The role of maternal homocysteine concentration in placenta-mediated complications: findings from the Ottawa and Kingston birth cohort

Women attending prenatal appointments and planning to deliver at The Ottawa Hospital, the Ottawa region, or the Kingston General Hospital were recruited to the Ottawa and Kingston (OaK) Birth Cohort from 2002 to 2009. Participants were included if between 12 and 20 weeks gestation of a viable singleton or twin pregnancy. For the present study, participants were excluded from the analytic data set if non-singleton, recruited before 12 or after 20 weeks gestation, if they withdrew, were lost to follow-up, or if the pregnancy was terminated.

Details of the cohort study have been previously reported [23]. Briefly, the baseline survey and post-partum follow-up consisted of an interviewer-administered questionnaire and hospital record abstraction. Maternal blood samples were collected at baseline and laboratory personnel were blind to outcomes. Blood samples for homocysteine and MTHFR were collected in K₂EDTA Vacutainer tubes (Becton Dickinson, Lincoln Park, NJ) and for serum folate in serum separator tubes (Becton Dickinson). Samples for plasma were immediately placed on ice and within 30 min centrifuged in 4 °C at 3000×g for 10 min, then aliquoted and stored at − 20 °C. Plasma homocysteine (μmol/L) was measured on the Abbott AxSYM II Immunoassay System (Abbott Laboratories, Abbott Park, IL) using fluorescence polarization immunoassay. Blood samples for serum were centrifuged at 3000×g for 10 min, then aliquoted and stored at − 20 °C. Serum folate (nmol/L) was measured using the Beckman Coulter Access 2 and Unicel DxI 800 immunoassay analyzers using manufacturer’s reagents (Beckman Coulter, Brea CA). Homocysteine and folate were measured within one month in batches.

The primary outcome was the composite of placenta-mediated complications (PMCs): small for gestational age (SGA), preeclampsia, placental abruption, and pregnancy loss. The secondary outcomes were individual components of the composite. Additionally we investigated severe SGA <5th percentile and severe preeclampsia.

A small for gestational age infant had a birth weight less than the 10th or 5th percentile of sex and gestational age-adjusted population standards [24]. Pregnancy loss was intrauterine death before 20 weeks or stillbirth. Preeclampsia was new onset hypertension with proteinuria. Hypertension was a diastolic blood pressure reading greater than or equal to 90 mm Hg and proteinuria of 2+ on a dipstick or proteinuria greater than 300 mg in a 24-h urine collection measured on two separate occasions of at least 6 h apart. Preeclampsia with delivery prior to 35 weeks gestation was considered severe [25]. Placental abruption was antepartum bleeding with objective evidence either on ultrasound or inspection of the placenta at birth or a pathologic examination of a retro-placental clot [26]. A committee of medical experts blind to participant exposure status independently examined information abstracted from medical records to adjudicate preeclampsia and placental abruption.

Potential confounders, risk factors, and interactions were specified prior to multivariable modeling using knowledge of the subject matter, our analyses of homocysteine determinants in OaK participants (unpublished), and previous work investigating PMCs [12‐14, 21, 27]. We considered the following factors: maternal age, race, education, income, nulliparity, smoking, diabetes, use of hormonal birth control prior to conception, chronic hypertension, history of PMCs, folic acid supplementation, and serum folate. Gestational age at recruitment was abstracted from hospital records.

We conducted multivariable logistic regression analyses to examine the association of plasma homocysteine concentration (as the primary exposure of interest) with PMCs, while adjusting for the identified potential confounders and risk factors. We did not transform homocysteine because log-transformation did not substantially normalize the distribution (Additional file 1: Figure A). Because serum folate is intermediate with respect to folic acid supplementation, analyses were conducted with and without serum folate to note any changes in effect estimates for folic acid supplementation. Changes were also noted for outliers in homocysteine concentration.

Multivariable model building was conducted as follows [22]. An initial model was fit including main effects for all independent variables. Continuous variables, i.e., homocysteine, gestational age, maternal age, BMI, and serum folate were modeled using restricted cubic splines with five knots set at the 5th, 27.5th, 50th, 72.5th, and 95th quantiles. Next, a plot of partial associations referred to as an ANOVA plot, corrected for the number of degrees of freedom, was generated to visualize strong and weak partial associations. The strengths of association informed a decision of how many degrees of freedom to allocate to each variable: strong associations were modeled with greater complexity than weak associations. For example, weaker partial associations in continuous variables were reduced to fewer knots or a linear term, while categorical and ordinal variables were collapsed. Akaike’s Information Criterion (AIC) and the Bayesian Information Criterion (BIC) were then used to confirm the allocations [22].

