Throughout the world, large ethnic disparities in the birth-weight distribution can be observed, with the lowest birth weights and highest proportion of intra-uterine growth restriction usually being found in minority populations(Reference Troe, Raat and Jaddoe1–Reference Shiono, Rauh and Park3). Existing evidence of nutritional differences between ethnic groups(Reference Suitor, Gardner and Feldstein4–Reference Rees, Doyle and Srivasta7) suggests that maternal nutrition may be one of the factors contributing to these disparities.
Among the nutritional factors considered relevant to fetal growth, the maternal n-3 and n-6 long-chain PUFA EPA (20 : 5n-3), DHA (22 : 6n-3), dihomo-γ-linoleic acid (DGLA; 20 : 3n-6) and arachidonic acid (AA; 20 : 4n-6) have increasingly gained interest in the past few decades. DGLA, AA and EPA are precursors of the PG 1, 2 and 3 series respectively, a series of hormone-like substances involved in a range of pregnancy-related processes that include placental blood flow, cervix-ripening and initiation of parturition(Reference Uauy, Treen and Hoffman8, Reference Allen and Harris9). DHA, as well as AA, are major components of neural tissue in particular(Reference Uauy, Calderon and Mena10, Reference Innis11).
Recently, we have shown that the maternal n-3 long-chain PUFA status (reflected by maternal plasma phospholipid n-3 long-chain PUFA concentrations) in early pregnancy was positively associated with fetal growth(Reference van Eijsden, Hornstra and van der Wal12), which was confirmed, for DHA, in a different cohort study(Reference Dirix, Kester and Hornstra13). For the n-6 long-chain PUFA AA and DGLA contrasting results were observed, with AA being negatively(Reference van Eijsden, Hornstra and van der Wal12, Reference Dirix, Kester and Hornstra13) and DGLA being positively associated with birth weight(Reference van Eijsden, Hornstra and van der Wal12). While this evidence suggests a role for the n-3 and n-6 fatty acids in the ethnicity-related disparities in perinatal health, few studies have investigated ethnicity-related differences in maternal fatty acid status. The present study is the first large-scale observational study to explore the within-country ethnic differences in maternal early pregnancy n-3 and n-6 fatty acid status (with a focus on the major functional long-chain PUFA DGLA, AA, EPA and DHA), taking into account maternal factors relevant to fatty acid metabolism(Reference Maniongui, Blond and Ulmann14–Reference Pawlosky, Hibbeln and Wegher20) and dietary intake (i.e. fish and fish oil consumption)(Reference van Houwelingen, Sørensen and Hornstra21–Reference Williams, Frederick and Qiu24). Data were derived from a large, unselected multi-ethnic cohort in Amsterdam, the Netherlands. In this population, considerable ethnic disparities in birth weight have been observed that could not be explained by conventional physiological and environmental risk factors, such as maternal age, height, parity, BMI and smoking habits(Reference Goedhart, van Eijsden and van der Wal25).
Methods
Study population and design
The Amsterdam Born Children and their Development (ABCD) study is a prospective community-based cohort study that examines the relationship between maternal lifestyle and psychosocial conditions during pregnancy and the child's health at birth as well as in later life. The essentials of the study design have been described previously(Reference van Eijsden, Hornstra and van der Wal12, Reference van Eijsden, van der Wal and Bonsel26). In short, between January 2003 and March 2004, all pregnant women living in Amsterdam were invited to enrol in the ABCD study during their first prenatal visit to their obstetric care provider (about the 12th week of gestation). They were requested to complete a questionnaire, covering sociodemographic data, obstetric history, lifestyle, dietary habits and psychosocial factors. The questionnaire was available in Dutch as well as in English, Turkish and Arabic for immigrant women. In addition, women were invited to participate in the ABCD biomarker study. For this study, an additional blood sample was taken during routine blood collection for screening purposes following the first prenatal check-up.
Of the 12 373 pregnant women invited to participate, 8266 returned the pregnancy questionnaire (response rate 67 %). Of these respondents, 53 % (n 4389) participated in the biomarker study. Approval of the study was obtained from the Medical Ethical Committees of participating hospitals and the Registration Committee of Amsterdam, and participants gave written informed consent.
