Fetal Growth Percentile Software: a tool to calculate estimated fetal weight percentiles for six standards

Objective

A supplement titled “Fetal Growth: Evaluation and Management” that featured six fetal growth standards was published in February 2018 in this Journal.¹ These standards were derived from different populations and were based on either fetal or neonatal data. Furthermore, these standards were proposed either as 1) one-fits-all, 2) customized by race/ethnic group and other maternal and fetal characteristics, or 3) fetus-specific (individualized). Achieving consensus about which standard is more suitable for clinical use requires evaluation of the predictive performance of available fetal growth standards with respect to clinically relevant outcomes.² However, such comparisons are hindered by the lack of a simple-to-use software tool that implements most of the growth standards in use. Therefore, the aim of this work was to develop a spreadsheet-based estimated fetal weight (EFW) percentile calculator and corresponding R software package to encompass six fetal growth standards. This collection of software tools is referred to herein as Fetal Growth Percentile Software (FetalGPS).

Study Design

Of the six fetal growth standards that have been implemented in FetalGPS, the INTERGROWTH-21^st, World Health Organization (WHO), National Institute of Child Health and Human Development (NICHD), and Perinatology Research Branch (PRB/NICHD) standards were previously discussed in the AJOG supplement.¹ In addition to these standards, we have also implemented the Hadlock et al.³ and Fetal Medicine Foundation (FMF)⁴ standards. The customized Gestation-Related Optimal Weight (GROW)⁵ standard was not re-implemented herein because it is already available as a spreadsheet calculator to be used for one pregnancy at a time (Individual Centile Calculator) or to assess up to 50,000 cases (Bulk Centile Calculator) (https://www.gestation.net/). Similarly, the Individualized Growth Assessment (IGA)⁶ approach was omitted because it requires longitudinal data and already has feature-rich, freely available implementation (https://igap.research.bcm.edu/).

For the standards based on normality assumptions (Hadlock et al,³ INTERGROWTH-21^st, and FMF) for which the mean and standard deviation of EFW at each gestational age were reported, exact percentiles were derived from the normal distribution. For the remaining three standards that were derived without assuming data normality (WHO, PRB/NICHD) or did not report mean and standard deviation values throughout gestation (NICHD), linear interpolation was used to derive more specific percentiles, as previously described.⁷ To determine the accuracy of interpolation, we considered the case of the PRB/NICHD standard for which the underlying data were available, and the equations of the 2.5^th, 5^th, 10^th, 25^th, 50^th, 75^th, 90^th, 95^th, and 97.5^th percentiles of EFW as a function of gestational age and maternal characteristics were previously reported and used herein as basis for interpolation. We compared the percentiles obtained by interpolation to the actual values determined by fitting additional intermediary centile curves (3.5^th, 7.5^th, 92.5^th and 96.5^th) by using the same methods and data utilized in the original study.⁸

FetalGPSX was implemented by using Visual Basic for Applications (VBA) in Microsoft Excel (Microsoft Corp., Santa Rosa, CA, USA), while the corresponding R software (www.r-project.org) library, FetalGPSR, was implemented in R to allow for command line use by those familiar with R.

Results

FetalGPSX and FetalGPSR require as input gestational age at the time of ultrasound examination and either EFW or primary fetal biometry measurements. For customized standards, maternal weight, height, parity, and race/ethnicity as well as fetal sex may be required. When individual ultrasound parameters are provided as input (instead of EFW), the software automatically selects the appropriate EFW calculation formula (a three- or four-parameter Hadlock equation for Hadlock, WHO, NICHD, PRB/NICHD, and FMF standards, and a two-parameter equation for INTERGROWTH-21^st). Given that two (WHO, PRB/NICHD) of the six standards do not rely on the data normality assumption, and that the NICHD study did not provide standard deviation values as a function of gestational age, we have used linear interpolation and extrapolation to derive the exact percentiles for these standards at any gestational age.

Two versions of the Excel spreadsheet-based calculator are available from the author’s website (https://bioinformaticsprb.med.wayne.edu/software/). FetalGPSX is designed to compute EFW percentiles for a few observations, typically collected from the same pregnancy, and to display EFW values against the desired standard (Figure 1, left panel). By contrast, FetalGPSXL calculates percentiles according to a specified standard for large datasets (up to 10,000 measurements).

Figure 1: FetalGPS software tools for calculation of estimated fetal weight percentiles.

The Excel spreadsheet-based calculator FetalGPSX (left panel) and the R package FetalGPSR (right panel) are shown. FetalGPSX requires selection of the desired growth standard (e.g. NICHD), gestational age at scan, and estimated fetal weight (EFW) or fetal biometry measurements to calculate EFW; the software may also possibly require additional covariates specific to each standard (e.g. maternal race/ethnicity for NICHD). The percentiles of EFW are calculated and plotted against the 5^th, 10^th, 50^th, 90th, and 95^th percentile curves. The FetalGPSR package can be installed from the GitHub repository in the R software using the install_github function. Using the package requires a) reading a table containing the same input data described above for FetalGPSX, and b) calculating the percentiles for all standards using the FetalGPS function.

The R package FetalGPSR contains a single function that takes as input the relevant covariates in a tabular format and returns additional columns with the percentiles according to all standards implemented. The output can then be easily saved and made available for downstream analyses in R or another statistical package (Figure 1, right panel).

Conclusion

FetalGPS software tools were introduced herein to streamline the assessment of fetal growth according to six fetal growth standards. Not only can observed EFW values be plotted against predefined centile lines of a given standard (e.g. 5th, 10th) but also interpolated percentile values are provided to enable custom analyses such as building Receiver Operating Characteristic curves for prediction of relevant outcomes. These tools are expected to facilitate reproducible research in the field of fetal growth assessment and to allow rapid evaluation of multiple standards based on the same input dataset.