The multivariable logistic regression models were fit using the rms (regression modelling strategies) package [33]. Odds Ratios for continuous variables modelled with restricted cubic splines were estimated comparing the 75th to 25th percentile. The logistic regression model was fit to each imputed dataset and results were combined across the 10 datasets using Rubin’s method which computes imputation-adjusted variances and average betas [30, 34]. Analyses of outcomes with a low frequency of events were conducted using penalized maximum likelihood. The best penalty factor was identified by tracing effective AIC for different penalties [22, 33].

Subgroup analyses examined the modifying effect of the MTHFR 677C>T genotype and high-risk pregnancy, defined as chronic hypertension, diabetes, history of a PMC, or BMI greater than 35. This was done by including interaction terms with homocysteine and each potential moderator. Subgroup analyses were not conducted for placental abruption and pregnancy loss due to a low number of events.

Homocysteine concentration was significantly higher in participants with the composite outcome of any PMC, small for gestational age (SGA) infants, and pregnancy loss (Additional file 1: Table A), also demonstrated visually by boxplots of the homocysteine distribution according to pregnancy outcome (Additional file 1: Figure A). In adjusted analyses, higher plasma homocysteine concentration was significantly associated with increased odds of any PMC (Table 2, p = 0.0007), SGA (Table 3, p = 0.0010), was marginally associated with preeclampsia (p = 0.07), and was not associated with placental abruption, and pregnancy loss (Table 3, 0.99, and 0.16). Additionally, homocysteine was associated with severe SGA (<5th percentile) and marginally with severe preeclampsia (delivery < 35 weeks gestation) (Table 3, 0.0012, and 0.0595). A 5 μmol/L change in homocysteine concentration, which is approximately 4 SDs of the homocysteine concentration, was associated with a 63% increased odds of any PMC (Odds Ratio (OR) 1.63, 95% Confidence Interval (CI) 1.23–2.16) and a 76% increased odds of SGA (OR 1.76, 95% CI 1.25–2.46). Serum folate and folic acid supplementation were not associated with the outcomes (Table 2 and Additional file 3: Tables C.1-C.4).

The results of the subgroup analyses are presented in Tables 4 and 5. In the subset of 4006 OaK participants with measured genotype, the interaction between MTHFR 677C>T and homocysteine was not statistically significant for preeclampsia (p = 0.84) but there was some evidence of interaction for any PMC (p = 0.12) and for SGA (p = 0.07) with associations qualitatively different within each subgroup. Associations in the CC/CT subgroups were positive and significant, whereas in TT subgroup associations were negative and not significant, with wide confidence intervals. Figure 2 illustrates these trajectories for the composite outcome with homocysteine. The interaction between high-risk pregnancy and homocysteine was not statistically significant for any PMC and SGA (p = 0.27 and p = 0.51 respectively), but there was some evidence of a moderating effect for preeclampsia (p = 0.14, Table 5). While the interaction effect (and hence, the difference in association between the high and low-risk subgroups) was not statistically significant at the conventional 5% level, the odds ratio was higher and statistically significant in the high-risk group (OR 2.84, 95% CI 1.19 to 6.79 for a 5 μmol/L change in homocysteine concentration) compared to a lower and non-statistically significant OR in the low-risk group (OR 1.31, 95% CI 0.74 to 2.30).

Different groups of risk factors were associated with SGA and preeclampsia (Tables 3, C.1-C.2). A high school or incomplete post-secondary education and smoking were associated with increased odds for any PMC and SGA. Chronic hypertension, diabetes, and history of experiencing a PMC were associated with increased odds for any PMC and preeclampsia.

We analyzed data from 7587 participants from the Ottawa and Kingston (OaK) Birth cohort. We found that maternal homocysteine concentration in the early to mid-second trimester was associated with increased odds of any placenta-mediated complication (PMC): a composite of small for gestational age (SGA), preeclampsia, placental abruption, and pregnancy loss, and was associated with increased odds of SGA and severe SGA and preeclampsia. In the high-risk subgroup homocysteine was associated with increased odds of preeclampsia.

To our knowledge, this is the largest cohort study to investigate the association between early to mid-second trimester maternal homocysteine concentration and the risk of placenta-mediated complications. We used multiple imputation to deal with missing values and conducted rigorous multivariable logistic regression analyses designed to explore flexible functional forms of association with homocysteine and other continuous factors, controlling for a wide range of potential confounders.

In most larger studies, homocysteine was dichotomized or grouped to investigate non-linearity and a threshold effect [12, 16]. However, the Hordaland Homocysteine study of several thousand participants suggested that for most conditions including pregnancy complications and adverse pregnancy outcomes, homocysteine exhibits a continuous concentration-response relation [1, 27]. Thus, one of the main strengths of our study is that we accounted for the functional form, i.e., shape of the association, of homocysteine and other continuous variables in relation to the outcomes of interest, which to the best of our knowledge has not been previously reported.

