This is a nested cased-control study based on the EDEN mother-child cohort (25) that was set up in 2003 in two university maternity clinics, Nancy and Poitiers, France. Participation in the EDEN mother-child cohort was proposed to all women visiting the prenatal clinic before 24 WG. Exclusion criteria were multiple pregnancies, known diabetes before pregnancy, the inability to read and understand French, or planning to move out of the region within the next three years. Among the 3,758 women invited to participate in the EDEN mother child-cohort, 2,002 (53%) were enrolled in the study (1,034 women from Nancy and 968 from Poitiers).

Data were collected from medical and obstetrical records and clinical examinations. Mothers underwent a clinical examination with blood sampling at 24 to 28 WG. At birth, the weight of the newborns was measured by midwives using an electronic scale (Seca Ltd). Data on maternal smoking during pregnancy and maternal and newborn characteristics were collected during pregnancy or at birth. Maternal weight before pregnancy was obtained by interview at inclusion and maternal height measured during the clinical examination. Both anthropometric measurements were used to calculate the BMI as the reported weight (kilograms) divided by the square of the measured height (meters). Information on ultrasound measurements and placental function measured at 26 WG were collected from obstetrical records: fetal head circumference, fetal abdominal circumference, femur length, amniotic fluid index, notch in Doppler flow velocimetry of the uterine artery, pulsatility index, and reverse flow on umbilical doppler. Information on the presence of a maternal vascular disease before pregnancy (chronic high blood pressure, sickle cell anemia, antiphospholipid or anticardiolipin antibody syndrome, or autoimmune disease) was also collected from obstetrical records.

We conducted a case-control study nested within the EDEN mother-child cohort. Cases were defined as mothers who gave birth to newborns with a birth weight < 10^th percentile according to customized fetal growth charts (26). For each case, two matched controls were recruited among mothers who gave birth to AGA (adapted for gestational age) newborns with a birth weight between the 25^th and 75^th percentile of the same customized references (26). Controls were matched for the maternity clinic (Nancy – Poitiers), sex of the child, maternal age, maternal body mass index (BMI) (4 groups: < 18.5 kg/m², 18.5-24.9 kg/m², 25-30 kg/m², > 30 kg/m²), and gestational age (GA) at blood collection. In total, 196 SGA mother-child pairs and 370 match control pairs were identified in the EDEN mother-child cohort.

SGA children were considered to be SGA with placental vascular dysfunction when they met at least one of the following criteria (1, 27): fetal head circumference > 10^th percentile and fetal abdominal circumference < 10^th percentile, presence of a notch in Doppler-flow velocimetry of the UtA, high-resistance UtA flow, defined as a mean pulsatility index > 95^th percentile, reverse flow on umbilical doppler, high-resistance UA flow, defined as a pulsatility index > 95^th percentile, or the presence of a maternal vascular disease before pregnancy (chronic high blood pressure, sickle cell anemia, antiphospholipid syndrome).

The EDEN mother-child cohort received approval from the ethics committee of the Kremlin Bicêtre Hospital (France) on 12 December 2002 and from the French data privacy institution CNIL. All participants were treated in accordance with French national ethics regulations.

Among the 193 identified cases of SGA, the DLK1 value was aberrant in the maternal serum of one SGA case (870.7 ng/mL vs a mean value of, for the other 562 samples, 29.5 ± 10.1ng/mL). After exclusion of this subject and her matched controls, 192 SGA and 368 matched controls were included in the analyses. Sixty-three cases of SGA (32.6%) were classified in the subgroup of SGA with placental vascular dysfunction. The distribution of DLK1 values is presented in Figure 1 .

Characteristics of the two groups are presented in Table 1 . The median gestational age at blood sampling was 26 WG. As expected, there were no differences between cases and controls concerning the matching factors. The proportion of women who declared smoking during pregnancy was no different between cases and controls.

There were also no differences in DLK1 levels according to the center of inclusion or the sex of the neonates ( Table 2 ). DLK1 levels in maternal blood serum at 26 WG were significantly lower for women who gave birth to SGA infants (27.7 ± 8.7 ng/mL) than for control women (30.4 ± 10.6 ng/mL, p = 0.001) ( Table 2 ).

Consideration of DLK1 as continuous variable in the multiple regression analysis showed a significant association of DLK1 levels with SGA: a 1ng/mL increase in DLK1 decreased the risk of SGA by 3%. Consideration of DLK1 as a categorical variable divided into quartiles showed an odds ratio for SGA of 1.98 [1.15 - 3.37] in the first quartile compared to the last ( Table 3 ).

Maternal circulating DLK1 levels were not significantly different in cases of SGA with signs of placental vascular dysfunction (n = 63, 27.1 ± 9.2 ng/mL) than in those without placental dysfunction (n = 129, 28.0 ± 8.5 ng/mL, p = 0.53). Indeed, they both showed significantly lower levels of maternal circulating DLK1 than controls (n = 368) (30.4 ± 10.6 ng/mL, p = 0.009) ( Table 2 ).

The ROC curve showed DLK1 levels to be less predictive of SGA (Area under the receiver operating characteristic curve – AUROC = 0.578) than the ultrasound alone model (AUROC = 0.658) ( Figure 2 ). Combining DLK1 levels with the prenatal ultrasound detection of SGA significantly increased the prediction of SGA than DLK1 alone (AUROC for ultrasound associated with maternal DLK1 levels = 0.675, p = 0.0002). However, DLK1 levels do not add much to the ultrasound model alone.

