Maternal blood pressure trajectories and associations with gestational age at birth: a functional data analytic approach

Abstract

Background: Research has revealed group-level differences in maternal blood pressure trajectories across pregnancy. These trajectories are typically constructed using clinical blood pressure data and multivariate statistical methods that are prone to bias and ignore the functional, dynamic process underlying a single blood pressure observation. The aim of this study was to use functional data analysis to explore blood pressure variation across pregnancy, and multivariate methods to examine whether trajectories are related to gestational age at birth. Methods: Clinical blood pressure observations were available from 370 women who participated in a longitudinal pregnancy cohort study conducted in Montreal, Quebec, Canada. Functional data analysis was used to smooth blood pressure data and then to conduct a functional principal component analysis to examine predominant modes of variation.Results: Three eigenfunctions explained greater than 95% of the total variance in blood pressure. The first accounted for approximately 80% of the variance and was characterized by a prolonged-decrease trajectory in blood pressure; the second explained 10% of the variance and captured a late-increase trajectory; and the third accounted for approximately 7% of the variance and captured a mid-decrease trajectory. The prolongeddecrease trajectory of blood pressure was associated with older, and late-increase with younger gestational age at birth. Conclusion:Functional data analysis is a useful method to model repeated maternal blood pressure observations and many other time-related cardiovascular processes. Results add to previous research investigating blood pressure trajectories across pregnancy through identification of additional, potentially clinically important modes of variation that are associated with gestational age at birth.

Association between perinatal mental health and reproductive and neonatal complications: a retrospective birth cohort study

“…Similar to results from our study, prenatal anxiety and depressive disorders have also been shown to significantly increase a person’s risk for a HDP diagnosis, and in some instances result in a shorter gestational period (37). Results from a prospective Mexican cohort showed that anxiety, sleep or social dysfunction, and acute somatic symptoms at 20 weeks of gestation were associated with a higher risk of developing HDP (38).…”

scikit-fda: A Python Package for Functional Data Analysis

“…Functional data analysis (FDA) is the branch of statistics that deals with observations varying over a continuous parameter, such as curves, surfaces, and other types of functions (Cuevas 2014;Wang, Chiou, and Müller 2016). These types of data appear in many different fields, such as biology (Cremona, Xu, Makova, Reimherr, Chiaromonte, and Madrigal 2019), demographics (Hyndman and Shahid Ullah 2007), economics (Frois Caldeira, Gupta, Suleman, and Torrent 2020), energy security (Gong, Wang, and Lin 2021), genomics (Leng and Müller 2006;Chen, Cremona, Qi, Mitra, Chiaromonte, and Makova 2020), medicine (Sørensen, Goldsmith, and Sangalli 2013;Ferrando, Ventura-Campos, and Epifanio 2020;Horsley, Ramsay, Ditto, and Da Costa 2021), meteorology (Beyaztas and Yaseen 2019), oceanography (Assunção et al 2020), traffic control (Wagner-Muns, Guardiola, Samaranayke, and Kayani 2018;Hu, Yuan, Zhu, Yang, and Xie 2019), and other areas of application (Ullah and Finch 2013). The functional nature of these data and, in particular, their continuous structure, entails important differences with respect to the classical multivariate statistics.…”