Table 4.
Net 2-year d3mfs caries increment and estimated effects of intervention on children's net 2-year d3mfs caries increment
| Extension of a priori model adjusting for | |||||
|---|---|---|---|---|---|
| a priori model (1) | Community factors (2) | Child factors (3) | Loss to followup (4) | Varnish dose response (5) | |
| Number of communities/number of children in analysis | |||||
| Control group | 15/262 | 15/262 | 15/262 | 15/322 | 15/262 |
| Intervention group | 15/281 | 15/281 | 15/281 | 15/344 | 15/281 |
| Unadjusted net d3mfs increment per child: mean (95% CI) | |||||
| Control group | 10.1 (8.9, 11.4) | 10.1 (8.9, 11.4) | 10.1 (8.9, 11.4) | 9.8 (8.8, 10.1) | 10.1 (8.9, 11.4) |
| Intervention group | 7.7 (6.8, 8.5) | 7.7 (6.8, 8.5) | 7.7 (6.8, 8.5) | 8.0 (7.2, 8.7) | 7.7 (6.8, 8.5) |
| Adjusted net d3mfs increment per child: mean (95% CI) | |||||
| Control group | 9.9 (8.5, 11.3) | 9.7 (8.5, 10.9) | 9.4 (8.0, 10.8) | 9.6 (8.5, 10.7) | 9.9 (8.5, 11.3) |
| Intervention group | 6.9 (5.5, 8.2) | 6.2 (5.0, 7.4) | 7.0 (5.6, 8.3) | 7.3 (6.2, 8.4) | |
| 0–3 varnish applications versus control | 7.1 (4.4, 9.8) | ||||
| 4 varnish applications versus control | 6.2 (4.2, 8.2) | ||||
| 5 varnish applications versus control | 7.1 (5.3, 8.9) | ||||
| 6–8 varnish applications versus control | 8.6 (3.7, 13.5) | ||||
| Effect estimates: difference in adjusted net d3mfs increment per child: mean (95% CI) | |||||
| Efficacy of intervention versus control group | −3.0 (−4.9, −1.2) | −3.5 (−5.1, −1.9) | −2.4 (−4.3, −0.6) | −2.3 (−3.7, −0.8) | |
| Effect of additional 1ppm F | −4.3 (−7.0, −1.6) | ||||
| Effect of age (years) | −0.3 (−0.3, −0.2) | ||||
| Effect of baseline d3mfs (no. of surfaces) | 0.5 (0.4, 0.6) | ||||
| Prevented fraction | 31% | 36% | 26% | 24% | |
| Intra-cluster correlation coefficient | 0.02 | 0.00 | 0.02 | 0.01 | 0.02 |
(1) A priori model is complete case, intent-to-treat analysis using multi-level linear regression model adjusted for fixed effect of stratum and random effect of communities.
(2) Addition to a priori model of fluoride concentration in drinking water as fixed effect covariate.
(3) Addition to a priori model of child's age and baseline d3mfs as fixed effect covariates.
(4) A priori model applied to 666 subjects by adding regression-imputation values of d3mfs increment for 123 children lost to follow-up. This produced an imputed intent-to-treat analysis of all randomized subjects.
(5) Replacement of binary study group from a priori model with dummy variables for control group and four categories of fluoride varnish applications.