Blood collection and analytical methods
For each participant of the biomarker study, a blood sample was taken in a 10 ml EDTA (K2) Vacutainer (Becton and Dickinson BV, Alphen aan de Rijn, the Netherlands) and sent to the Regional Laboratory of Amsterdam for processing. The samples were sent by courier or overnight mail in special envelopes, enabling processing within 28 h of sampling. A previous study of our group demonstrated that this delay did not compromise the validity of the biomarkers measured(Reference van Eijsden, van der Wal and Hornstra27). At the laboratory, plasma was prepared by centrifugation (1600 g for 10 min at room temperature) and stored as 1 ml samples at − 80°C until analysis.
Fatty acid analysis was performed at the Analytical Biochemical Laboratory (Assen, the Netherlands) using a previously described methodology(Reference Al, van Houwelingen and Kester28, Reference Otto, van Houwelingen and Hornstra29). In short, after the addition of an internal standard (1,2-dinonadecanoyl-sn-glycero-3-phosphocholine) and 10-heptadecenoic acid (17 : 1) to check for carry-over of NEFA during the isolation procedure, plasma lipid extracts were prepared by a modified Folch extraction method(Reference Hoving, Jansen and Volmer30) and phospholipids were isolated by solid-phase extraction on aminopropyl-silica columns (500 mg/3 ml; Varian, Palo Alto, CA, USA)(Reference Kaluzny, Duncan and Merritt31). Subsequently, the phospholipids were hydrolysed and the resulting fatty acids methylated with boron trifluoride–methanol(Reference Morrison and Smith32). Finally, the fatty acid methyl esters were quantified by capillary GC with flame ionisation detection (HP5890 series II; Hewlett Packard, Palo Alto, CA, USA), using both a polar and a non-polar column (BPx70 and BP1 respectively; SGE Analytical Science Pty Ltd, Ringwood, Victoria, Australia). The oven temperature was programmed to begin at 160°C for 4 min, and then to increase to 200°C by 6·0°C/min. After 3 min, the temperature was further increased to 260°C at a rate of 7°C/min, and kept constant for 2·34 min. The injector temperature was kept at 250°C and the detector temperature at 300°C.
Absolute amounts of fatty acids (mg/l plasma) were quantified on the basis of recovery from the internal standard and calculated in relative values (percentage of total plasma phospholipid-associated fatty acids). The present study focused only on the n-3 and n-6 fatty acids and included the n-3 fatty acids α-linolenic acid (18 : 3n-3), eicosatetraenoic acid (20 : 4n-3), EPA (20 : 5n-3), docosapentaenoic acid (22 : 5n-3) and DHA (22 : 6n-3), and the n-6 fatty acids linoleic acid (18 : 2n-6), DGLA (20 : 3n-6), AA (20 : 4n-6), adrenic acid (22 : 4n-6) and Osbond acid (22 : 5n-6). Most fatty acids were determined by analysis using the apolar column; for 18 : 3n-3, 20 : 5n-3 and 22 : 5n-6, data obtained with the polar column were used. Inter-assay CV for the fatty acids varied from ≤ 22 % for 20 : 4n-3 (the fatty acid with the lowest relative concentration) to ≤ 2 % for 18 : 2n-6 (the fatty acid with the highest relative concentration). Stearidonic acid (18 : 4n-3) and γ-linolenic acid (18 : 3n-6) were not included, as their concentrations were < 0·1 % of total fatty acids.
Questionnaire measurements
The pregnancy questionnaire provided information on ethnic group as well as covariables considered relevant to maternal fatty acid status. We defined ethnic group by the woman's country of birth or that of her mother, so that second-generation women could also be included(Reference Goedhart, van Eijsden and van der Wal25). In total, seven ethnic groups were defined: Dutch (reference group), Surinamese, Antillean, Turkish, Moroccan, Ghanaian, and other ethnic origin.
