Department of Community Health Sciences, Faculty of Medicine, University of Calgary, 3330 Hospital Drive NW, Calgary, Alberta, T2N 4N1, Canada

Abstract

Background

The Babson and Benda 1976 "fetal-infant growth graph" for preterm infants is commonly used in neonatal intensive care. Its limits include the small sample size which provides low confidence in the extremes of the data, the 26 weeks start and the 500 gram graph increments. The purpose of this study was to develop an updated growth chart beginning at 22 weeks based on a meta-analysis of published reference studies.

Methods

The literature was searched from 1980 to 2002 for more recent data to complete the pre and post term sections of the chart. Data were selected from population studies with large sample sizes. Comparisons were made between the new chart and the Babson and Benda graph. To validate the growth chart the growth results from the National Institute of Child Health and Human Development Neonatal Research Network (NICHD) were superimposed on the new chart.

Results

The new data produced curves that generally followed patterns similar to the old growth graph. Mean differences between the curves of the two charts reached statistical significance after term. Babson's 10^{th }percentiles fell between the new data percentiles: the 5th to 17th for weight, the 5th and 15th for head circumference, and the 6th and 16th for length. The growth patterns of the NICHD infants deviated away from the curves of the chart in the first weeks after birth. When the infants reached an average weight of 2 kilograms, those with a birthweight in the range of 700 to 1000 grams had achieved greater than the 10^{th }percentile on average for head growth, but remained below the 3^{rd }percentile for weight and length.

Conclusion

The updated growth chart allows a comparison of an infant's growth first with the fetus as early as 22 weeks and then with the term infant to 10 weeks. Comparison of the size of the NICHD infants at a weight of 2 kilograms provides evidence that on average preterm infants are growth retarded with respect to weight and length while their head size has caught up to birth percentiles. As with all meta-analyses, the validity of this growth chart is limited by the heterogeneity of the data sources. Further validation is needed to illustrate the growth patterns of preterm infants to older ages.

Background

Growth monitoring is a part of the medical and nutritional assessment of preterm infants. Growth charts provide the basis for this assessment by providing a comparison to a reference which allows for a visual picture of both the infant's achieved size and growth trajectory. Parents and health professionals like to know whether a preterm infant is able to maintain growth velocity or achieve catch-up in growth compared to the fetus and the term infant. Intrauterine growth charts allow this comparison for infants before term. Babson and Benda (Babson)

The growth chart developed by Babson is still recommended for use in Neonatal Intensive Care Units

Although commonly used, Babson's growth chart has limitations. The X axis begins at 26 weeks of gestation thus limiting its usefulness to plot younger preterm infants from birth. The Y axis is made in 500 gram increments which make precise plotting difficult. The sample size of the data behind Babson's chart was small. There were only 45 infants in their sample that were 30 weeks or younger

The purpose of this project was to prepare an updated fetal-infant Babson-type growth chart for use in neonatal intensive care units. This chart will allow a comparison for preterm infants as young as 22 weeks of gestation, first with intrauterine and then with post term references.

Methods

Literature review and selection

A search of the literature was conducted on three databases (Pub Med, the Cochrane Library, EMBASE from 1980 to June 2002) using the subject headings: infant, (premature, very low birthweight), anthropometry, growth, birthweight, head, cephalometry, gestational age, newborn, and reference values. Articles selected included surveys of intrauterine and post term growth. Reference lists of relevant articles were searched.

To improve on the Babson graph, two types of data were needed: infant size measured at the time of birth for the intrauterine section and term infant measurements for the post-term section. Population studies with large sample sizes were preferred to improve generalizability. The World Health Organization has recommended that gestational age of infants be described as completed weeks

Intrauterine data

Three recent large population based ^{th }and 90^{th }percentiles for 32 weeks in Alexander's corrected data were 2.2 and 3.2 kilograms (kg) respectively. The 50^{th }and 90^{th }percentiles for 32 weeks in Kramer's corrected data were the more plausible 1.9 and 2.3 kg, respectively. Kramer et al used a statistical method to correct for misclassification of term infants incorrectly labeled as preterm. Table

Details of the Data Sources

Kramer

Niklasson

Beeby

CDC

Data used

Birthweight ≤ 40 weeks

Head and length measures ≤ 40 weeks

Head and length measures ≤ 40 weeks

Weight, head and length measures after term

Sample size (n)

676,605

376,000

Head circumference: 29090 and length: 26973

Birth data :weight: 82 million, length: 900,000, head circumference: 400. First year of life: 2200 to 38,000 infants.

n < 30 weeks

4101

0

274

N/A

Gestational age range of study

22 to 43 weeks

28.5 to 42.5 weeks

22 to 43 weeks

Post term

Inclusion criteria

All births

"Healthy" newborn infants

Singleton livebirths

NHANES surveys

Exclusion criteria

Ontario (province) was excluded due to problems with data quality.

