Evaluating Anthropometric Growth Endpoints with Z-Scores and

Preparing to load PDF file. please wait...

0 of 0
Evaluating Anthropometric Growth Endpoints with Z-Scores and

Transcript Of Evaluating Anthropometric Growth Endpoints with Z-Scores and

PharmaSUG 2021 - Paper SA-155
Evaluating Anthropometric Growth Endpoints with Z-Scores and Percentiles
Snehal Sanas, Pradeep Umapathi, Covance by Labcorp
Disease conditions in children can potentially affect a child's growth and development, hence an effective treatment for them is required to sufficiently demonstrate improvement, not only in disease condition but also in growth pattern. Unlike adults, assessment of drug safety and efficacy in children requires specific consideration of the impact of treatment on child growth.
Measuring overall child development can be a 'complex task', as it involves evaluating anthropometric measures like weight, height, head circumference and mid-upper arm circumference at different stages of development. Clinical trials involving pediatric populations need to have composite efficacy endpoints to account for the impact on growth. WHO and CDC have established worldwide anthropometric data standards against which a child's growth can be validated and conclusions can be drawn as to whether the child is growing correctly or has any growth issues.
This paper aims to understand the rationale behind data collected by WHO and CDC, what are Z-scores and percentiles, how to calculate Z-scores and percentiles, how to include this reference data into programs for statistical analysis and also how these are interpreted to evaluate growth. These standards have been used in clinical trials to demonstrate growth related drug efficacy, across various therapeutic areas as unique as rare congenital enzyme deficiencies to pediatric cardiology. This paper attempts to describe a case for implementing Z-scores and percentiles into clinical trials using a macro in SAS and also to touch base on how this approach is useful, not only in clinical trials but also in general assessment of a child’s growth by pediatricians and health care providers.
Pediatric illnesses have additional aspects to be addressed by treatments, compared to those in adults. Given the developing age of patients these treatments can potentially impact a child’s growth and development. With children, physiology is constantly altering and upgrading. This changing physiology causes differences in the body’s response to medication, compared with an adult. Thus, these treatments not only need to heal the pathophysiology, but also not to interrupt the growth processes. Pediatric treatments are required to sufficiently demonstrate improved or unaffected growth.
For such clinical trials, ICH Guidance E11 (Clinical Investigation of Medicinal Products in the Pediatric Population) and E11 (R1) (Addendum to ICH E11) has emphasized on determining possible effects on skeletal, behavioral, cognitive, sexual and immune maturation and development in pediatric patients. As a result, pediatric trials are required to have appropriate growth indicators as part of the composite efficacy endpoints to account for impact on growth.
The scope of this paper includes understanding and using Anthropometric Growth Indicators in Pediatric trials for assessing the effects on skeletal development of pediatric patients.
The term ‘Anthropometric’ comes from the Greek words Anthropos meaning ‘Human’ and matron meaning ‘Measure’ respectively.
Anthropometric measurements can be broadly categorized into measurements of Structural dimensions and measurements of Functional dimensions. Structural dimensions (Figure 1) are static body measurements such as height or weight. Whereas Functional dimensions (Figure 2) are dynamic measurements which gauge a person’s ability to perform a particular action, such as the angle of shoulder rotation. The following are images illustrating these measurements:
© Copyright 2021 Covance Inc.

