Plasma long‐chain omega‐3 fatty acid status and risk of recurrent early spontaneous preterm birth: a prospective observational study

A 2018 Cochrane review found that omega‐3 supplementation in pregnancy was associated with a risk reduction of early preterm birth of 0.58; prompting calls for universal supplementation. Recent analysis suggests the benefit may be confined to women with a low baseline omega‐3 fatty acid status. However, the contemporary omega‐3 fatty acid status of pregnant women in the UK is largely unknown. This is particularly pertinent for women with a previous preterm birth, in whom a small relative risk reduction would have a larger reduction of absolute risk. This study aimed to assess the omega‐3 fatty acid status of a UK pregnant population and determine the association between the long‐chain omega‐3 fatty acids and recurrent spontaneous early preterm birth.


| INTRODUC TI ON
Globally, preterm birth is the leading cause of death in children under 5 years old. 1 Previous preterm birth is the strongest risk factor for subsequent preterm delivery. 2 A 2018 Cochrane review concluded that omega-3 supplementation was an effective strategy to prevent preterm birth, with a 42% risk reduction (from 46 to 27 per 1000 births; 95% CI, 23-56) for preterm birth less than 34 weeks. 3 A subsequent randomized controlled trial 4 with secondary analysis 5 suggested that the benefit may be confined to women with a low baseline total long-chain omega-3 fatty acid level. Worryingly, within the secondary analysis, 5 supplementing women with higher total long-chain omega-3 fatty acid status was associated with increased rates of early preterm birth.
The fatty acid components with the strongest evidence of preterm birth prevention are eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA), 3 which are collectively referred to as long-chain omega-3 fatty acids. These nutrients are predominantly obtained from oily fish and seafood and are associated with a more affluent diet. The long-chain omega-3 intake in pregnancy in the UK has been estimated from food frequency questionnaires as low, 6,7 or adequate 8 in three studies between 1991 and 2007.
Liverpool Women's Hospital has a tertiary referral preterm birth prevention clinic that serves the fourth most deprived local authority area in England (out of 343). 9 Based on the Cochrane review findings 3 we offered omega-3 supplementation to these high-risk women from February 2019. 10 However, we were unsure whether this would offer benefit because of the unknown baseline long-chain omega-3 status in our population. Plasma levels of omega-3 in the UK pregnant population have not been assessed to our knowledge.
The Danish National Birth Cohort 11,12 showed that women in the lowest quintile of plasma EPA+DHA (<1.42% of total fatty acids), in the second trimester, had a 2.13 times increased risk (95% CI 1.18-3.79) of spontaneous preterm birth (sPTB) before 34 weeks of gestation compared with women in quintiles 3-5. The association between omega-3 and preterm birth was not present with levels in the third quintile and above. This is consistent with Simmonds et al 5 and suggests that the main benefit of supplementation is in pregnancies with a lower baseline long-chain omega-3 status.
Importantly there has been a recent corrigendum 12 to the original research within the Danish National Birth cohort 11 based on the effect of thawing of stored samples before analysis of long-chain omega-3 fatty acids; this was therefore addressed within our analysis too.
We had two objectives. First, to determine the expected distribution of long-chain omega-3 fatty acids within "healthy pregnancies" in our locality; low-risk pregnant women who delivered at ≥39 weeks without preterm prelabor rupture of membranes (PPROM). Second, to assess the relation between long-chain omega-3 status and recurrent early (under 34 +0 weeks) sPTB and PPROM in our region.

| MATERIAL AND ME THODS
Women with singleton pregnancies were enrolled at Liverpool Women's Hospital from April 1, 2012 until December 31, 2017 as part of the "Development of novel biomarkers for prediction of preterm labor in a high-risk population" study. Participants were invited to two visits at approximately 16 (15 +1 -18 +6 weeks) and 20 (19 +0 -23 +0 ) weeks of gestation. For the purposes of this analysis, the first sample available was used (single sample per participant).
A flowchart of selection entry from two different obstetric populations is shown in Figure 1. A "high-risk" population consisted of women with a history of sPTB or PPROM at 16 +0 -33 +6 weeks of gestation. Low-risk women were parous women with all previous births at ≥37 +0 weeks of gestation. Full details of the recruitment process, inclusion criteria, and careful pregnancy outcome classification criteria are given in the Supporting Information Appendix S1.
Participants were excluded from the statistical analysis if omega-3 supplements had been used in pregnancy.
To describe the expected distribution of omega-3 fatty acids in our population, low-risk women that delivered at ≥39 +0 weeks were selected (low-risk population sample).
Recurrent early sPTB/PPROM was defined as high-risk participants who had a late miscarriage, PPROM, or sPTB at 16 +0 -33 +6 weeks of gestation. High-risk women who gave birth at ≥37 +0 weeks of gestation without PPROM were allocated to the high-risk term birth group. no association with risk of recurrent early spontaneous preterm birth. This could be because our population levels were too low to show benefit in being omega-3 "replete"; or else omega-3 levels may be of lesser importance in recurrent early preterm birth.

