gigs: A package for standardizing fetal,neonatal,and child growth assessment with extensions to egen

Abstract

In this article, we describe gigs, the Guidance for International Growth Standards package for Stata. gigs contains multiple egen functions for converting between fetal biometry or anthropometric measurements and z scores or centiles in both the World Health Organization child growth standards and the INTERGROWTH-21st standards. We also describe an additional Stata command that wraps these functions to provide growth outcome classification for newborns and infants up to five years of age using a suite of international growth standards. Clear and consistent commands with the functionality described in this article have not been available to Stata users prior to the release of this article. These features will be instrumental for standardizing methods for growth assessment in fetal, newborn, and child health research, as well as implementation, clinical practice, and population-level comparisons.

1 Introduction

Clinicians, researchers, and other key stakeholders require robust tools for statistical description and classification of infant and child growth. Prescriptive growth standards, which describe how a child “should” grow given optimal conditions, should be used to identify nutritionally at-risk infants and inform targeting of therapeutic interventions (Bertino et al. 2007; Ohuma and Altman 2019a). Currently, two sets of global prescrip­tive standards exist: the World Health Organization (who) child growth standards for term infants from birth through five years of life (World Health Organization 2006, 2007) and the International Fetal and Newborn Growth Consortium for the 21st Cen­tury (lntergrowth-21st) standards. However, there is no clear guidance on when and how to use these standards, which has implications on clinical decision-making and estimation of growth indicators at the national and global levels.

The Guidance for International Growth Standards (gigs) project aims to provide guidance on when to use these standards and statistical software tools that implement these standards. The gigs package for Stata can transform raw anthropometric data into z scores or centiles and vice versa. It also includes gigs_classify_growth, a command for categorizing newborn and infant growth data into commonly assessed growth outcomes.

1.1 INTERGROWTH-21st standards

intergrowth-21st was a multicenter, multiethnic, population-based project designed to study fetal, preterm, and newborn growth. Its primary aim was to develop a suite of prescriptive growth standards, each of which represented possible growth under optimal conditions (Uauy et al. 2013). These standards were intended to complement and extend existing growth standards from the who Multicentre Growth Reference Study (mgrs) (Villar et al. 2013).

1.1.1 Newborn size

We have included seven intergrowth-21st newborn size standards in gigs, which here incorporate the standards for very preterm (Villar et al. 2016) and preterm or term newborn size (Villar et al. 2014a) in weight, length, and head circumference for gestational age (ga). We also include the body composition standards (Villar et al. 2017). Table 1 describes the range of gas for which we implemented conversion functions in each standard.

Table 1.

intergrowth-21st newborn size standards

acronym	Description	Unit	gest_days() range
wfga	Weight for ga	kg	168 to 300 days
lfga	Length for ga	cm	168 to 300 days
hcfga	Head circumference for ga	cm	168 to 300 days
wlrfga	Weight-for-length ratio for ga	kg/cm	168 to 300 days
ffmfga a	Fat-free mass for ga	kg	266 to 294 days
bfpfga a	Body fat % for ga	%	266 to 294 days
fmfga a	Fat mass for ga	kg	266 to 294 days

a

Outputs may deviate slightly from published values because we had to reconstruct these standards from published centile tables using polynomial regression.

The preterm or term newborn size standards were based on singleton birth data from 20,486 mothers from sites in Brazil, China, India, Italy, Kenya, Oman, the uk, and the us, in which all newborn growth was unlikely to be constrained by environmental factors or by unmet maternal nutritional or health needs (Villar et al. 2014a,b). The very preterm newborn size standards were based on a smaller population of 408 very preterm (168 to 231 days’ ga) neonates, with some risk factors for fetal growth restriction but no evidence of congenital malformations or ultrasound evidence of fetal growth restriction before birth (Villar et al. 2016).

Implementation details for each standard vary based on the measurement type and ga (Ohuma and Altman 2019b). For the weight, length, and head circumference for ga standards in very preterm newborns (168 days < ga < 231 days), fractional poly­nomial regression was used to fit equations for the mean and standard deviation of each measurement at different gas (Villar et al. 2016; Ohuma and Altman 2019b; Royston and Altman 1994). For these standards in preterm or term newborns (231 days < ga < 300 days), the growth curves are modeled using the skew t type 3 distribution (Jones and Faddy 2003) with four parameters—mu (μ), sigma (σ), nu (v), and tau (τ)—for each ga and sex (Villar et al. 2014a; Ohuma and Altman 2019b).

