Background to methods
In a small pilot study of 32 couples, we found that women who had an unexplained elevation of MSAFP and a normal midtrimester fetal ultrasound, and their partners, had a significantly increased C677T MTHFR frequency compared to Manitoba newborns (RR 1.42, 95% CI 1.08–1.85, p = 0.012, two tailed) . The newborn study that examined 977 anonymous consecutive neonatal screening blood spots showed that 36% of Manitoba newborns were heterozygous and 7% were homozygous for C677T MTHFR  (q = 0.25). Subsequently, on evaluation of the pregnancy outcomes of our pilot study women, we noted that, among eight women who had gone on to experience complications of pregnancy, the odds ratio for having the C677T MTHFR allele was 2.3 times higher than in the Manitoba population. However, the result was not statistically significant (p = 0.151, two tailed) indicating the frequency was increased but, this could have been a random result.
Ascertainment and recruitment of study population
All pregnant women in Manitoba are eligible for routine serum screening through the voluntary MMSSP. In Manitoba, an elevation of MSAFP is defined as 2.3 multiples of the median (MOM) or greater. Candidates for inclusion in this study were women with an unexplained MSAFP elevation (i.e. not due to fetal anomalies, incorrect estimation of gestational age, previously unrecognized fetal demise, or multiple gestation) with either a complicated or uncomplicated pregnancy outcome. After appropriate approvals had been obtained from The University of Manitoba Health Research Ethics Board, review of the screening records began in 1999 and took three years. For a study using a two step consent to participate methodology administered by mail, the expected response rate (after excluding lost to follow-up) would be 20% . Our goal was 1000 invitations. We anticipated this would result in approximately 120 participants. This would be double the minimum number of participants suggested by the power analysis we had conducted for the pilot study. To increase our response rate further, we added telephone follow-up for invited potential participants who were non-responders .
All screening records from 1995–1999 were reviewed, accounting for 783 invitations. Records for 2000–2002 were reviewed systematically as outcome information on each pregnancy became available to MMSSP. Records for 1990–1994 were then reviewed systematically in order to bring the total up to 1000. If a woman had more than one pregnancy with an elevation of MSAFP screened by the MMSSP, only the first pregnancy encountered in the retrospective review was used for the study. Previous or subsequent pregnancies were not included. Women with preexisting conditions known to influence pregnancy outcome, such as essential hypertension, and mothers of babies with major congenital anomalies were excluded. Eight women who had relinquished their babies for adoption or whose babies were placed in foster care were also excluded.
Women who met the inclusion criteria were divided into two groups for analysis. Cases were defined as women with pregnancies complicated by one of the complications previously shown to be associated with an unexplained elevation of MSAFP at midtrimester . These include: intrauterine growth restriction (IUGR) (<10th percentile), pregnancy induced hypertension, preeclampsia, eclampsia, postpartum hemorrhage, retained placenta requiring manual delivery, abruptio placenta, premature delivery (<36 weeks gestation or requiring specialized neonatal care for prematurity) and unexplained fetal demise. Study controls were women with normal outcomes which were defined as those with delivery at term ≥ 36 weeks gestation), no complications of pregnancy, a normal placenta and a healthy baby. Definition of complications was based on ICDC-9 codes in the MMSSP outcome charts for each patient  which are then confirmed later by chart review for all those with a positive MMSSP result. All women ascertained as having unexplained MSAFP elevations and who fit the inclusion criteria above, were invited by letter to participate. The previously reported newborn study provided population control group data .
Women who agreed to participate in the study were mailed the appropriate questionnaires and blood requisitions. The questionnaire included a semi-quantitative food frequency questionnaire (FFQ) based on standard methodology but, modified to suit Manitoba residents and previously validated for this population by biochemical analysis during the pilot study [11, 16]. The survey included questions on vitamin supplement intake to determine preconceptional or prenatal supplementation as well as current use of vitamins. Dietary intake of folate and folic acid from supplements, and intake of the cofactors B12 and B6, were calculated from the FFQ for intake both during pregnancy and at the time of the study. A correction of an additional 0.1 mg for folic acid fortification that began in Canada in 1998 was included for pregnancies that began after fortification . FFQ analyses were performed with the researcher blinded as to the status of the mother.
Total plasma homocysteine, red blood cell folate, and serum folate were determined using established methodology [18, 19]. Samples were processed on site with clotting and separation by spinning. Sera was stored at 4°C during shipping to the central laboratory and until processing. DNA was extracted from whole blood and C677T MTHFR genotyping was performed using previously established methodology [11, 20, 21]. Genotyping and biochemical analyses were performed also blinded.
Chi-squared analysis (one tailed unless otherwise noted) was used for allele frequency. Comparisons of potentially confounding factors between the case group and the study control group were undertaken. Parametric data were analyzed with the Student's t test for difference between means with Bonferroni correction for multiple comparisons. Data not normally distributed were analyzed using the nonparametric Mann-Whitney Rank Sum Test. Linear regression was used to test the validity of the dietary survey. A multivariate analysis included age, smoking, maternal weight at the time of MSAFP testing, presence of C677T MTHFR, gender and weight of infant, biochemical parameters, and FFQ results for folate, B12 and B6, both at the time of the survey and for during the pregnancy was undertaken. In order to avoid convergence due to the large number of variables, the analysis was completed in subsets of six variables. Variables with the higher association scores from these analyses were then combined for further testing in various combinations using stepwise multiple linear regression. Also linear regression analysis of each continuous variable with genotype results was performed. Corrections for multiple comparisons were included. Software used was NCSS Statistical Systems for Windows .