Birth weight is strongly associated with longevity of offspring as well as parents. Extremes of birth weight are associated with increased perinatal,1,2 infant,2 and adulthood mortality.3–6 Low birth weight is associated with increased all-cause mortality in adulthood,3–5 whereas adults who were large at birth have an increased risk of cancer-related death.3,6 The association between birth weight and adult mortality persists after adjusting for genetic effects of parental life span and socioeconomic factors.4
Birth weight is also associated with parental life span, irrespective of socioeconomic status. Mothers of small for gestational age infants experience a twofold increase in mortality due to cardiovascular disorders,7–9 whereas mothers of large for gestational age neonates have a higher incidence of breast cancer.10
An individual’s birth weight is the sum of physiologic and pathologic influences operating during pregnancy. To define abnormalities of birth weight, these two influences must be separated. The sum of physiologic influences in an individual determines the limits of normal variability in birth weight. Birth weight departure outside these limits is assumed to be due to pathologic influences.
Conventionally, birth weight is assessed by comparison to population or ultrasound norms. Population norms include all pregnancies and assign percentiles based on distribution of birth weight for each gestational age. They are traditionally used for the evaluation in the nursery. Population norms include pregnancies with pathologic conditions affecting fetal growth and birth weight11 and hence provide flawed correction for physiologic variables. Ultrasound norms are based on serial ultrasound estimations of fetal weight in uncomplicated pregnancies. They assign percentiles based on distribution of estimated fetal weights for each gestational age. Ultrasound norms are used for ultrasound estimation of fetal growth. By averaging individual trajectories of fetal growth, ultrasound norms remove physiologic variability, undermining individualization of growth.12–17
Customized norms are not available for the U.S. population to date. They predict optimal term birth weight for each pregnancy based on four maternal characteristics, maternal weight, height, parity, and race, and infant sex. This term optimal weight is extrapolated to earlier gestational ages. Customized norms better detect aberrations of fetal growth than population norms.18–21 However, their accuracy in comparison with ultrasound norms is uncertain.22 Many factors known to affect birth weight are not considered by customized norms.23 They also do not account for factors with dual physiologic and pathologic effect on birth weight and for early variation in fetal growth.24,25 Consequently, their accuracy in defining normal birth weight variability and thus fetal growth abnormality is limited.
Here we identify a set of physiologic influences affecting birth weight in a population with normal outcome of pregnancy. We show that adjusting for those physiologic influences in the rest of the population improves identification of complicated pregnancies compared with existing methods.
MATERIALS AND METHODS
The study population consisted of patients recruited during First And Second Trimester Evaluation of the Risk for Aneuploidy (FASTER) trial,26 an observational study conducted in 15 centers and representing all major geographic regions of the country. Between 1999 and 2002, 43,267 patients were approached for enrollment, and 38,033 patients who met criteria were enrolled in the study. The inclusion criteria were a singleton pregnancy and a gestational age of 10 weeks 3 days through 13 weeks 6 days at the entry to the study, based on ultrasound measurement of fetal crown–rump length. Multiple gestations, fetal anencephaly, and women aged less than 16 years were excluded.
Extensive information on patients’ demographics, medical and obstetrical history, socioeconomic status measures, and exposures during pregnancy were recorded. Copies of medical records were reviewed in all cases where parents reported a possible neonatal medical problem, in all Down syndrome screen-positive cases without karyotype results, and in a 10% random sample of all other enrolled cases. Institutional review board approval was obtained in all participating centers, and all participants gave written informed consent.
Uncomplicated pregnancies were defined as those where none of the following were documented: miscarriage, abortion, perinatal death, maternal diabetes, hypertension, exposure to medications or illicit drugs, preeclampsia, pregnancy-induced hypertension, gestational diabetes, preterm labor, preterm premature rupture of the membranes, placental abruption, placenta previa or other placental conditions, chromosomal abnormalities, congenital malformations, neonatal complications, neonatal imaging or blood testing regardless of the test results and readmission to hospital, and, parental history of genetic, chromosomal, or congenital abnormalities. A total of 15,680 pregnancies met these criteria (Fig. 1).
