Fetal growth restriction represents the failure of a fetus to reach its predetermined growth potential and is associated with an increased risk of stillbirth, obstetric complications, neonatal morbidity, and neonatal death. In contrast, fetuses are classified as small for gestational age (SGA) if their estimated fetal weight is below a given percentile in a gestational age–specific population-based distribution of birth weight. Although a subset of SGA fetuses truly are affected with fetal growth restriction, many SGA fetuses are constitutionally small and have no increased risk of morbid outcomes stemming from fetal growth restriction. Improved ability to distinguish fetal growth restriction from constitutional smallness could improve resource allocation toward pregnancies at increased risk for adverse outcomes and reduce the burden associated with increased surveillance in pregnancies with small but appropriately grown fetuses, such as patient anxiety, testing that is not indicated, and iatrogenic preterm delivery.
Customized birth weight standards are intended to account for the variation in fetal growth that is attributable to differences in maternal characteristics. These growth curves are adjusted to reflect fetal gender, as well as maternal height, weight, race, and parity. Customization of fetal growth standards has been purported to improve the ability to identify pregnancies at increased risk for morbid perinatal outcomes1 – 5 and to enable providers to distinguish between fetal growth restriction and constitutional smallness.4,6,7 Customized standards of fetal growth have been endorsed by the British Royal College of Obstetrics and Gynaecology as the preferred clinical standard.8
Despite increasing use of customized fetal growth standards, the validity of claims that they improve the ability to distinguish between physiologically and pathologically small fetuses has been challenged.9 – 13 With increasing clinical application, it is imperative that the clinical utility of customized standards of fetal growth be rigorously tested and investigated. The purpose of this retrospective study was to estimate and compare the risk of morbid perinatal outcomes in pregnancies identified as SGA with customized compared with conventional standards of fetal growth.
MATERIALS AND METHODS
The methods used to generate an intrauterine growth trajectory and customized fetal growth standard have been described previously.14 Briefly, data were collected from the Magee Obstetric Medical and Infant database, which routinely collects comprehensive maternal, fetal, and neonatal outcomes from electronic and medical record data of all women delivering at Magee-Womens Hospital, University of Pittsburgh Medical Center, Pittsburgh, PA. A subset of individuals included in the Magee Obstetric Medical and Infant database had ultrasound data available for analysis. Data from 7,510 ultrasound-derived estimates of fetal weight obtained during clinically indicated sonograms in pregnancies that resulted in delivery of a live singleton at term were analyzed. We excluded individuals with diabetes (gestational or pregestational), chronic hypertension, gestational hypertension, preeclampsia, or congenital anomalies. Using quadratic polynomial regression, the gestational age at the time of ultrasound examination was used to predict the log-transformation of estimated fetal weight.
Using methods established by Gardosi et al,3,15,16 mulitvariable linear regression was then used to generate a predictive model of birth weight using demographic and clinical variables as covariates. Women who delivered live full-term singletons without congenital anomalies at Magee-Womens Hospital between March 2004 and May 2007 were evaluated (n=5,072). Individuals with diabetes, hypertension, or preeclampsia were not excluded from birth weight analysis. Physiological variables (maternal height, weight, parity, race, fetal gender, and gestational age) and pathological factors (smoking, illicit drug use, diabetes, and hypertension or preeclampsia) were included in the regression model. Covariates were selected by backward elimination with a significance level of 0.05.
After generating a predictive model of birth weight, an ideal birth weight for 40 weeks of gestational age was calculated, based exclusively on nonpathologic variables (maternal height, weight, parity, race, and fetal gender) and a standardized gestational age of 40 weeks. After calculating an ideal birth weight at 40 weeks, cutoff values for the 10th and 90th percentiles were generated. These values were based on the coefficient of variation multiplied by the Z score for the 10th and 90th percentiles (−1.28 and 1.28, respectively). The mean birth weight for the study population was 3,417 g, with a standard deviation of 485 g, yielding a coefficient of variation of 0.14. Thus, the resulting formula for calculation of the 10th and 90th percentiles was ideal birth weight at 40 weeks of gestation±(ideal birth weight at 40 weeks×1.28×0.14).