Villar et al.’s (2014a) standards for newborn weight, length, and head circumference were fit using the gamlss (general additive models for location, scale, and shape) package for R (Rigby and Stasinopoulos 2005; Ohuma and Altman 2019a; R Core Team 2023). In R, the gamlss.dist package provides the qST3() and pST3() functions for conversion between measurements and centiles in a skew t type 3 distribution, but no similar function exists for Stata. We therefore adapted the methods used in these R functions for use in our extensions to egen.

The weight-for-length ratio standard was also fit using fractional polynomial re­gression, and the equations for the mean and standard deviation at different GAs are available in an Excel file from the intergrowth-21st project website (Villar et al. 2017). However, the same equations are not available for the body composition stan­dards (fat-free mass, body fat percentage, and fat mass for GA; ffmfga, bfpfga, and fmfga, respectively). Therefore, we derived new equations for these standards by apply­ing polynomial regression to the published ffmfga, bfpfga, and fmfga centile tables in R 4.3.1 (R Core Team 2023). Consequently, the ffmfga, bfpfga, and fmfga standards available in gigs deviate slightly from the published values when converting between measurements and z scores or centiles. When we reconstructed the centile tables for each standard based on our own equations, the maximum differences across all centiles and gas were less than six grams for ffmfga, less than 0.11% for bfpfga, and less than two grams for fmfga).

1.1.2 Postnatal growth

The intergrowth-21st postnatal growth standards used longitudinal data from 201 live singletons born between 26 and less than 37 weeks’ gestation without congenital malformations, fetal growth restriction, or severe postnatal morbidity from the same geographic range as the population used to construct the intergrowth-21st fetal standards (Papageorghiou et al. 2014b; Villar et al. 2015). These standards were con­structed by modeling the mean and standard deviation across different postmenstrual ages (pmas) using fractional polynomial regression (Royston and Altman 1994). It was found that by 64 weeks’ pma, these standards overlapped with the who child growth standards. Table 2 describes the acronym, unit, and xvar() range of each intergrowth-21st Postnatal Growth standard included in gigs.

Table 2.

INTERGROWTH-21st postnatal growth standards for preterm infants

acronym	Description	Unit	xvar() range
wfa	Weight for age	kg	27 to 64 weeks PMA
lfa	Length for age	cm	27 to 64 weeks PMA
hcfa	Head circumference for age	cm	27 to 64 weeks PMA
wfl a	Weight for length	kg	35 to 65 cm

a

The postnatal weight-for-length standards for preterm infants included in gigs were derived from INTERGROWTH-21st data and provided by the INTERGROWTH- 21st team (Eric Ohuma); they have not been previously published by the INTERGROWTH-21st project.

1.1.3 Fetal standards

The intergrowth-21st fetal standards included in gigs are from several publications, covering 21 individual standards (table 3). The standards are based on data collected as part of the intergrowth-21st Fetal Growth Longitudinal Study (fgls), which is described fully in Villar et al. (2013, 2014b). In brief, data from healthy singleton pregnancies in educated, affluent women were collected from eight geographically defined urban populations and pooled to construct fetal growth standards.

Table 3.

intergrowth-21st fetal standards

acronym	Description	Unit	xvar() range
hcfga	Head circumference for ga	mm	98 to 280 days
bpdfga	Biparietal diameter for ga	mm	98 to 280 days
acfga	Abdominal circumference for ga	mm	98 to 280 days
flfga	Femur length for ga	mm	98 to 280 days
ofdfga	Occipito-frontal diameter for ga	mm	98 to 280 days
efwfga	Estimated fetal weight for ga	grams	154 to 280 days
hefwfga	Hadlock estimated fetal weight for ga	grams	154 to 280 days
sfhfga	Symphisis-fundal height for ga	mm	112 to 294 days
crlfga	Crown-rump length for ga	mm	58 to 105 days
gafcrl	ga for crown-rump length	days	15 to 95 mm
gwgfga	Gestational weight gain for ga	kg	98 to 280 days
pifga	Pulsatility index for ga		168 to 280 days
rifga	Resistance index for ga		168 to 280 days
sdrfga	Systolic or diastolic ratio for ga		168 to 280 days
tcdfga	Transcerebellar diameter for ga	mm	98 to 280 days
gaftcd	ga for transcerebellar diameter	days	12 to 55 mm
poffga	Parietal-occipital fissure for ga	mm	105 to 252 days
sffga	Sylvian fissure for ga	mm	105 to 252 days
avfga	Anterior horn of the lv for ga	mm	105 to 252 days
pvfga	Atrium of the PH of the lv for ga	mm	105 to 252 days
cmfga	Cisterna magna for ga	mm	105 to 252 days

notes: GA, gestational age; LV, lateral ventricle; PH, posterior horn.

The exact populations used to construct the intergrowth-21st fetal standards differ for each set of standards associated with a single publication. For the ultrasound- based growth standards, ultrasound measurements were obtained in 4,321 singleton pregnancies every 5 ± 1 weeks from 98 to 294 days (14⁺⁰ to 42⁺⁰ weeks) of gestation (Papageorghiou et al. 2014b).

The estimated fetal weight (efw) standard used data from 2,404 babies born within 14 days of an ultrasound scan to relate data from these scans with their weight at birth (Stirnemann et al. 2017), and the symphisis-fundal height (sfh) standard was based on 20,566 total sfh measurements taken from 4,239 healthy pregnancies every 5 ± 1 weeks from 98 days (14⁺⁰ weeks) to birth. More recent work suggests that Hadlock’s three-parameter formula is more accurate than Stirnemann et al.’s (2017) efw equation. Therefore, the intergrowth-21st team developed a new Hadlock efw standard using fgls data, which we recommend that end users apply when performing efw analyses (Stirnemann, Salomon, and Papageorghiou 2020).

The crown-rump length (crl) and ga from crl standards used crl measurements from 4,625 pregnancies where crl was measured between 63 and 91 days (9⁺⁰ and 13⁺⁰ weeks) (Papageorghiou et al. 2014b). The fetal Doppler standards were based on data taken during phase 2 of the fgls and used 1,243 fetal Doppler measures sourced from 431 individual pregnancies from 154 days (22⁺⁰ weeks) until birth (Drukker et al. 2020). Finally, the fetal brain development standards are based on 3,016 3D ultrasound fetal head volumes, taken every 5 ± 1 weeks from 1,130 children who later had a normal developmental assessment at 2 years (Rodriguez-Sibaja et al. 2021; Napolitano et al. 2020).

The gestational weight-gain standard stands out from the other fetal standards be­cause it represents a maternal phenomenon, namely, expected weight changes across a singleton pregnancy in women who had a body mass index between 18.50 and 24.99 in their first trimester (Cheikh Ismail et al. 2016). It is based on 24,977 weight measure­ments taken from 3,097 normal-weight pregnant women and is included in gigs as a fetal standard to eliminate the need for a separate egen function that includes only the gestational weight-gain standard.

For most of the fetal standards, second-degree fractional polynomial regression was used to model the mean and standard deviation of each measure across a range of x variables (Ohuma and Altman 2019b). This was the approach taken for the fetal growth (Papageorghiou et al. 2014b), sfh (Papageorghiou et al. 2016), crl (Papageorghiou et al. 2014a), transcerebellar diameter (Rodriguez-Sibaja et al. 2021), and fetal brain development (Napolitano et al. 2020) standards. The gestational weight-gain (Cheikh Ismail et al. 2016) standard was modeled similarly but on a logarithmic scale to stabilize variance.

For the nonnormally distributed measures of fetal growth, second-degree fractional polynomials were used with more complex models. The efw standard uses Cole’s (1990) lms method. The authors provide equations that calculate lambda (λ), mu (μ), and sigma (σ) coefficients for a given ga (Stirnemann et al. 2017). These can then be used for conversion between values and z scores using the methods in Cole (1989, 1990) and Cole and Green (1992). For the fetal Doppler standards, the exponential normal distribution was used, with equations described by Drukker et al. (2020) for mu (μ), sigma (σ), and lambda (λ) at different gas. These coefficients are equivalent, respectively, to the μ_T, σ_T, and λ_T coefficients described in section 3.1.6 of Royston and Wright (1998).

1.2 WHO child growth standards

The who child growth standards describe multiple facets of infant growth (de Onis et al. 2004; Group and de Onis 2006). They are the result of the Multicentre Growth Reference Study (mgrs), which was undertaken by the who to address issues with the then “gold standard” 1977 National Center for Health Statistics Growth Reference (Garza and de Onis 2004). These issues included the disproportionately middle-class populations used, limited breastfeeding in the reference population, and inclusion of only Caucasian babies. The mgrs addressed this by constructing their standards with data from 6,669 healthy breast-fed infants from the us, Oman, Norway, Brazil, Ghana, and India, with gas ranging from 37 to 42 weeks (Group and de Onis 2006).

The who standards were constructed using general additive models for location, scale, and shape with the Box-Cox power exponential distribution, smoothed with cubic splines (Borghi et al. 2006; Group and de Onis 2006). The final models simplified to Cole’s (1990) lms model (World Health Organization 2006, 2007). Sex-specific curves were obtained by deriving different lms values for either sex. Table 4 describes the who child growth standards made available in gigs, as well as the unit and xvar() range for each standard.

Table 4.

WHO child growth standards

acronym	Description	Unit	xvar() range
wfa	Weight for age	kg	0a to 1856 days
bfa	Body mass index for age	Kg/m2	0a to 1856 days
lhfa	Length or height for age	cm	0a to 1856 days
hcfa	Head circumference for age	cm	0a to 1856 days
wfl	Weight for length	kg	45 to 110 cm
wfh	Weight for height	kg	65 to 120 cm
acfa	Arm circumference for age	cm	91 to 1856 days
ssfa	Subscapular skinfold for age	mm	91 to 1856 days
tsfa	Triceps skinfold for age	mm	91 to 1856 days

a

Because the standards were constructed using infants with gas ranging from 37 to 42 weeks, 0 days here refers to a newborn born between 37 to 42 weeks.

1.3 Classifying growth

For growth classification, we selected cutoffs for different growth outcomes based on existing approaches. For example, our size-for-ga classifications are based on existing cutoffs: < 10th centile is small for ga (sga), and > 90th centile is large for ga (lga) (Battaglia and Lubchenco 1967; World Health Organization 1995). The < 3rd centile cutoff for severe sga is based on guidelines from the Royal College of Obstetricians and Gynaecologists (2013). We also implement small vulnerable newborn (SVN) classifica­tions, which stratify size-for-ga classifications based on whether a newborn is preterm or term (Ashorn et al. 2023; Lawn et al. 2023). The size-for-ga and svn cutoffs and categories are detailed in tables 5 and 6, respectively. Our z score cutoffs for stunting, wasting, and weight-for-age (underweight) classification are the same as those used by the Demographic and Health Surveys Program of the us Agency for International De­velopment (Croft, Marshall, and Allen 2020) and are described in tables 7, 8, and 9. The head-size cutoffs, detailed in table 10, were sourced from Victora et al. (2016) and Accogli et al. (2021). Finally, the “outlier” cutoffs for length or height for age (stunting), weight for length or height (wasting), and weight for age (underweight) are taken from Mei and Grummer-Strawn (2007).

Table 5.

Values and labels in the sfga and sfga_severe variables generated by gigs_classify_growth

Value	Label	Centile range	z score range
−2	Severely small for gestational age	< 3rd	< −1.88
−1	Small for gestational age (SGA)	< 10th	< −1.28
0	Appropriate for gestational age (AGA)	10th to 90th	−1.28 to 1.28
1	Large for gestational age (LGA)	> 90th	> 1.28

Table 6.

Values and labels in the svn variable generated by gigs_classify_growth

Value	Label	Term status	Centile range
−4	Preterm SGA	Preterm (< 37 weeks)	< 10th
−3	Preterm AGA	Preterm (< 37 weeks)	10th to 90th
−2	Preterm LGA	Preterm (< 37 weeks)	> 90th
−1	Term SGA	Term (≥ 37 weeks)	< 10th
0	Term AGA	Term (≥ 37 weeks)	10th to 90th
1	Term LGA	Term (≥ 37 weeks)	> 90th

Table 7.

Values and labels in the stunting and stunting_outliers vari­ables generated by gigs_classify_growth

Value	Label	z score range
−2	Severe stunting	−6 to −3
−1	Stunting	−3 to −2
0	No stunting	−2 to 6
999	Outlier	< − 6 or > 6

Table 8.

Values and labels in the wasting and wasting_outliers variables generated by gigs_classify_growth

Value	Label	z score range
−2	Severe wasting	−5 to −3
−1	Wasting	−3 to −2
0	No wasting	−2 to 2
1	Overweight	2 to 5
999	Outlier	< −5 or > 5

1.4 Implementation details

For growth standards based on coefficient tables (that is, the lms-based who child growth standards or multistage nonstochastic-based intergrowth-21st newborn size standards), coefficients are available only for discrete x values. Where users provide x values between two discrete x values, we use linear interpolation to estimate the “expected” coefficients at this value of x. This provides more accurate z scores than simply rounding nondiscrete x values to the nearest discrete x value in the LMS or multistage nonstochastic coefficient table in use (Kiger and Taylor 2016).

Though Vidmar, Cole, and Pan (2013) use cubic interpolation for their growth coefficients, they also use coarser growth coefficients for the who standards; that is, they use a coefficient per week whereas we use daily coefficients. Because we have this extra information, we can opt for the simpler linear interpolation and not lose information. From looking at the zanthro() source code, we also note that Vidmar, Cole, and Pan (2013) did not apply a z score-constraining procedure for weight-based who standards (World Health |Organization 2006, 2007). This leads to calculation errors with zanthro() in the who growth standards for 0-5 year olds if |z_WHO| > 3, though the number of observations with this extreme a z score is very low in practice.

We perform our linear interpolation in Mata using the mm_ipolate() function from the moremata module (Jann 2005). Consequently, the who_gs() function in gigs re­quires around 430 milliseconds to perform z score or centile conversion in 100,000 ob­servations: around 4.5 times less than the zanthro function, which requires over two seconds (Vidmar, Cole, and Pan 2013).

1.5 Confirming the accuracy of gigs

Within the GitHub repository for gigs is a folder with do-files containing unit tests for each egen function and command in gigs. Most of these tests generate reference growth data for each implemented growth standard and compare these with published z score or centile curves from the who and intergrowth-21st projects. They also ensure that the conversion functions are consistent with each other—for example, that identical observations are converted into z scores and centiles that correspond to one another. Other tests compare the outputs of the classification command against an R implementation, which itself has a thorough test suite. Using this test suite, we have confirmed that gigs accurately converts between values and z scores or centiles and can classify growth according to the cutoffs described in this article.

2 Conversion functions

The first set of functions introduced in this package is used to convert between measured anthropometric values and z scores or centiles in the intergrowth-21st standards for fetal growth (ig_fet()), newborn size (ig_nbs()), postnatal growth of preterm infants (ig_png()), or who child growth standards (who_gs()).

2.1 Syntax

by cannot be used with these functions.

2.2 Functions

ig_fet(varname, acronym, conversion) converts between values and z scores or centiles in the intergrowth-21st fetal growth standards for ga values in the ranges defined by table 3. It has three arguments:

varname is the variable name in your dataset that you want to convert to a z score, a centile, or an anthropometric measure (for example, headcirc_mm or weight_g).

acronym defines the intergrowth-21st fetal growth standard by which to con­vert the values in varname and should be one of the acronyms listed in table 3.

conversion defines the type of conversion to perform on values in varname and should be one of "v2c" (value to centile), "v2z" (value to z score), "c2v" (centile to value), or "z2v" ( z score to value).

ig_nbs(varname, acronym, conversion) converts between values and z scores or centiles in the intergrowth-21st newborn size standards for ga values in the ranges defined in table 1. It has three arguments:

varname is the variable name in your dataset that you want to convert to a z score, a centile, or an anthropometric measure (for example, weight_kg or fat_mass_kg).

acronym defines the intergrowth-21st newborn size standard by which to con­vert the values in varname and should be one of the acronyms listed in table 1.

conversion defines the type of conversion to perform on values in varname and should be one of "v2c" (value to centile), "v2z" (value to z score), "c2v" (centile to value), or "z2v" (z score to value).

ig_png(varname, acronym, conversion) converts between values and z scores or centiles in the intergrowth-21st postnatal growth standards for pma values in the ranges defined by table 2. It has three arguments:

varname is the variable name in your dataset that you want to convert to a z score, a centile, or an anthropometric measure (for example, weight_kg or headcirc_cm).

acronym defines the intergrowth-21st postnatal growth standard by which to convert the values in varname and should be one of the acronyms listed in table 2.

conversion defines the type of conversion to perform on values in varname and should be one of "v2c" (value to centile), "v2z" (value to z score), "c2v" (centile to value), or "z2v" (z score to value).

who_gs(varname, acronym, conversion) converts between values and z scores or cen­tiles in the who child growth standards. The range of acceptable xvar() values for each acronym is defined in table 4. It has three arguments:

varname is the variable name in your dataset that you want to convert to a z score, a centile, or an anthropometric measure (for example, weight_kg or arm_circ_cm).

acronym defines the who child growth standard by which to convert the values in varname and should be one of the acronyms listed in table 4.

conversion defines the type of conversion to perform on values in varname and should be one of "v2c" (value to centile), "v2z" (value to z score), "c2v" (centile to value), or "z2v" (z score to value).

2.3 Options

xvar(varname) specifies the x variable used to standardize the measure of interest for the intergrowth-21st postnatal growth and fetal standards. For ig_png(), this is usually pma in weeks but may also be length (in centimeters) if using the wfl standard. For ig_fet(), this is usually ga in days but can be crown-rump length in millimeters or transcerebellar diameter in millimeters if you are using the gafcrl or gaftcd standard, respectively. For who_gs(), this is usually age (in days) but may also be length or height (in centimeters) if using the wfl or wfh standard. See tables 2, 3, and 4 for appropriate x variables for each possible acronym value. Any x variable values outside the ranges described in tables 2, 3, and 4 will return missing values.

gest_days(varname) specifies ga in days for the intergrowth-21st newborn size standards. Any value outside the range of valid gas as specified by the acronym argument (see table 1) will return a missing value.

sex(varname) specifies the sex of each observation. This variable should be an integer, byte, or string. The codes for male and female must be specified by the user in the sexcode() option.

sexcode(male = code, female = code) specifies the codes for male and female. The sex can be specified in either order, and the comma is optional. You should not place the codes in quotes even if your sex variable is a string.

3 Classification command

3.1 Syntax

outcomes is text input. Set to all to get all implemented growth outcomes, or specify as many of sfga, svn, stunting, wasting, wfa, and headsize as you wish to run, as long as each word is separated by spaces. The command will tell you whether it lacks the data required to assess growth patterns, for example, if you specify all or headsize but not the headcirc_cm() option.

by cannot be used with this command.

3.2 Description

Classifications for six growth outcomes are implemented. These are size for ga (sfga), svn status (svn), stunting (stunting), wasting (wasting), weight for age (wfa), and head size (headsize).

The command generates new variables with the continuous and categorical growth measures—an error will be issued if the names of variables already in your dataset clash with the new variables it wants to generate, or you can have gigs_classify_growth replace these variables by specifying the replace option. Also, gigs_classify_growth requires specific data for each growth category. For example, if you request a head size (headsize) but do not specify the headcirc_cm() option, the analysis will not be run, and a message will be printed to the console.

Specific growth standards are applied for each observation based on the value of the variables supplied by gest_days() and age_days(). Observations are first assessed as to whether they were taken “at birth” or “postnatally”. An observation is considered to be taken “at birth” if age_days = min(age_days) and age_days <3.

All “at birth” observations with a ga within 24⁺⁰ weeks (168 days) and 42⁺⁶ weeks (300 days) are assessed with the intergrowth-21st newborn size standards. The command will issue a warning if the intergrowth-21st newborn size standards are used for an observation with age_days() > 0.5, so you can assess how many infants in your dataset have “at birth” z scores derived from data taken > 12 hours after birth. For “at birth” observations where gest_days() > 300 (42⁺⁶ weeks), a who z score is calculated and a warning issued.

For “postnatal” observations, infants are classed as “preterm” or “term”, where term infants have a ga > 37⁺⁰ weeks (259 days). For preterm infants, the intergrowth- 21st postnatal growth of preterm infants standards are applied from 27⁺⁰ to 64⁺⁰ weeks’ pma, after which the who growth standards are used. For postnatal observations of preterm infants where PMA < 27⁺⁰, no z score is calculated, because no growth standard is available for these infants. For “term” infants, the who standards are used for all postnatal measurements.

These rules differ slightly for some of the growth indicators. For the newborn-specific categories (sfga and svn), birthweight centiles and categories are computed only for “at- birth” observations. For wasting, there is no intergrowth-21st newborn standard for weight for length, so weight-for-length z scores and wasting categories are computed only for postnatal observations.

The variables generated for each growth outcome are described below:

sfga generates three new variables: birthweight_centile, with centiles calculated using the intergrowth-21st weight-for-ga standards; and sfga or sfga_severe, with size-for-ga classifications in newborns using the values and labels described in table 5.

svn generates two new variables: birthweight_centile, with centiles calculated using the intergrowth-21st weight-for-ga standards; and svn, with svn classifications. Our svn classifications follow the framework of The Lancet’s 2023 Small Vulnerable Newborn Series (Ashorn et al. 2023; Lawn et al. 2023), and the resulting variable has the values and labels specified in table 6.

stunting generates three new variables: lhaz, with length- or height-for-age z scores calculated with an appropriate growth standard for each observation as described above; and stunting or stunting_outliers, with stunting classifications using the values and labels described in table 7.

wasting generates three new variables: wlz, with weight-for-length z scores calculated with an appropriate growth standard for each observation as described above; and wasting or wasting_outliers, with wasting classifications using the values and labels described in table 8.

wfa generates three new variables: waz, with weight-for-age z scores (wazs) calculated with an appropriate growth standard for each observation as described above; and wfa or wfa_outliers, with weight-for-age (underweight) classifications using the values and labels described in table 9.

Table 9.

Values and labels in the wfa and wfa_outliers variables generated by gigs_classify_growth

Value	Label	z score range
−2	Severely underweight	−6 to −3
−1	Underweight	−3 to −2
0	Normal weight	−2 to 2
1	Overweight	2 to 5
999	Outlier	< − 6 or > 5

headsize generates two new variables: hcaz, with head-circumference-for-age z scores calculated with an appropriate growth standard for each observation as described above; and headsize, with head-size classifications using the values and labels de­scribed in table 10.

Table 10.

Values and labels in the headsize variable generated by gigs_classify_growth

Value	Label	z score range
−2	Severe microcephaly	< −3
−1	Microcephaly	−3 to -2
0	Normal head size	−2 to 2
1	Macrocephaly	2 to 5
2	Severe macrocephaly	>2

3.3 Options

gest_days(varname) specifies the ga in days for each observation. gest_days() is required.

age_days(varname) specifies the age in days for each observation. age_days() is re­quired.

sex(varname) specifies the sex of each observation. varname should be an integer, byte, or string. The codes for male and female must be specified by the user in the sexcode() option.

sexcode(male = code, female = code) specifies the codes for male and female. The sex can be specified in either order, and the comma is optional. You should not place the codes in quotes, even if your sex variable is a string. sexcode() is required.

weight_kg(varname) specifies the weight in kilograms for each observation.

lenht_cm(varname) specifies the length or height in centimeters for each observation. For observations with an age of less than 731 days (2 years), it will be assumed that recumbent length was recorded. For observations where age > 731 days (2 years), it will be assumed that standing height in centimeters was recorded.

headcirc_cm(varname) specifies the head circumference in centimeters for each obser­vation.

id(varname) specifies the ID of each infant in the dataset. varname should be an integer, byte, or string. You must specify this if you are supplying data from multiple infants, or gigs_classify_growth will identify only one birth measure across the whole dataset and therefore return incorrect results.

replace specifies whether gigs_classify_growth should replace existing variables that use names it applies for its new variables. If this option is not specified, the command will issue an error if any variables in your dataset need to be renamed. If this option is specified, the command will issue messages to the console about replaced variables.

4 Examples

For these examples, we are using a small subset of data from the Low-Birthweight Infant Feeding Exploration (life) study (Vesel et al. 2023), supplied within gigs as life6mo.dta. This dataset contains anthropometric data from 300 singletons, taken from birth to six months at visits to a clinic at 0, 1, 2, 4, 6, 10, 14, 18, and 26 weeks of chronological age (with a small window of variation for each visit). In the life dataset, the columns contain an id per infant (id), ga at birth in days (gestage), sex (sex; coded male = 1 and female = 2), the visit week of each observation (visitweek), pma in days (pma), the age in days at each visit (age_days), and infant weight in grams recorded at each visit (weight_g).

4.1 Conversion functions

To generate accurate wazs, we need to use different standards for different measurement times and term statuses. We will first generate wazs at birth using the intergrowth- 21st newborn size standards, using the fact that the first “visit” for each child was a birth assessment:

Next we calculate wazs for term infants after birth using the who weight-for-age standard:

We then generate wazs after birth for preterm infants using the intergrowth-21st postnatal growth standards for preterm infants.

Collapsing these variables into a single variable manually could be prone to error. Fortunately, we can use gigs_classify_growth to pick growth standards for us instead. We specify the wfa outcome to request a weight-for-age analysis:

Note how the command lets you know which new variables it has generated and the outcomes you have listed.

We can visualize the wazs for all infants based on whether they were preterm. Let’s generate a new variable to denote term status (preterm) and plot the wazs against visitweek (figure 1):

Figure 1.

WAZs against visit week for term (solid line) and preterm (dashed line) infants

In our plot, the term infants consistently have lower wazs than the preterm infants because of the sample used in the life study—almost all term participants were sga at birth, but preterm participants were a mix of sga, aga, and lga (Vesel et al. 2023).

4.2 Classification functions

We could also categorize the newborn measurements in this study according to their size for ga, using the intergrowth-21st newborn size standards. To do this, we would use the gigs_classify_growth command. We first reload life6mo.dta. We then run gigs_classify_growth to get size-for-ga categories for each birth measure and graph the proportions of each size-for-ga category in the dataset (figure 2):

Figure 2.

Size-for-ga classification percentages in the life data extract. ga, gestational age; sga, small for ga; aga, appropriate for ga; lga, large for ga.

Because our sample has a mixture of term and preterm infants, we can disambiguate these size-for-ga categories into svn classifications. This mode of classification is more useful than size for ga alone because neonatal mortality risks are greater in nonterm sga babies than in term sga babies (Lawn et al. 2023). We use gigs_classify_growth to get svn categorizations. Note how the command is running in replace mode, so it tells you which variable it is replacing.

Once this is done, you can then graph the svn categories (figure 3). The results shown here are again due to the sampling process for life—the preterm cases display a mix of size-for-ga types, but term cases are almost entirely sga.

Figure 3.

Small vulnerable newborn classification percentages in the life data extract. svn, small vulnerable newborn; ga, gestational age; sga, small for ga; aga, appropriate for ga; lga, large for ga.

Conclusions

We have illustrated how the new gigs package for Stata can be used for growth analysis in newborns and children by either the generation of z scores or centiles from measured anthropometric indices or classification of observations into specific growth categories.

Supplemental Material

sj-dta-1-stj-10.1177_1536867X251365448 - Supplemental material for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen

Supplemental material, sj-dta-1-stj-10.1177_1536867X251365448 for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen by Simon R. Parker, Linda Vesel and Eric O. Ohuma in The Stata Journal

Supplemental Material

sj-dta-10-stj-10.1177_1536867X251365448 - Supplemental material for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen

Supplemental material, sj-dta-10-stj-10.1177_1536867X251365448 for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen by Simon R. Parker, Linda Vesel and Eric O. Ohuma in The Stata Journal

Supplemental Material

sj-dta-11-stj-10.1177_1536867X251365448 - Supplemental material for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen

Supplemental material, sj-dta-11-stj-10.1177_1536867X251365448 for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen by Simon R. Parker, Linda Vesel and Eric O. Ohuma in The Stata Journal

Supplemental Material

sj-dta-12-stj-10.1177_1536867X251365448 - Supplemental material for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen

Supplemental material, sj-dta-12-stj-10.1177_1536867X251365448 for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen by Simon R. Parker, Linda Vesel and Eric O. Ohuma in The Stata Journal

Supplemental Material

sj-dta-13-stj-10.1177_1536867X251365448 - Supplemental material for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen

Supplemental material, sj-dta-13-stj-10.1177_1536867X251365448 for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen by Simon R. Parker, Linda Vesel and Eric O. Ohuma in The Stata Journal

Supplemental Material

sj-dta-14-stj-10.1177_1536867X251365448 - Supplemental material for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen

Supplemental material, sj-dta-14-stj-10.1177_1536867X251365448 for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen by Simon R. Parker, Linda Vesel and Eric O. Ohuma in The Stata Journal

Supplemental Material

sj-dta-2-stj-10.1177_1536867X251365448 - Supplemental material for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen

Supplemental material, sj-dta-2-stj-10.1177_1536867X251365448 for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen by Simon R. Parker, Linda Vesel and Eric O. Ohuma in The Stata Journal

Supplemental Material

sj-dta-3-stj-10.1177_1536867X251365448 - Supplemental material for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen

Supplemental material, sj-dta-3-stj-10.1177_1536867X251365448 for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen by Simon R. Parker, Linda Vesel and Eric O. Ohuma in The Stata Journal

Supplemental Material

sj-dta-4-stj-10.1177_1536867X251365448 - Supplemental material for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen

Supplemental material, sj-dta-4-stj-10.1177_1536867X251365448 for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen by Simon R. Parker, Linda Vesel and Eric O. Ohuma in The Stata Journal

Supplemental Material

sj-dta-5-stj-10.1177_1536867X251365448 - Supplemental material for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen

Supplemental material, sj-dta-5-stj-10.1177_1536867X251365448 for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen by Simon R. Parker, Linda Vesel and Eric O. Ohuma in The Stata Journal

Supplemental Material

sj-dta-6-stj-10.1177_1536867X251365448 - Supplemental material for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen

Supplemental material, sj-dta-6-stj-10.1177_1536867X251365448 for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen by Simon R. Parker, Linda Vesel and Eric O. Ohuma in The Stata Journal

Supplemental Material

sj-dta-7-stj-10.1177_1536867X251365448 - Supplemental material for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen

Supplemental material, sj-dta-7-stj-10.1177_1536867X251365448 for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen by Simon R. Parker, Linda Vesel and Eric O. Ohuma in The Stata Journal

Supplemental Material

sj-dta-8-stj-10.1177_1536867X251365448 - Supplemental material for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen

Supplemental material, sj-dta-8-stj-10.1177_1536867X251365448 for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen by Simon R. Parker, Linda Vesel and Eric O. Ohuma in The Stata Journal

Supplemental Material

sj-dta-9-stj-10.1177_1536867X251365448 - Supplemental material for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen

Supplemental material, sj-dta-9-stj-10.1177_1536867X251365448 for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen by Simon R. Parker, Linda Vesel and Eric O. Ohuma in The Stata Journal

Supplemental Material

sj-orig-2-stj-10.1177_1536867X251365448 - Supplemental material for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen

Supplemental material, sj-orig-2-stj-10.1177_1536867X251365448 for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen by Simon R. Parker, Linda Vesel and Eric O. Ohuma in The Stata Journal

Supplemental Material

sj-txt-1-stj-10.1177_1536867X251365448 - Supplemental material for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen

Supplemental material, sj-txt-1-stj-10.1177_1536867X251365448 for gigs: A package for standardizing fetal, neonatal, and child growth assessment with extensions to egen by Simon R. Parker, Linda Vesel and Eric O. Ohuma in The Stata Journal