Multiple Failure-time Survival Analysis Impact on Results About Caries Sealing

Objectives: Survival analyses are useful for analyzing time-related events. Currently, failures/events related to dental treatments have been considered until the 1^stfailure happens. However, in the real life, treatments may fail more than once and might demand reinterventions after that. Thus, this study aimed to evaluate what was the actual impact of using a multiple failure-time survival analysis when testing the efficacy of caries sealing compared to restorations in occlusal moderate caries lesions (ICDAS 3-4) in primary teeth.
Methods: This study is part ofa non-inferiority clinical trial (NCT03005405) comparing sealing (no caries removal) vs. restorations both using high viscous glass ionomer cement (HVGIC). Children were followed-up for 24months and treated teeth assessed biannually. Major failures (need of replacements) was set as the endpoint. Failed sealants/restorations were replaced using the same baseline intervention, whenever necessary. Survival analyses were performed to verify the efficacy of sealing caries lesions instead restoring them. The Cox regression analyses were performed using single first failures (shared frailty models) and multiple ordered failures for 24months (conditional risk set model). Hazard Ratios (HR) were calculated. Independent variables were used to adjustment.
Results: 103 restorations and 101 sealants were included. 18% of restorations and 27% of sealants required replacement after the first intervention, but no difference was observed when only the single 1^stfailure was considered (HR=1.50; p=0.10) and similar mean time (t=23months) until 1^stfailure was observed among treatments. Including reinterventions, the failure rates increased to 22% for restorations and 37% for sealants, evidencing a statistically significant difference between groups (HR=1.97; p=0.04).
Conclusions: Using multiple failure-time events, the sealing of moderate caries lesions on occlusal surfaces of primary molars is inferior to restorations in terms of need of replacement, what cannot be evidenced when single failures analysis is used. That is why multiple events, when applicable, should be considered when testing efficacy of treatments.