Once ideal birth weight at 40 weeks of gestation and the corresponding 10th and 90th percentiles were calculated, the ideal birth weight and corresponding cutoffs for the 10th and 90th percentiles could be extrapolated for any gestational age using the intrauterine growth curve previously generated. With this method, a customized ideal birth weight and a customized standard for suboptimal or excessive growth could be generated for any pregnancy at any gestational age.
After generating a model of customized fetal growth, we compared rates of morbid outcomes relating to fetal growth restriction in participants classified as SGA using a customized standard (SGAcust) and those classified as SGA using a population-based standard (SGApop). A separate population of 32,070 pregnancies that resulted in delivery at Magee-Womens Hospital from 2003 to 2008, distinct from that used to generate a fetal growth curve or customized birth weight standard, was used for this analysis. This population included all women who delivered liveborn neonates beyond 24 weeks of gestation with complete records that were not included in the predictive birth weight model, and included individuals with multiple gestations, congenital anomalies, and preterm delivery. We classified pregnancies as SGAcust if the birth weight was below the customized cutoff for the 10th percentile. Pregnancies were classified as SGApop if the recorded birth weight was below the 10th percentile for gestational age according to the national standards published by Alexander et al.17 The demographic composition of this population has been reported previously.14 Multivariable logistic regression was used to compare the risk of a 5-minute Apgar score less than 4, placental abruption, and neonatal death in the SGApop and SGAcust cohorts. Multinomial logistic regression was used to compare rates of specific hypertensive disorders in SGApop compared with SGAcust individuals.18 Covariates were selected by backward elimination with a significance level of 0.05. The cohort of pregnancies not classified as SGA by any method served as the referent group for all regression analyses.
Continuous variables were compared using the Kruskal-Wallis rank test. Categorical variables were compared using the χ2 test. All calculations were performed using Stata (10 and 11).
Two-thousand eight-hundred forty-five neonates were classified as SGApop (8.87%), whereas 2,579 neonates were classified as SGAcust (8.04%) (Fig. 1). As demonstrated in Tables 1 and 2, both the SGApop and SGAcust cohorts had an increased risk of morbid outcomes and hypertensive disorders of pregnancy.
As illustrated in Figure 1, there was a substantial degree of overlap between the SGAcust and SGApop cohorts. To investigate the population of individuals who would be overlooked with the exclusive use of a single classification method, we created four strata for analysis: 1) individuals who were not identified as SGA by any standard; 2) individuals identified as SGA only by the customized standard (SGAcust only); 3) individuals identified as SGA only by the conventional birth weight standard (SGApop only); and 4) individuals identified as SGA by both methods (SGAboth). The demographic characteristics of these cohorts are listed in Table 3.
In the unadjusted analyses, individuals identified as SGAcust only were at increased risk for a 5-minute Apgar score less than 4, placental abruption, neonatal death, and every hypertensive disorder analyzed (Tables 4 and 5). With the exception of mild preeclampsia, pregnancies identified as SGApop only were at no greater risk for adverse perinatal outcomes than non-SGA pregnancies.
The average gestational age at delivery in the SGAcust only cohort was more than 4 weeks earlier than that of the SGApop only cohort; 65% of neonates identified as SGAcust only were delivered before 37 weeks of gestation, whereas only 4.8% of SGApop only individuals delivered preterm (Fig. 2). To control for confounding stemming from differences in rates of prematurity between cohorts, we included gestational age at delivery as a covariate in our logistic regression model. Table 4 illustrates that after adjustment for differences in gestational age at delivery, the magnitude of risk of adverse outcomes in the SGAcust only cohort was greatly reduced from that obtained in our unadjusted analyses. After adjustment, the point estimates of the odds ratio of a 5-minute Apgar score less than 4, placental abruption, and neonatal death in the SGApop only cohort were higher than those in the SGAcust only cohort. Table 5 shows the calculated relative risk ratios of specific hypertensive disorders after adjustment for gestational age at delivery. With the exception of mild preeclampsia, the risk of all hypertensive disorders of pregnancy remained increased in the SGAcust only cohort after adjustment for gestational age at delivery. However, this adjustment led to a reduction in the relative risk ratios of all preeclamptic disorders in the SGAcust only cohort, with little effect on the calculated risk ratios for the SGApop only cohort. Exclusion of multiple gestations from the analysis did not change the significance of our findings (data not shown).
