Abstract
OBJECTIVE. The objective of this study was to assess whether low maternal serum cholesterol during pregnancy is associated with preterm delivery, impaired fetal growth, or congenital anomalies in women without identified major risk factors for adverse pregnancy outcome.
METHODS. Mother-infant pairs were retrospectively ascertained from among a cohort of 9938 women who were referred to South Carolina prenatal clinics for routine second-trimester serum screening. Banked sera were assayed for total cholesterol; <10th percentile of assayed values (159 mg/dL at mean gestational age of 17.6 weeks) defined a “low total cholesterol” prenatal risk category. Eligible women were aged 21 to 34 years and nonsmoking and did not have diabetes; neonates were liveborn after singleton gestations. Total cholesterol values of eligible mothers were adjusted for gestational age at screening before risk group assignment. The study population included 118 women with low total cholesterol and 940 women with higher total cholesterol. Primary analyses used multivariate regression models to compare rates of preterm delivery, fetal growth parameters, and congenital anomalies between women with low total cholesterol and control subjects with mid–total cholesterol values >10th percentile but <90th percentile.
RESULTS. Prevalence of preterm delivery among mothers with low total cholesterol was 12.7%, compared with 5.0% among control subjects with mid–total cholesterol. The association of low maternal serum cholesterol with preterm birth was observed only among white mothers. Term infants of mothers with low total cholesterol weighed on average 150 g less than those who were born to control mothers. A trend of increased microcephaly risk among neonates of mothers with low total cholesterol was found. Low maternal serum cholesterol was unassociated with risk for congenital anomalies.
CONCLUSIONS. Total serum cholesterol <10th population percentile was strongly associated with preterm delivery among otherwise low-risk white mothers in this pilot study population. Term infants of mothers with low total cholesterol weighed less than control infants among both racial groups.
- cholesterol
- preterm birth
- birth defects
- congenital defects
- birth weight
- embryonic and fetal development
- lipoproteins
Maternal cholesterol is essential for both the hormonal and physical changes of early pregnancy.1–8 Circulating low-density lipoprotein cholesterol (LDL-C) is the chief substrate for placental progesterone biosynthesis.9,10 Subclasses of high-density lipoprotein cholesterol also participate in placental cholesterol balance.11 Cholesterol in plasma membranes is a bulk constituent of decidual tissue critical to implantation and uteroplacental vascularization.12–15 Alterations in placental cholesterol concentrations have been associated with changes in placental transport functions during gestation.16 Gestational disorders such as preeclampsia and gestational diabetes, as well as maternal alcohol consumption, have been associated with lipoprotein changes.11,17–21
Maternal cholesterol is probably transported to the early embryo and would be available for development of both embryonic and placental tissues.22–28 Longitudinal studies have documented specific changes in total and fractionated cholesterol levels in maternal serum as gestation progresses; total cholesterol increases substantially during the second and third trimesters.29–35 This physiologic hypercholesterolemia of later pregnancy suggests an adaptive function for pregnancy maintenance or fetal growth.36 Decreased levels of maternal total cholesterol (TC) and LDL-C have been reported in association with intrauterine growth restriction (IUGR).37,38 Conversely, maternal hypercholesterolemia is suspected to be injurious, because concentrations >300 mg/dL have been linked to increased cholesterol deposition in the fetal aorta39; the “fetal origins hypothesis” links this phenomenon to subsequently increased risk for cardiovascular disease in the adult offspring.40–42 It is not known whether optimal levels of maternal serum cholesterol during pregnancy can be defined. We are not aware of data addressing any potential risk to the pregnancy when maternal serum cholesterol falls below a lower bound. This pilot study investigated the premise that low maternal serum cholesterol (LMSC) during gestation may be associated with adverse pregnancy outcomes. In view of the evidence for maternal transport of cholesterol to the embryo in early pregnancy, we reasoned that maternal cholesterol may have a disproportionate impact during critical periods for placentation and early neuroepithelial expansion.5,6,43–48 Specifically, we hypothesized a priori that LMSC may increase the risk for preterm delivery, microcephaly, or other central nervous system anomalies.7,49–53 Secondarily, we hypothesized an increased risk for fetal growth restriction and congenital anomalies.7,50,54
Our study population was retrospectively sampled from a population-based cohort of 9938 women whose blood had been banked concurrently with routine serum screening performed at prenatal clinics that serve western South Carolina. This screening occurred over a broad spectrum of gestational ages (GAs) throughout the second trimester (range: 13–23 weeks). We derived the population distribution of maternal TC from all banked sera and defined a low-TC risk exposure as TC <10th percentile of this screened population (159 mg/dL at a mean GA 17.6 weeks, or 4.11 mmol/L at that date). For facilitation of comparisons among cholesterol values sampled at different GAs, all TC values in the study population were adjusted for GA at date of screening using linear regression methods. Women with higher levels of TC formed the pool of potential control subjects; their adjusted TC values were separated into mid-TC and high-TC categories encompassing the 10th to 90th and >90th population percentiles, respectively (90th population percentile at 17.6 weeks' GA was 261 mg/dL, or 6.75 mmol/L). Eligibility criteria were designed to minimize identifiable pregnancy risk factors or potential confounders; among other characteristics, participants were aged 21 to 34; had no history of smoking, type 1 diabetes, or other medical or gestational risk conditions; and had ultrasound-dated singleton pregnancies that resulted in a live birth. The study hypotheses were tested by assessing risk ratios for specific adverse outcomes in infants of mothers with low TC compared with mid-TC control mothers, using multivariate regression models.
METHODS
Study Population Ascertainment
The Greenwood Genetic Center (GGC) in Greenwood, South Carolina, has an established protocol for collecting baseline maternal data and banking serum samples from all consenting women who are referred for routine second-trimester serum screening to clinics that serve urban and rural areas of western South Carolina. Institutional review boards of participating institutions approved this study protocol. Serum samples and clinical data were identified to National Institutes of Health (NIH) investigators only by a study number assigned at GGC. Banked sera from all women who were consecutively screened between 13 and 23 weeks' gestation (mean: 17.6 weeks; SD: 1.5 weeks) during 1996–2001 were assayed for TC (n = 9938). Although women in this cohort delivered primarily at 2 hospitals, for this pilot investigation, ascertainment of hospital charts was limited to women who delivered at the hospital closest to GGC (composing ∼40% of the total screened cohort). We defined LMSC pregnancies as those in which maternal TC fell below the 10th percentile of the screened cohort (<159 mg/dL, after adjusting the assayed TC value for the recorded GA at which it was obtained). Hospital charts were ultimately reviewed for approximately half of the low-TC group and one third of the much larger control pool. GGC staff remained blinded to maternal cholesterol status. Study exclusion criteria (Table 1) eliminated approximately two thirds of the ascertained cohort subjects in both low-TC and control groups. The majority of exclusions in each group were because of age <21 years, positive smoking history, and/or pregnancy dating by last menstrual period rather than by ultrasound. The final cohort available for analysis included 118 mothers with low-TC and 940 comparison mothers classified as having mid-TC or high-TC.
Ascertainment of Low-TC Risk Group and Higher-TC Control Group
Serum Cholesterol Assays
Frozen sera (−80°C) were shipped on dry ice from GGC to the NIH. TC in serum was analyzed enzymatically by the cholesterol oxidase method on a Hitachi 917 analyzer using reagents from the manufacturer (Roche, Indianapolis, IN). Both the interassay and intra-assay coefficients of variation were <5%.
Determination of Eligibility
Baseline data recorded by participants and clinic staff at the time of serum screening were sent to the NIH in anonymized files identified by study codes linked to the serum samples. Eligibility criteria were designed to increase uniformity among the study population and to exclude those who were at elevated risk for adverse pregnancy outcome: we selected participants who were aged 21 to 34 years at screening and had a singleton gestation, race/ethnicity identified as black or white non-Hispanic only, no history of type 1 diabetes, no current or previous smoking, no other described substance use, and no reports of previous abnormal pregnancy history. Medical charts for the index pregnancy and its outcome were sought for potentially eligible women. Mothers who were evaluated for >1 eligible pregnancy during the screening period had data included for 1 randomly selected pregnancy only (n = 112).
