Subjects

The children who were evaluated were enrolled in the National Institute of Child Health and Human Development Maternal-Fetal Medicine Units (MFMU) Network, which is a randomized, double-masked, placebo-controlled, multicenter clinical trial of single- vs multiple-course antenatal corticosteroids. The primary aim of this trial was to assess the clinical efficacy and safety of repeated doses of antenatal corticosteroids in pregnancies that were at risk for preterm birth. The details of the trial, which was conducted between 2000 and 2003, have been reported previously.

A secondary aim of the MFMU Network antenatal corticosteroid trial was the correlation of steroid regimen with neurodevelopmental outcomes as assessed by the Bayley Scales of Infant Development, second edition, at age 2 years. The Bayley Scales consist of mental and psychomotor developmental indices (MDI and PDI). A Bayley score of 85 is 1 SD below the mean and consistent with mild neurodevelopmental delay. A Bayley score of 70 is 2 SD below the mean and consistent with more significant delay. Placental samples and fetal cord serum were collected in a subset of subjects who were enrolled in the MFMU Network antenatal corticosteroid trial per study protocol.

Our secondary analysis aimed to evaluate the association between SNPs in fetal inflammation and coagulation pathway genes and neurodevelopmental delay at age 2 years. Inclusion criteria consisted of (1) maternal enrollment in the MFMU Network antenatal corticosteroid trial, (2) Bayley scores at age 2 years, and (3) DNA extracted from placental samples or fetal cord serum. Cases were children with mental and/or psychomotor delay that had been defined by a Bayley MDI and/or PDI score of <85 at age 2 years. CP cases (as defined in the primary trial) were excluded from the analysis to test our hypothesis of an association between inflammation and coagulation gene SNPs and neurodevelopmental delay at age 2 years in the absence of CP. Control subjects were children with normal neurodevelopment that had been defined by Bayley MDI and PDI score of ≥85.

The study group is illustrated in the Figure . From the original steroid trial, 583 children were discharged alive. Of these, 4 children died between discharge and follow-up evaluation, and 23 children were not available because they were born at a center that discontinued participation in the MFMU Network in 2001. Of the remaining 556 children who were available for follow-up evaluation, 459 children (83%) had Bayley scores at age 2-3 years. After exclusion of CP cases (n = 7), 452 children were eligible for participation in our analysis. There were 40 twin pairs in the original steroid trial. For this study, 1 twin from each pair (n = 40) was excluded randomly to avoid the issue of including related individuals in the analysis. Of the remaining 412 infants, DNA from placental tissues and/or fetal serum was available for 272 infants (125 cases and 147 control subjects).

Identification of cases and control subjects

From the original steroid trial, 583 children were discharged alive. Of these, 4 children died between discharge and follow-up examination, and 23 children were not available because they were born at a center that discontinued participation in the Maternal-Fetal Medicine Units Network in 2001; cases and control subjects were selected from the remaining 556 children who were available for follow-up evaluation.

CP , cerebral palsy; MDI , mental developmental indices; PDI , psychomotor developmental indices.

Clark. Fetal inflammation and coagulation pathway gene polymorphisms. Am J Obstet Gynecol 2010.

After being reviewed, this study was determined to be exempt from institutional review board approval procedures because of the deidentification of data and study samples before this analysis.

DNA Extraction

DNA was extracted from placental samples and/or umbilical cord serum that had been collected at the time of delivery. Placental samples were either fresh (n = 145) or formalin-fixed and paraffin-embedded (n = 92). DNA was extracted from approximately 0.7 g of fresh placental tissue with the PureGene DNA Purification System (Qiagen, Valencia, CA), as per the manufacturer’s protocols. DNA from the formalin-fixed paraffin-embedded samples was extracted from 3 to 6 5-μm sections that were obtained from the paraffin block and isolated with the PureGene DNA Purification System, as per the manufacturer’s protocols. DNA was also extracted from umbilical cord serum samples (n = 190). Briefly, 250-1000 μL of cord serum was centrifuged at 10,000 g for 10 minutes to pellet any cells in the sera. The supernatant was discarded, and DNA was extracted with the same procedure used for placental samples.

Selection of SNPs for genotyping

We selected 46 SNPs in 27 inflammation and coagulation genes based on the available literature and hypothesized causal pathways. We also included SNPs that have been reported to be associated with CP. The SNPs that were included in our custom multiplex assay are shown in Table 1 .

Genotyping

The 48-Plex GenomeLab SNPStream Genotyping System and accompanying automated genotype calling software (Beckman Coulter, Fullerton, CA) were used for genotyping. Polymerase chain reaction (PCR) assays and extension primers for these SNPs were designed with the use of Beckman Coulter’s Autoprimer multiplex primer design engine ( www.Autoprimer.com ). PCR and extension reactions were performed according to the manufacturer’s instructions. Control subjects were included appropriately in each 384-well plate. Researchers and laboratory personnel were blinded to the case/control status of the biologic samples.

When DNA was available from both placenta and serum, the placental samples were used preferentially secondary to higher quality DNA. In subjects with a placental tissue genotyping success rate of <85%, the serum samples were also genotyped with a goal of increasing the number of SNP results. In 25 subjects with both placental tissue and sera genotyped, manufacturer’s automated allele calling resulted in at least 1 discordant SNP between the 2 samples. Comparison of data quality allowed a correct genotype to be easily identified in most cases. If the subject’s correct genotype could not be identified easily, no genotype was reported for that SNP. In all but 1 case, the serum data were questionable because of the low intensity of the genotype signal caused by the poor quality and low concentration of the extracted DNA. In 3 subjects, the data for both serum and placental samples was strong, but many SNPs were discordant. These 3 subjects were removed from the analysis. Because the serum samples tended to have lower DNA quality, the automated genotype results were individually reviewed. If data quality did not allow for a clear genotype determination for a given serum sample, no genotype was reported.

Statistics

Demographic and clinical characteristics of cases and control subjects were compared with the use of the χ ² or Fisher’s exact test for categoric variables and the Wilcoxon Rank Sum test for continuous variables. Differences in allele and genotype frequencies between cases and control subjects were tested for each SNP with the use of the χ ² or Fisher’s exact tests, as appropriate. SNPs that were found to be significantly related to an MDI and/or PDI score of <85 in the univariate analyses ( P < .05) were analyzed further by multiple logistic regression modeling. Genotypes were included as covariates in a series of regression models that assumed an additive, dominant, or recessive genetic pattern. The additive model assumes that having 2 copies of minor allele has twice the effect of having 1 copy; the dominant model assumes that having at least 1 copy of the minor allele is sufficient for disease, and the recessive model assumes that 2 copies of the minor allele are needed for disease. Other covariates included gestational age at delivery, small-for-gestational-age status (based on a birthweight <10th percentile of published standards), sex, steroid treatment group, race, smoking, and maternal level of education. Of the 3 ways of examining genotype, the regression model that had the highest likelihood score was considered to be the best-fitting model for the respective SNP. The sample size and the rarity of some genotypes and phenotypes limited further modeling of interactions.

Tests for Hardy-Weinberg equilibrium (χ ² or Fisher’s exact) were performed for control subjects for each SNP that was significant in the univariate analysis. Because this was an exploratory study, no adjustments were made for multiple comparisons, and all comparisons are reported. A probability value of < .05 was considered to be statistically significant. All calculations were performed with SAS statistical software (SAS Institute, Inc, Cary, NC).