor (RR=0.47, P<0.001) but attenuated results when imputation was stratified by treatment category (RR=0.62, P=0.008). Observed caries clinical measures were 35% and 49% (P=0.12) for intervention and control, respectively, and 28% and 42% by MI (P=0.05).
Conclusions: Multiple imputation produced similar conclusions as the observed data for this CAMBRA-PBRN clinical trial (marked reductions in caries risk and clinical caries measures). Numeric findings were sensitive to the MI method. ClinicalTrials.gov NCT01176396.
This abstract is based on research that was funded entirely or partially by an outside source:
PacifiCare/UnitedHealthcare, DentaQuest, UCSF and California Dental Association.
The submitter must disclose the names of the organizations with which any author have a relationship, the nature of the relationship, and the clinical or research area involved. The following is submitted: NONE