As clinical data were reviewed at the NIH, additional patients were excluded when charts indicated significant intercurrent infections or other illness, preeclampsia or other gestational disorders except idiopathic IUGR, substance use, or a genetic syndrome diagnosed in the neonate. We also excluded patients with conflicting data regarding GA at delivery or other key outcome variables. Posthoc, we excluded neonates with a structural cardiac defect, because such infants had extremely high rates of IUGR and preterm delivery.
Predictor, Covariate, and Outcome Variables
In addition to analyzing primary data, we generated specific predictor, covariate, and outcome variables to model our results. Because our chief hypothesis concerned LMSC as a predictor of adverse pregnancy outcome, we sought to isolate intrinsic differences in maternal TC levels from the introduced component of TC attributable to GA when samples were drawn. It is well established that, in general, maternal TC rises progressively with GA from the second trimester forward in pregnancy, with various percentage increases of LDL-C and TC described during the second and third trimesters by several authors.33,34,55,56 For investigation of whether the increase of TC with GA in our cohort could be expressed as a percentage rise in TC per unit of increase in GA, a model would be required to use logarithms of TC rather than arithmetic values themselves in a linear regression equation. Such log transformations are commonly applicable to biological variables and have been used in other applications of human TC data, generally to correct for the nonnormal distribution of TC.57–60 The model that best fit our study data is as follows: logTC(adjusted) = logTC(raw) + slope(17.6 − GA). This equation yields a downward adjustment of raw TC values where GA > 17.6 and an upward adjustment where GA < 17.6. The slope values for GA (0.026 for white women, 0.032 for black women) were significant at P < .0001 in the linear regression model. All TC values that were adjusted via this model represent the expected value of an individual's TC if it had been sampled at the mean GA of 17.6 (Table 2 gives distributions of raw and adjusted TC). Adjusted TC was examined as a continuous and an ordinal variable; however, on the basis of our a priori definition of a low-cholesterol exposure, most of our analyses used TC <10th percentile as the risk variable. The outcome of preterm delivery was defined as delivery before 37 completed weeks of gestation and term delivery as 37 to 41 completed weeks. Inspection of the relationship between preterm birth and maternal cholesterol on an ordinal scale (as in Fig 1) revealed a nonlinear relationship with increased prevalence at both tails of the cholesterol distribution. Because previous reports suggested a possibly increased risk for prematurity associated with very high maternal cholesterol,61 we dichotomized cholesterol risk as <10th percentile versus 10th to 90th percentiles (159–261 mg/dL). We further defined a variable dichotomizing cholesterol values >90th percentile (261 mg/dL) versus values in the 10th to 90th percentile range to characterize this posthoc association of elevated serum cholesterol with preterm delivery. The low-TC risk group for prematurity was subdivided into progressively lower levels of TC (less than fifth percentile, less than third percentile).
Prevalence of preterm birth within percentiles of maternal cholesterol level, according to route of delivery and maternal race.
Distribution of Maternal Serum Cholesterol Values According to Percentile Ranges, as Derived From Screened Pregnant Population and Applied to Study Population
We tested potential covariates on the basis of the literature of prematurity and availability of data. Season and year of delivery were evaluated in initial models. Maternal weight at screening was used as a surrogate adjustment for effects that are known to be associated with extremes of BMI, prepregnancy weight, and/or weight gain in pregnancy.62,63 Maternal weight was not correlated with GA at screening (P = .56), so the measured value at screening was used. Maternal age at screening was used in adjusted models as a categorical variable (ages 21–24, 25–29, and 30–34). Gender of infant and maternal race were covariates in many models.
We dichotomized outcome variables reflecting fetal growth. Infants who were born at each GA were classified as less than or more than the 5th or 10th percentile for weight, length, and head circumference.64–66 We then defined IUGR as weight and length both <10th percentile for GA, regardless of head circumference. Infants were classified as having microcephaly when the head circumference was >2 SD below the mean for GA (less than fifth percentile), whereas the length was >10th percentile.