Stillbirths, twins, complications during pregnancy with potential effects on fetal growth and significant malformations

Multiple births

Infants with birthweights < 1500 grams

Dates

1994 to 1996

1977 to 81

1982 to 1995

Primarily 1963 to 1994

Completed gestational weeks

yes

yes

yes

N/A

Method to assess gestational age

"early ultrasound has increasingly been the basis for gestational age assessments in recent years"

Last menstrual period was used when it was in agreement (+/- 2 weeks) with the obstetric assessment. If not, the latter was used. When either estimate was missing the gestational age was considered unknown.

In the majority of cases the gestational age was based on first trimester ultrasound, or when early ultrasound was not available, last menstrual period dates were used. In < 1% of cases, the Ballard assessment was used when neither dates nor ultrasound were available.

N/A

Outliers

Assumed a log normal distribution of birthweight at each gestational age and compared the probabilities of accurate versus misclassification of infant's gestational age.

The mother's medical record was examined and following verification of the gestational age and other data, outliers were either accepted as real or corrected.

Curves were smoothed with a variety of parametric and non-parametric procedures

Two population based studies that met the a priori criteria included head circumference and length, one from Sweden by Niklasson et al

To develop the head circumference and length curves, the numerical data from Niklasson

There was remarkable consistency between the head and length results from the Australian and Swedish studies. Between 31 and 33 weeks; the differences were less than 0.3 centimeter (cm) for 50^{th }percentiles and 0.4 cm for the 10^{th }and 90^{th }percentile curves. The magnitudes of the differences were greater for data at 29 weeks and greater than 36 weeks which may to be due to the way the Swedish data was reported. These data were expressed as a third degree polynomial function, so their curves were curved downward at both ends. At the ends of the curves the differences increased to 0.4 cm at both ends of the curve for head circumference and to 0.8 cm (29 weeks) and 1.0 cm (40 weeks) for length. (This problem was dealt with under Smoothing of the Curves.) The consistency in the mid sections of the curves provides confidence regarding the combining of the results from these two studies.

Post 40 week data

In selecting the term infant data for the post 40 weeks section of the fetal-infant chart the Center for Disease Control (CDC)

Chart development

The data for both genders were averaged together for the 3^{rd}, 10^{th}, 50^{th}, 90^{th }and 97^{th }percentiles to create one growth chart. Separate gender charts were not produced since the gender differences were considered not important enough to warrant separate charts. The sample sizes for the birthweight data are very large, so significant differences between the genders could be seen for most ages even when the differences were small and of minor practical importance. Specifically, the differences between the male and female weights were statistically significant at all ages above 23 weeks for the 50^{th }percentile (beginning with a difference of 32 grams) and above 24 weeks for the 3^{rd }percentile (beginning with a difference of 44 grams). None of the head and length or the CDC data were reported in a manner that permitted a statistical comparison by gender.

The largest differences between the genders were confined to late gestation and after term. Along the 3^{rd }percentile, there was only one weight difference between the genders that was greater than 100 grams (40 weeks) and none of the head differences were greater than 0.4 cm. The differences between the genders along the 10^{th }percentile were limited to greater than 36 weeks for weight differences greater than 100 grams and only 1 head difference was greater than 0.6 cm. Along the 50^{th }percentile the only differences between the genders for weight more than 100 grams were greater than 35 weeks and the only head differences greater than 0.4 cm were after term. For each of these three percentiles, only 1 length measure was greater than 1 cm and all of these were after term.

A large-scale grid was used to aid accuracy of plotting. The increments were 100 grams for weight, one cm for head circumference and length, and one-week intervals for time. The chart was extended to 50 weeks since the majority of preterm infants are discharged home by this age. Spaces were made at the bottom of the chart to note the date of measurements. The 3rd, 10th, 50^{th}, 90^{th }and 97th percentile curves for weight, head circumference and length were plotted on the grid. There were disjunctions between the data sets at 40 weeks. For example, the CDC 50^{th }percentile for weight was 2% lower than the Kramer value, while the 3^{rd }percentile was 14% lower. The disjunctions between the pre and post-term sections are likely due to artifacts of the original data sets and the processing methods.

