Competing‐risks model for prediction of small‐for‐gestational‐age neonate from maternal characteristics and serum pregnancy‐associated plasma protein‐A at 11–13 weeks' gestation

Abstract

This study describes a new competing-risks model based on a combination of maternal characteristics and medical history with serum pregnancy-associated plasma protein-A (PAPP-A) at 11-13 weeks' gestation for prediction of a small-for-gestational-age (SGA) neonate. PAPP-A likelihood was expressed as a continuous function of both gestational age at delivery and birth-weight Z-score in the same model. What are the clinical implications of this work? Addition of serum PAPP-A improves the performance of screening for a SGA neonate achieved by maternal factors alone and demonstrates the methodology for incorporation of further biomarkers into a single model that can be used numerous times during the course of pregnancy to predict SGA of any severity of smallness and degree of prematurity.

Competing‐risks model for prediction of small‐for‐gestational‐age neonate from maternal characteristics, serum pregnancy‐associated plasma protein‐A and placental growth factor at 11–13 weeks' gestation

“…The strengths of this study are, first, the large number of prospectively collected data as a part of a screening program, second, use of a folded-plane model that best describes the distribution of PlGF and PAPP-A, third, use of a continuous joint probability model that allows estimation of patient-specific risks for any desired Table 6 Comparison of detection rate (DR), at fixed false-positive rate of 10%, of all small-for-gestational-age (SGA) (birth weight < 3 rd percentile) cases, SGA with pre-eclampsia (PE) or SGA without PE, with delivery at < 32, < 37 or ≥ 37 weeks' gestation, in screening by different combinations of maternal factors (MF), pregnancy-associated plasma protein-A (PAPP-A) and placental growth factor (PlGF) SGA definition, and, fourth, use of Bayes' rule in an update process that can be repeated at any stage of pregnancy. The new model is stable and better than other screening methods, as we have demonstrated previously 20,21 . Therefore, we opted not to carry out an internal validation in this study.…”

“…The strengths of this study are, first, the large number of prospectively collected data as a part of a screening program, second, use of a folded‐plane model that best describes the distribution of PlGF and PAPP‐A, third, use of a continuous joint probability model that allows estimation of patient‐specific risks for any desired SGA definition, and, fourth, use of Bayes' rule in an update process that can be repeated at any stage of pregnancy. The new model is stable and better than other screening methods, as we have demonstrated previously 20,21 . Therefore, we opted not to carry out an internal validation in this study.…”

“…We fitted likelihood functions for each biomarker, conditional to birth‐weight Z ‐score and GA at delivery. We adopted a folded‐plane regression method, as an extension of the broken‐stick regression, described in detail in a previous study 21 . The combination of PAPP‐A and PlGF was achieved by a multivariable Gaussian distribution that was fitted to the log 10 multiples of the median (MoM) values of the biomarkers, assuming a constant covariance matrix.…”

“…The new FMF charts, by simply recognizing the self-evident truth that birth weight and ultrasonographically estimated fetal weight are measuring the same quantity in the same person, revealed that prematurity and smallness are related. In the new model, SGA is now expressed as a continuous two-dimensional outcome, consisting of GA at delivery and birth weight, which are associated and jointly distributed 20,21 . Another manifestation of this association is the progressively lower level for PAPP-A and PlGF for both lower birth-weight Z-scores and GAs, described by the likelihoods of the model.…”

“…We have developed a new method for SGA prediction, analogous to the competing-risks model in the assessment of the risk for pre-eclampsia (PE) [15][16][17][18][19] . The first step of this novel methodology is a competing-risks approach, based on the joint distribution of birth-weight Z-score and gestational age (GA) at delivery, according to maternal factors, and the second step is the process of updating this distribution by including the likelihood of biomarkers according to Bayes' theorem 20,21 . The new model has clinical merits and it is supported by new methodological concepts: first, the severity of smallness and the burden of prematurity, the two dimensions of SGA that define its outcome [22][23][24][25] , are expressed continuously and jointly in the same model; second, prediction of SGA for any chosen cut-off is feasible and SGA is regarded as a two-dimensional spectrum disorder rather than a fragmented disease 20,21 ; third, the capacity to add more biomarkers in the same model according to Bayes' theorem is feasible; and, fourth, the new model is superior and more stable compared to the RCOG Green-top guideline and the logistic regression models, a fact that has been proven in previous studies through a process of vigorous internal validation 20,21 .…”