One of the main limitations of our study is that we did not examine the effect of Vitamins B₆ and B₁₂. In a survey of the Canadian population from 2007 to 2009, Vitamin B₁₂ was the main determinant of elevated homocysteine concentration in the folate replete [35]. We would expect OaK participants to have adequate Vitamin B₁₂ levels because 85% were supplementing with multivitamin or prenatal vitamin supplements, which likely contained Vitamin B₁₂. Caffeine or coffee consumption is another homocysteine determinant we did not examine [9], though pregnant women tend to consume less caffeine during pregnancy.

Other studies have reported an increased risk of SGA associated with elevated homocysteine [12, 36]. In Bergen et al.’s cohort of 5085 participants recruited in the Netherlands from 2002 to 2006, early second trimester homocysteine in the upper versus lower quintile was associated with an increased risk of SGA (<5th centile) [12]. In Dodds et al.’s cohort of 2119 participants recruited in Nova Scotia, Canada from 2002 to 2005, a study time frame similar to our OaK birth cohort, a homocysteine concentration greater than the 90th percentile was not associated with a risk of SGA [13]. The mean homocysteine concentration in the latter study was lower than Bergen et al. [12], but comparable to our participants’ mean homocysteine concentration of 4.8 μmol/L. In another study of 65 SGA cases and 358 controls recruited in a Chinese textile factory from 1996 to 1998, pre-conception homocysteine above the 90th percentile was not associated with an increased risk of SGA [16].These latter two studies dichotomized homocysteine concentration using cut-offs that are not necessarily physiological, which may limit comparability with our findings and those of Bergen et al. [12]

The MTHFR 677C>T genotype qualitatively modified the association of homocysteine with any PMC and SGA. However uncertainty around the negative effect of the TT genotype, likely due to fewer participants the TT subgroup, makes it difficult to interpret the modifying effect. Higher folate intake and status are known to mitigate but not eliminate the effect of this polymorphism on homocysteine concentration [19, 37, 38]. Although approximately 95% of our sample was consuming a folic acid supplement and white flour and other cereal products in Canada have been fortified with folic acid since the late 1990s, homocysteine concentration was associated with the MTHFR genotype in the OaK cohort. Given the variations in clinical testing for MTHFR genotype in current obstetric practice [39], our study findings suggest testing may be warranted, particularly in high-risk subgroups, for example, those who exhibit comorbidities or a history of complications.

We found that elevated homocysteine was marginally associated with an increased risk of preeclampsia, and that the association was significant in the high-risk subgroup. Others have found no association [12, 14]; for example, Kahn et al.’s nested case-control study with 113 preeclampsia cases and 443 controls recruited in Montreal from 1999 to 2004 [14]. Studies of repeated homocysteine measurement during pregnancy have reported increasing homocysteine concentration during the course of preeclampsia [40]. In a longitudinal analysis of 252 women of whom 49 developed preeclampsia, homocysteine increased in the preeclampsia group independent of B-vitamin (B₆, B₁₂, and folate) and obesity status, while concentrations in the uncomplicated group remained steady [41]. Therefore early homocysteine measurement, as in our study, may pre-date onset of disease and changes in homocysteine. Our finding of a stronger association of homocysteine in severe preeclampsia, marginally significant with fewer outcome events, also suggests a role of homocysteine earlier in pregnancy when severity is increased.

Some studies have, however, reported an increased risk of preeclampsia associated with higher homocysteine concentrations [13, 42‐45]. Many of these studies were conducted during a time period before routine peri-conceptional folic acid supplementation and/or mandatory folic acid fortification (i.e., the early to mid-1990s) [42‐44], or were conducted in countries without mandatory folic acid fortification [45‐47]. This suggests that where folate intake is lower, homocysteine levels would tend to be higher and therefore homocysteine might play a greater role, earlier on, in the development of preeclampsia.

We report the association of the range of potential confounders with the composite and individual outcomes. Our results, although exploratory in this regard, demonstrated a greater role of diabetes and chronic hypertension in the development of preeclampsia compared to SGA. In a clinical opinion paper, Ness and Sibai [48] hypothesized that the maternal syndrome observed in preeclampsia develops in the presence of abnormal placentation that interacts with maternal metabolic syndrome, and that fetal growth restriction develops in the absence of metabolic syndrome. Our findings of increased odds of preeclampsia in the high-risk subgroup also lend support to this hypothesized role of developing endothelial dysfunction in preeclampsia.

In our study elevated homocysteine concentration was not associated with an increased risk of placental abruption and pregnancy loss, but a greater number of events would be necessary to confirm an association. Some studies have found no association of homocysteine with early pregnancy loss [15, 17, 49]. However, one-time early pregnancy losses are characteristically different from recurrent early pregnancy losses and losses throughout pregnancy [10].