Omega-3 fatty acid supplement use and oxidative stress levels in pregnancy

Oxidative stress is a biological imbalance in reactive oxygen species and antioxidants. Increased oxidative stress during pregnancy has been associated with adverse birth outcomes. Omega-3 fatty acid (n-3 FA) supplementation may decrease oxidative stress; however, this relationship is seldom examined during pregnancy. This study assessed the association between n-3 FA supplement use during pregnancy and urinary oxidative stress biomarker concentrations. Data came from The Infant Development and the Environment Study (TIDES), a prospective cohort study that recruited pregnant women in 4 US cities between 2010–2012. Third trimester n-3 FA intake was self-reported. Third trimester urinary 8-iso-prostaglandin F2α (8-iso-PGF2α) was measured as an oxidative stress biomarker. Additionally, we measured the major metabolite of 8-iso-PGF2α and Prostaglandin F2α (PGF2α) and utilized the 8-iso-PGF2α to PGF2α ratio to calculate the change in 8-iso-PGF2α reflecting oxidative stress versus inflammation. Adjusted linear models were used to determine associations with control for confounding. Of 725 women, 165 reported n-3 FA supplement use in the third trimester. In adjusted linear models, n-3 FA use was associated with 10.2% lower levels of 8-iso-PGF2α (95% Confidence Interval [CI]: -19.6, 0.25) and 10.3% lower levels of the metabolite (95% CI: -17.1, -2.91). No associations were observed with PGF2α. The lower levels of 8-iso-PGF2α appeared to reflect a decrease in oxidative stress (percent change with supplement use: -18.7, 95% CI: -30.1, -5.32) rather than inflammation. Overall, third trimester n-3 FA intake was associated with lower concentrations of 8-iso-PGF2α and its metabolite, suggesting a decrease in maternal oxidative stress during pregnancy.


Introduction
Oxidative stress, a biological imbalance in reactive oxygen species and antioxidant levels, results in an excess of free chemical radicals that can cause lipid peroxidation and other outcomes leading to cell damage [1]. In pregnancy, elevated oxidative stress levels have been associated with prevalent adverse birth outcomes in the United States, including placental aging [2], intrauterine growth restriction [3,4], preterm labor [5][6][7][8], and preeclampsia [9,10]. These outcomes have been associated with increased risks of morbidity and mortality for mothers and infants [11]. An accessible method to lower oxidative stress levels during pregnancy could help prevent these adverse birth outcomes.
Omega-3 fatty acid (n-3 FA) supplementation has the potential to reduce damage caused by oxidative stress. One possible biological mechanism by which this may occur is through the actual replacement of arachidonic acid with n-3 FA in cell membranes. Then, when an excess of reactive oxygen species causes oxidation of lipids in those membranes, the cleaved products of n-3 FA are hypothesized to be less damaging to other cells than those of arachidonic acid [12,13].
A number of studies have examined the relationship between n-3 FA supplementation and oxidative stress in humans [14][15][16]. However, only two of these studies have evaluated associations with maternal oxidative stress levels during pregnancy [17,18]. In one randomized trial of pregnant women with gestational diabetes, women who were given n-3 FA supplements in pregnancy had decreased concentrations of plasma malondialdehyde (MDA), a biomarker of oxidative stress, compared to those who did not [17]. Another trial evaluated this association in healthy pregnancies by measuring maternal plasma thiobarbituric acid-reactive substance (TBARS) levels as an indicator of oxidative stress, and observed a surprising positive association [18]. Of note, these trials were restricted by smaller samples sizes (N = 54-270) and limitations in oxidative stress biomarker measurement [19][20][21][22][23]. Specifically, MDA has been considered unreliable [20] while TBARS assays are often non-specific and results may represent other processes aside from lipid peroxidation [24]. Both biomarkers are known to be susceptible to cross-reactions with other existing biochemicals [20].
We sought to examine the association between prenatal n-3 FA supplementation and maternal oxidative stress using data from The Infant Development and Environment Study (TIDES), a multi-center US pregnancy cohort. We examined the relationship between maternal self-reported n-3 FA supplement use and urinary concentrations of 8-iso-PGF 2α , its major metabolite (metabolite 2,3-dinor-5,6-dihydro-15-F 2 T-isoprostane), and Prostaglandin F 2α (PGF 2α ). These well-established biomarkers of oxidative stress are direct products of lipid peroxidation that represent lipid damage through specific pathways, are stable during pregnancy, and easily detectible in urine [6,8,25,26]. Furthermore, we used the novel 8-iso-PGF 2α / PGF 2α ratio to investigate whether changes in 8-iso-PGF 2α were attributable to either upregulation of inflammatory pathways or to true oxidative stress [27]. Thus, utilizing these biomarkers represents a significant improvement on previous research on this relationship in pregnant women.

