Breast milk n-3 long-chain polyunsaturated fatty acids and blood pressure: an individual participant meta-analysis

We performed an individual participant data (IPD) meta-analysis. Using the https://​www.​birthcohorts.​net website, we identified six studies that had collected breast milk samples and had at least one measurement of BP during childhood. All six eligible birth cohorts (COPSAC2000 [26], COPSAC2010 [27], EDEN [28], KOALA [29], OCC [30], PIAMA [31]) agreed to participate. The principal investigators of the studies or their representatives were invited to share data on breast milk fatty acid composition, BP, and covariates of their study. Through the PIs of the earlier mentioned cohorts, one additional study (CU Trial) [32] was identified. This was an intervention study designed to supplement lactating mothers with fish oil, which assessed the resulting effect on breast milk fatty acid composition. The supplement in this trial provided 1.5 g/day of n-3 LC PUFA, which resembled the upper range in the intake from dietary sources in the Danish population. Other studies were observational in nature, or with an intervention that was unrelated to this research question (e.g. mattress covers in PIAMA for prevention of allergies), with the exception of COPSAC2010. In this study, women were randomized to supplementation during the third trimester of pregnancy with either 2.4 g/day fish oil (55% EPA and 37% DHA) or olive oil. Thus, we included seven studies from Denmark, France, and the Netherlands representing a total of 2188 participants. All studies obtained ethical approval from local institutional review boards and have therefore been performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments. The data for this meta-analysis were stored in a single center on a secured data server with access for the first author (LvR) only.

Gas chromatography with flame ionization was used to determine the fatty acid composition of all the milk samples, although there were some differences in sampling, preparation and fatty acid analysis (please see Supplemental Table 1 for details). In general, these methods can produce relatively similar results although the use of acid-based catalysts can result in higher fatty acid recoveries [33], although at an increased risk of degrading of conjugated linoleic acid (a fatty acid of interest in dairy products) [34]. Expressing the fatty acid data as a relative weight percentage (wt%) of the total fatty acid content helps standardize fatty acid data [35], and we focused on DHA and EPA levels as well as DHA/AA ratio and not conjugated linoleic acid.

The outcomes of interest were Systolic blood pressure (SBP) and Diastolic blood pressure (DBP). All studies used an automated oscillatory device to measure BP. The specific devices of each cohort are listed in Supplementary Table 1 of Online resource 1. Protocols were fairly similar, including a 5–10 min rest before measurements, an appropriately adjusted cuff size, measurement on the non-dominant or left arm, and a sitting or supine position. Furthermore, most studies (expect one) took several measurements which were averaged (with one study excluding the first measurement). We used BP in mmHg in the main analyses, adjusting for height, age, and sex in each analysis as BP in children is a function of length/height, age, and sex. In addition, for descriptive purposes, we show BP z-scores to be able to compare BP across ages. z-scores were derived from the Pediatric Hypertension Guidelines, updated from The Fourth Report on the Diagnosis, Evaluation, and Treatment of High BP in Children and Adolescents [36].

All cohorts obtained information on covariates; most covariates were based on self-reported questionnaires and some covariates were derived from medical records. These covariates were maternal age (years), maternal educational level (low, medium, high), maternal place of birth (European, non-European), maternal pre-pregnancy body mass index (BMI) (kg/m²), smoking status during pregnancy (yes, no), infant’s birth weight (grams), and gestational age (weeks) as the main potential confounders in the statistical analysis. All covariates were included a priori in the model, because of their known association with blood pressure.

As the included studies measured BP at different ages, we assumed that statistical heterogeneity was present. Therefore, we applied a random effects meta-analysis which estimates the average association between the determinant and the outcome [37, 38]. We performed a two-stage meta-analysis. In the first stage, we ran two models for each study. Model 1 included the association between fatty acid level (DHA, EPA, or the DHA/AA ratio) and BP, adjusted for age, sex, and height. For this model, we fitted several models for height, by adding “height²” and “height³” to the model, and ran the model with the best fit for height, as determined by the -2log likelihood. For most studies, height was used as a single term. Six out of the seven included studies had repeated measurements of BP. For these studies a mixed model was used, to account for the intra-cohort correlations between the measurements. For the study with the single outcome measurement (i.e. KOALA), we used a linear regression model. In addition to height, age and sex, model 2 was further adjusted for the above-mentioned potential confounders and breastfeeding duration. In the second stage, the estimates of each study were pooled with a mixed model, where each cohort was a random effect, and each cohort was assigned a weighting factor (1/variance). Estimates represent the difference in BP (mmHg) per 0.1 wt% increase in breast milk fatty acids levels, which would be a realistic increase given the range of fatty acids in the studies.

From studies in adults on the association between fatty acids and BP, it has been suggested that the association may not show a dose–response relationship [39]. Therefore, we also examined associations with BP with breast milk DHA dichotomized at the median for each study.

We examined the interactions between breast milk and child sex and age in relation with BP by adding the appropriate interaction terms (e.g. DHA × sex) into the models.

As child’s height and child’s BMI could be an intermediate or modifying factor in the association between DHA and BP, given evidence in the literature on an association between DHA and height [40, 41] and DHA and BP in overweight children [42], we performed a sensitivity analysis where we used height and BMI as the outcome instead of BP. For those studies that had an intervention component by design (related or unrelated), we checked whether the associations were similar for the intervention and control arms.