Figure 1: Height Measurement

Figure 2: Shoulder Rotation Measurement

Measuring ‘Anthropometric’ growth is a complex task, as it involves monitoring multiple interdependent parameters. For example, let us consider a 3-year-old boy who weighs 10 kg and is 65 cm in height, however for his age the average weight of someone similar is approximately 11-18 kg and average height is approximately 89-103 cm. This apparently indicates underdevelopment compared with their expected height and weight, however is his current weight appropriate to his current height, is his weight properly distributed across his body, is his weight appropriate for his muscle and fat content, is his skeletal development proportionate, and how many children in a similar population are showing such a growth pattern. Answering all these question can help in concluding if the given boy is overall underdeveloped or his weight is appropriate but skeletal development is slow and so on.
To tackle this challenge, efforts have been made to provide guidance standards and reference databases so that growth can be compared and evaluated. In 1977 NCHS (National Center for Health Statistics, US) developed growth charts from reference data as a tool for pediatricians, so that child growth can be evaluated. In 2000 CDC (Center for Disease Control) further revised and upgraded these charts with additional parameters and based on the data collected from a wider national population. These charts were also adopted by WHO (World Health Organization) for international use, however there was a need to build up databanks considering various population groups around the globe, for having better accuracy in assessment of growth. The WHO conducted a Multicenter Growth Reference Study (MGRS) from July 1997 to December 2003. This study was carried out at 6 sites worldwide, Davis, California, USA; Muscat, Oman; Oslo, Norway; Pelotas, Brazil; Accra, Ghana; South Delhi, India. It involved a sample size of 8440 children who lived in socioeconomic conditions favorable to growth and where mobility was low. Data was recorded for the following anthropometric parameters; weight, recumbent length or height, head circumference, arm circumference, triceps skinfolds (aged ≥ 3 months), and subscapular skinfolds (aged ≥ 3 months).
After in-depth quality checks, data was analyzed and detailed validation reports, descriptive statistics tables and plots were produced. Growth indicators such as length/height-for-age, weight-for-age, weight-forlength, weight-for-height, body mass index-for-age, head circumference-for-age, arm circumference-forage, triceps skinfold-for-age and subscapular skinfold-for-age were established from the statistical analysis of the collected database. Growth curves or growth charts with z-scores and percentiles for the above indicators were developed for multiple age groups gender-wise. The database and resulting growth curves and growth charts were released as standards by WHO in 2006. The following are some example of growth charts (Figure 3, Figure 4):

© Copyright 2021 Covance Inc.

Figure 3: Z-scores Growth Chart
Figure 4: Percentiles Growth Chart
The z-score is the deviation from the median value of the reference population, divided by the standard deviation of the reference population. Z-scores describe how far a measurement is from the median (average) measurement of the reference population. A percentile is the value below which a percentage of the data falls. A percentile for an individual’s measurement (such as height) indicates what percent of the reference population the individual’s measurement would equal or exceed. For every z-score there is a corresponding percentile and vice versa. The following is a hypothetical data distribution plotted to illustrate z-scores and cumulative percentiles (Figure 5):
© Copyright 2021 Covance Inc.

Figure 5: A normal bell shaped curve cut into Z-score segments The z-score for a body measurement in this distribution is calculated as follows: z-score = (observed value) - (median reference value)/ z-score of the reference population For example, assuming that the above distribution is for weights of girls aged 12 to 24 months with median reference value of weight as 9.5 kg and reference z-score as 1; for a 15-month old girl weighing 8.5 kg ‘weight-for-age z-score’ is calculated as; z-score = (8.5-9.5) / 1 = -1 Z-score of -1 indicates weight related growth for the girl is less than average for her age and corresponding percentile of 34.1 indicates her weight-for-age growth equals or exceeds 34.1% of the studied reference population. The analyzed data from WHO’s MGRS when presented as a graph form a very user friendly tool to access growth for pediatricians, nurses and parents; below are some illustrative examples:
Figure 6: Interpretation of Z-score on Growth Chart
© Copyright 2021 Covance Inc.

The above example (Figure 6) uses measurements of weight and height to get an approximate idea of the z-score and an overview clinical impression that the boy’s weight for his height is over the average.
Figure 7: Interpretation of Percentile on Growth Chart The above example (Figure 7) uses measurements of weight and height to get an approximate idea of the percentile and an overview clinical impression that the girl’s weight for her height is between 15% and 50% of the reference population.
As discussed above it is essential for clinical trial protocols to include anthropometric indicators as additional efficacy end points wherever appropriate. The following section describes methodologies to implement anthropometric evaluation as clinical end points. For a hypothetical clinical trial of medicinal product ‘X’ being evaluated for indication ‘Y’ in a pediatric population falling in the age group of up to 2 years; the study protocol is designed to evaluate safety and efficacy of the medication for improving a pathological condition. Along with this the study also evaluates the impact of medication on the weight as well as height development of the participating population. Data is collected for weight and height of the patient over the course of treatment at the following time points: Day0/Baseline, every 15 days up to months 2, every other month up to month 20, and the end of treatment visit at month 24. In order to utilize collected data of weight and height, z-scores can be calculated by using WHO’s standard reference database and the resulting values can be summarized and presented using any necessary tables, listings and figures. For the purpose of processing any of the collected measurements from WHO’s standard reference database, WHO has also provided a utility set containing a SAS executable macro ‘igrowup_standard’ and all required documentation to be able to implement this macro. This macro processes the collected data against data sets containing the WHO Child Growth Standards and produces the z-scores output files. The following flow chart provides the flow of data using this macro to generate zscores of the collected data (Figure 8):
© Copyright 2021 Covance Inc.

Figure 8: Data flow using WHO’s ‘igrowup_standard’ macro
This macro has 15 parameters which indicate which anthropometric measurements are to be processed and calculates z-scores for the respective anthropometric measurements. The following is a sample call to invoke the macro followed by a description of the macro parameters: %igrowup_standard (reflib= REFDATALIB,
datalib= OUDATALIB, datalab= _AP, sex= SEXN, age= AGE_M, ageunit= months, weight=WEIGHT_Kg, lenhei=HEIGHT_cm, headc=, armc=, triskin=, subskin=, measure=, oedema=, sw=); reflib: SAS Library/Directory with reference data. datalib: SAS Library/Directory for input data / output data set. datalab: Prefix for output files. sex: Name of a variable containing sex information. If it is a numeric variable, its values must be, 1 for males and 2 for females. And if it is a character variable, it must be, "m" or "M" for males and "f" or "F" for females. age: Name of a numeric variable containing age information. Age can be in either days or months.
© Copyright 2021 Covance Inc.

ageunit: Unit of the age variable. It must be specified as either "days" or "months" (they are case sensitive).
weight: Name of a numeric variable containing body weight information, which must be in kilograms.
lenhei: Name of a numeric variable containing length (recumbent) or height (standing) information, which must be in centimeters.
headc: Name of a numeric variable containing the head circumference information, which must be in centimeters.
armc: Name of a numeric variable containing the mid-upper arm circumference information, which must be in centimeters.
triskin: Name of a numeric variable containing the triceps skinfold information, which must be in millimeters.
subskin: Name of a numeric variable containing the subscapular skinfold information, which must be in millimeters.
measure: Name of a character variable indicating whether recumbent length or standing height was measured. The values of this variable must be "l" or "L" for recumbent length, and "h" or "H" for standing height.
oedema: Name of the character variable containing oedema information. The values of this variable must be "n" or "N" for non-oedema, and "y" or "Y" for oedema.
sw: Name of a numeric variable containing the sampling weights.
Users must code any missing values as "." for numeric variables and as " " for character variables.
The output data sets from macro retains all the records and variables from input data set and adds on the following 19 variables derived by the macro (Table 1).