The covariables included maternal age (years)(Reference Maniongui, Blond and Ulmann14), parity (0, 1, ≥ 2)(Reference Al, van Houwelingen and Hornstra15, Reference Levant, Ozias and Carlson16), educational attainment (number of years after primary school)(Reference Sontrop, Campbell and Evers17), pregravid BMI (kg/m2) as based on self-reported height and weight(Reference Warensjö, Örhvall and Vessby18), smoking and alcohol consumption during pregnancy (self-reported previous week's behaviour, recoded into yes, no)(Reference Simon, Fong and Bernert19, Reference Pawlosky, Hibbeln and Wegher20) and fish and fish oil consumption(Reference Hjartåker, Lund and Bjerve22, Reference Williams, Frederick and Qiu24). For height and weight, missing values (3·3 and 9·8 % respectively) were imputed by means of a random imputation procedure using linear regression(Reference Allison33), which accounted for the differences among the ethnic groups. Fish and fish oil consumption was assessed by four frequency questions adapted from the Danish fish frequency questionnaire of Olsen & Secher(Reference Olsen and Secher34). Women were asked to indicate how often during their pregnancy they had eaten (a) fish as part of a hot meal, (b) bread or toast with fish, (c) salad (such as a green salad or pasta salad) with fish, and (d) fish oil. Predefined response categories were: never, less than once per month, 1–3 times per month, 1–2 times per week, 3–6 times per week, and every day, assuming to correspond to 0, 0·5, 2, 4, 20 and 28 servings per 28 d respectively. To avoid uncertainties associated with differences in the average n-3 long-chain PUFA contents of various fish species and consequently with the calculated average n-3 long-chain PUFA intake(Reference Racine and Deckelbaum35), we computed a cumulative frequency measure to define overall fish consumption. This cumulative (or total) number of fish and fish oil servings per week was arrived at by simple combination of the responses on the four questions. Subsequently, we defined three frequency groups in such a way that each group was of a reasonable size and that intake frequency increased progressively: (1) < 1 serving per week, (2) 1–1·9 servings per week, and (3) 2 or more servings per week.
Statistical analysis
Fatty acid results were available for 4336 of the 4389 participants. We excluded all respondents with known diabetes (n 26), hypertension (n 152) or unknown gestational age at blood sampling (n 15), as well as all respondents with missing information on fish or fish oil consumption (n 25) or on any of the above-mentioned covariables (n 15). Restriction to the six main ethnic groups (Dutch, Surinamese, Antillean, Turkish, Moroccan and Ghanaian) provided the final sample for the analysis of 3254 subjects.
First, differences in the distribution of relevant maternal characteristics and the habitual fish and fish oil intake between the Dutch and ethnic minority groups were described and tested with the χ2 test for categorical variables and ANOVA for continuous variables. Second, ethnic differences in relative fatty acid concentrations were explored using the general linear model function in SPSS (SPSS, Inc. Chicago, IL, USA), which tests both the overall association between ethnic origin and fatty acid concentrations as well as the fatty acid-specific ethnic differences(Reference Tabachnik and Fidell36). Finally, the ethnic differences in the proportion of the major functional long-chain PUFA (EPA, DHA, DGLA and AA) in maternal plasma phospholipids were further explored, by stepwise extension of the basic general linear model (model 1, including only ethnicity). In the first step, we included gestational age at blood sampling (based on ultrasound or, if unavailable, on the time of the last menstrual period) to assess the influence of changes in the fatty acid contents of plasma phospholipids that normally occur during pregnancy(Reference Al, van Houwelingen and Kester28) (model 2). In the next step, we added the potential confounders (model 3) and, in the final step, we included the fish frequency measure (model 4).
We used the model-estimated means to calculate the relative (i.e. percentage) differences between the ethnic minority groups and the Dutch reference group and assumed the changes in relative differences across the models to indicate the contribution of the covariables to these differences. When necessary, transformations were applied to obtain more symmetrical distributions and improve the normality of the residuals in the various models (18 : 3n-3, 20 : 4n-3, 22 : 5n-6, square root transformation; EPA (20 : 5n-3), log transformation). For 22 : 4n-6, 22 : 5n-6, 18 : 3n-3, 20 : 4n-3 and EPA, the measurements included zero values ( < 0·5 % of cases), which were replaced by half of the value of the lowest measured value.
