Background: This work aims to study the spatio-temporal evolution of a woman's age at menarche in the central region of Portugal. One of the concerns of the study is early or late menarches; thus, we consider percentile regression to build the respective curves as opposed to the more traditional mean regression approach.
Methods: We analysed the data from [Formula: see text] women born in the period 1920-1973 who attended a free breast cancer screening program between 1990 and 2019. Distributional regression models inside the package GAMLSS in R were considered. These methods allowed us not only to model the location (mean) of the specific probability distribution of the age at menarche, but also allowed for the scale (variance) parameter of this distribution to depend on covariates. Additionally, a spatial random-effect was considered in order to capture the correlation at the regional level. The obtained clustered spatial effects were analysed to assess geographical differences among the percentiles of the age at menarche by year of birth.
Results: A decreasing trend in the age at menarche (about 1.5 years in 5 decades) and regional differences for all the considered percentiles were found. Women living in the north-central areas of the central region of Portugal tend to have menarche at older ages.
Conclusion: We obtained percentile estimates for the age at menarche by year of birth and region of residence and demonstrated that these two explanatory variables have an impact on the explanation about the decreasing trend in age at a woman's first menstruation.
Keywords: Distributional regression; GAMLSS; Menarche; Portugal; Spatial; Temporal.
© 2023. BioMed Central Ltd., part of Springer Nature.