Materials and Methods

Approval for this investigation was obtained from the Institutional Review Board of Columbia University Medical Center.

This study was a retrospective analysis of a subset of participants in an ongoing prospective cohort designed to observe the relationship between β ₂ AR genetic variation and preterm labor and delivery. Gravidas with singleton pregnancies presenting for routine obstetrical care between 8-20 weeks’ gestational age were approached for participation. Informed written consent was obtained from each subject to allow investigators to obtain and evaluate data and genetic information (genomic DNA) related to pregnancy. A 5-mL venous blood sample was obtained from each participant for genetic analysis. Demographic, medical, and obstetrical information was collected from all subjects.

Inclusion for analysis in the study cohort required vaginal delivery at Columbia University Medical Center with a gestational age ≥34 weeks. We selected 34 weeks as a cutoff due to an institutional practice of withholding tocolytic medications to suppress late-preterm labor. Exclusion criteria included multiple gestations, major fetal malformations, and fetal demise.

For all qualifying subjects, antenatal, intrapartum, and maternal and neonatal outcomes data were collected following delivery through a review of the Columbia University Medical Center electronic medical record system (Eclipsys XA; Eclipsys Corp, Atlanta, GA). Data were automatically extracted from archived template-based narrative notes and intrapartum flow sheets by a biomedical informatician (C.W.). To validate the electronic extraction, all physician and nursing notes and flow charts were manually reviewed by a single investigator (R.S.M.).

Subjects were excluded from analysis if there was insufficient, incomplete, or inconsistent intrapartum data, or if genotyping studies were unsuccessful. Subjects lacking a documented cervical dilatation <10 cm in active labor were also excluded, as a minimum of 2 data points in active labor are necessary to estimate a rate of cervical dilatation by linear regression. Active labor was defined as the period from 4-10 cm in a patient with regular contractions effecting cervical change.

Pregnancy dating was confirmed or revised using a standardized institutional ultrasound dating policy. All ultrasound records were reviewed by a single investigator (R.S.M.).

DNA was extracted and purified with a Puregene extraction kit (Gentra, Minneapolis, MN) and tested for quantity, purity, and quality by optical densitometry (ratio, 260/280 nm) and gel electrophoresis. For the identifications of the polymorphisms of the ADRB2 gene, 60 ng DNA was amplified by polymerase chain reaction (96-well microtiter plate block; Biometra, Gottingen, Germany) using specific and validated primers for single-copy DNA regions surrounding polymorphisms Arg16Gly and Gln27Glu located in the single exon of ADRB2, as described in previous work. ADRB2 genotype of both single-nucleotide polymorphisms at codons encoding amino acids 16 (rs1042713) and 27 (rs1042714) was determined by Sanger sequencing reaction and electrophoresis on a fluorescent DNA fragment analyzer apparatus (ABI3100; Applied Biosystems, Foster City, CA).

Subjects were categorized into 2 genotype groups: those with homozygous genotype encoding for Arg/Arg16 and a control group consisting of all other genotypes (encoding for either Arg/Gly16 or Gly/Gly16). Groups were compared for selected demographic and obstetrical variables. The χ ² or Fisher exact test was used to compare categorical variables and the unpaired Student t test was used to compare means of continuous variables between groups. STATA 10 (StataCorp LP, College Station, TX) was used for analysis of demographic and obstetrical data.

For a given genotype group, all timed intrapartum cervical examinations were plotted, with the ordinate representing cervical dilatation (in centimeters from 4-10 unless otherwise specified) and the abscissa representing hours before full dilatation. Linear regressions were created using the least squares fitting method with 95% confidence intervals using Origin 8.0 software (OriginLab Corp, Northampton, MA). Labor rates were considered to be distinct between genotype groups if the 95% confidence bounds did not overlap. Statistical differences in labor rates among groups were calculated in PLT Tools (PLTsoft, San Francisco, CA) for NONMEM (version 6; ICON Development Solutions, Ellicott City, MD). For each data set, the timed cervical examinations were fit with a linear equation as a naïve pooled data fit (assumption of no intraindividual variability). Random noise was treated as an additive term. Genotype was introduced as an additional variable. The model with genotype was compared to an identical model without it using the likelihood ratio test. If inclusion of the covariate decreased the log likelihood by >3.84 (χ ² distribution, P < .05 with 1 degree of freedom), then the covariate was considered statistically justified.

Genotype groups were also created and compared as described above for genotype at codon 27. Lastly, subjects were categorized into putative ADRB2 haplotype groups based on genotype at the 2 studied polymorphic sites encoding for codons 16 and 27. Haplotype groups were compared as described above for genotype groups, except that analysis of variance was used for comparisons of means of continuous variables among haplotype groups.

With respect to haplotype assignments, it is well established that there is strong linkage disequilibrium between ADRB2 codons encoding amino acids 16 and 27, with Arg16 almost never occurring in the presence of Glu27. Because of this relationship, haplotype can be imputed in the vast majority of cases from the genotypes at the 2 codons.

The sample size was determined by the number of subjects enrolled into the preterm labor trial from May 2006 through August 2007 meeting the inclusion criteria specified above. Post hoc evaluation of effect size (defined as difference in rate of active labor based upon β ₂ AR genotype) was estimated with a bootstrap analysis with replacement using 1000 iterations of our final model that included a difference in active labor rate between Arg/Arg16 subjects and subjects with all other genotypes using the available data set in PLTtools ( Plessthan.com ) and NONMEM (ICON Development Solutions).