Study population
TIDES is a multi-center prospective cohort of women enrolled during pregnancy [28,29]. Women were recruited between August 2010 and August 2012 at clinics in the University of California, San Francisco, CA; University of Rochester Medical Center, NY; University of Minnesota, MN; and University of Washington-Seattle Children's Hospital, WA. Cohort eligibility included a minimum age of 18 years, the ability to read and write in English, and a maximum gestational age of 13 weeks at the time of recruitment, to reflect first trimester exposures in relation to neonatal outcomes. At three prenatal appointments, targeted at one per trimester, participants completed questionnaires and provided urine samples (protocols described in detail elsewhere) [29]. TIDES was approved by the institutional review board at each study site and each participant provided signed informed consent prior to data collection.
All questionnaire data was collected at each of the four study sites and is held centrally at the Icahn School of Medicine at Mount Sinai. Use of this data was deemed exempt by the NIEHS IRB, and was transferred to NIEHS under a data use agreement between the two institutions. The oxidative stress biomarker data was generated at Vanderbilt University Medical Center and transferred directly to NIEHS.
The present analysis was a cross-sectional assessment of the associations between n-3 FA supplementation, as reported by questionnaire, and urinary oxidative stress biomarker concentrations. From the overall study population we included women who provided a urine sample and responded to the n-3 FA survey question at the third study visit. From the overall study population (N = 971), these restrictions excluded 210 women who were missing a urine sample at the third visit and 36 women who did not respond to the n-3 FA survey question, for a final sample size of N = 725.

n-3 FA supplement use
Participants completed self-administered questionnaires at each visit to provide information on lifestyle, demographic characteristics, and health. Supplement use was assessed with a list; participants were asked to check a box next to any supplement that they had consumed daily for at least one consecutive week in their current trimester. The list included n-3 FA supplements, listed as "fish oil supplements", prenatal vitamins, multivitamins, and other individual supplements.

Biomarkers of oxidative stress
Biomarkers of oxidative stress were measured in third visit urine samples at the Vanderbilt Eicosanoid Core Laboratory. Samples were collected at a mean of 32.6 weeks gestation (range = 25.7-41.1 weeks gestation). Samples were analyzed via gas chromatography-negative ion chemical ionization mass spectrometry for several compounds derived from arachidonic acid: 8-iso-PGF 2α ; its primary metabolite 2,3-dinor-5,6-dihydro-15-F 2 T-isoprostane; and PGF 2α . The metabolite of 8-iso-PGF 2α may be more sensitive than the parent compound; it is generated in the lungs rather than the kidneys and thus may be a better indicator of oxidative stress occurring throughout the entire body [30]. PGF 2α is a reliable marker of inflammation [31].
8-iso-PGF 2α can be generated through upregulation of inflammatory pathways or through chemical oxidative stress [27]. To distinguish the source in our study, we used the novel 8-iso-PGF 2α / PGF 2α ratio which evaluates the proportion of 8-iso-PGF 2α produced from enzymatic synthesis by prostaglandin-endoperoxide synthases (PGHS) (i.e., inflammation) as compared to chemical lipid peroxidation (i.e., oxidative stress). Higher levels of the ratio indicate a greater contribution to 8-iso-PGF 2α from chemical lipid peroxidation, i.e., oxidative stress, as compared to enzymatic synthesis, i.e., inflammation, and it can be used to calculate the proportion of 8-iso-PGF 2α originating from each source [27]. This ratio has been examined and validated in other studies as well [32][33][34]. Thus, we examined a total of five markers: three that were measured (8-iso-PGF 2α , 8-iso-PGF 2α metabolite, PGF 2α ) and two that were derived (8-iso-PGF 2α chemical and 8-iso-PGF 2α enzymatic).
For examining distributions of oxidative stress biomarkers, we corrected each concentration for urinary specific gravity because this correction has been found suitable in accounting for the hydration status of pregnant women when analyzing urinary biomarkers. Compared to creatine, specific gravity correction is more reproducible within-person with less systematic variance [35]. We used the following formula for correction: where O c represents the specific gravity-corrected oxidative stress biomarker concentration, O represents the measured biomarker concentration, 1.014 is the median specific gravity of all TIDES samples, and Sg is the specific gravity level measured in that sample. For statistical models, raw concentrations were modeled and specific gravity was included as a covariate.