Our results show an association between DLK1 levels in pregnancy during the 2^nd trimester of gestation and SGA, which was first described by Cleaton et al. for DLK1 levels measured during the 3^rd trimester of gestation (20). They reported lower levels of DLK1 at 36 WG for women who gave birth to SGA children in a study that included 45 women who delivered a SGA newborn and 43 matched controls. Recently, McDonald et al. confirmed lower DLK1 levels at 36 WG for women with SGA children in a cohort of 105 SGA and 245 controls. They also reported that DLK1 levels were already significantly reduced at 28 WG in women who delivered a SGA infant in a sub-group of 84 SGA cases and 132 controls (24). With a larger cohort and earlier in pregnancy than Cleaton and McDonald, our study confirms the association of DLK1 levels with SGA. The demonstration of such an association at an early stage in pregnancy should be beneficial for the management of newborns.

However, it is not well established whether the reduction of circulating maternal DLK1 levels during pregnancy causes fetal growth restriction or whether the low weight of the fetus and the placenta diminishes maternal DLK1 levels. An important clue to address this question would be provided by determining the source of maternal DLK1 during pregnancy. Circulating DLK1 levels are very low in non-pregnant women and substantially increase during pregnancy (20, 24, 28). Maternal DLK1 levels have been reported to be significantly higher in twin pregnancies than normal singleton pregnancies after 17 WG (29). Overall, these data suggest that maternal DLK1 measured during the 2nd and 3rd trimester originate from the fetus or the placenta. Cleaton et al. hypothesized that the source of maternal circulating DLK1 is the fetus, and not the placenta, based on data from murine models with conditional deletion of Dlk1 (20). However, more recent data from humans have shown that DLK1 is present in the extra-embryonic tissue as early as 5 WG and remains detectable until birth (18, 30). Other studies have reported that placental expression of DLK1 is associated with fetal growth. One study reported a two-fold increase in the placental expression of DLK1 in large for gestational age newborns (birth weight > 90th percentile) (31) and another reported significantly lower DLK1 expression in the placentas of mothers who gave birth to SGA than those who gave birth to AGA infants (32). MacDonald et al. showed a positive correlation between maternal circulating DLK1 levels and placental weight and that DLK1 is secreted by the human placenta in in-vitro studies (24). Overall, these data at least partially support a placental origin of maternal circulating DLK1 in humans.

Data in the literature are contradictory on the association between maternal circulating DLK1 levels and placental vascular dysfunction. Cleaton et al. reported highly reduced DLK1 levels only in cases of SGA with placental vascular dysfunction, but no significant difference in maternal circulating DLK1 levels between SGA due to other causes and controls. MacDonald et al. found no relationship between circulating DLK1 levels and ultrasound measures of placental vascular dysfunction in a cohort of 63 women at 36 WG. Unlike Cleaton et al., we did not observe a significant difference in DLK1 levels between SGA with placental vascular dysfunction and SGA due to other causes in our cohort, in which 32.6% of SGA cases showed signs of placental vascular dysfunction. But we confirmed the results of MacDonald et al. in a larger cohort. However, the definition of SGA with placental vascular dysfunction is not consensual and differs widely depending on the study. In the study of Cleaton et al., SGA with placental vascular dysfunction was defined as SGA with high-resistance uterine artery flow, high-resistance umbilical artery flow, or low abdominal circumference growth velocity. Fifty-three percent of SGA cases were defined as SGA with placental vascular dysfunction in the study of Cleaton et al. In contrast, McDonald et al. reported only 6.3% of SGA with placental vascular dysfunction, defined as a low cerebroplacental ratio.

Regardless of birth weight, the association between DLK1 levels and placental vascular dysfunction has been documented by several studies. In a study that included 102 pregnant women, the median DLK1 serum concentration during pregnancy was significantly lower for women with preeclampsia than healthy pregnant women, even after adjustment for gestational age at blood sampling and birth weight (30). Schrey et al. reported lower placental expression of DLK1, primarily in the endothelium and cytotrophoblast, following immunohistochemistry of 40 placental biopsies of healthy pregnant women and speculated that aberrant cytotrophoblast development and endothelial dysfunction in preeclampsia may contribute to reduced placental expression of DLK1 in SGA fetuses.

Using the combination of ultrasound data with a blood test that indicates placental dysfunction is a promising approach for the prediction of SGA (1). The measurements in maternal blood of various proteins or hormones produced by the placenta which are modified in the case of placental vascular dysfunction (pregnancy-associated plasma protein A - PAPP-A, soluble fms-like tyrosine kinase-1 -, sFLT1, placental growth factor - PlGF …) have been tested to identify women at risk of this adverse pregnancy outcome leading to FGR, stillbirth, and preeclampsia. But to date, none is used in clinical practice to predict the risk of SGA with or without FGR (1).

In our study, DLK1 level cannot be considered as a biomarker of SGA, as previously reported by Cleaton et al. and McDonald et al. (20, 24). DLK1 level is less predictive of SGA than ultrasound alone, and DLK1 level does not add much to the predictive value of ultrasound model alone.

The accuracy of detecting SGA for each woman could be enhanced by using ultrasound and following individual trends in the level of DLK1 during pregnancy. Indeed, we have observed a perfect exponential increase in the level of DLK1 when measured each four weeks during pregnancy in a woman who gave birth to an AGA child (unpublished data), offering promising perspectives.

Our study is based on a well-characterized prospective cohort, which allowed us to measure maternal circulating DLK1 levels in a large sample of 193 cases of SGA and 370 matched controls. Maternal levels of DLK1 remained lower in women who delivered SGA than controls when considering the two centers of inclusion separately, showing that our results are robust. We found no association between maternal DLK1 level and placental vascular dysfunction for SGA children. This may have been due to the limited number of subjects with placental vascular dysfunction and the resulting lack of power. Nonetheless, the DLK1 levels for these subjects were very close to those of SGA due to other causes, making this hypothesis unlikely. Also, due to the design of the EDEN cohort and the information collected, data about the placental histopathology were not available.