Improved method for revising the Israel birthweight references

Introduction

Materials and methods

Results

There were 1,761,884 live births for the years 2010–2019. Of these, 66,556 (3.8%) births were excluded: 60,982 due to missing data (gestational age, gender, parity or weight), 5,412 infants were excluded for major congenital malformations and 162 live births had a gestational age either less than 22 weeks or greater than 44 weeks. An additional 243 infants were excluded for blatantly implausible birthweights for the reported gestational age using visual mapping. Using criteria of more or less than five interquartile ranges another 785 (0.05%) with reported weights that were too low, and 4,604 (0.3%) with reported weights that were too large were excluded (Figure 1). In sum, 72,188 births, 4.1% of the total, were excluded. 1,689,696 (95.9%) live births for the years 2010–2019 were included in the analysis. There were 829,624 male singletons, 785,080 female singletons, 37,915 male multiple births and 37,077 female multiple births. Of the total live born infants, 4.4% were from multiple births. The percent of multiple gestation deliveries was 2.2%. The proportion of multiple births declined from 2.4 to 2.0% over the study period, as did the proportion of triplets (from 0.16 to 0.11%).

Figure 1:

Sample selection process.

Figures 2 and 3 present the birthweight curve of the singleton infants by gender with the distribution of the number of births for each gestational age superimposed. Curves for multiple births by gender are available in the Supplemental Material Appendices C and D. Among the singleton infants, the average gestational age was 39.1 weeks, which was significantly longer than the 35.6 weeks found among the 74,992 multiples. The mean birthweight of singletons (3,251 g) was 947 ± 4 g (2 SD for the difference of the means) higher than that of multiples (2,304 g). Birthweights for male infants were greater than for females.

Figure 2:

Birthweight percentiles by gestational age female singletons, 2010–2019.

Figure 3:

Birthweight percentiles by gestational age, male singletons, 2010–2019.

The percentiles of birthweights for each gestational age week by gender in singleton vs. multiple deliveries are available in Supplemental Material Appendices E–H. There were several differences between singletons and multiples in our population. The birthweight of the male multiple births began to differ from that of the singletons at 30 weeks whereas that of the female multiple births began to deviate at 31 weeks (Figure 4). At 40 weeks the difference in weight for a full term multiple birth infant was on the average a full 490 g less than for a singleton, whether male or female. The increase in birthweight of singletons leveled after 42 weeks and those born after 43 weeks were actually lighter than infants born earlier. The median GA for multiples was 37 weeks. The graphs for the multiple births continue only until 41 weeks due to the small number of births after week 41 for multiples. Among singleton deliveries, 5.2% of infants were born preterm (less than 37 weeks) and 5.4% were born with low birthweight (<2,500 g). In contrast, the preterm delivery rate among multiples was significantly higher at 53.6% and low birthweight was significantly higher at 60.4%.

Figure 4:

Median birthweight of singletons and multiples by gender.

Figure 5 compares the curves for all singletons from the present study (years 2010–2019) to those previously reported from the years 1993–2001. While the median is similar for both, the 90th and 97th percentiles between 28 and 37 weeks are much higher for the previously reported cohort. The updated 3rd and 10th percentiles for weeks 22–24 are lower than those for the previously reported cohort, while they are consistently higher beginning from week 32.

Figure 5:

Comparison birthweight percentiles by gestational age, singletons 1993–2001 to 2010–2019.

Discussion

We describe the birthweight by gestational age curves of live born infants in Israel between 2010 and 2019. These curves update those previously reported from the years 1993–2001 and are based upon data that is both more complete and accurate. The present curves include 95.9% of live births in the country as compared to the previous sample which included 73.4% of live births. Completeness of the database was improved by active and periodic cross checking with the independently run Israel Neonatal Network Registry to ensure that all preterm and very low birthweight infants were included. The introduction of the use of validations with upper and lower limits for data entry beginning in 2010 contributed to the accuracy of the measures and reduced the rate of exclusions for implausible birthweights for gestational age. The implementation of real time data transfer of birthweight and gestational age to the MOH registry beginning in 2012 eliminated the need for separate entry of gestational age data and contributed to the improved coverage and accuracy. Finally, we were able to exclude births with significant congenital malformations, cognizant of association between congenital malformations and small for gestational age. Therefore, the present exclusion rate due to missing weight, gestational age, gender and multiplicity of birth data was much lower.

In addition to the changes in the data base management, the present analysis cleaned the data of implausible birthweights for gestational age using adaptations of accepted algorithms after initially cleaning the data of obvious and extreme outliers. In the absence of individual level information regarding both the date of the last menstrual period (LMP) as well as the first trimester ultrasound dating, we could only exclude but not correct implausible birthweight-gestational age values using an adaptation of the Basso-Wilcox algorithm [9]. The use of five interquartile ranges as the exclusion criteria is more robust than the use of standard deviations from the mean since the distribution of birthweights is not normal or even symmetric and this method does not use the values of the outliers to determine what is an outlier, reducing the probability of including outliers with extreme and implausible weights for gestational age. Talge et al., in their study of birthweights in Michigan using the above noted algorithm report relying upon the LMP for 85% of the births and upon the obstetric-ultrasound assessment for the remaining 15% [10]. They corrected 9% of the gestational ages thus enabling them to include 99% of the sample. This is indeed less than our 4% exclusion rate, however the high proportion of women receiving first trimester ultrasound in Israel for dating probably contributed to overall accuracy of dating and assessments of gestational age in our study. The vast majority of women who enroll for antenatal care in the first trimester undergo an early ultrasound examination which is included in Israel’s basket of services.