| Omega-3 fatty acid analysis
Maternal blood samples were taken in 10 mL BD vacutainer ® tubes (Becton Dickinson) containing dipotassium ethylenediaminetetraacetic acid, placed on ice immediately and processed within 1 hour of sampling. Tubes were centrifuged at 1200 × g for 10 minutes at 4°C.
Plasma was aspirated and stored in cryovials at −80°C. A total of 30 µL was transferred to blood spot cards that were coated in antioxidants and chelating agents so as to minimize oxidation of polyunsaturated fatty acids. 13 The dried blood spot cards were transported by post to the South Australian Health and Medical Research Institute where the plasma spots were transesterified and distributions of fatty acids were determined by capillary gas chromatography. 13 The laboratory team was blinded to the pregnancy status of the samples. The distributions of long-chain omega-3 fatty acids within the lowrisk population sample were calculated and used to define quintiles of total omega-3, DHA, and EPA for our population.

| Statistical analyses
Histograms were used to show the distribution of long-chain omega-3 fatty acids within the high-risk group according to whether the participant did, or did not, have recurrent early sPTB/PPROM. Twoterm fractional polynomials were then used to visualize the expected non-linear association between long-chain omega-3 fatty acids levels and risk of recurrent early sPTB/PPROM within the high-risk group.
High-risk participants in the early sPTB/PPROM and high-risk term birth groups were assigned to the quintiles based on the distribution of total omega-3, DHA, and EPA within the low-risk population sample and to the quintiles described by Olsen et al. 11,12 Binomial logistic regression was used to calculate the odds ratios of early sPTB/PPROM compared with term birth per quintile. Quintiles 3-5 were combined and used as the reference group based on previous work. 11,12 Analysis was performed unadjusted and adjusted for covariates that were selected based on biological plausibility. The chosen covariates were: maternal age at study participation; maternal body mass index; maternal smoking at time of study visit (binary outcome of yes/no); and index of multiple deprivation (IMD). Age and body mass index were converted to quadratic terms because of the bimodal relations between these variables and risk of preterm birth. IMD was obtained using the woman's home postcode on the UK government website. 14 The IMD ranks every neighborhood in England from 1 (most deprived) to 32844 (least F I G U R E 1 Participant selection. Two obstetric populations were used to recruit women. The first was women at high-risk of sPTB based on their history of previous sPTB. The second population was used to represent "normality" and consisted of women with a history of term birth only. Two stages of analysis were performed. The first was to visualize the relation between long-chain omega-3 status and the occurrence of sPTB or PPROM under 34 weeks in the high-risk cohort. The second analysis used etiological modeling to assess the contribution of long-chain omega-3 to recurrent early preterm birth, for this analysis a clear "split" in the preterm and term cases was desired, and so births 34 +0 -36 +6 weeks were excluded from this analysis. "Cases" consisted of women with recurrent sPTB or PPROM <34 +0 weeks of gestation. sPTB, spontaneous preterm birth, PPROM, preterm prelabor rupture of membranes deprived). 9 The IMD is a collective score summarizing income deprivation, employment deprivation, health deprivation and disability, education skills and training deprivation, barriers to housing and services, living environment deprivation, and crime. IMD scores were used as continuous variables within the logistic regression.
Adjusted odds ratios for early sPTB/PPROM are presented both for the participants with all covariates available, and for all participants using imputation for missing covariates. Multiple imputation using chain equations was used to account for missing data as this allows for binary covariates (such as smoking). The proportion of total sampling variance due to missing data for IMD was 48%, therefore, as recommended, 50 imputations were performed. 15 The variables used in the imputation model were all of the covariates described above as well as: pregnancy outcome (birth at term or early sPTB/ PPROM); quintile of total omega-3, EPA, DHA, and DHA plus EPA; and quintile according to Olsen et al. 11,12 No auxiliary variables were identified.
During the course of this work concern was raised that previous thawing may alter plasma long-chain omega-3 fatty acid analysis. 12 We therefore undertook three further pieces of statistical analysis. First, assessment was performed of the long-chain omega-3 fatty acid status by number of previous freeze-thaw cycles of the sample. Second, the number of freeze-thaw cycles was included as a covariate in the logistic regression described above.
Finally, the binomial logistic regression was repeated using only samples that had undergone pevious freeze-thaw cycles, and those without.