Figure 3 shows the customized SGA curves for individuals whose customized ideal birth weight were at the 1st, 5th, 10th, and 50th percentiles of our study population, along with the curve reflecting the 10th percentile of the birth weight data of Alexander. The customized cutoff for SGA in individuals in the lowest percentile of the study population was higher than the SGA threshold derived from the Alexander curve in the early portion of the third trimester.
Our data illustrate the confounding effect of the discrepancy in gestational age at delivery between the SGAcust only and SGApop only cohorts. Prematurity is a strong predictor of neonatal death and morbidity, and rates of these adverse outcomes decline dramatically with increasing gestational age.19 In the current study, neonates identified as SGAcust only were far more likely to be delivered preterm than SGApop only neonates. A similar discrepancy in rates of prematurity was reported in previous studies.1,3,5 This discrepancy is unrelated to the customization process itself. As a result of the differing contours of the SGAcust and SGApop fetal growth trajectories, small neonates delivered at early gestational ages are classified as either SGAcust only or SGAboth, but never as SGApop only (Fig. 3). Consequently, SGAcust only individuals have far higher rates of premature delivery than SGApop only individuals.
Proponents of customized standards of fetal growth have cited the significant increase in risk of adverse outcomes in the SGAcust only cohort, as well as the absence of an increased risk in the SGApop only cohort, as evidence of their superiority over population-based standards.1,6,20 With an alpha of 0.05, our analysis had 80% power to detect a 2.3-fold difference between SGAcust only and SGApop only individuals regarding odds of neonatal death. After adjustment for gestational age at delivery, the point estimate of risk of neonatal death was higher in the SGApop only cohort, and no significant difference was detected. Our findings demonstrate that correction for differences in gestational age at delivery negates the discrepancy in risk of adverse perinatal outcomes between SGAcust only and SGApop only pregnancies. An assessment of the utility of a fetal growth standard must include consideration of the effect of confounders on rates of morbid outcomes. Analyses that do not account for these differences in gestational age at delivery are fraught with systematic bias.
In addition to recognition of and correction for the effect of gestational age at birth, our study offers a number of strengths. Our analyses of morbid outcomes were performed on a population distinct from that used to generate our predictive model of birth weight. Furthermore, we distinguished clinically distinct subtypes of pregnancy-related hypertension and used multinomial regression to account for multiple competing outcomes.
Our study is limited by its retrospective nature. The effect of altered obstetric management associated with clinical concern for growth restriction cannot be fully accounted for in our analyses. The data used for outcome analysis were collected outside of a research protocol, and the accuracy of pregnancy dating is prone to the same limitations as found in clinical practice. As with previous studies of customized growth standards, the applicability of our findings to routine practice may be limited. In routine obstetric practice, a fetus is classified as SGA based on ultrasound findings, typically in the third trimester. In contrast, our analyses are based on newborn birth weights recorded at the time of delivery. The introduction of customized standards of fetal size into clinical obstetric practice would likely effect the interpretation of ultrasound-derived estimated fetal weights during antenatal screening for growth abnormalities. From 24 to 36 weeks of gestation, use of a customized standard of SGA would lead to an increase in the estimated fetal weight threshold below which a fetus would be considered SGA (Fig. 3). Consequently, with the use of a customized standard for SGA, the proportion of fetuses classified as SGA during third-trimester ultrasound screening would dramatically increase. Whether the increased number of positive screens for fetal growth restriction along with the associated increased utilization of medical resources, patient anxiety, and possible iatrogenic preterm delivery would actually result in improved perinatal outcomes cannot be determined in a retrospective study of birth weight data.