Statistical analysis
All statistical analyses were performed using SAS 9.4 (Cary, NC]. First, we examined demographic characteristics associated with n-3 FA use using chi-squared tests. We then assessed associations between n-3 FA consumption and urinary oxidative stress biomarker concentrations. Finally, we created crude and adjusted linear models to evaluate the associations between n-3 FA intake and each oxidative stress marker. For all statistical models, urinary oxidative stress biomarker concentrations were natural log-transformed. Crude models included gestational age at urine sample collection and specific gravity as covariates. For adjusted models, confounders were identified using a Directed Acyclic Graph developed following a literature review. Potential confounders were then empirically evaluated within our dataset for their impact on effect estimates. Final models included gestational age at sample collection (weeks, continuous], maternal age (years, continuous), specific gravity (continuous), race (White/ Black/other), education (college degree vs. none), and study center. Additional variables such as income, smoking status, prenatal vitamin use, pre-pregnancy BMI and fish intake were considered but not included due to low sample size or little influence on effect estimates (less than 10% change). All effect estimates and 95% confidence intervals (CIs) were scaled to present the percent change in oxidative stress biomarker in association with n-3 FA supplementation in pregnancy, for interpretability.

Sensitivity analyses
We ran several additional analyses to test the robustness of our findings. Because the decision to take n-3 FA is strongly associated with socioeconomic status, we examined models stratified by education level [36]. Second, because individuals in our sample who supplement with n-3 FA were more likely to take prenatal vitamins which may also influence oxidative stress levels as well, we restricted our analytic population to women who reported prenatal vitamin use in the third trimester. Use of other vitamin supplementation was uncommon in our population. Of the 725 women in our study, the most common supplements, aside from n-3 FA, were vitamin D (n = 148), iron (n = 141), and calcium (n = 96). Other individual supplements, including amino acids, herbal supplements, and vitamins A, C, E or K were used much less frequently (n = 1-6). These compounds were not associated with oxidative stress biomarkers in preliminary analyses and thus were not examined in subsequent models.

Results
Our study included 725 women with urinary oxidative stress biomarker measurements and n-3 FA supplement questionnaire responses at the third study visit. Participants were primarily White, married or living with their partner, and did not smoke or drink alcohol during pregnancy (Table 1). On average, the women were 31.7 years old and had a pre-pregnancy BMI of 25.6 kg/m 2 . In our sample, 165 (23%) women reported taking an n-3 FA supplement for at Table 1

Characteristic n-3 FA Supplementation (n = 165) n (%) No n-3 FA Supplementation (n = 560) n (%)
Age ( least a week in the third trimester. These women were more likely to be White, non-smokers, and married or living with their partner, compared to those who did not take n-3 FA supplements. Women who took n-3 FA supplements were also older and had higher incomes and education levels then women who did not take n-3 FA supplements. Specific gravity was higher in women who did not use N-3 supplements, reflecting more concentrated urine. The median (IQR) in users was 1.011 (1.007, 1.016) and 1.015 (1.009, 1.021) in non-users.
Concentrations of both 8-iso-PGF 2α and the 8-iso-PGF 2α metabolite were lower among women who took n-3 FA in the third trimester compared to those who did not (Table 2). In adjusted models, n-3 FA consumption was associated with 10.2% lower levels of 8-iso-PGF 2α (95% confidence interval [CI]: -19.6, 0.25) and 10.3% lower levels of the 8-iso-PGF 2α metabolite (95% CI: -17.1, -2.91) ( Table 3). We did not observe an association between n-3 FA use and PGF 2α levels. When we examined associations with the chemical versus the enzymatic fractions of 8-iso-PGF 2α , n-3 FA use was associated with 18.7% lower levels in the chemical fraction of 8-iso-PGF 2α (95% CI: -30.1, -5.32), reflecting oxidative stress, but was not associated with the enzymatic fraction of 8-iso-PGF 2α, reflecting inflammation. Results were greater in magnitude in crude models (Table 3); all covariates included in the model meaningfully changed effect estimates compared to the crude model.  Overall, our sensitivity analyses produced results fairly similar to our primary results. First, in a stratified analysis, there was no evidence of statistical interaction by education level (S1 Table). However, associations were greater in magnitude among women with no college degree. Second, we restricted our sample to women who reported third trimester prenatal vitamin use (n = 622). Effect estimates were similar to those observed in the original analytic sample (S2 Table).