Among this group, we excluded pregnancies delivering outside the range 246 and 298 days (which distribution of birth weight was nonnormal) and where the disparities between expected and ultrasound dates were greater than 7 days24,25 (because differences outside this range more likely reflect error in menstrual dating). Variability of the first trimester growth affects final birth weight. To avoid bias associated with first trimester growth we included the variable deltaGA in the model. DeltaGA is a measure of first trimester growth.24,25 The observed size of the fetus by ultrasound-measured crown–rump length in the first trimester was related to the expected size on the basis of the date of conception (estimated as 14 days after the first day of last menstrual period in pregnancies conceived spontaneously and precisely known in pregnancies conceived with assistance of in vitro fertilization or intrauterine insemination) and expressed as equivalence to days of gestational age (Fig. 1). The normal population is not a population of patients with normal pregnancy outcome, rather a subset of this population with actual and potential complications of pregnancy affecting birth weight excluded.
All pregnancies not meeting criteria for uncomplicated pregnancies constituted the complicated population, which included chromosomal and congenital malformations (Fig. 1). In both normal and complicated populations excluded were also participants with missing birth weight, gestational age, and physiologic determinants of birth weight data (Fig. 1).
This model included variables which predate pregnancy, and variables measured in the first or early second trimester. Preconceptionally known variables, selected based on literature, included maternal height and weight,18 self-reported race and ethnicity,18 age, education,27 marital status,28 length of folic acid supplementation before conception,29 number of prior term and preterm pregnancies,30 miscarriages and abortions, smoking and drinking status, number of cigarettes smoked and alcoholic drinks consumed daily in the first trimester,29 and altitude of the area of residence.30 Seven centers were less than 100 feet, three between 100 and 1,000 feet, one at 4,226 feet, and four at 5,260 feet; variables measured in the first trimester (10 3/7 to 13 6/7 weeks) included fetal heart rate, nuchal translucency,33 deltaGA (the difference between the observed and expected size of the fetus),25 maternal serum concentrations of placental proteins pregnancy-associated plasma protein A (PAPP-A), and the free β-hCG.34,35
Variables measured in early second trimester (15 0/7 to 18 6/7 weeks) were maternal serum concentrations of placental proteins hCG, inhibin A, alpha fetoprotein (AFP), and unconjugated estriol, which maternal concentrations are associated with placental function.36–38
A model predicting optimal birth weight was identified using multivariable regression analysis. The final model included all variables significantly associated with the birth weight in the normal population. Due to complexity of the model, interaction terms were not included, because expending model to incorporate all significant interactions between all physiologic determinants in the model would increase its complexity exponentially. Our objective was to identify a parsimonious model predicting optimal birth weight.
Fetal growth potential was calculated in four steps. First, individual optimal weight at 280 days was calculated using the final model of birth weight in the normal population. Second, individual optimal growth trajectory was calculated by assigning a proportion of individual optimal weight at 280 days to other gestational ages by interpolating fetal growth based on ultrasound growth equations. For each fetus, six individual optimal growth trajectories were calculated, one for each growth equation from six longitudinal ultrasound studies.12–16 Third, standard deviations (SDs) for those individual optimal growth trajectories were calculated as proportions of the standard deviation at 280 days. Fourth, percentiles of individual optimal fetal growth, the growth potential, were calculated using the Gaussian distribution. Thus, the calculated percentile describes the departure of actual birth weight from an individually calculated optimal weight.
Six growth potential norms were calculated using six fetal growth equations. Those equations describe change of fetal weight with gestational age. Selection of the equation for the growth potential norm was determined by whether the percentiles fitted a classification into three groups below the 10th, above the 90th or between the 10th and the 90th percentile expected from the Gaussian distribution.
Growth potential norms were compared with population,11 ultrasound,12 and customized norms18 in the normal population, the complicated population, pregnancies with diabetes or hypertensive disorders and neonatal morbidity. Customized norms are not available for the U.S. population and do not include coefficients for Hispanics. Moreover, their coefficients were based on pregnancies dated using late second-trimester ultrasonography. Thus, to allow fair comparison of norms, we fitted customized norms with coefficients developed in our normal population for variables used by those norms. Specifically, we used the set of original variables but replaced their coefficients with the coefficients (the parameters in the regression equation describing strength of the relationship between birth weight and given variable) derived from our normal population. Diabetes and hypertension disorders are a subset of the complicated population with pregestational, gestational diabetes, chronic hypertension or preeclampsia. Those conditions are strongly associated with fetal growth aberrations and make up a majority of indications for prenatal fetal growth assessment. Neonatal morbidity was defined as conditions in the neonatal period requiring admission to the hospital.