Because of the multiple comparisons made, associations were considered significant at P < 0·01. All analyses were conducted in SPSS (version 15.0; SPSS, Inc.).
Results
In the present analysis, 25 % of the study population belonged to a non-Dutch ethnic group. The baseline characteristics of the ethnic groups are presented in Table 1. Overall, non-Dutch women had their prenatal check-up including the blood sampling at a later gestational age than the Dutch women ( ≥ 14·4 v. 12·9 weeks). Between the ethnic groups, significant differences existed in the distribution of maternal age, parity, educational attainment, BMI, smoking habits, alcohol consumption and fish consumption (P < 0·001). Dutch women were generally older and more highly educated than women of non-Dutch background; on average, Turkish women were the youngest mothers and Ghanaian women the least educated mothers. Antillean and Dutch women were more often nulliparous and had a lower BMI on average. Women from Ghana were more often multiparous (i.e. parity ≥ 2) and had the highest BMI. Turkish women reported smoking during pregnancy more often than Dutch or other non-Dutch women, while alcohol consumption was most common among Dutch women. Fish consumption, lastly, was similarly low among Dutch, Surinamese and Antillean women; in these groups 17 % of women reported consuming a serving of fish or fish oil at least twice per week. Consumption was lower among Turkish women (13 %) and higher among Moroccan women (33 %). The highest consumption was found among Ghanaian women, of whom 56 % reported consuming a serving of fish or fish oil at least twice per week.
(Mean values and standard deviations or percentages)
Mean value or percentage was significantly different from that of the Dutch ethnic group: *P < 0·01, **P < 0·001.
† Test for differences between groups: χ2 statistic for categorical variables; ANOVA for continuous variables.
Table 2 shows the fatty acid composition in maternal plasma phospholipids for each ethnic group separately; for those fatty acids with a skewed distribution, the median and interquartile range are given instead of the mean and standard deviation. For both n-3 and n-6 fatty acids, considerable ethnicity-related differences were observed (Pillai's trace criterion for multiple outcomes; P < 0·001).
(Mean values and standard deviations or medians and interquartile ranges for skewed distributions)
DGLA, dihomo-γ-linolenic acid; AA, arachidonic acid.
Mean or median value was significantly different from that of the Dutch ethnic group: *P < 0·01, **P < 0·001.
† General linear model; ANOVA statistics for fatty acid-specific difference between groups following the significant overall association between ethnicity and fatty acid concentrations (Pillai's trace criterion; P < 0·001). For skewed distributions, statistics are based on transformed data as described in the Methods section.
‡ After standardisation for gestational age at blood sampling, mean value became significantly different from that of the Dutch ethnic group (P < 0·01).
§ After standardisation for gestational age at blood sampling, mean or median value became non-significantly different from that of the Dutch ethnic group (P ≥ 0·01).
Compared with the Dutch women, all ethnic minority groups had lower proportions of α-linolenic acid (18 : 3n-3) and its derivative eicosatetraenoic acid (20 : 4n-3). For Surinamese, Antillean and Moroccan women, proportions of EPA (20 : 5n-3) and docosapentaenoic acid (22 : 5n-3) were also (significantly) lower. The lowest proportions of EPA and docosapentaenoic acid, as well as DHA (22 : 6n-3) (the three fatty acids found primarily in fish and fish oil) were observed for the Turkish women; in contrast, Ghanaian women were found to have the highest proportions.
Ethnic patterns were more complex for the n-6 fatty acids. Surinamese, Antillean, Turkish and Moroccan women all showed higher proportions of linoleic acid (18 : 2n-6) and its longer-chain derivatives AA (20 : 4n-6), adrenic acid (22 : 4n-6) and Osbond acid (22 : 5n-6), but lower or comparable proportions of DGLA (20 : 3n-6). Interestingly, the Turkish women showed the largest deviation from the Dutch group for linoleic acid, Osbond acid and adrenic acid, but the smallest deviation for AA and DGLA. The pattern for Ghanaian women was again different; they showed lower proportions of DGLA as well as linoleic acid, adrenic acid and Osbond acid, but higher proportions of AA.