Discussion
In a healthy cohort of US pregnant women, we found that n-3 FA supplementation was associated with lower levels of 8-iso-PGF 2α and its metabolite after controlling for relevant confounders, indicating lower levels of oxidative stress. Our examination of the fractions of 8-iso-PGF 2α originating from chemical oxidative stress as compared to inflammation suggested that changes observed were attributable to lower chemical oxidative stress levels. Although these estimates are adjusted for relevant confounders based on substantive and statistical evidence, these results should be interpreted with caution as additional, unmeasured health behaviors may influence this association.
Overall, randomized trials primarily suggest that n-3 FAs are effective in reducing oxidative stress in both animals and humans [14][15][16][37][38][39][40]. Although, one study has suggested that prolonged use in animals may actually increase oxidative stress levels [41]. Previous animal and human studies that have explored the association between prenatal n-3 FA supplementation and maternal oxidative stress in pregnancy have been inconclusive. Most, but not all, findings from animal studies align with our results that n-3 FA supplementation during gestation is associated with lower maternal oxidative stress biomarker concentrations [42][43][44]. The two previous human studies that have assessed this relationship in pregnant women have produced mixed results [17,18]. Our results align with those from Jamilian et al. who found that n-3 FA supplementation was associated with decreased oxidative stress levels in women with gestational diabetes [17]. Finally, although not directly related, this evidence is also consistent with studies that have examined associations between prenatal n-3 FA supplementation and cord blood or neonatal oxidative stress levels [45][46][47][48]. Animal studies have also observed that maternal n-3 FA supplementation is associated with lower levels of oxidative stress biomarkers in placenta and offspring [42][43][44][49][50][51][52][53][54].
Previous studies of n-3 FA supplementation in pregnancy were limited by the sample sizes and characteristics of their study population as well as the biomarkers used assess oxidative stress, which could explain some of the differences observed. The only other study to evaluate the impact of n-3 FA supplementation on oxidative stress during healthy pregnancy observed an elevation in oxidative stress, as indicated by TBARS levels, in association with fish oil supplementation [18]. However, TBARS has a number of limitations as a biomarker of oxidative stress. TBARS assays can be unreliable in providing direct measures of lipid peroxidation. These assays are sensitive to cross-reactions with other compounds, which can lead to inaccurate results [20,24]. Another study which assessed associations between n-3 FA supplementation and oxidative stress in pregnant women, but which used seafood portions instead of pill supplements, also had differing findings from ours in that they observed no differences in plasma or urinary measurements of 8-iso-PGF 2α levels in the treatment and control groups [55]. However, the differences in results between that study and ours could be attributed to the fact that use of oral supplements likely delivers higher n-3 FA amounts than the seafood portions that were administered. Our findings are more consistent with the aforementioned study of prenatal supplementation and oxidative stress in pregnancy by Jamilian et al., which observed decreased levels in association with supplement use, despite the fact that they also used a less reliable oxidative stress biomarker, malondialdehyde [20,24].
Our findings additionally add to this body of literature by improving the ability to distinguish the pathway by which n-3 FA supplementation decreases urinary 8-iso-PGF 2α concentrations. Previous studies have noted that supplementation is also associated with lower levels of inflammation biomarkers, such as C-reactive protein [16,17]. Because inflammation can lead to a generation of oxidative stress biomarkers, it can be difficult to distinguish which mechanism is impacted by supplementation in these studies. We did not observe associations with PGF 2α in our study population. This compound is produced through enzymatic peroxidation of arachidonic acid by the cyclooxygenase enzymes which are induced under an inflammatory state [31]. Because we observed associations with 8-iso-PGF2α, which is largely produced through chemical lipid peroxidation, but not with PGF 2α , we interpret our findings to indicate that the associations observed between n-3 FA supplementation and 8-iso-PGF2α are attributable to a decrease in oxidative stress rather than inflammation. Furthermore, we examined associations between n-3 FA supplementation and 8-iso-PGF 2α generated from each source, by utilizing a novel ratio of 8-iso-PGF 2α and PGF 2α . These results also showed that n-3 FA supplementation was associated with a change in the chemical fraction of 8-iso-PGF 2α , which indicates a specific association with oxidative stress.