The decision to implement use of a particular fetal growth standard should be based on the ability to positively affect health outcomes with acceptable societal cost. We have demonstrated the effect of systematic bias leading to elevated rates of prematurity in the cohort of individuals exclusively identified as SGA with a customized standard. Previous studies of customized standards have not addressed the effect of differences in gestational age on rates of morbid outcomes. Furthermore, adoption of a customized standard of SGA would lead to a universal increase in the threshold for SGA in the early third trimester. Retrospective studies of customized newborn birth weight standards cannot assess the effect of altered obstetric management that would result from the antenatal classification of a fetus as SGA. Without consideration of the consequences of labeling a fetus as SGA, it cannot be determined if adoption of customized criteria for fetal SGA would result in improved outcomes. Before implementation of customized growth standards, it is imperative that they be evaluated in an unbiased prospective fashion that reflects the true consequences of their introduction to clinical practice. Calls for their use should be received with caution.
1. Odibo AO, Francis A, Cahill AG, Macones GA, Crane JP, Gardosi J. Association between pregnancy complications and small-for-gestational-age birth weight defined by customized fetal growth standard versus a population-based standard. J Matern Fetal Neonatal Med 2011;24:411–7.
2. Clausson B, Gardosi J, Francis A, Cnattingius S. Perinatal outcome in SGA births defined by customised versus population-based birth weight standards. BJOG 2001;108:830–4.
3. Gardosi J, Francis A. Adverse pregnancy outcome and association with small for gestational age birth weight by customized and population-based percentiles. Am J Obstet Gynecol 2009;201:28e1–8.
4. Gardosi J, Clausson B, Francis A. The value of customized percentiles in assessing perinatal mortality risk associated with parity and maternal size. BJOG 2009;116:1356–63.
5. Figueras F, Figueras J, Meler E, Eixarch E, Coll O, Gratacos E, et al.. Customised birth weight standards accurately predict perinatal morbidity. Arch Dis Child Fetal Neonatal Ed 2007;92:F277–80.
6. Figueras F, Gardosi J. Intrauterine growth restriction: new concepts in antenatal surveillance, diagnosis, and management. Am J Obstet Gynecol 2011;204:288–300.
7. Gardosi J. Clinical strategies for improving the detection of fetal growth restriction. Clin Perinatol 2011;38:21–31.
8. The investigation and management of the small-for-gestational-age fetus. Royal College of Obstetricians and Gynaecologists Guideline No. 31. London (UK): Royal College of Obstetricians and Gynaecologists; 2002.
9. Hutcheon JA, Zhang X, Cnattingius S, Kramer MS, Platt RW. Customised birth weight percentiles: does adjusting for maternal characteristics matter? BJOG 2008;115:1397–404.
10. Hutcheon JA, Zhang X, Platt RW, Cnattingius S, Kramer MS. The case against customised birth weight standards. Paediatr Perinat Epidemiol 2011;25:11–6.
11. Zhang X, Platt RW, Cnattingius S, Joseph KS, Kramer MS. The use of customised versus population-based birth weight standards in predicting perinatal mortality. BJOG 2007;114:474–7.
12. Hutcheon JA, Walker M, Platt RW. Assessing the value of customized birth weight percentiles. Am J Epidemiol 2011;173:459–67.
13. Hemming K, Hutton JL, Bonellie S. A comparison of customized and population-based birth-weight standards: the influence of gestational age. Eur J Obstet Gynecol Reprod Biol 2009;146:41–5.
14. Larkin JC, Speer PD, Simhan HN. A customized standard of large size for gestational age to predict intrapartum morbidity. Am J Obstet Gynecol 2011;204:499.e1–10.
15. Gardosi J, Francis A. A customized standard to assess fetal growth in a US population. Am J Obstet Gynecol 2009;201:25e1–7.
16. Gardosi J, Chang A, Kalyan B, Sahota D, Symonds EM. Customised antenatal growth charts. Lancet 1992;339:283–7.
17. Alexander GR, Himes JH, Kaufman RB, Mor J, Kogan M. A United States national reference for fetal growth. Obstet Gynecol 1996;87:163–8.
18. Hosmer DW, Lemeshow S. Applied logistic regression. 2nd ed. New York (NY): Wiley; 2000.
19. Garite TJ, Clark R, Thorp JA. Intrauterine growth restriction increases morbidity and mortality among premature neonates. Am J Obstet Gynecol 2004;191:481–7.
© 2012 The American College of Obstetricians and Gynecologists
20. Figueras F, Gardosi J. Should we customize fetal growth standards? Fetal Diagn Ther 2009;25:297–303.