Smoothing of curves

To produce a working chart, the disjunctions between the pre and post term sections needed to be smoothed. It was difficult to merge the pre- and post-term data sets at their boundaries. The goal of the smoothing was to produce a bridge between the pre-term and the post-term curves without the typical deceleration seen in intrauterine curves just before term. Initial approaches to numerically smooth the boundary produced obvious undesirable results. Therefore manual methods were used to smooth the disjunction from the deceleration point of the pre-term curves to a point prior to the first measurement after term (at 2 months).

For weight the smoothing was done between 36 and 46 weeks. (Figure ^{th }percentile needed the least modifying since both the Kramer and CDC weights at term were approximately 3.5 kg. The 3rd percentile required smoothing up to 49 weeks. The smoothing for head circumference and length was continued back to 22 weeks to level the disjunction between the two data sources at 29 weeks. All of the curves had merged with the CDC curves prior to 50 weeks.

Illustration of smoothed curves superimposed on the raw data curves

Illustration of smoothed curves superimposed on the raw data curves. Solid curves are from the original data, dotted curves are the smoothed and final version.

The final curves (Figure

A new fetal-infant growth chart for preterm infants developed through a meta-analysis of published reference studies

A new fetal-infant growth chart for preterm infants developed through a meta-analysis of published reference studies.

Comparisons with the Babson chart

Three comparisons were made between Babson's chart and the new data. First, a graphical comparison was prepared of Babson's mean, 3rd and 97^{th }percentiles superimposed on the new raw data (Figure ^{th }percentiles were equal to the mean. This comparison was done using t-test at 28, 30, 32, 36, 40 weeks and 1.5 and 2.5 months. Standard deviations were estimated using a least squares fit of the percentiles to the normal curve. It was not possible to test for differences prior to 28 weeks since Babson did not include percentile curves until 28 weeks. A probability of p < 0.05 was considered statistically significant and adjustment was made for multiple comparisons using the Bonferroni method.

The dotted curves are Babson and Benda's 1976 fetal-infant growth curves superimposed on the raw data curves

The dotted curves are Babson and Benda's 1976 fetal-infant growth curves superimposed on the raw data curves.

The 10th percentile is a frequent cut point for assessment of size for age. The third comparison was a calculation of the percent of new chart values below Babson's 10^{th }percentile. This was done by first calculating Babson's 10^{th }percentile using the standard deviation estimate. Then z scores were calculated for Babson's 10^{th }percentile compared to the new distributions and an estimate was made of the percent of new data below Babson's 10^{th }percentile. This comparison was done at 28, 30, 32, 36 and 40 weeks and at 1 and 2 months.

How does the growth of preterm infants appear on the chart?

The data from the large multicenter cohort study done by the National Institute of Child Health and Human Development Neonatal Research Network (NICHD)

The postnatal growth data of 4 of the infant cohorts from the National Institute of Child Health and Human Development Neonatal Research Network superimposed on the new chart

The postnatal growth data of 4 of the infant cohorts from the National Institute of Child Health and Human Development Neonatal Research Network superimposed on the new chart.

Results

Comparisons with the Babson chart

The graphical comparison of Babson's mean, 3rd and 97^{th }percentiles ^{th }percentile (Figure ^{rd }and 97^{th }percentiles.

When the means of Babson's and the new data were compared, head circumference and length measures were significantly different (p < 0.00005) after term. The small numbers in the sample used by Babson ^{th }percentile, with differences for weight of 260 grams at 36 weeks, for head circumference of 1.1 cm at 32 weeks and for length of 1.8 cm at both 30 and 32 weeks.

The percent of values below Babson's 10^{th }percentile ranged from the 5th to 17th for weight, the 5th and 15th for head circumference, and the 6th and 16th for length (data not shown).

How does the growth of preterm infants appear on the chart?

The growth curves from the NICHD for weight, length and head circumference showed clear deviations away from the intrauterine curves in the first weeks after birth. (Figure ^{th }percentile, and the smaller three groups remained below the 3^{rd }percentile. The average growth of head circumference recovered the original birth percentiles except for the smallest group. Only the smallest group's average head size remained below the 10^{th }percentile. The average growth in length also recovered from the post birth deceleration but like weight, only the largest group had an average size above the 3^{rd }percentile.

Discussion

This new fetal-infant chart is an updated Babson-type growth chart for use in neonatal intensive care units. This chart will allow a comparison for preterm infants as young as 22 weeks of gestation first with intrauterine and then with post term references and it can replace the one developed by Babson ^{th }percentile of this chart is accurate to the source data prior to 36 weeks; therefore it could be used for the assessment of size for gestational age for infants smaller than 2 kilograms.

