Compensation Method for Missing Wear Data in Long-term Clinical Trials

Introduction:  In the few existing long-term clinical trials on posterior composites, patient attrition (typically 5-10% per year) substantially limits quantity and true quality of results.   Suspicions exist that many missing values underestimate actual changes.  Objective:   Assess the impact of missing values using a novel estimation method for missing values in clinical trials of posterior composite wear over 10-years.  Methods :  Two composites (1981-1991 FulFil [F70], conventional; 1998-2008, SureFil, packable; Dentsply-Caulk) were placed previously with conventional techniques, recalled (0.5-1-2-3-5-10y; 0.5-1-2-4-5-10y), and monitored with impressions for wear (Leinfelder method, baseline-subtracted, 3 highly-calibrated clinicians). Restorations were Class I and II in first or second molars with at least half of all possible wear values for inclusion.  Regression of occlusal wear for each restoration versus log time (y=m(ln(x)+b) was determined to calculate missing values.  Artificially completed data sets (W2) were compared to missing-value sets (W1) by one-way ANOVA (p≤0.05, Bonferroni).  Results :  Means (X) and standard deviations (sd) are reported below with no significant differences (>0.05) between W1 and W2 for any recall for either composite. Surprisingly, increased values had little effect. While insignificant, the 10y W2 packable trend indicated a potential overestimate.     Wear (µm)   Conventional Packable Recall Period W1 (missing values) [n], X±sd W2 (completed) [n], X±sd W1 (missing values) [n], X±sd W2 (completed) [n], X±sd 0.5y [31], 80±53a [32], 79±50a [59] 18±24a [51] 18±25a 1y [36], 132±62a [32], 131±63a [56], 16±22a [51], 17±22a 2y [28], 150±64a [32], 154±61a [45], 25±31a [51], 31±34a 3y [28], 179±60a [32], 181±57a [45], 68±41a [51], 65±41a 4y ------ ------ [36], 52±46a [51], 49±42a 5y [29], 203±72a [32], 193±75a ------ ------ 10y [16], 224±63a [32], 224±68a [17], 145±75a [51], 98±74a           W2 average r2 fit ------ 0.73±0.28 ------ 0.57±0.28 Conclusion:   Added values from estimations for long-term trials did not affect results. Acknowledgment: Supported indirectly by Dentsply-Caulk.