Analysis
We performed 3 levels of analysis. First, we directly tested our a priori hypotheses, with adjustment for anticipated covariates. Results are reported from models that fit the data best. Next, we performed supplemental testing of alternative risk, covariate, and outcome variables to evaluate robustness of the original findings against potential sources of classification error and particularities of model selection. Finally, we conducted posthoc analyses to characterize patterns that were evident in the data but had not been anticipated. All analyses were performed using procedures from SAS 8.2 (SAS Institute, Inc, Cary, NC).67 Statistical significance was assessed by nominal P value for a priori hypotheses. Supplemental testing was not designed to test alternative hypotheses, so these confidence intervals and P values were interpreted as descriptive measures characterizing patterns in our data and robustness of primary hypotheses.
Relationship of maternal cholesterol to preterm delivery was assessed by logistic regression models that estimated the odds ratios (ORs), 95% confidence intervals (CIs), and P values associated with preterm birth among infants of low-TC mothers compared with those with maternal cholesterol values in the mid-TC 10th to 90th percentile range. Models were adjusted for maternal race, age category, weight in pounds, infant gender, and presence of IUGR. Secondary analyses tested robustness of OR estimates by modifying definitions of key variables. These measures included using as controls only pregnancies with maternal TC between 25th and 75th percentiles (179–231 mg/dL), examining risks for delivery at ≤35 weeks compared with the more homogeneous term weeks 39 to 41, replicating key analyses with unadjusted cholesterol values, and incorporating different covariates. We further characterized maternal subgroups with TC less than fifth and less than third population percentiles (147 and 138 mg/dL, respectively). Postterm deliveries were excluded from analyses.
Direct measurements of fetal growth at term were tabulated, comparing both low-TC and high-TC groups with the mid-TC reference group. Both univariate and multivariate adjusted models were used to compare size of term infants, excluding those who were classified as growth-restricted or macrosomic (weight <10th or >90th percentile, examined separately). Multivariate logistic regression models were used to estimate ORs associated with microcephaly, IUGR, or size <10th percentile for GA.
The rate of congenital anomalies among cohort infants was compared between low-TC and control groups by univariate χ2 and multivariate regression analysis. This was also done for category of anomaly by organ system.
RESULTS
The 1058 mother-infant pairs that were retained for data analysis represent just >10% each of all LMSC and higher cholesterol pregnancies that were screened for the original cohort (Table 1). Mothers with LMSC were more often excluded for age <21 years, were more likely to smoke, and were less likely to have had sonographic gestational dating. Mothers with LMSC in the study group were somewhat younger, heavier, and more likely to be of black race than were control mothers (Table 3).
Maternal Baseline Characteristics According to Serum Cholesterol Level and Race
The prevalence of preterm delivery in this highly selected study cohort (6.6%) was considerably less than that observed in the background South Carolina singleton population (∼12% in the sampled counties).68 All analyses that tested the association of LMSC with preterm delivery showed statistically significant elevation of risk in analyses that compared mothers with low-TC with mothers with mid-TC (Table 4). Subgroup analyses indicated that this association was present exclusively among white mothers; black mother-infant pairs showed no association between LMSC and preterm delivery in any model (Fig 1, Table 4). Black women showed an excess of preterm delivery exclusively among the high-TC group. This substantial qualitative difference between the races in patterns of prematurity led us to perform additional analyses of this and other birth outcomes separately by race, as posthoc findings. The unadjusted OR among white mothers with TC <10th population percentile was 5.22 (95% CI: 2.52–10.8; P < .0001), which increased slightly to 5.63 after adjustment for infant IUGR and gender, maternal age subgroup, and maternal weight >90th percentile (Table 4). The same regression model when used to compare the risk for birth at ≤35 gestational weeks against birth at weeks 39 to 41 gave similar results (adjusted OR: 6.77; 95% CI: 2.93–15.6; P < .0001). All secondary analyses that tested robustness of preterm findings among white mother-infant pairs gave results comparable to the primary analyses. In addition, analyses consistently indicated that maternal cholesterol values less than the third percentile had higher associated risk ratios than the corresponding 5th or 10th percentile thresholds for LMSC (Table 4).
Prevalence of Preterm Birth and Adjusted Associations of Preterm Birth With Maternal Serum Cholesterol Level According to Maternal Race
Infants who were born at 37 to 41 weeks to mothers with low TC weighed on average 124 g less than those who were born to control mothers (P = .021, 1-way analysis of variance; Table 5). This difference increased to 147 g with adjustment for specific GA within the term interval, race, infant gender, maternal age group, and maternal weight (P = .0006, linear regression, excluding outliers to the normal birth weight distribution). The distributions of infant length and head circumference did not differ in any comparisons of LMSC and control groups. Macrosomia (weight >90th US population percentile for GA at term66) was not associated with maternal TC level (P = .83, χ2 test for trend). Among preterm infants, GA was the only significant predictor of size.
Comparative Measurements of Term Infants by Maternal Serum Cholesterol Level and Race
Infants in the left tail of the size distribution, measuring <10th percentile for their GA in weight, length, or head circumference or meeting the composite definition of IUGR were no more likely to be born to mothers with LMSC than to control mothers (Table 6) . This was true whether the thresholds for small size were derived from 1 of several published distributions64,66,69 or classified by internal cohort percentiles. However, maternal TC less than the third percentile was consistently associated with low birth weight. Low maternal weight was the most significant predictor of low birth weight for GA. Microcephaly was approximately twice as prevalent among infants of mothers with LMSC as among infants of control mothers with mid-TC; significance testing indicated a trend in adjusted logistic regression (Table 6).
Adjusted Associations of Small Birth Size for GA With Maternal Serum Cholesterol Level According to Race
Low-TC pregnancies in our study were not more likely than control pregnancies to produce an infant with a congenital anomaly. Infants with anomalies did have markedly increased frequency of either very low or very high maternal TC or maternal weight, as well as all combinations of these extremes (7 of 8 infants with cardiac anomalies and 11 of 15 with other major anomalies were born to mothers in the <5% or >90% tails of maternal cholesterol and/or maternal weight distribution [P < .0001 in posthoc χ2 analysis]; 7 of these infants also weighed >90th percentile for GA). There was no trend toward preterm delivery or IUGR as maternal weight increased in this study group; however, neonatal weight >4000 g was strongly associated with increasing maternal weight (P < .0001, χ2 test for trend).
DISCUSSION
Our data indicate a substantially elevated risk for preterm delivery among white women whose TC was among the lowest 10% of the background population as sampled at 17 to 18 weeks' GA. This finding was statistically significant in the initial pooled analysis as well and so does not represent exclusively a posthoc subgroup result. Factors that support the validity of this finding include the exceptionally high point estimate for risk for prematurity among mothers with very low maternal cholesterol (138 mg/dL, the third population percentile of the original screened cohort) and the consistent finding that LMSC was the strongest predictor of preterm delivery in every model tested. We considered possible sources of both systematic and random error: confirmatory analyses were conducted to address possible classification error in the assignment of GA, to test the robustness of low-TC and higher-TC definitions, and to examine multivariate models for sensitivity to covariate selection and other factors.70–72 All such tests of robustness yielded consistent, statistically significant risk estimates for prematurity among white mothers with LMSC. Whether this finding can be replicated or extended to other populations of pregnant women will be essential to examine.
Our a priori hypothesis that risk for microcephaly would be increased among infants who are born to mothers with LMSC was not confirmed statistically; however, the presence of a statistical trend and an estimated twofold increase in risk is provocative and likewise deserves additional scrutiny. These data did not support the hypothesized associations of LMSC with size <10th percentile for GA, with the exception of an excess of infants with low weight for GA among mothers with TC in the lowest subgroup (less than third percentile). There was also a shift toward lower birth weights still within the reference range among term infants who were born to mothers with LMSC. As a posthoc finding, this must be regarded with caution; the additional relative decrease in birth weight among term infants of mothers with LMSC after adjustment for multiple factors that are known to affect birth weight does support its potential validity. The hypothesized association of LMSC with congenital anomalies was not supported.
The observed association between LMSC and preterm delivery requires validation. If real, then such an association could be mediated directly by restricting the availability of substrate for hormonal and nutritional support of early pregnancy, including an effect on placentation.12,18 LMSC could also indicate altered lipoprotein fractions or apolipoprotein profiles, which have been associated with altered pregnancy outcome in various scenarios implicating both maternal and fetal genotypes.3,23,24,45,73–76 It would be of great interest to know how commonly TC in the moderately low 5% to 10% range as measured in midpregnancy correlates with a substantially lower value in early pregnancy, which could have affected first-trimester events, or how commonly LMSC represents a relatively normal periconceptional cholesterol level that failed to rise as expected in the second trimester, affecting later events, such as week of delivery. As pregnancy progresses, a reduction in physiologic gestational hypercholesterolemia could affect aspects of lipid homeostasis in pregnancy (eg, fetal erythrocyte membranes normally have altered cholesterol concentrations in later pregnancy, with concomitant alterations in the function of membrane-based ion channels36). The complex pathways that are associated with onset of labor may also potentially be affected by altered steroid balance.49
Although low serum cholesterol level is known to be correlated with poor nutritional status, there was no correlation between maternal TC and maternal weight in this cohort. However, micronutrient deficiencies may be more common among the low-TC risk group studied here and could account for the observed adverse outcomes. Many such nutritional deficiencies have been studied as predictors of preterm delivery or low birth weight.77–80 Why any of these potential mechanisms might manifest among white but not black mothers would be critical to address should these findings be replicated.81,82 By contrast, the extremely high risk ratios observed among mothers with TC below the third population percentile suggests a severe and persistent dyslipidemia, which might exert complex effects throughout pregnancy.4,83 Our data also indicate a significantly increased risk for preterm delivery among mothers with TC > 90th population percentile (261 mg/dL at 17.6 weeks' GA). Therefore, the concept of an optimal range for maternal serum cholesterol during pregnancy may have merit.
Limitations are inherent in our methods and findings. The specific numeric thresholds for low and high TC are based on a mean sampling GA of 17.6 weeks and cannot be generalized to other populations with differing profiles of gestational risk. The ascertainment of potential study subjects was incomplete, and the LMSC group differed in baseline characteristics from the control group: the specific effects of any resulting selection biases cannot be predicted. We did not have access to some sociodemographic variables that are known to correlate with birth outcome, and we collected a subset of relevant clinical data commensurate with the goal of determining whether comprehensive prospective studies are indicated. This pilot study had several a priori hypotheses, reducing somewhat the statistical impact of the observed associations.
Our strict inclusion criteria preclude extrapolation of these results to the general population, yet this design strategy has allowed us to discern an important preliminary finding. The strength and consistency of the elevated risk for preterm delivery among mothers with LMSC in this study should spur the prospective, longitudinal characterization of maternal cholesterol and lipoprotein profiles in subsequent investigations of preterm birth.
Acknowledgments
This research was supported by the Division of Intramural Research, National Human Genome Research Institute, NIH, Department of Health and Human Services.
We thank our GGC colleagues Ericka Strickland, RN, for extraction of clinical data from hospital records, Kim Stewart for preparation of coded serum samples, and Karen Buchanan for compilation of anonymized data sets. We thank our NIH colleagues Robert T. Long for maintenance of serum samples and testing logs, Maureen Sampson for performance of all TC assays, and Julia Fekecs for drafting the figure and tables.
Footnotes
- Accepted March 14, 2007.
- Address correspondence to Maximilian Muenke, MD, Medical Genetics Branch, National Human Genome Research Institute, National Institutes of Health, Department of Health and Human Services, 35 Convent Dr, MSC 3717, Building 35, Room 1B-203, Bethesda, MD 20892-3717. E-mail: mmuenke{at}nhgri.nih.gov
The authors have indicated they have no financial relationships relevant to this article to disclose.
REFERENCES
- Copyright © 2007 by the American Academy of Pediatrics
In this issue
Jump to section
Related Articles
Cited By...
- The Echo of Pulmonary Tuberculosis: Mechanisms of Clinical Symptoms and Other Disease-Induced Systemic Complications
- Window of Opportunity: Targeting ANGPTL4 Improves Triglyceride Levels in Maternal Obesity During Pregnancy
- Preterm Delivery and Future Risk of Maternal Cardiovascular Disease: A Systematic Review and Meta-Analysis
- Normalizing Metabolism in Diabetic Pregnancy: Is It Time to Target Lipids?
- Pregnancy Outcomes in Familial Hypercholesterolemia: A Registry-Based Study
- Transport of maternal cholesterol to the fetus is affected by maternal plasma cholesterol concentrations in the Golden Syrian hamster
- The Changing Face of Preterm Birth