Maternal lipid profile in pregnancy and embryonic size: a population-based prospective cohort study

This study was embedded in the Generation R Study, a large multi-ethnic population-based prospective cohort study in the city of Rotterdam, the Netherlands [29, 30]. All methods were carried out in accordance with the Declaration of Helsinki. The study protocol has been approved by the Medical Ethics Committee of the Erasmus University Medical Centre (Erasmus MC), Rotterdam (MEC-2007-413). Written informed consent was obtained from all participants. We excluded women that were not pregnant when included in the Generation R study (n = 925). Next we excluded twin pregnancies, abortions, fetal deaths and women lost to follow-up (n = 343). From the 8633 pregnancies resulting in a live singleton birth, we excluded pregnancies without information on lipid concentrations (n = 2416), pregnancies without information on CRL (n = 4719), pregnancies complicated by (gestational) diabetes [18] and pregnancies in which cholesterol regulating medication was used [6]. The study population comprised 1474 women with a live born singleton and of whom information was available on lipid measurements and ultrasonic assessment of embryonic size (Figs. 1 and 2).

Non-fasting blood was sampled early in pregnancy (median 12 + 3 weeks of gestation, 90% range [10 + 1–13 + 6 weeks]) by trained research nurses. Details of the processing procedures have been described earlier [30]. After thawing, the total cholesterol (mmol/L) and HDL-c (mmol/L) concentrations were determined using standard laboratory methods. Concentrations of LDL-c were calculated using the Friedewald equation [31]. This calculation is not valid when the triglyceride level is ≥5 mmol/L. In this study population, there are no women with triglycerides above 5 mmol/L. Remnant cholesterol was calculated as the total cholesterol minus LDL-c and minus HDL-c ([total cholesterol – LDL-c] – HDL-c). Non-HDL-c was calculated by subtracting HDL-c from total cholesterol (total-cholesterol – HDL-c). The TG/HDL-ratio was calculated by TG divided by the HDL-c concentration (TG/ HDL-c) (Additional Table 1). Both cholesterol, TG and glucose (mmol/l) were measured with the c702 module on a Cobas 8000 analyzer (Roche, Almere, The Netherlands). Results on maternal lipid levels in this cohort have previously been published [18].

Embryonic size was assessed by ultrasound examinations using an Aloka model SSD-1700 (Tokyo, Japan) or the ATL-Philips Model HDI 5000 (Seattle, WA, USA). Ultrasound examinations for this study were performed by dedicated ultrasonographers at each prenatal visit to the designated research centers [29]. The crown-rump length (CRL) was measured in a true mid-sagittal plane with the genital tubercle and the spine longitudinally in view, according to standard procedures [27, 32, 33]. Intra-class correlation coefficients for intra-observer and inter-observer reproducibility of crown to rump length measurements were 0.998 and 0.995 [32]. Gestational age (GA) adjusted standard deviation scores (SDS) were constructed for the CRL measurements. Gestational age of the embryo’s was expressed in weeks of gestational age. These scores were based on reference growth curves from the whole study population and represent the equivalent of Z-scores [34]. We obtained information on birth weight from midwifery and obstetric medical records. Gestational-age-adjusted SDS for birth weight were constructed using North European growth standards as the reference growth curve and represent the equivalent of z-scores [34, 35].

The gestational age is the most important determinant of embryonic and fetal growth. In clinical practice, pregnancy dating is based on the CRL. However, for the purpose of analyses with CRL as the outcome, gestational age should be based on the LMP [36]. In this study, pregnancy dating was thus based on the last known menstrual period in women with a regular menstrual cycle (28 days, range 24–32 days) [27]. The first day of the last menstrual period was derived from the referral letter of the community midwife or hospital [27]. At the ultrasound visit, we checked this date with the mother and obtained additional information on the regularity and duration of the menstrual cycle.