Data analysis was conducted with SAS software version 9.4 (SAS Institute, Inc., Cary, NC, USA).

This individual participant meta-analysis of seven studies representing 2188 children with BP assessment in the full range from infancy to adolescence, observed no overall association between systolic or DBP and increasing levels of breast milk DHA ranging from mean of 0.22–0.76 wt%. There were also no associations between BP and increasing levels of EPA or of DHA/AA ratio in breast milk, and no signs of age or sex differences in the associations.

Our result is in line with four earlier publications that reported no association between early life n-3 LC PUFAs intake and BP in childhood [17, 19, 21, 23], with an age at outcome measurement ranging from 2.5 to 9 years. These studies directly supplemented infants (280 mg/day DHA as ethyl-ester form from birth to 6 months) [19] or infant formula with n-3 LC PUFA (0.30 wt% DHA as triglyceride and phospholipid forms from birth to 2 months) [17], and one of the studies [21, 23] supplemented lactating mothers with fish oil. Conversely, two other studies reported that n-3 LC PUFAs were associated with a lower BP [16, 18]. One of which was a multicentric European RCT that supplemented infant formula with 0.15–0.25 wt% of DHA (as egg phospholipid form) found a lower BP in 6-year-old children (− 3.6 mmHg for DBP; not significant for SBP) [18]. The second one was our previous observational study, where we found that children who received breast milk with a DHA ≥ 0.18 wt% had a lower BP (− 4.79 mmHg for SBP and − 2.47 mmHg for SBP) at the age of 12 compared to children who received infant formula with no DHA (with adjustment for child’s sex) [16]. Another report found a lower BP with higher mother’s milk n-3 LCPUFAs (ranging 0.24–0.47 wt%, with 75% being DHA) but only in boys (SBP and DBP) at 4 months of age, then only in girls (SBP) at 18 months of age, and no association regardless the infant sex at 36 months of age [43]. One study found a higher BP in boys whose mothers were supplemented with fish oil during lactation at the age of 7 years [20], with replication of the findings at the age of 13 [22]. Our meta-analysis did not show clear sex differences for the association between DHA and BP, though the associations seemed to be more beneficial for boys, especially in early age and adolescence. The differences in the associations between n-3 LC PUFA and BP in the earlier studies cited herein does not seem to be related to study design, age at BP assessment, method of administration, or level of n-3 LC PUFA. This meta-analysis adds to the literature by including data from several unpublished studies with data on breast milk DHA, and by harmonizing the data to enhance comparability. Relevant issues to consider in future studies are potential interference from other components in infant feeding or lifestyle factors, the latter being important to study in the context of an RCT design.

The strength of our study is that we included all eligible studies, irrespective of whether these studies had published the results. Moreover, most studies did not have the association of interest in this paper as a primary research question, which prevented publication bias [44]. Also, the studies included in the meta-analysis had a wide range of breast milk DHA, and our associations represent increments that can be expected by supplementing women with the recommended dose of DHA [45, 46]. However, our study has some limitations that should be taken into account when interpreting the results. First, all cohorts had loss-to-follow-up in the period between breast milk collection and BP measurements. This could have led to selection bias if those with a low or high DHA level and a low or high BP were selectively lost to follow up. Second, when converting the BPs to z-scores by the percentiles of “The Fourth Report on the Diagnosis, Evaluation, and Treatment of High BP in Children and Adolescents”, we noticed relatively high BP z-scores in all studies included in this meta-analysis, but especially in younger children. This is possibly related to the method of measurements (e.g. number of repeats) as well as the difficulty of measuring BP in young children. Although this would not have affected validity, associations between DHA and BP could be less precise in younger children. Indeed, absolute BPs were comparable with an earlier study investigating the association between DHA and BP in children, but SEs were smaller in this study, and statistically significant associations were detected [42]. Third, most studies used a single sample of breast milk to determine the amount of breast milk n-3 LCPUFAs, which may have led to misclassification of the exposure as breastmilk n-3 LCPUFA vary from day to day based on the mothers recent intake of fish [47]. Thus, the latter two limitations may have prevented us to find a real association between n-3 LC PUFAs and BP, if present.

The results reported in our study are per 0.1 wt% increase in breast milk fatty acids, which is a relatively small increase, and results in a small potential change in BP which is not clinically relevant. Though statistical significance and precision remains unchanged, we illustrate implications for clinical relevant changes in BP if breast milk DHA increases to a larger, but feasible, extent. Per SD increase in DHA, DHA is associated with a 3.24 mmHg lower (95% CI − 8.59, 2.10) SBP, and a 1.11 mmHg lower (95% CI − 5.68, 3.46) DBP. If breast milk DHA increases with 0.8 wt% (which is the increase that was measured after supplementing women with fish oil during lactation), DHA is associated with a 9.52 mmHg lower (95% CI − 26.51, 7.47) SBP, and a 3.50 mmHg lower (95% CI − 20.59, 13.58) DBP.

Although hypertension is rare in childhood [48], BP tends to track from childhood to adulthood from a very young age [49]. Therefore, the less increased the BP in childhood, the lower the risk of hypertension in adulthood, which implies that prevention of hypertension should start early in life. Our results do not support a beneficial effect of breast milk n-3 LC PUFAs on childhood BP. However, the current global recommendations for pregnant and lactating women of 200 mg/day of DHA intake (obtained for example by one portion of oily fish a week) still apply for other health outcomes [50].