This meta-analysis does not represent the actual growth of preterm infants for three reasons. First, the initial parts of the curves are based on the size of fetuses at birth, which do not show the change in weight that occurs after birth. This is followed by curves based on the growth of term infants who have not had the growth depressing effect of prematurity

Second, the validity of a meta-analysis is affected by methodological quality of the individual studies and the extent of heterogeneity in the studies and their results. Data from different studies were combined to prepare the graphs. It would have been preferable if all the data could have been obtained from one study with uniform methodology. The a priori criteria did limit the studies to those that were population based, had large sample sizes, and recorded gestational age in completed weeks. The studies used were all from developed countries where the majority of women are well nourished and the three studies used for the preterm section have publicly funded universal health care. The studies have slight differences in the inclusion and exclusion criteria (Table

Could a secular trend make the combining of the data invalid? There has been a slight trend towards an increase in birthweight of term infants in the United States and in Canada of 1 to 3 grams per year on average over the time frame that the data were collected

Third, the smoothing that was applied to bridge the pre and post term curves may or may not be a valid estimate of the growth of preterm infants. Further validation studies will determine the soundness of the smoothed disjunction between the data sets. The ideal validation study would compare the growth of a population of healthy preterm infants followed prospectively from birth though 50 weeks post-conceptual age. Healthy infants should be used to decrease the chance of bias due to illness of prematurity, however it would be useful to also document the growth patterns of unwell infants on this growth chart.

The superimposed NICHD growth curves (Figure ^{th }percentile for head growth, but remained below the 3^{rd }percentile for weight and length.

This finding is similar to that seen in previous studies of the growth of preterm infants on this type of growth chart

There were agreements and differences between this newer data with that of Babson. The agreements suggest that the Babson curves had fairly accurate depictions of infant size which may account for the continued popularity of this chart. The differences may reflect the small sample sizes of the early chart and the use of only maternal dates for the gestational age

The fetal-infant chart encourages the use of age that is adjusted for prematurity after 40 weeks. Once an infant surpasses 50 weeks, the regular CDC growth charts could be used, with continued adjustment for prematurity. Some authors have suggested that age be adjusted for preterm infants to at least 3 years

The data behind the new chart are predominantly cross sectional. Data of this type can be criticized since they are based on attained size of individuals and not on actual growth patterns

The selection of the CDC data for the post 40 week section may not be ideal since the database excluded preterm infants with a birth weight less than 1.5 kg. These infants grow differently than those of higher birthweight and this exclusion make the CDC charts more like a growth standard

It has been suggested that the distribution of weights among fetuses that are born preterm may be smaller than those that remain in utero and are delivered at term

The use of growth charts is only as accurate as the measurements that are made of the infants. Measurements of weight on electronic scales

Unfortunately the literature that accompanied the population based data sources does not include a description of the scales and tapes used to measure the infants, nor of the accuracy of these instruments. Since the data sources were population based, it is likely that a variety of precision occurred.

Head growth is important to monitor, since growth that deviates from the intrauterine rate could indicate problems

At term age, infants born prematurely tend to have a different body composition than infants born at term, with lower weight, lean tissue and bone mineral content and a higher percent body fat

Conclusions

This updated fetal-infant chart supports growth monitoring of preterm infants from as early as 22 weeks gestational age to 10 weeks post term age. It allows a comparison of an infant's growth with first the fetus and then the term infant and therefore allows an evaluation of catch up growth. Although Babson's "fetal-infant growth graph" had reasonably accurate estimates along the 50^{th }percentile, the larger sample sizes and more accurate gestational age assignments used here may provide better confidence in the extreme percentiles. As with all meta-analyses, the validity of this growth chart is affected by the methodological quality of the data sources and the heterogeneity of their results.

Comparison of the growth of the NICHD infants at attained weight of 2 kilograms provides evidence that, on average, preterm infants are growth retarded with respect to weight and length while, except for the smallest infants, their average head size has caught up to birth percentiles. Further validation work is needed to illustrate the growth patterns of preterm infants to older ages.

List of Abbreviations

kilogram (kg)

centimeter (cm)

Center for Disease Control (CDC)

National Institute of Child Health and Human Development Neonatal Research Network (NICHD)

Competing interests

None declared.

Authors' contributions

TF carried out the literature search, data selection, grid design, chart development, statistical comparisons and manuscript preparation.

Acknowledgements

Thanks to: Pat Fenton for assistance with the computer graph preparations, Doug McMillan for scientific and editorial advice, Debbie McNeil for statistical advice and Sari Czink and Suzanne Tough for editorial assistance.

Pre-publication history

The pre-publication history for this paper can be accessed here: