Comparison of national cross-sectional breast-feeding surveys by maternal education in Europe (2006–2016)

Objective Breast-feeding is an important determinant of health of mothers and their offspring. The present study aimed to compare breast-feeding rates across Europe disaggregated by maternal education and establish what proportion achieves at least 50 % exclusive breast-feeding (EBF) at 6 months. Design/Setting Secondary analysis of national or sub-national studies’ breast-feeding data for EU Member States plus Norway and Iceland, published in 2006–2016. Nineteen EU Member States plus Norway reported rates of EBF and any breast-feeding disaggregated by maternal education, of which only thirteen could be matched to the International Standard Classification of Education. Participants Mothers and their infants aged 0–12 months. Results Data on EBF rates at 6 and 4 months were found in only four and six countries, respectively. At 6 months, EBF rates of 49 % in Slovakia and 44 % in Hungary were closest to WHO’s target of at least 50 % EBF. At 4 months, mothers with high education level in Denmark, the Netherlands and Germany had the highest EBF rates (71, 52 and 50 %, respectively). Mothers with low education level were less likely to initiate breast-feeding and cessation occurred early. The inequality gap ranged from 63 % in Irish mothers to no gap or very low levels of inequality in Poland, Sweden and Norway. Conclusions More mothers with high, compared with low, education initiate breast-feeding and practise EBF for longer. More European policies should be targeted to protect, support and promote breast-feeding, especially among mothers with only mandatory education.

WHO Member States have agreed six Global Targets for Nutrition (1) . One target is to increase the rate of exclusive breast-feeding (EBF) in the first 6 months to at least 50 % by 2025 and another target is to halt the increase in childhood overweight. Globally, breast-feeding rates have not improved substantially over recent years and rates of EBF at 6 months are well below the 2025 target in most countries. In low-and middle-income countries reliable monitoring systems report breast-feeding rates less than 40 % EBF in the first 6 months (2) , but the average rates in low-and middleincome countries are high compared with those in highincome countries. Interestingly, it appears in low-and middle-income countries that women with low socioeconomic status (SES) breast-feed longer than those with high SES, in contrast to high-income countries where the trend appears to be in the opposite direction (3) . Mothers of low SES in Europe appear less likely to initiate breastfeeding and cessation occurs early compared with those of high SES (4)(5)(6)(7) . Moreover, infants who are predominantly formula-fed compared with those being exclusively breastfed for the first 6 months are two-and-a-half times more likely to be obese at 24 months (8) and cessation of EBF before 4 months increases the risk of childhood obesity (9,10) . Indeed, the prevalence of obesity is also greater among low-compared with high-SES families (8,(11)(12)(13) and obese mothers especially have to overcome more barriers when trying to breast-feed (14,15) . It has been suggested that an improvement in EBF rates among low-SES mothers could help reduce health inequalities (16,17) related to obesity (16) and resulting morbidity (3,17,18) .
Because of the lack of published studies comparing the rates of breast-feeding across socio-economic groups in Europe (19,20) , the second recommendation in the report Diet, Nutrition and Obesity: Infant Feeding by Socioeconomic Status stated that: (Bulgaria 2010, Lithuania 2016, Portugal 2007, Slovenia 2010 and Norway 2010) were found. One peer-reviewed study was found for each of EBF at 4 months in France (2016) and breast-feeding at 1 week in Sweden (2016).

Maternal education level
Socio-economic position was represented by maternal education level. Education levels were harmonized according to the International Standard Classification of Education (ISCED) (23) and defined in three categories: low = levels 0-2 (compulsory, i.e. primary or lower secondary education); middle = levels 3-4 (upper secondary and post-secondary non-tertiary education); and high = levels 5-6 (tertiary education; Table 1).

Breast-feeding variables
For the purpose of comparison, the following breastfeeding variables used in national surveys were grouped together into one variable called 'any breast-feeding (ABF)': breast-feeding within the first hour (recommended by WHO (24) ); ever breast-fed; breast-feeding at all; breast-feeding at birth; and breast-feeding at discharge. In addition, rates on 'breast-feeding at 1 week' from Sweden and 'breast-feeding at <1 month' from Finland were included in the ABF variable. Three additional ABF variables were created (Table 2) to group the different time frames used in national studies to allow their comparison: 'ABF at ≤2 months', which included ABF rates from between 1 and 8 weeks; 'ABF at ≤4 months', which included ABF rates from between 8 and 16 weeks; and 'ABF at 6 months', which included ABF rates reported at 6 months. Similarly, the wide range of variables used in national surveys to report rates of EBF were grouped together into four variables (Table 3). These were based on the different definitions for time frame used in national studies. The variable 'EBF at ≤1 month' included all EBF rates from 1 to 4 weeks. The variable 'EBF at ≤3 months' included all EBF rates from 5 to 12 weeks. The variable 'EBF at 4 months' included all EBF rates relating to 4 months (e.g. EBF up to 4 months and EBF at least until 4 months). The fourth variable was 'EBF at 6 months'.

Inequality gap
The inequality gap analysis was carried out by calculating the difference between the breast-feeding rates of mothers with the highest minus the lowest level of education, and expressed as a percentage of the rate in the highest education level (25) . The breast-feeding rates of mothers with the highest level of education were used as the highest possible achievable rate for that specific country when calculating the inequality gaps. Table 1 Levels of maternal education defined according to ISCED classification (low = ISCED 0-2; middle = ISCED 3-4; high = ISCED 5-6)

Country
Indicators used for level of maternal education in national surveys or studies Classification based on ISCED Austria (26,27) Compulsory education Low  Apprenticeship  Middle  Intermediate vocational school  Secondary school certificate  Secondary school certificate, nursing school  University degree  High  Belgium No data by maternal education Bulgaria* (28,29) No education Low Primary Elementary Secondary Middle Croatia (30,31) Primary school Low Secondary school Middle Higher education High Cyprus No data by maternal education Czech Republic No data by maternal education Denmark (32)(33)(34) Primary school Low Vocational education Middle General or vocational upper secondary education Short or intermediate higher education High Long higher education Estonia No data by maternal education Finland (35,36) Basic Low Upper secondary level Middle Lower university degree or lowest level of tertiary education High Highest university degree or postgraduate education France (37,38) Less than Baccalauréat level Low Baccalauréat level Middle Above Baccalauréat level High Germany (39) Low Low Medium Middle High High Greece (40,41) No school at all Low Some primary classes Primary school graduate Third high school class Middle High school graduate Graduate technological education institutions (TEI) High University graduate Hungary No data by maternal education Ireland (42,43) Lower secondary or less Low Leaving certificate Middle Sub-degree Degree or third level High Italy (44,45) Compulsory education Low Diploma Middle Graduation High Latvia (I Pudule, unpublished results) (46) Elementary Low Secondary Middle Vocational education Higher education High Lithuania No data by maternal education Luxembourg † (47,48) Primary school education or less Low Professional education Middle Secondary or higher education High Malta (49,50) Primary and secondary education Low Post-secondary and non-tertiary education Middle Tertiary education High Netherlands (51,52) Low No data by maternal education Romania (50,53) Without primary school Low With primary school Medium studies Middle Superior studies High   (61) 96 Denmark 2016 (2014) (32) 90 Estonia 2016 (2015) (62) 88 Finland 2012 (2010-2011) (

Statistical analysis
All data were inserted and cleaned in MS Excel (Microsoft Office 2016). Relative breast-feeding rates with reference to high level of maternal education were computed for middle and low levels of maternal education. For statistical testing, breast-feeding percentage values in the groups by level of maternal education were converted to arcsine values to apply parametrical testing. For comparisons between multiple groups, ANOVA with the post hoc Tukey honest significant difference test was applied and P < 0·05 was considered to indicate statistical significance. Breastfeeding cessation rates were evaluated with the matchedpair t test. The statistical software package IBM SPSS Statistics version 24 was used for statistical testing.

Results
The levels of maternal education reported are listed in Table 1   . No breast-feeding rates by level of education were found for one-third (n 10) of the countries (Belgium, Cyprus, Czech Republic, Estonia, Hungary, Lithuania, Portugal, Slovak Republic, Spain and Iceland). The levels of mothers' education were grouped into high, medium or low categories using the ISCED classification for those twenty countries that did report breast-feeding rates by level of education. Only eighteen countries could be compared based on the ISCED classification since education levels in Bulgaria and Luxembourg could not be matched to the ISCED classification. Bulgaria's and Luxembourg's highest level of education included 'secondary' or 'secondary or higher', respectively, whereas the ISCED levels 5-6 (i.e. high) include only tertiary education plus bachelor or equivalent and higher but not secondary.
Any breast-feeding Data on average rates of ABF were found for all except one country (Slovak Republic; Table 2 (26,29,30,32,36,38-40, 43,45,47,49,52,53,55-71)    reporting their ABF rates. The time frames ranged from infant ever breast-fed, initiation at birth, within the first hour, at discharge and from breast-feeding at a range of weeks and/or months, up to 6 months. This resulted in about seventy different variables for ABF. In an attempt to compare the wide range of time frames they were grouped into four: any breast-feeding (ABF); ABF at ≤2 months (1-8 weeks); ABF at ≤4 months (8-16 weeks); and ABF at 6 months (Table 2).  (38) 30 Any breast-feeding disparities Fifteen countries reported rates of ABF disaggregated by education level but the data from Bulgaria and Luxembourg could not be compared with the others. Among the thirteen countries that were compared, there is a significant difference in ABF rates between mothers with high, medium and low education levels. The proportion of mothers with high education levels, practising ABF, is on average 20 % higher (P = 0·035) than those with low levels. Figure 1 illustrates a clear tendency towards a social gradient in breast-feeding rates between the mothers according to high, middle and low level of education. However, some countries have less steep social gradients as shown by the very small percentage inequality gaps in the Polish, Norwegian and Swedish mothers with high and low education levels (Fig. 2). In Poland, Norway and Sweden, more than 90 % of mothers are breast-feeding regardless of their level of education and so the inequality gaps are minor. Small disparities are observed in Italy and Croatia (<10 %), with relatively larger inequality gaps in Finland and Austria (<20 %) and in France, the Netherlands and Germany (<30 %). The largest inequality gaps are observed in the UK (31 %), Malta (41 %) and Ireland (63 %; Fig. 2). In five countries (Greece, Poland, Sweden, Luxembourg and Netherlands), where rates are reported at more than one time point during the first 6 months, there is a marked rapid cessation after initiation. The cessation gradient tends to be steeper in mothers with low level of education as they stop breast-feeding earlier (mean 71 (SD 17) %), compared with those having high education level (mean 62 (SD 17) %; P = 0·036). Similarly, at 6 months, the inequality gap in ABF of Swedish mothers increases to 38 % as more mothers with low education level stop breast-feeding earlier compared with those with higher education. Low levels of maternal education are not only associated with low rates of initiation but also with a higher probability of early cessation.

Exclusive breast-feeding disparities at 6 and 4 months
Only four countries, Latvia, Luxembourg, Poland and Romania, reported rates of EBF at 6 months by level of maternal education (Table 3). In Poland and Latvia, two to three times more mothers with high compared with low education level practised EBF at 6 months. In both Romania and Luxembourg there was little difference in EBF rates at 6 months between mothers with high compared with low education level (Table 3). Six countries reported EBF rates at 4 months by level of maternal education (Table 3): more Danish and German mothers with high compared with low education level practised EBF and the inequality gaps were 45 and 60 %, respectively. Similarly, a greater proportion of Dutch, French, Croatian and Luxembourg mothers with high compared with low education level practised EBF at 4 months and the respective inequality gaps were 23, 27, 34 and 44 % at 4 months.
Finally, Table 4 lists the wide range of different time frames used to define breast-feeding rates, for example: initiation of breast-feeding at birth; initiation of breast-feeding within the first hour; any breast-feeding; ever breast-feeding; breast-feeding at discharge from hospital; breast-feeding at 1 week; and breast-feeding at or less than 1, 2, 3 and 4 months. This wide range of non-standardized time frames makes comparison between countries impractical.

Discussion
The current paper presents information comparing the rates of breast-feeding in Europe in mothers of different SES, defined by level of maternal education. The results confirm what other authors have stressed concerning the need for standardized methods to monitor breast-feeding rates in Europe to allow comparison between countries and recommendations to reduce inequalities (75,76) .

Any breast-feeding and disparities
The results confirm previous studies and provide new comparative evidence that disparities exist, not only within but also between countries (6,77) . As indicated, based on surveys from thirteen countries, one-fifth more mothers with high compared with low education level breast-fed. Whereas the initial disparities in the Nordic, Baltic and Central European countries are relatively small, the inequality gaps found in Ireland, Malta, the UK, Germany, the Netherlands and France are high and should give cause for concern to the respective governments (Fig. 2). When breast-feeding rates are reported only as national averages, these average figures can conceal a steep gradient between mothers with high compared with low education level. For example McAndrew et al. (56) found that the incidence of ABF was 91 % among UK mothers with high education level compared with 63 % among those with low education level. Similarly, five countries (Greece, Poland, Sweden, Luxembourg and Netherlands) reported rates over two or three time periods during which the trend is that the inequality gaps get wider. Even Nordic countries, for example Sweden, report that by 6 months the inequality gap increases considerably as more Swedish mothers with low education stop breast-feeding earlier than those with higher education. This is confirmed by Wallby and Hjern (78) , who found that low income levels are a strong predictor for early cessation of breast-feeding and recommended extra protection for low-income Swedish mothers. Similarly Finnish (10,79,80) and Danish studies (81) reported that mothers with low compared with high education were more likely to introduce foods too early. Indeed, even Norwegian authors recently reported that very few (7 %) Norwegian mothers with low education level still exclusively breast-fed by 5 months compared with three times more mothers (22 %) with high level (82) .

Exclusive breast-feeding and disparities
The results confirm that on average the EBF rates in most countries, except the Slovak Republic (49 %) and Hungary (44 %), are well below governments' 2025 target of at least 50 % EBF at 6 months (1) . Victora et al. (2) confirm that Nordic countries have high breast-feeding initiation rates of over 90 %. Similarly, Baltic countries and those in Central Europe report high initiation rates, with over 90 % in Bulgaria, Croatia, Czech Republic, Hungary, Poland, Slovak Republic and Slovenia. Even at 6 months about one-third of Baltic (Latvian, Lithuanian and Estonian) mothers are still practising EBF (Table 3). Results from smaller studies (83,84) report that, on average, about onehalf of Lithuanian mothers exclusively breast-feed for 6 months, but only one-fifth with low compared with fourfifths of those with high level of education continue until  6 months. Slovakia and Hungary, along with the Baltic and Nordic countries, report the highest EBF rates (Table 3) and interestingly these countries provide the best statutory entitlement of paid parental leave, closely followed by Czech Republic and Bulgaria (85) . In contrast, French mothers can only expect 6 weeks off work before and 10 weeks after birth, and even those with high education level have one of the lowest breast-feeding initiation rates (71 %) in Europe (although dramatically increased since 1995) (86,87) . Parental leave payment rates are also low in Ireland and the UK, where fully paid maternity leave lasts only 9 and 12 weeks, respectively (85) . Furthermore it appears that French, Irish and British mothers are subjected to aggressive marketing and many violations are reported against the International Code of Marketing of Breast Milk Substitutes ('the Code' hereafter) in France (88) , Ireland (89) and the UK (90) . Robust policies, better maternity leave and protection against marketing are needed not only to achieve the global target of at least 50 % EBF in the first 6 months by 2025 (1) but also to achieve the EU Council's aim (91) to halt the rise in childhood obesity, which is closely associated to poor infant feeding practices. For example, parents have the right to receive information on infant feeding practices, from health professionals, that is independent of commercial interests and free from conflicts of interest. Low levels of breast-feeding have been attributed to lack of knowledge about breastfeeding and incorrect advice on infant feeding from health professionals in Lithuania (92) , Malta (93) and Romania (94) .

Implications for public health policy
A recent German study (95) is among the first to demonstrate a widening socio-economic disparity in breastfeeding practices using a time-trend analysis. This disparity was largely explained by the proportion of German mothers with low SES who were likely to be obese and/or to smoke. Both these risk factors are well known to be associated with low breast-feeding rates (96,97) . Breastfeeding cessation among Dutch mothers with low education level also increases rapidly (98) and so in the Netherlands there is increased focus on how to reduce health inequalities in pregnant women and their offspring (99) . Similar calls are being made in Italy, where authors stress the need for policies that specifically support mothers with low level of education (100,101) . In the future it is crucial to carry out regular monitoring of trends of breast-feeding rates along with the relative prevalence of obesity. The latest data (2015-2017) from the WHO Childhood Obesity Surveillance Initiative (102) show that Southern European countries have the highest rate of child obesity. In Cyprus, Greece, Italy, Malta and Spain, approximately one in five boys is obese, whereas Nordic and Baltic countries are among the countries with the lowest rates. Obesity in young children is associated with low initiation rates of breast-feeding along with early cessation (103,104) . This both negatively affects long-term health and increases the risk of obesity over the next generations (105) . However, data on both national breast-feeding rates and obesity prevalence, disaggregated by level of maternal education, are needed to better investigate the complex relationship between obesity, breast-feeding and SES. A recent review (106) tried to separate independent outcomes but a clearer understanding of the several different mechanisms involved, and their interrelationships, is of utmost importance. The European Blueprint for Action on Breast-feeding (75) especially recommends: full implementation of the Code and subsequent relevant World Health Assembly resolutions; maternity protection legislation that enables all working mothers to exclusively breastfeed their infants for 6 months; establishing standards for best practice within public places as well as in all maternity and childcare services; and harmonized collection, at least every 3-5 years, of disaggregated breast-feeding data that correspond to the recommended standardized indicators and definitions.

Limitations
As these comparisons are derived from data gathered using different methods over different time periods, reported rates and their comparison should be treated with caution (107) . Breast-feeding rates were collected on different dates over a long time period and a wide range of methods was used in different countries to collect breast-feeding rates (Table 4). It was attempted to harmonize education levels into three categories according to the ISCED (108) ; however, our categorizations could be flawed. Furthermore, reported breastfeeding rates were not always nationally representative and so comparisons should be interpreted with caution. The inequality gap analyses are useful but limited in that they only reflect the difference between low level of maternal education compared with high. The analysis can be potentially affected by extreme values for each of these groups and does not necessarily reflect the experience of the entire population due to the different distribution of mothers' education levels (109) .

Conclusion
Overall, there is less initiation of breast-feeding and shorter duration of EBF by mothers with low compared with high education level. These findings can help governments to investigate why these disparities exist and why some countries such as the Nordic, Baltic and Central European countries have higher breast-feeding rates and lower disparities compared with other European countries. National policies common to Nordic, Baltic and Central European countries are that they have paid maternity leave regulations which go beyond the minimum stipulated by EU law. In addition, Nordic countries have been efficient in minimizing violations of the Code and until recently Baltic and Central European countries have not been as exposed to high levels of marketing by companies selling breast milk substitutes. The results highlight disparities in breast-feeding which can only be reduced if governments develop harmonized monitoring systems that are carried out regularly and where the data are disaggregated by SES to develop appropriate health promotion policies. tor, Baby Friendly Initiative). Financial support: This work was carried out under a 'Pilot project related to reducing health inequalities: building expertise and evaluation of actions' (tender number SANTE/2014/C4/032; service contract number 2015-C4-03 (SI2.724119)), funded by the European Commission, DG Health and Food Safety, Unit C-4 Health Determinants and Inequalities. The European Commission, DG Health and Food Safety, Unit C-4 Health Determinants and Inequalities had no role in the design, analysis or writing of this article. The information and views set out in this publication are those of the authors and do not necessarily reflect the official opinion of the Commission. The Commission does not guarantee the accuracy of the data included in this publication. Neither the Commission nor any person acting on the Commission's behalf may be held responsible for use which may be made of the information contained herein. Conflict of interest: None. Authorship: M.S. collected and analysed data, wrote the initial version of the manuscript and contributed to the final version. A.P. supported statistical analysis, contributed to the preparation of the manuscript and provided critical comments. A.R. supervised the data collection, its analysis and the writing of the manuscript, and was responsible for contacting experts on breastfeeding policy in the countries investigated. All authors reviewed and approved the manuscript. Ethics of human subject participation: Not applicable.