| Ethical approval
The study was approved by the North West Research Ethics Committee, Liverpool Central, reference 11/NW/0720 on November 4, 2011. All participants provided written informed consent.

| RE SULTS
We recruited 296 high-risk participants and 271 low-risk participants. Of 283 high-risk women with data suitable for analysis, 51 (18%) had a recurrent early sPTB or PPROM and 178 (63%) had term births (≥37 weeks) without PPROM ( Figure 1). Of 271 low-risk participants, 188 gave birth at ≥39 +0 weeks without PPROM, and had samples suitable for analysis. We selected the first 100 of these participants to send samples for laboratory analysis. Four of these participants were subsequently noted to have used omega-3 supplementation, and so the remaining 96 participants formed the low-risk population sample.
The baseline characteristics and pregnancy outcomes are broadly similar across the pregnancy groups (Table 1), except for known risk factors for sPTB/PPROM. Compared with the low-risk population sample, more of the high-risk participants smoked (9.7% of lowrisk vs 24.0% of high-risk group), and the high-risk participants had slightly lower IMD scores (more social deprivation). Preterm birth prevention treatment was offered in accordance with UK national guidelines. 16 None of the low-risk women required an intervention but 32.8% (93/283) of the high-risk women did.
The low-risk population sample was used to define the expected distribution of omega-3 fatty acid levels in our population ( Table 2).
Levels of total omega-3, DHA, and EPA within the high-risk group show similar distributions in women who have an early sPTB/ PPROM, and those who do not (Figure 2A-D). The risks of recurrent early sPTB by total omega-3, DHA, and EPA levels are visualized in Figure 2E-H. Visually, it appears that there could be a weak relation between higher levels of EPA, DHA, and total omega-3 and preterm birth, but the wide confidence intervals are also consistent with no correlation.
The high-risk group was then split into three groups according to total omega-3, DHA, and EPA quintiles obtained from the lowrisk population sample: quintiles 1, 2, and 3-5 (reference group) ( although none of these differences reached conventional statistical significance (p < 0.05) ( Table 3).
We performed the same analysis adjusting for covariates of smoking, maternal age, body mass index, and IMD, both restricting the analysis to participants with all variables available and using multiple imputation to account for missing variables (Table 3). These results also showed no association between long-chain omega-3 fatty acids and early sPTB/PPROM, and the non-significant trend towards higher risk of preterm birth with higher levels.
Omega-3 fatty acid levels were universally lower in our population than in the Danish National Birth Cohort 11,12 (  (Table 4).
Before our analysis, samples from 17% (16/96) of our low-risk population sample, 9.0% (16/178) of our high-risk term birth group, and 57% (29/51) of our high-risk early sPTB/PPROM groups had undergone three freeze-thaw cycles (Table S1). The remainder of samples had undergone no previous freeze-thaw cycles. We found no statistically significant difference in DHA, EPA, or DHA+EPA levels when comparing samples with and without previous freeze-thaw cycles, but within the high-risk reference group there was a trend for slightly lower omega-3 fatty acid levels in samples that had undergone previous freeze-thaw cycles. When the logistic regression described in Table 3 was repeated in samples both without previous freeze-thaw cycles (Table S2) and with previous freeze-thaw cycles (

