Page 106 - Craniomaxillofacial Implant Surgery - Jeroen P.J. Dings

104
Chapter 6
Mixed models were used to determine the influence of appreciation of the different anatomical subunits on the overall view of the reconstructed nose or ear. A proportion of the autologous and prosthetic nasal reconstructions (94.7% and 23.1%, respectively) comprised partial reconstructions. The non-reconstructed anatomical subunits were not included in the questionnaires. Multiple imputation was used to handle missing data in the case of partial reconstruction, by selecting a random score for a control patient given by the same respondent. This procedure was repeated 10 times, resulting in 10 complete data sets. Multiple imputation was used instead of single imputation to reflect the uncertainty in the estimation of the distribution, resulting in unbiased estimates with correctly estimated standard errors and confidence intervals. The results of the analyses of each dataset were combined 19. A stepwise selection procedure was used to find a sparse but sufficiently accurate mixed model to describe the influence of separate anatomical subunits on the overall view of the reconstruction. A model with all anatomical subunits and all interactions between type of reconstruction and anatomical subunit was fitted, and the non-significant interactions (P > 0.05) were individually removed from the model. Anatomical subunits that did not show a significant interaction with type of reconstruction were also removed from the model. The final model is described in more detail below.
Interquartile ranges (P25–P75) were used to calculate the influence of a nasal or auricular anatomical subunit on the score for the general appearance of a nasal or auricular reconstruction. The influence of the anatomical subunit on the overall appearance was assessed by calculating the difference in the mean (predicted) score of a subject with an anatomical subunit value equal to the third quartile minus the mean score of the anatomical subunit equal to the first quartile, taking into account the variation of that anatomical subunit (the larger the difference, the larger the influence). Comparisons returning a P value < 0.05 were considered statistically significant. Differences of 10 units on the VAS scale and 0.5 units on the Likert scale were considered clinically meaningful. All analyses were performed using SAS® version 9.4 (SAS institute Inc., Cary, NC, USA).
RESULTS
The mean values obtained for the questions on general appearance and characteristics were evaluated for each reconstruction group, according to the anatomical location. These data were analysed separately for each respondent group, as summarised in Tables 3 and 4. A mixed model analysis revealed the differences between these scores (Table 5).