In a consensus meeting (DG, AP, ES, JRvL), we identified confounders for the association between maternal lipid concentrations and embryonic size. This resulted in a Directed Acyclic Graph (DAG) (Supplementary Fig. 1) [37]. The identified confounders were: maternal age (continuous), pre-pregnancy BMI (continuous), parity (nulliparous, multiparous), educational level (no education finished, lower education, middle education, higher education), ethnicity (Dutch and Western, Turkish and Moroccan, African, Asian), smoking (never smoked during pregnancy, smoked until pregnancy was known, continued smoking in pregnancy), folic acid supplement use (started preconceptionally, started in first 10 weeks of pregnancy, no folic acid supplement intake) and glucose concentrations (continuous). Maternal questionnaires at study enrolment provided information on maternal age, pre-pregnancy body mass index, parity, educational level, ethnicity, smoking habits and folic acid supplement use.

We collected information about pre-pregnancy weight by questionnaire, and measured height and weight at enrollment. Questionnaire based weight and measured height were then used to calculate body mass index (BMI)(kg/m²). The correlation of pre-pregnancy weight obtained by questionnaire and weight measured at enrollment was high (β = 0.97, P < 0.01) [38]. Normal weight was defined as a BMI < 25.0 kg/m² and overweight was defined as a BMI ≥ 25.00 kg/m².

First, baseline characteristics and the distribution of the covariates were determined. We examined potential differences in baseline characteristics between women included and excluded from the analysis. Differences in continuous variables with a normal distribution (mean, SD) were analyzed with Students t-test, and variables with a skewed distribution (median, 90% range) with the Mann-Whitney U test. Categorical variables were analyzed with chi-square tests (Additional Table 2).

Second, multivariate linear regression analyses were performed to study the association between differences in embryonic size for the lower and upper tertiles of the maternal lipid concentrations, compared to the middle tertile. We carried out tests for trends based on multiple linear regression models with the maternal lipid concentrations as a continuous variable. To allow mutual comparison of the lipid measures, we constructed Multiple of the Median (MoM) scores of all lipid measures. This MoM score indicates of how far the individual measure deviates from the median in the study population. The crude model was the univariate analysis of maternal lipid concentrations and embryonic size. In the adjusted model, we additionally corrected for the previously determined confounding factors.

Since lipid changes may not only be the consequence of higher glucoses, but may also cause disturbances of glucose metabolism, there is an interactive effect of the maternal lipid status and maternal glucose status [39]. This changes the direction in the DAG graph, indicating that there may be a confounding or mediating effect of maternal glucose level. We examined whether maternal glucose concentrations mediated the association of maternal lipid concentrations with size growth by adding it to our models (fully adjusted model).

We aimed to investigate the effect of the switch in nutritional source of the embryo, from uterine glands and yolk sac to the placenta, which occurs at around week 12 of gestation. Therefore, sensitivity analyses were performed. Associations between the maternal lipid status and embryonic size were separately investigated in the period of 10 to 12 weeks GA versus 12 to 14 weeks GA, with gestational-age adjusted MoM’s (Additional Table 3). With other sensitivity analyses, we tested the effect of the lowest lipid concentrations by assessing the cases with the lowest 5% of the lipid concentrations (Additional Table 4). Results of all linear regression analyses are presented as SDS with a 95% confidence interval (CI).

The following confounders had missing values: pre-pregnancy BMI (14.3%), parity (0.4%), educational level (4.3%), ethnicity (2.1%), smoking (8.8%), folic acid supplement use (19.3%) and glucose (2.7%). To prevent bias associated with missing data, we used multiple imputations for covariates with missing values. We imputed missing data on the basis of the correlation of missing variables with other participant characteristics, according to the Markov Chain Monte Carlo method [40]. Ten datasets were created and analyzed together. A sensitivity analysis was performed to observe differences in observed and expected values of confounders before and after imputation (Additional Table 5). Lastly, complete case analyses were conducted, in which datasets were used without the imputed missing values (data not shown).

We showed that both maternal triglycerides and remnant cholesterol in early pregnancy are positively associated with embryonic size, especially in overweight women and even after adjustment for glucose concentrations [41].