Multivariable regression analysis was used to identify factors associated with birth weight in the normal population. The regression equation of those physiologic determinants of birth weight was used to predict individual optimal birth weight. Backward elimination was used for selection of variables. The model was constrained to include known clinically important variables (gestational age at delivery, race, sex, and maternal weight and height). When eliminating continuous variables, first the highest-order coefficient, the cubic coefficient, was removed. If a higher-order coefficient was significant (for example cubic), then the lower order coefficient (for example, square) was kept in the model, whether significant or not. Thus the final equation consisted of the lowest-order significant coefficients for variables, with the exception of those variables where a higher order coefficient was significant and lower order not, in which case all coefficients were included regardless of significance.
The regression’s square root of the residual mean square was used to calculate SD at 280 days. The SD for the entire range of the individual optimal growth trajectory was then calculated as a proportion of the SD at 280 days. Thus, the relationship between individual optimal weight for gestational age and individual optimal weight at 280 days is the same as between the SD for the given gestational age and the SD at 280 days and is defined by the fetal growth equation. The actual birth weights were then converted into z-scores, and percentiles were calculated based on the relationship between z-scores and percentiles existing in the Gaussian distribution.
In the normal population, classification of percentiles into three classes (lower 10%, middle 80%, and upper 10%) was compared with the classification expected from the Gaussian distribution using χ2 test for goodness of fit. In this context, Gaussian distribution relates to distribution of percentiles among three categories: below 10th, 10th to 90th, and above the 90th percentile. In the Gaussian distribution, 10%, 80%, and 10% of participants would be expected in the three categories, respectively. The significance of the difference in classification into those three categories between Gaussian distribution and observed with application of each birth weight norm was tested using χ2 test for goodness of fit.
In the complicated population and the population with diabetes or hypertensive disorders, proportions of pregnancies with growth aberrations below the 10th and above the 90th percentile identified by traditional norms were compared with the proportion identified by growth potential. The difference between the two proportions was calculated and related to performance of traditional norms. This difference was shown as a percent of proportion of pregnancies identified by traditional norm. Confidence intervals for the difference in the proportions of pregnancies identified with aberrant growth were derived by bootstrapping the differences with 1,000 replications of the data set. Thus, the confidence intervals for the differences were calculated by estimating those differences in 1,000 randomly selected samples of the population with replacement. The differences in identification of aberrant growth between traditional norms and growth potential with confidence intervals not crossing 0 were statistically significant with P<.05. This method of comparison between tests has previously been used successfully.26 The proportion of pregnancies classified as aberrantly grown between growth potential and traditional norms was also tested using the McNemar test.
There were 9,818 pregnancies in the normal population, and 14,229 in the complicated population with all data available and deltaGA between –7 and +7 days (Fig. 1). Pregnancies in the normal population had lower maternal age, weight, and height. They were more likely to be parous and less likely to have had a previous preterm delivery, miscarriage, or abortion. Mothers in the normal population were less likely to reside at higher altitudes, smoke or drink alcohol, and conceive with assistance of reproductive technologies. Normal pregnancies in the first trimester had smaller nuchal translucency, slower fetal heart rate, higher maternal concentrations of PAPP-A, and lower of free β-hCG. In the second trimester, normal pregnancies had lower concentrations of inhibin A and hCG and higher estriol. Pregnancies in the normal population had longer duration, higher birth weight, and a lower proportion of male infants (Table 1).
Birth weight in the normal population was significantly and independently associated with gestational age, maternal characteristics, measures of socioeconomic status, outcomes of prior pregnancies, environmental exposures, ovulation induction, fetal sex, measures of growth and development in the first trimester of pregnancy, and biomarkers of placental function (Table 2). The regression model of birth weight using those variables had a coefficient of determination R2=0.358. Thus more than 35% of the variability in the birth weight can be explained by this regression model. Analysis of the distribution of residuals found no evidence of violation of assumptions of linear regression.
