Anti-Caries Agents and Dental Caries Among High-Risk Adults
Anti-Caries Agents and Dental Caries Among High-Risk Adults
We calculated doubly-robust adjusted estimates for caries outcomes according to categories of anti-caries therapy using g-computation and inverse probability treatment weighting in a combined approach. This technique has been described approachably in recent publications. We fitted regression models for caries outcomes (negative binomial model for DFT increment and logistic model for DFT increment >0), where anti-caries therapy category was the exposure variable and baseline covariates were age, sex, race/ethnicity, payer type, baseline number of teeth, baseline number of decayed teeth, calendar year, provider program, provider years in training, and follow-up time. Models were used to predict adjusted marginal caries outcomes under each category of anti-caries therapy received, setting follow-up time to 18-months (548 days), the mean value in the follow-up sample. Regression models were weighted using inverse probability treatment weights to enhance robustness to model misspecification and using inverse probability censoring weights to account for losses to follow-up from the baseline sample. We multiply imputed missing baseline data (0.2 % of covariate data among eligible participants) and averaged point estimates over 25 imputations. Results were unchanged in a sensitivity analysis restricted to cases with complete baseline covariate data.
Estimates represent the expected DFT increment associated with each level of anti-caries therapy under the same distribution of participant characteristics that was observed in the baseline population and with equal follow-up time (18 months). As measures of association, we computed the difference in DFT increment and ratio in the percentage of patients with DFT increment >0, given single or repeated delivery of anti-caries therapy, as two separate pair-wise comparisons, each with respect to no therapy received. We used the percentile bootstrap method (3000 bootstrap re-samples) to obtain 95 % confidence intervals (CI) and considered results to be statistically significant at the 0.05 level if the 95 % CI for measures of association excluded the null value. Analyses were performed using statistical software (Stata 13.1, StataCorp LP, College Station, United States and R 3.1.2, R Foundation for Statistical Computing, Vienna, Austria). Study reporting followed the STROBE statement (Additional file 1 http://www.biomedcentral.com/1472-6831/15/111/additional) .
Statistical Approach
We calculated doubly-robust adjusted estimates for caries outcomes according to categories of anti-caries therapy using g-computation and inverse probability treatment weighting in a combined approach. This technique has been described approachably in recent publications. We fitted regression models for caries outcomes (negative binomial model for DFT increment and logistic model for DFT increment >0), where anti-caries therapy category was the exposure variable and baseline covariates were age, sex, race/ethnicity, payer type, baseline number of teeth, baseline number of decayed teeth, calendar year, provider program, provider years in training, and follow-up time. Models were used to predict adjusted marginal caries outcomes under each category of anti-caries therapy received, setting follow-up time to 18-months (548 days), the mean value in the follow-up sample. Regression models were weighted using inverse probability treatment weights to enhance robustness to model misspecification and using inverse probability censoring weights to account for losses to follow-up from the baseline sample. We multiply imputed missing baseline data (0.2 % of covariate data among eligible participants) and averaged point estimates over 25 imputations. Results were unchanged in a sensitivity analysis restricted to cases with complete baseline covariate data.
Estimates represent the expected DFT increment associated with each level of anti-caries therapy under the same distribution of participant characteristics that was observed in the baseline population and with equal follow-up time (18 months). As measures of association, we computed the difference in DFT increment and ratio in the percentage of patients with DFT increment >0, given single or repeated delivery of anti-caries therapy, as two separate pair-wise comparisons, each with respect to no therapy received. We used the percentile bootstrap method (3000 bootstrap re-samples) to obtain 95 % confidence intervals (CI) and considered results to be statistically significant at the 0.05 level if the 95 % CI for measures of association excluded the null value. Analyses were performed using statistical software (Stata 13.1, StataCorp LP, College Station, United States and R 3.1.2, R Foundation for Statistical Computing, Vienna, Austria). Study reporting followed the STROBE statement (Additional file 1 http://www.biomedcentral.com/1472-6831/15/111/additional) .
Source...