| DISCUSS ION
Contrary to previous findings, we did not demonstrate a relation between long-chain omega-3 levels and spontaneous preterm birth.
This was despite comparing plasma total omega-3, DHA, and EPA levels with both "healthy" pregnancies in our population, and to levels in Danish pregnant women that have previously been associated with preterm birth. 11,12 In our population, both women at high and low risk of preterm birth had lower levels of plasma DHA plus EPA than those described in the Danish population. 11,12 The plasma long-chain omega-3 levels within our population could have been so low that we did not have enough "replete" participants to show the benefit in preterm birth reduction with adequate levels. However, our results show a non-significant trend in the opposite direction to previous literature (ie, a higher risk of preterm birth with a higher level of omega-3, DHA, and EPA), and no biological gradient.
Women with a previous preterm birth are often highly motivated to avoid recurrence, and could have become aware of evidence to support increased omega-3 intake 17,18 during their pregnancy.
Omega-3 fatty acids may have a rapid effect on risk of preterm birth. 19,20 If a substantial number of the women in our study actually did increase their omega-3 fatty acid intake during pregnancy, this may have confused the relation between omega-3 measured in early second trimester and subsequent risk of early preterm birth.
To our knowledge this is the fourth analysis relating blood DHA and EPA levels in the second trimester to preterm birth risk. Previous A strength of this study is that our preterm group included only recurrent sPTB, or PPROM, before 34 +0 weeks of gestation. We aimed to achieve as pure a "phenotype" of spontaneous preterm birth as possible. Previous studies into the association between omega-3 levels and preterm birth have included all births under 34 weeks, 5 or 37 weeks, 21 or only excluded cases of preeclampsia before 34 weeks. 11 It is possible that the benefit of omega-3 to prevent preterm birth is confined to preterm births that are medically indicated by conditions such as preeclampsia and growth restriction. 22 However, the most recent Cochrane review shows no impact of omega-3 supplementation upon these conditions. 3 In keeping with this, our initial visualization of the relation between longchain omega-3 status and early preterm birth in the whole high-risk group, including those with late medically indicated preterm births ( Figure 2), did not show an association between long-chain omega-3 levels and all preterm births. Alternatively the impact of omega-3 upon preterm birth prevention may be within the low-risk population that was not assessed for preterm birth risk in this study.
A limitation of this study is that 56% of sPTB/PPROM samples had undergone previous freeze-thaw cycles, compared with only 9% of the high-risk reference group. However, we found higher than expected levels of omega-3 fatty acids in the sPTB/PPROM group, and freeze-thaw cycles might be expected to lower the expected levels of omega-3 fatty acids. 12 We therefore do not feel that this has materially impacted our findings.
TA B L E 2 Normal distribution of plasma long-chain omega-3 fatty acids in the low-risk population sample Note: Women were parous, with all previous births at term and birth ≥39 +0 weeks in the index pregnancy.

Mean (SD) Range
Abbreviations: DHA, docosahexaenoic acid; EPA, eicosapentaenoic acid; SD, standard deviation. All values are percentage of the total plasma fatty acids. We acknowledge that plasma levels of omega-3, DHA, and EPA

F I G U R E 2
were measured on samples from participants that were not fasted; however, the previous study to find an association between plasma levels of DHA and EPA and preterm birth used samples taken by GPs at routine visits and no mention is made of fasting in the description. 11,23 This was a pragmatic study based on a biomarker study that had finished recruiting at the time of study inception. As such, no formal power calculation has been performed, and we did not have a predefined a priori level at which we are able to accept/reject our null hypothesis of no association between long-chain omega-3 levels and recurrent spontaneous early preterm birth. Nevertheless, we feel that knowledge of the low baseline levels of long-chain omega-3 fatty acids within pregnant women in the UK, and also no indication of an association between long-chain omega-3 fatty acids TA B L E 3 Relation between quintile of long-chain omega-3 (as defined by the low-risk population sample) and pregnancy outcome in the high-risk group. recurrent preterm births are able to "overpower" any contribution of long-chain omega-3 status. We suggest that future research should include baseline long-chain omega-3 fatty acids testing on a large scale, and evaluate the influence of these levels on other risk factors of preterm birth. This would be relevant to both women with, and without, identifiable risk factors for preterm birth, and may be achieved by an individual patient data meta-analysis of already conducted work.

| CON CLUS ION
We found low plasma omega-3, DHA, and EPA levels in the second trimester in women at high and low risk of preterm birth. The previously described association between low DHA and EPA and preterm birth was not replicated. We suggest that either plasma long-chain omega-3 fatty acids were so low in this population we did not have enough "replete" participants to show a benefit, or there are alternative mechanisms for recurrent early preterm birth in this setting.

ACK N OWLED G M ENTS
We thank members of the Harris-Wellbeing Patient and Public Engagement group. We would also like to thank Mrs Tracy Ricketts for administrative support with the study, and the Liverpool Women's Hospital for hosting the research.

CO N FLI C T O F I NTE R E S T
LG has received study support grants from Wellbeing of Women charity for this research; AC has received salary and study support