Gestational age and maternal height and weight in the first trimester were positively associated with birth weight in a S-shaped manner (Supplemental Fig. 1). Nonwhite mothers had lower birth weight in the following descending order: Hispanics, Asian, Native American, other ethnicities, and African Americans. Being a single mother (P=.006) and incomplete high school education (P=.06), had independent negative effects on birth weight (Table 2). Number of prior abortions showed a weak trend for higher birth weight (Supplemental Fig. 1).
The number of prior term pregnancies, but not prior preterm deliveries, was strongly positively associated with birth weight. The effect plateaued after a second prior term pregnancy (Fig. 2).
Altitude and smoking had significant negative effects on birth weight. Comparing with mothers residing at altitudes of 100 to 1,000 feet, higher and lower altitudes had a negative effect on birth weight (Table 2). DeltaGA, nuchal translucency, and fetal heart rate had positive linear associations with birth weight.
Pregnancy-associated plasma protein A, hCG, estriol, and inhibin A were positively associated with birth weight, whereas AFP was negatively associated. Free β-hCG was only weakly associated with birth weight (Fig. 3). The center where the patient was recruited did not materially affect the coefficients of birth weight determinants.
Hadlock’s equation of fetal growth12 was only one of the six fetal growth formulae used in the calculation of growth potential percentiles that resulted in a Gaussian distribution of percentiles. Hence, this equation was selected for calculation of growth potential percentiles and validation.
Growth potential norms were superior to traditional norms in normal and complicated populations. In the normal population, only growth potential percentiles classification into three classes was not significantly different from classification expected from Gaussian distribution (P =.4). Population, ultrasound, and customized percentiles distributions into three classes were significantly nonnormal (P<.001 for all) (Table 3). The distribution of normal pregnancies among three categories, below the 10th, 10th to 90th, and above the 90th percentile, by traditional norms are significantly different from expected from the Gaussian distribution 10%, 80%,and 10%, respectively.
In the complicated population, growth potential percentiles classified significantly more pregnancies below the 10th or above the 90th percentile than population, ultrasound, and customized norms (Table 4). Growth potential classified outside the normal range 15.9%, 41.4%, and 51.7% more pregnancies than customized, ultrasound, and population norms, respectively, P<.001 for all (Table 4).
Additionally, the distribution of growth potential percentiles was evenly distributed between categories below the 10th and above the 90th percentile. The distributions of population, ultrasound, and customized norms were skewed toward lower percentiles (Table 4). Traditional norms classified more pregnancies below the 10th percentile, but above the 90th percentile they classified less than 10% expected even from the Gaussian distribution of normal population.
Among 1,518 pregnancies with diabetes or hypertensive disorders, growth potential percentiles classified more pregnancies as being outside the normal range than population, ultrasound, and customized norms (Table 4). Growth potential identified 11.2%, 33.2%, and 64.6% more pregnancies as aberrantly grown, than customized, ultrasound, and population norms, respectively; P<.001 for all (Table 4).
In 1,347 pregnancies with neonatal morbidity, growth potential norms classified more pregnancies as aberrantly grown than population, ultrasound, and customized norms (Table 4). Growth potential classified below 10th or above 90th percentile 11.7%, 33.7%, and 69.0% more pregnancies than customized, ultrasound, and population norms, respectively; P<.001 for all (Table 4).
In this article, we have developed a method for classification of birth weight in relation to a presumed normal range. The methodologic advantages of the approach used are as follows: First, we included a large number of demographic, biochemical, and biometric factors that influence fetal growth. This was possible by using a large, prospective cohort study of women who were followed longitudinally. Second, we derived the average and normal limits of variation of factors associated with fetal growth from a population that had a completely normal outcome of pregnancy using strict criteria that would effectively eliminate pregnancies with overt or covert abnormalities of fetal growth. Third, for each fetus, we interpolated fetal growth trajectory between the beginning of pregnancy and the individually predicted weight at 280 days. This addresses the problem that impaired fetal growth is associated with increased rates of preterm delivery and delivery at earlier weeks of term. Fourth, we were able to identify factors with both physiologic and pathologic effect on birth weight in different parts of their range. Finally, we accounted for early growth aberrations by dating all pregnancies using first trimester ultrasonography. As a consequence of the superiority of the technical approach, we believe that the growth percentiles described in this article most accurately reflect normal and the limits of normal of human fetal growth.
This conclusion is supported by a number of observations. First, natural variation in fetal growth would be expected to be reflected in a Gaussian distribution of observed values in a healthy population. When we examined our healthy population, we observed a Gaussian distribution of values using our percentile method, but not using other methods previously described (Table 3). We interpret their nonnormal results as likely indicating flawed estimation of achieved fetal growth in healthy pregnancies. Second, we applied our method to women with complicated pregnancies. We hypothesized that a better method for characterizing abnormal growth would result in a greater proportion of women with complicated pregnancies being classified as having aberrant fetal growth. We observed this to be the case (Table 4). It may be argued that this finding simply reflects the fact that we developed a model that was fitted to our data. This cannot, however, explain the improved performance, because the model used to characterize growth among women with complicated pregnancies was derived from a different subset, namely, women with uncomplicated pregnancies.
Some of the factors affecting birth weight have both physiologic and pathologic effects, depending on their value. Maternal body mass index (BMI) has a positive association with birth weight.18 However, at the extremes of its range, malnutrition or obesity, BMI is associated with the well recognized adverse effects on fetal growth and perinatal mortality.39,40 Smoking is also associated with low birth weight and perinatal mortality.31,41 Paradoxically, however, among low birth weight infants smokers have lower fetal mortality than nonsmokers. This may be explained by a leftward shift of gestational age–specific curves of perinatal mortality as a function of birth weight observed in smokers. Such curves have characteristic U-shapes. Therefore a leftward shift observed in smokers results in more pregnancies being classified as low birth weight, most of them with low perinatal mortality.
Thus, effects of smoking on birth weight and perinatal mortality are, at least in part, separate.31,42 This may reflect different ways smoking interacts with genotype polymorphisms43 or complications of pregnancy like preeclampsia, which is less prevalent in smokers.44 Although the effect of smoking on birth weight in the absence of pathologic conditions is not strictly physiologic, it needs to be accounted for in the prediction of optimal weight to be achieved in the absence of adverse outcomes. We defined physiologic effects of those factors by identifying their effect on birth weight in the normal population in the absence of confounding pathologic conditions and adverse outcomes.
The study population is a good representation of the U.S. population. Maternal age, body mass index, and nulliparity were very similar between the two populations. The prevalence of maternal age younger than 35 years (79% compared with 86%), BMI 18.5–25 (58% compared with 50%), and nulliparity (45% compared with 40%) did not materially differ between the study population and the general U.S. population. However, the study population had fewer African Americans, more whites, and a similar proportion of Hispanics (5% compared with 14%, 67% compared with 57%, and 22% compared with 21%, respectively). Unmarried mothers and mothers with less than 12 years of education were also less common in the study population than in U.S. population (21% compared with 36% and 26% compared with 52%, respectively).
This analysis reveals insights into the importance of abnormal fetal growth in complicated pregnancies. Overall, women with complicated pregnancies had 26% of fetuses classified as outside the 10–90th percentiles (Table 4). The expected figure is 20%. Hence, between 5% and 10% of complicated pregnancies seem to have some degree of growth disorder. The complicated pregnancies group is by its definition heterogeneous and does not identify pregnancies with adverse outcomes. Rather, it constitutes a subpopulation of pregnancies with various degrees of complication serving to compare growth potential with traditional norms. Among pregnancies with diabetes or hypertensive disorders, 37.3% of birth weights were classified as outside the 10–90th percentiles (Table 4). This suggests that almost 20% of those pregnancies are associated with abnormal fetal growth. Diabetes and hypertensive disorders constitute the most common indications for evaluation and monitoring of fetal growth. Thus, in such a population growth potential and other norms would be most likely applied. Importantly, a third of pregnancies with severe neonatal morbidity had abnormal fetal growth.
Although growth potential norms are a multimarker test, they can be applied immediately clinically. Many of the variables used are routinely collected during pregnancy, both for routine prenatal care and multimarker Down syndrome screening. The variables can also be added as they become available, starting with only birth weight and gestational age. Finally, growth potential calculation can be performed using a Web-based program.
In conclusion, we have identified factors affecting fetal growth under physiologic conditions, allowing better understanding and further study of this fundamental human process. We have shown that growth potential norms based on the physiologic determinants of birth weight are a better discriminator of aberrations of fetal growth than traditional norms. Those findings have a potentially immediate research and clinical applicability.
1. McIntire DD, Bloom SL, Casey BM, Leveno KJ. Birth weight in relation to morbidity and mortality among newborn infants. N Engl J Med 1999;340:1234–8.
2. Schoendorf KC, Hogue CJ, Kleinman JC, Rowley D. Mortality among infants of black as compared with white college-educated parents. N Engl J Med 1992;326:1522–6.
3. Kajantie E, Osmond C, Barker DJ, Forsen T, Phillips DI, Eriksson JG. Size at birth as a predictor of mortality in adulthood: a follow-up of 350 000 person-years. Int J Epidemiol 2005;34:655–63.
4. Andersen AM, Osler M. Birth dimensions, parental mortality, and mortality in early adult age: a cohort study of Danish men born in 1953. Int J Epidemiol 2004;33:92–9.
5. Barker DJ, Osmond C, Forsen TJ, Kajantie E, Eriksson JG. Trajectories of growth among children who have coronary events as adults. N Engl J Med 2005;353:1802–9.
6. Ahlgren M, Melbye M, Wohlfahrt J, Sorensen TI. Growth patterns and the risk of breast cancer in women. N Engl J Med 2004;351:1619–26.
7. Davey Smith G, Hart C, Ferrell C, Upton M, Hole D, Hawthorne V, et al. Birth weight of offspring and mortality in the Renfrew and Paisley study: prospective observational study. BMJ 1997;315:1189–93.
8. Smith GC, Pell JP, Walsh D. Pregnancy complications and maternal risk of ischaemic heart disease: a retrospective cohort study of 129,290 births. Lancet 2001;357:2002–6.
9. Pell JP, Smith GC, Walsh D. Pregnancy complications and subsequent maternal cerebrovascular events: a retrospective cohort study of 119,668 births. Am J Epidemiol 2004;159:336–42.
10. Cnattingius S, Torrang A, Ekbom A, Granath F, Petersson G, Lambe M. Pregnancy characteristics and maternal risk of breast cancer. JAMA 2005;294:2474–80.
11. Alexander GR, Himes JH, Kaufman RB, Mor J, Kogan M. A United States national reference for fetal growth. Obstet Gynecol 1996;87:163–8.
12. Hadlock FP, Harrist RB, Martinez-Poyer J. In utero analysis of fetal growth: a sonographic weight standard. Radiology 1991;181:129–33.
13. Gallivan S, Robson SC, Chang TC, Vaughan J, Spencer JA. An investigation of fetal growth using serial ultrasound data. Ultrasound Obstet Gynecol 1993;3:109–14.
14. Jeanty P, Cantraine F, Romero R, Cousaert E, Hobbins JC. A longitudinal study of fetal weight growth. J Ultrasound Med 1984;3:321–8.
15. Marsal K, Persson PH, Larsen T, Lilja H, Selbing A, Sultan B. Intrauterine growth curves based on ultrasonically estimated foetal weights. Acta Paediatr 1996;85:843–8.
16. Schwarzler P, Bland JM, Holden D, Campbell S, Ville Y. Sex-specific antenatal reference growth charts for uncomplicated singleton pregnancies at 15-40 weeks of gestation. Ultrasound Obstet Gynecol 2004;23:23–9.
17. Johnsen SL, Rasmussen S, Wilsgaard T, Sollien R, Kiserud T. Longitudinal reference ranges for estimated fetal weight. Acta Obstet Gynecol Scand 2006;85:286–97.
18. Gardosi J, Chang A, Kalyan B, Sahota D, Symonds EM. Customised antenatal growth charts. Lancet 1992;339:283–7.
19. Clausson B, Gardosi J, Francis A, Cnattingius S. Perinatal outcome in SGA births defined by customised versus population-based birthweight standards. BJOG 2001;108:830–4.
20. Ego A, Subtil D, Grange G, Thiebaugeorges O, Senat MV, Vayssiere C, et al. Customized versus population-based birth weight standards for identifying growth restricted infants: a French multicenter study. Am J Obstet Gynecol 2006;194:1042–9.
21. Efkarpidis S, Alexopoulos E, Kean L, Liu D, Fay T. Case-control study of factors associated with intrauterine fetal deaths. MedGenMed 2004;6:53.
22. Owen P, Ogah J, Bachmann LM, Khan KS. Prediction of intrauterine growth restriction with customised estimated fetal weight centiles. BJOG 2003;110:411–5.
23. Kramer MS. Intrauterine growth and gestational duration determinants. Pediatrics 1987;80:502–11.
24. Smith GC, Smith MF, McNay MB, Fleming JE. First-trimester growth and the risk of low birth weight. N Engl J Med 1998;339:1817–22.
25. Bukowski R, Smith GC, Malone FD, Ball RH, Nyberg DA, Comstock CH, et al. Fetal growth in early pregnancy and risk of delivering low birth weight infant: prospective cohort study. BMJ 2007;334:836.
26. Malone FD, Canick JA, Ball RH, Nyberg DA, Comstock CH, Bukowski R, et al. First-trimester or second-trimester screening, or both, for Down’s syndrome. N Engl J Med 2005;353:2001–11.
27. Arntzen A, Nybo Andersen AM. Social determinants for infant mortality in the Nordic countries, 1980-2001. Scand J Public Health 2004;32:381–9.
28. Raatikainen K, Heiskanen N, Heinonen S. Marriage still protects pregnancy. BJOG 2005;112:1411–6.
29. Scholl TO, Johnson WG. Folic acid: influence on the outcome of pregnancy. Am J Clin Nutr 2000;71:1295S–303S.
30. Wilcox MA, Chang AM, Johnson IR. The effects of parity on birthweight using successive pregnancies. Acta Obstet Gynecol Scand 1996;75:459–3.
31. Wilcox AJ. Birth weight and perinatal mortality: the effect of maternal smoking. Am J Epidemiol 1993;137:1098–104.
32. Yip R. Altitude and birth weight. J Pediatr 1987;111:869–76.
33. Kelekci S, Yilmaz B, Savan K, Sonmez S. Can increased nuchal translucency in the first trimester of pregnancy predict gestational diabetes mellitus. J Obstet Gynaecol 2005;25:579–82.
34. Smith GC, Stenhouse EJ, Crossley JA, Aitken DA, Cameron AD, Connor JM. Early-pregnancy origins of low birth weight. Nature 2002;417:916.
35. Dugoff L, Hobbins JC, Malone FD, Porter TF, Luthy D, Comstock CH, et al. First-trimester maternal serum PAPP-A and free-beta subunit human chorionic gonadotropin concentrations and nuchal translucency are associated with obstetric complications: a population-based screening study (the FASTER Trial). Am J Obstet Gynecol 2004;191:1446–51.
36. Yaron Y, Cherry M, Kramer RL, O’Brien JE, Hallak M, Johnson MP, et al. Second-trimester maternal serum marker screening: maternal serum alpha-fetoprotein, beta-human chorionic gonadotropin, estriol, and their various combinations as predictors of pregnancy outcome. Am J Obstet Gynecol 1999;181:968–74.
37. Huang T, Alberman E, Watt HC, Wald NJ. Using Down syndrome serum screening results to predict low birthweight. Prenat Diagn 2003;23:420–6.
38. Dugoff L, Hobbins JC, Malone FD, Vidaver J, Sullivan L, Canick JA, et al. Quad screen as a predictor of adverse pregnancy outcome. Obstet Gynecol 2005;106:260–7.
39. Rantakallio P, Laara E, Koiranen M, Sarpola A. Maternal build and pregnancy outcome. J Clin Epidemiol 1995;48:199–207.
40. Cnattingius S, Bergstrom R, Lipworth L, Kramer MS. Prepregnancy weight and the risk of adverse pregnancy outcomes. N Engl J Med 1998;338:147–52.
41. Hoyert DL, Mathews TJ, Menacker F, Strobino DM, Guyer B. Annual summary of vital statistics: 2004 [published erratum appears in Pediatrics 2006;117:2338]. Pediatrics 2006;117:168–83.
42. Cnattingius S, Haglund B, Kramer MS. Differences in late fetal death rates in association with determinants of small for gestational age fetuses: population based cohort study. BMJ 1998;316:1483–7.
43. Wang X, Zuckerman B, Pearson C, Kaufman G, Chen C, Wang G, et al. Maternal cigarette smoking, metabolic gene polymorphism, and infant birth weight. JAMA 2002;287:195–202.
44. Castles A, Adams EK, Melvin CL, Kelsch C, Boulton ML. Effects of smoking during pregnancy. Five meta-analyses. Am J Prev Med 1999;16:208–15.