Adjustment for gestational age at blood sampling did not alter the size of the model coefficients (results not shown), but did affect the significance levels for five of the forty-four associations (Table 2).
Table 3 compares the observed differences in EPA, DHA, DGLA and AA across the ethnic minority groups. Results are presented as relative differences (i.e. percentage differences) from the Dutch proportions, with models 1 to 4 representing the increasing levels of adjustment. For the fish fatty acids EPA and DHA the Table reveals the extreme position of Ghanaian and Turkish women v. Dutch women, showing differences of +93 and − 55 % for EPA, and differences of +41 and − 21 % for DHA respectively (model 1). For Surinamese, Antillean and Moroccan women, adjustment for gestational age at blood sampling and maternal characteristics attenuated the EPA differences but augmented the DHA differences (models 2 and 3 v. model 1). In contrast, adjustment attenuated both EPA and DHA differences for Turkish women, but augmented these for Ghanaian women. For the n-6 fatty acids, differences across the ethnic minority groups and across models were less pronounced. The changes across models 3 and 4 show a modest attenuation of the Dutch–Ghanaian differences in EPA and DHA proportions after adjustment for fish consumption, but only minor changes in other group comparisons.
Differences (%) in EPA, DHA, dihomo-γ-linolenic acid (DGLA) and arachidonic acid (AA) in maternal plasma phospholipids for the five main ethnic minority groups compared with the Dutch reference group†
*P < 0·01, **P < 0·001.
† Differences are expressed as percentage differences from values in the reference group (i.e. Dutch ethnic group).
‡ Model 1, crude (not adjusted); model 2, as model 1+ adjustment for gestational age at blood sampling; model 3, as model 2+ adjustment for maternal age, parity, educational attainment, pregravid BMI, smoking and alcohol consumption; model 4, as model 3+ adjustment for fish and fish oil consumption.
Discussion
In the present observational study, we found distinct patterns of fatty acid proportions in maternal plasma phospholipids across ethnic groups, with the Ghanaian and Turkish ethnic groups deviating most from the Dutch reference group. Ghanaian women, who reported the highest fish intake, had generally higher proportions of n-3 and lower proportions of n-6 fatty acids, while for Turkish women, who reported the lowest fish consumption, the opposite was observed. Attenuation of the differences in relative EPA, DHA, DGLA and AA concentrations after adjustment for maternal physiological and lifestyle-related variables and fish and fish oil consumption was, however, modest.
To our knowledge, only one previous study has examined ethnicity-related differences in maternal fatty acid status in early pregnancy(Reference Otto, van Houwelingen and Antal37), applying a cross-country rather than a within-country comparison and not adjusting for potential explanatory factors such as fish consumption(Reference van Houwelingen, Sørensen and Hornstra21–Reference Williams, Frederick and Qiu24). Yet, in the present study, fish consumption did not appear to explain the ethnic variation in long-chain PUFA status to a relevant degree. This was perhaps to be expected for the ethnic groups that reported intake levels comparable with the Dutch, but not for those who deviated. However, the fish FFQ was relatively short, and may have been too crude to adequately assess EPA and DHA intake. Alternatively, being developed in Denmark, it may have had a lower validity in this multi-ethnic context. When we examined the Spearman rank correlation between the relative EPA concentrations and the number of fish servings per week, the overall correlation (0·33) was well in the range of previous questionnaire–concentration correlations (0·19–0·50)(Reference Hjartåker, Lund and Bjerve22), but the correlations for the Ghanaian (0·13) and Moroccan (0·15) groups in particular were indeed lower. As these were the groups reporting the highest fish consumption, this could also reflect a lower power of the questionnaire to explain EPA variability at higher intakes. Given these uncertainties, we cannot make any final conclusions about the contribution of differential fish consumption to ethnic disparities in maternal early pregnancy fatty acid status at this stage. Further study is required, not only to gain more insight in ethnicity-related differences in maternal fatty acid intake, but also to explore potential alternative explanations that are beyond the scope of the present paper, such as the possible inter-ethnic differences in the consumption of the parent essential fatty acids and their conversion to long-chain PUFA(Reference Emken, Adlof and Gulley38–Reference Schaeffer, Gohlke and Muller41), and potential differences in post-conceptional long-chain PUFA mobilisation(Reference Al, van Houwelingen and Kester28, Reference Otto, van Houwelingen and Badart-Smook42) or synthesis(Reference Burdge, Sherman and Ali43).
