Application of Latent Growth Curve Modelling in Dental Research

Objectives: Many dental research questions require repeated measurements of variables. Nevertheless, the statistical analysis of longitudinal data requires advanced, sophisticated methods to explore the complexity of information within the data. Currently, the most often used advanced statistical methodology in dental research is multilevel modelling (MLM). The aim of this study is to use two examples to introduce an alternative methodology known as latent growth curve modelling (LGCM). Methods: The first example is repeated measures of ramus length in a group of boys aged from 8 to 10. We show the similarity between LGCM and MLM, and the advantages of LGCM over MLM in terms of the flexibility of statistical modelling. Another example is repeated measures of distance between Sella and Pterygomaxillary Fissure in children from age 8 to 14. We show how to use to test research hypothesis that can be difficult to do using traditional statistical methods. Results: The first example shows that when the residual variances of repeated measures of ramus length are constrained to be constant, LGCM and MLM will give rise to similar results. However, it is easy to relax this constraint in LGCM to overcome the problem of inconstant residual variances, e.g. the measurement errors are not constant. The second example shows that when there is multiple groups, LGCM can estimate different coefficients and covariance structure for each group and test whether this will give a better fit to the data than constraining these parameters to be equal across groups. Conclusions: LGCM provides a flexible framework of statistical modelling and allow dental researchers to test their research hypothesis in a novel way. Further extension of LGCM includes latent class growth analysis and general growth mixture modelling, and these recently developed methodologies have a great potential to provide new directions for future research.