Lipids such as triglycerides and cholesterol reach the developing embryo or fetus through different mechanisms, which change over the course of pregnancy. In the first 12 weeks of pregnancy, the placenta is developing and not fully functional [42]. In this period, the developing embryo is dependent on the yolk sac and uterine glands for the storage and transport of nutrition [43, 44]. The yolk sac transports maternal lipids into the vitelline vessels that are connected with the circulation of the embryo [45]. Animal studies showed that as the maternal serum lipid concentrations increased, so did the concentrations in the yolk sac, and consequently the secretion by the yolk sac into the embryo [46]. This indicates that the lipid transport to the embryo is dependent on maternal serum lipid concentrations. For triglycerides to pass the yolk sac membrane, they have to be hydrolyzed into free fatty acids by placental lipases [47]. From animal studies it is known that during embryonic growth, approximately 90% of the total energy requirement is derived from yolk lipid fatty acid oxidation [48]. This indicates triglycerides have an important role as energy source in the development of an embryo, supporting our positive association between triglycerides and embryonic growth. In the performed sensitivity analyses, we were not able to verify the increased dependency of the embryo on the maternal lipid concentrations in the first 12 weeks of pregnancy (analyses split for gestational age 10–12 weeks and 12–14 weeks). Neither did we find stronger associations when we investigated the group of women with the lowest 5% lipid concentrations compared to the higher lipid levels in relation to embryonic size. This could be explained by the fact that by stratifying, the groups are smaller, which lowers the statistical power to detect statistically significant differences. However, the observed effect estimates in these sensitivity analyses are very similar from the main analyses.

Since triglycerides and remnant cholesterol only make up a small part of the total cholesterol content, it could explain why we did not find a positive association between total cholesterol concentrations with embryonic growth. Moreover, based on previous studies, it is postulated that only very low maternal total cholesterol levels are related to fetal and newborn size [13, 49]. In a study demonstrating a negative association between total cholesterol levels and newborn birth weight [13], the mean level of total cholesterol was 3.6 mmol/L. By contrast, the mean level in this study population was 4.8 mmol/L. This possibly explains the absence of an association in this study population. Additionally, we did not find an association between HDL-c levels and embryonic size. This is in line with earlier studies examining the association between HDL-c levels and postnatal measures of fetal growth (i.e. LGA), which also did not observe a significant association [50, 51]. Lastly, we did not find an association between LDL-c and embryonic size. Previous epidemiological studies observed reduced serum LDL-c levels in pregnancies complicated by fetal growth restriction [52]. A suggested underlying mechanism for the association between LDL-c levels and impaired fetal growth is an increased lipid oxidation of LDL-c. These modified LDLs are not recognized nor taken up by the LDL receptor, after which the modified LDL accumulates outside the receptor and initiates plaque formation in the maternal spiral arteries of the placenta. This contributes to artery occlusion, a disturbed perfusion of the placenta and has fetal growth restriction as a result [53, 54]. Because the spiral arteries and placenta are not formed yet in the first trimester of pregnancy, we propose (oxidized) LDL-c does not have the same negative effect on embryonic size.

Strikingly, the demonstrated associations between maternal serum lipid concentrations and embryonic size were most prominent in overweight women. Adiposity is associated with metabolic and endocrinologic changes, and they differ depending on different BMI [55]. The findings of a previous study investigating maternal lipid levels and fetal birthweight, suggest that the metabolic lipid pathways affecting fetal growth may be substantially different in overweight women, compared to the normal weight women [56]. We also find associations with embryonic size in this group of overweight women to be more profound, which may be explained by the possible positive association between maternal BMI and embryonic size [57]. Second, this could be explained by the strong association between both obesity and insulin resistance, and insulin resistance and remnant cholesterol [58‐60]. Therefore, the effect of higher lipid levels may be stronger if the woman is insulin resistant. However, since the gold standard for the assessment of insulin resistance is the hyperinsulinemic-euglycemic clamp, and this was not utilized, we were not able to verify this in this study, [61]. Though, the analyses were additionally adjusted for maternal glucose levels, which is an indirect measure of insulin resistance. Apart from small changes in effect estimates, we could not substantiate an insulin dependent effect in overweight women. However, again, we did not make use of the gold standard for insulin resistance. The yolk sac is important for nutrient transport to the embryo. Hypothetically, hyperglycemia in early pregnancy injures the development of the yolk sac, which impairs embryonic growth and development. Next, there appears to be a decrease in glucose transport in embryos exposed to a hyperglycemic environment, which may lead to programmed cell death, and therefore impaired growth and development [62]. However, the analyses in which we additionally adjust for maternal glucose concentrations did not reveal a glucose dependent effect on embryonic size. This is in line with both in vivo and in vitro studies, that could not verify this negative association of maternal glucose levels on early embryonic growth [63, 64]. In this study population, there was a relatively small number of women with high levels of glucose or hyperglycemia, which might have decreased the possible interactive effect on the association between the maternal lipids and embryonic size. Therefore, the observed associations may be stronger among higher-risk populations.

