Nonlinear Regression on Growth Curves for Placental Parameters in R

Growth charts play a crucial role in the evaluation and surveillance of paediatric populations, serving as indispensable tools for paediatricians and public health researchers. The development of these growth charts for fetal parameters has been extensively used in recent decades. However, investigation of placental parameters and their relationship with obstetric outcome has been relatively neglected, resulting in a significant gap in understanding their biological significance. This study presents an alternative approach for constructing reference growth curves specific to the placental parameter, Diameter 2. Our methodology uses the generalized additive models for location, scale, and shape (GAMLSS), offering distinct advantages over traditional quantile regression methods. One of the key advantages of GAMLSS is its flexibility to accommodate any statistical distribution, allowing for the modelling of various parameters that characterize the distribution of the response variable. Through the application of our proposed methodology, we demonstrated that by using P-splines as a smoothing function and Box-Cox t (BCT) as a distribution, we can achieve a representative growth curve for the Diameter 2 of the placenta throughout gestational age (GA). The resulting models demonstrated high representativeness, with \(R^2\) values of 0.7608 and 0.7673 and AIC scores of 7953 and 7946 for the best two models, respectively. Moreover, our approach has the ultimate goal to facilitate early diagnosis of fetal complications, thereby providing valuable assistance to healthcare professionals.