_agedays _clenhei _cbmi _zwei _zlen _zwfl _zbmi _zhc _zac _zts _zss

: calculated age in days for deriving z score : converted length/height (cm) for deriving z score

: calculated bmi=weight / squared(_clenhei)

: Weight-for-age z-score


: Length/height-for-age z-score


: Weight-for-length/height z-score


: BMI-for-age z-score


: Head circumference-for-age z-score _fhc

: Arm circumference-for-age z-score _fac

: Tricpes skinfold-for-age z-score


: Subscapular skinfold-for-age z-score _fss

: Flag for _zwei < -6 or _zwe i> 5 : Flag for _zlen < -6 or _zlen > 6 : Flag for _zwfl < -5 or _zwf l> 5 : Flag for _zbmi < -5 or _zbmi > 5 : Flag for _zhc < -5 or _zhc > 5 : Flag for _zac < -5 or _zac > 5 : Flag for _zts < -5 or _zts > 5 : Flag for _zss < -5 or _zss > 5

Table 1: Z-scores output Data Set Variable name: Variable label

The variables containing z-scores are to be used to calculate percentiles. The z-scores and percentiles can be further listed, summarized and plotted as per the Statistical Analysis Plan. The following are example summary outputs:

© Copyright 2021 Covance Inc.

Listing of z-scores and percentiles displayed subject wise can provide information on how subject’s weight and height have changed over the course of treatment (Figure 9).
Figure 9: Listing of Z-scores and Percentiles Summary tables with descriptive statistics on z-scores and percentiles can provide understanding about the drug’s effect on the overall study population (Figure 10).
Figure 10: Summary table of Z-scores and Percentiles
© Copyright 2021 Covance Inc.

Percentile line plots by subject can visualize an individual subject’s growth (Figure 11).
Figure 11: Line plot of Percentiles Scatter plot of percentiles for study population can visualize trends of effect on growth of study population (Figure 12).
Figure 12: Scatter plot of Percentiles MERITS
• Recording anthropometric measurements is an inexpensive approach to monitor effects on skeletal growth. Some of these measurements are routine in a clinical trial, such as height, weight.
• The tools required for data collection are easy to use and do not require any specialized skill.
© Copyright 2021 Covance Inc.

• With these relatively easy to record measurements entire study population can be assessed for a given end point.
• WHO’s standard database eliminates any need of multiple age or sex based normal ranges to be tabulated in data sets.
• As WHO’s MGRS was conducted including global sample size and multiple groups, comparative assessment is possible.
LIMITATIONS • As the WHO’s MGRS was conducted in study populations living in socioeconomic conditions favorable to growth, over diagnosis of growth failure in children from developing countries is a possibility.
• This can also lead to incorrect diagnosis in children with non-supportive growth conditions.
• Regular updates to the WHO’s database are required with further wider categorization of data on the basis of various population groups.
• More local growth databases specific to particular populations are required for more accurate assessment.
Following are some country specific databases utilized; USA: CDC, WHO growth chart based assessment up 2 years UK: WHO growth chart based assessment up 4 years India: WHO growth chart based assessment, other local databases like Indian Academy of Pediatrics (IAP) are also available.

Z-scores and percentiles are useful for assessing growth related drug efficacy across various therapeutic areas. Z-scores and percentiles are also useful for the general assessment of child growth by pediatricians, health care providers, as well as for diagnosis of growth failure.
For children in developing age, z-scores and percentiles for growth indicators are sensitive to changes in pathological conditions and thus are an important tool to evaluate growth related efficacy of a medicinal product in pediatric trials.


Training Courses and Instruction documents on Child Growth Assessment, World Health Organization. http://www.who.int/childgrowth/training/en/.

Multicenter Growth Reference Study http://www.who.int/childgrowth/publications/en/





WHO Child Growth Standards STATA igrowup package. http://www.who.int/growthref/tools/en/ and http://www.who.int/childgrowth/software/en/

Frequently Asked Questions About the https://www.cdc.gov/growthcharts/growthchart_faq.htm





Daniel K. Benjamin, Jr., Steven Hirschfeld, Coleen K. Cunningham and Ross E. McKinney, Jr. 2004. “Growth as a part of the composite endpoint in paediatric antiretroviral clinical trials.” Journal of Antimicrobial Chemotherapy, 54, 701–703.

© Copyright 2021 Covance Inc.