In the present study a comparison was made between the Dutch ethnic group and the five main ethnic minority groups in the Netherlands. It should be noted though that this sample may reflect a relatively healthy subpopulation. Indeed, compared with the group of non-respondents (neither questionnaire nor biomarker study), women in the present analysis were more often of Dutch ethnic background, more highly educated and started prenatal care earlier in pregnancy (data not shown). However, differences in lifestyle factors, such as BMI, smoking habits and fish consumption between the ABCD study respondents who did and who did not donate blood were small (data not shown), which suggest only a minimal selection bias. It should also be kept in mind that our measure of ethnicity (country of birth), though commonly used in Dutch research and prescribed by the Netherland Organisation for Health Research (www.zonmw.nl/en), does not capture heterogeneity in genetic make-up, history, culture or dietary preferences within ethnic groups, such as exists between the Creole and Hindustani populations of Surinam. However, to the extent that information for these subpopulations was available, no significant differences in fatty acid status were observed, allowing for inclusion of both in one Surinamese group.
Interestingly, the observed high proportions of EPA and DHA among the Ghanaian women coincided with high AA, but low DGLA, content of plasma phospholipids in this group. A similar contrast between the relative AA and DGLA concentrations was observed for the Surinamese and Antillean women, who, together with the Ghanaians, form the ethnic minority women at highest risk of adverse perinatal outcomes (fetal growth restriction and preterm birth)(Reference Goedhart, van Eijsden and van der Wal25, Reference Goedhart, van Eijsden and van der Wal44). Over the past years, the main focus in perinatal fatty acid research has been on the positive effects of the n-3 fatty acids. Our present observations, in conjunction with previous results showing significant but contrasting associations of DGLA (positive) and AA (negative) with birth weight(Reference van Eijsden, Hornstra and van der Wal12), however, suggest distinct roles of the n-6 fatty acids in perinatal health as well. More detailed intervention studies would be worthwhile to investigate if and how adaptation of the maternal fatty acid profile can affect perinatal health, and ethnic disparities therein. Specific groups could potentially benefit from GLA and EPA supplements, a combination shown to improve DGLA and EPA concentration without raising AA(Reference Barham, Edens and Fonteh45).
In conclusion, the present results suggest the presence of considerable ethnic disparities in the maternal n-3 and n-6 fatty acid status during pregnancy, which, in view of the existing and newly emerging evidence of the role of these nutrients in human growth and development, may be relevant to ethnic disparities in perinatal health. Given the limitations of this observational study, further research into these distinct ethnicity-related fatty acid patterns is warranted, particularly to elucidate the role of fatty acid intake v. metabolic differences, and to explore the potential benefits of fatty acid supplementation.
Acknowledgements
The research described in this paper was performed at the Public Health Service of Amsterdam and the Academic Medical Centre (University of Amsterdam) as part of the ABCD study.
Financial support for the ABCD cohort study was granted by the Netherlands Organisation for Health Research and Development (ZonMw) in The Hague, the Public Health Service and Municipal Council of Amsterdam, the Academic Medical Centre, and Nutricia Research BV in Zoetermeer.
We thank the hospitals, obstetric clinics and general practitioners for their assistance in the implementation of the ABCD study, and all pregnant women who participated for their cooperation.
M. F. W. and G. J. B. conceived of and designed the ABCD study; G. H. was advisor in study design. M. E. and M. F. W. collected and processed the data. M. E. conducted the statistical analysis, with G. J. B. providing supervision. All authors contributed to the interpretation of the results and writing of the manuscript, and have seen and approved the final version. None of the authors had a financial or personal conflict of interest related to the content of the study.