Elevated levels of oxidative stress biomarkers in pregnancy, including 8-iso-PGF 2α , have been associated with adverse birth outcomes such as preeclampsia and preterm labor [4,6]. Thus, theoretically, increased n-3 FA through supplementation could have the potential to mitigate common adverse birth outcomes by reducing oxidative stress levels. However, randomized control trials evaluating the impact of n-3 FA supplements or dietary interventions to increase n-3 FA intake on these outcomes generally report a null effect [56][57][58][59][60][61][62][63]. Examining the association between n-3 FA supplementation and oxidative stress markers may be a more direct way of capturing the biological impact on the body during pregnancy, and could have implications for future studies examining the impact of supplementation on adverse birth outcomes. For example, focusing on this biological mechanism could improve the understanding which of subpopulations (e.g., demographic groups or individuals with co-morbidities] are more or less responsive to the effects of supplementation. It could also provide insight on coexposures or supplements that either exacerbate or detract from the physiologic effects of supplementation.
Our study used a cross-sectional design to assess associations between n-3 FA supplementation and oxidative stress biomarker levels. However, this approach may be ideal for this research question. n-3 FA levels rise quickly within the body upon consumption of n-3 FA supplements [64] and 8-iso-PGF2α is a direct reflection of lipid peroxidation at the time it is measured [25,26]. Therefore, it was important that both exposure and outcome measurements were captured at the same timepoint. It should be carefully noted, however, that associations observed in the third trimester may not be generalizable to the rest of pregnancy. Maternal plasma n-3 FA levels generally decrease throughout gestation as a result of increased maternal transfer of n-3 FAs to the fetus [65]. Therefore, the impact of supplementation on maternal oxidative stress levels may be lesser in this time period than in non-pregnant women or in other trimesters.
Our study had several limitations related to assessing n-3 FA. First, we did not have biomarkers for FA status. Thus, we assessed n-3 FA supplementation by questionnaire, and selfreport could have led to recall bias and consequently exposure misclassification. However, we would such bias to be non-differential with regard to outcome. Recall bias would most likely be nondifferential because questionnaires were administered in close proximity to time of supplementation intake, and participants did not know their oxidative stress levels [66]. Second, we only had information on whether the participant had taken a n-3 FA supplement for one week during the third trimester and not information on which week; supplementation within the last 1-2 weeks would have been the most relevant [67]. Third, we were unable to assess supplement dosage, which could have led to a lack of precision in our effect estimates and limits our ability to comment on whether or not intake was consistent with current n-3 FA dietary recommendations for pregnant women [68]. Fourth, we were unable to assess dietary patterns or consumption of specific foods, such as fish intake, which are also important sources of n-3 FA. While dietary sources of n-3 FA and other micronutrients could influence oxidative stress levels, we would not expect this to confound our results since it is not apparent that n-3 FA supplementation is associated with this type of dietary intake [69]. Lastly, n-3 supplementation is strongly influenced by demographic factors and health behaviors, some of which were unmeasured in TIDES. Thus, residual confounding is possible and our results should be interpreted with caution.
Our study also had many strengths. We utilized 8-iso-PGF 2α and its major metabolite, which are well-established biomarkers of oxidative stress that are direct products of lipid peroxidation and are stable during pregnancy [6,25]. These markers are also easily detectible in urine, minimally influenced by fasting or diurnal fluctuations and stable over pregnancy [6,25]. Additionally, we were better able to assess the mechanism underpinning the association between n-3 FA supplementation and 8-iso-PGF 2α levels by application of the 8-iso-PGF 2α / PGF 2α ratio. Finally, our sizable study population, larger than most other studies assessing this association, drew from 4 diverse city study sites across the US allows for considerable generalizability of our findings.

Conclusions
We observed lower levels of 8-iso-PGF 2α and its primary metabolite in association with n-3 FA intake in pregnancy, which were attributable to decreases in chemical oxidative stress. n-3 FA supplementation may be an easily implemented strategy to decrease maternal oxidative stress during pregnancy; however, our results need to be interpreted with caution as residual confounding is possible. Additional research is warranted as these maternal oxidative stress biomarkers have been linked to adverse birth outcomes and n-3 FA supplements are often affordable and easy to obtain. However, the appropriate dose, clinically relevant benefit, and any adverse effects of supplementation need to be more carefully.
Supporting information S1 Table. Adjusted percent change (95% confidence intervals) in urinary oxidative stress levels in association with omega-3 fatty acid supplement use in the 3 rd trimester of pregnancy stratified by education level. a. Model includes gestational age at sample collection, specific gravity, maternal age, race, education, and study center. Table. Adjusted a percent change (95% confidence intervals) in urinary oxidative stress levels in association with omega-3 fatty acid supplement use in the 3 rd trimester, restricted to women who used prenatal vitamins in 3 rd trimester. a. Model includes gestational age at sample collection, specific gravity, maternal age, race, education, and study center.