Guiliani et al. conducted a systematic analysis of the methodology used to construct neonatal anthropometric curves and lend significant weight to prospectively conducted studies using uniformly performed measurements [11]. In so doing, the number of preterm infants for whom data is collected and analyzed is severely limited. Even in the INTERGROWTH-21 report, the data for females born at 33 weeks included only 17 observations [12]. Using a data base that spanned 10 years we were able to analyze data for considerable numbers of very low gestational age infants. The smallest number of observations upon which our data was analyzed was for 38 males born at 22 weeks. This improved accuracy is of import in light of the clinical implications of assessing the adequacy of fetal growth at low gestational ages. In France and elsewhere analyses of large computerized data sets have been used to successfully create growth curves and we opine that the present study methodology more than compensates for its retrospective nature [13]. The smoothed and unsmoothed curves produced were nearly identical and we were able to present the unsmoothed curves.

The literature as to the appropriate curves to use to assess intrauterine growth and birth size is conflicting. The INTERGROWTH 21 consortium proposes the need for universal prescriptive standards based on their studies showing similarity of fetal growth in healthy populations, similar to the assumptions underlying the development of the WHO 2006 Growth standards [12]. The proposed universal curves permit international comparisons of rates for small-for-gestational age, intrauterine growth retardation (IUGR) and large-for-gestational age. Numerous studies however have shown that there are significant and meaningful differences between locally derived data and that reported by the INTERGROWTH-21 consortium. Moreover, even within the data upon which those prescriptive curves are based, significant between-site differences exist, further buttressing the need for curves which reflect the unique characteristics of the population being described and assessed [14].

The question of whether to use descriptive or prescriptive growth charts is relevant also to twins. Whether twins’ reduced growth in the third trimester should be considered pathological or physiological depends upon whether one views it as a manifestation of inadequate supply to increased demand or an adaptative response to excessive uterine distension which permits a delay in the onset of labor. A recent review cites recent data supporting the benign adaptive hypothesis and recommends the use of specific twin growth charts provided one takes into account assymetric growth between the twins [15]. We present separate growth curves for twins which show how birthweights of twins begin to differ from that of singletons beginning at 30 weeks, similar to what Sanklampi described [16].

Our finding of reduced birthweight after 43 weeks is similar to that described in the previous Israel study [6]. We have no reason to suspect greater error in estimation of gestational age for this group nor do we have a definitive explanation for this finding. Although it is known that some but not all infants born post term suffer from placental insufficiency this is probably an insufficient explanation for the finding. While the number of cases is small, the finding is similar to that previously described and we understand it to be valid for the Israeli population.

Even when basing the curves on a population similar to the one in question, there remains the issue of whether using birthweight data is appropriate for assessing the adequacy of intrauterine growth. Hutcheon and Platt note that “weight-for-gestational-age charts and definitions of ‘small-for-gestational-age’ based on the distribution of livebirths at a given gestational age (that) have conventionally been used to identify infants whose fetal growth is poor … have serious shortcomings at preterm ages due to missing data on the weights of fetuses still in utero.” [17] Insofar as preterm birth may be associated with restricted intrauterine growth, the use of birthweight data of infants born preterm includes an inherent risk of skewing the curves downwards with the concomitant effect of underestimating fetuses with IUGR. Using ultrasound measurements correlated with measured birthweights has been used to generate more accurate estimates of fetal weight [18]. Sapir et al. published in 2017 a single center study of ultrasound measurement based fetal growth curves for Israel, comparing them to the Dollberg curves based upon birthweights [19]. This study updated earlier curves for Israel proposed by Romano-Zelekha [20]. Sapir’s study found that the 50th percentile of sonographic EFW was consistently larger as compared to the 50th percentile of neonatal BW in the preterm period, with a discrepancy of up to 11% at the 29 weeks of gestation. Of note is their finding that the 90th percentile in their study was significantly lower than the sonographic estimated fetal weight percentile between weeks 31–34, with a maximum difference of 36% at 32 week of gestation. While the present study is insufficient to resolve this controversy, it is salient to note that when comparing our updated curves to the previously reported Dollberg birthweight curves, the 90th and 97th percentiles between weeks 28–37 are significantly lower. This further suggests the improved validity of the present study. Moreover, the ultrasound based curves are based upon a single center with a preponderance of ultra-orthodox population, and disproportionate number of multiparae and grand multiparae, which may not be representative of Israel’s overall population. A prospective study is needed to resolve the question of the need for ultrasound based curves.

Conclusions

We present an analysis of birthweights by completed gestational age Improved data collection and validation permitted inclusion of over 96% of births for analysis. Use of interquartile range distribution to exclude values of birthweight/gestational age that were implausible improves validity. In comparison to curves reported previously, changes were found in the distribution of birthweights for the upper and lower percentiles. Periodic updates of growth curve references are important.