Our findings demonstrate that the previously established associations between maternal lipids, fetal growth and adverse birth outcomes may already be present during the first trimester of pregnancy [17, 18, 28, 49, 65, 66]. [20].

It is also in line with the Developmental Origins of Health and Disease (DOHaD) theory, which states that adverse influences in early pregnancy have the potential to affect the change of adverse birth outcomes [24]. The finding that specifically both triglycerides and remnant cholesterol are associated with embryonic size are not surprising, as plasma triglycerides and remnant cholesterol are highly correlated [67]. Remnant cholesterol is the cholesterol content of triglyceride-rich lipoproteins. In a clinical setting, triglycerides are even proposed as a surrogate marker of remnant cholesterol [68, 69]. Additionally, our results emphasize the potential of triglycerides and remnant cholesterol as markers for first trimester size.

To unravel the mechanisms of nutrient transport from mother to embryo, and especially lipid transport, more fundamental research is needed. Also changes in nutrient transport due to the switch from yolk sac and uterine glands to the placenta as main nutrient transporter is interesting. Second, due to the small measures of the CRL, the measurement ranges are also small. Lastly, research with repeated CRL measurements would make it possible to investigate embryonic growth patterns.

To our knowledge, this is the first study which investigates the association between the maternal lipid profile and embryonic size in early pregnancy. One limitation is that maternal blood samples were obtained in a non-fasting state and not on a specific time of the day, while the levels are sensitive towards intake. This could have led to an underestimation of the observed associations, due to non-differential misclassification of high or low lipid levels. Moreover, multiple studies have demonstrated that plasma lipids only change a little in response to food intake [70‐77]. Therefore, non-fasting lipids levels can used to evaluate the serum lipid status of pregnant women instead of fasting lipids. As an exception, only fasting blood samples should be considered if non-fasting plasma triglycerides are above 5 mmol/L. [75] However, in our study population, there were no women with non-fasting triglycerides that exceeded 5 mmol/L.

A second limitation, is that embryonic size was measured only once. Therefore, no patterns of embryonic growth could be assessed. Also no information on (changes in) pre-pregnancy lipid concentrations was available. We therefore cannot investigate the effect of preconceptional lipid concentrations on embryonic size. Next, the use of MoM’s in the analyses makes it harder to clinically interpret the associations. However, these MoM’s enable to compare the different lipid concentrations to each other. The effect sizes for the association between triglycerides and remnant cholesterol and embryonic size are comparable.

Moreover, there might be the issue of response bias or self-selection, which is known to happen in cohort studies. Indeed, the median BMI of 22.6 within our study population is within the healthy range and the majority of women did not smoke during pregnancy (73.9%) (Additional Table 1). Indeed, most of the measured maternal lipid concentrations are within the recommended ranges for the first trimester of pregnancy [78]. The selection of a relatively healthy study population did thus not allow to investigate the associations of extreme dyslipidemia. This might imply that effects in the general population with more and severe dyslipidemia may be even larger, and thus has affected the generalizability of our results. Finally, the observational nature of this study does not allow for inference of causality.