Article Text
PDF
Abstract
Background Discrimination is associated with obesity, but this may differ according to the type of obesity and ethnic group. This study examines the association of perceived ethnic discrimination (PED) with general and abdominal obesity in 5 ethnic minority groups.
Methods We used cross-sectional data from the HELIUS study, collected from 2011 to 2015. The study sample included 2297 Ghanaians, 4110 African Surinamese, 3021 South-Asian Surinamese, 3562 Turks and 3868 Moroccans aged 18–70 years residing in Amsterdam, the Netherlands. Body mass index (BMI) was used as a measure for general obesity, and waist circumference (WC) for abdominal obesity. PED was measured using the Everyday Discrimination Scale. We used linear regression models adjusted for sociodemographics, psychosocial stressors and health behaviours. In additional analysis, we used standardised variables to compare the strength of the associations.
Results In adjusted models, PED was significantly, positively associated with BMI in the South-Asian Surinamese (β coefficient 0.338; 95% CI 0.106 to 0.570), African Surinamese (0.394; 0.171 to 0.618) and Turks (0.269; 0.027 to 0.510). For WC, a similar pattern was seen: positive associations in the South-Asian Surinamese (0.759; 0.166 to 1.353), African Surinamese (0.833; 0.278 to 1.388) and Turks (0.870; 0.299 to 1.440). When stratified by sex, we found positive associations in Surinamese women, Turkish men and Moroccan men. The strength of the associations with BMI and WC was comparable in the groups. Among the Ghanaians, no significant associations were observed.
Conclusions Ethnic and sex variations are observed in the association of PED with both general and abdominal obesity. Further research on psychosocial buffers and underlying biological mechanisms might help in understanding these variations.
- ETHNICITY
- OBESITY
- PSYCHOSOCIAL FACTORS
- SOCIAL EPIDEMIOLOGY
Statistics from Altmetric.com
Introduction
Obesity is considered one of the most important risk factors for morbidity and mortality worldwide.1 Evidence indicates important ethnic differences in obesity.2 ,3 In the USA, for example, African-Americans are more obese than white Americans and Asian Americans.3 In the Netherlands, as compared with the Dutch, abdominal obesity is less prevalent among African Surinamese men, but more prevalent among African Surinamese women and South-Asian Surinamese men and women.2
These ethnic differences in obesity might be explained by differences in demographics, socioeconomic status and health behaviours, as well as by psychosocial stressors.4 An important psychosocial stressor is perceived discrimination.5 There is some evidence indicating a positive association of perceived discrimination with obesity.6 ,7 This association might work through depressive symptoms,8 ,9 health behaviours10 and direct physiological stress pathways (eg, hypothalamic–pituitary–adrenal axis).11–13
Several studies investigated the association between perceived discrimination and obesity in ethnic minority groups in the USA, yielding mixed results.6 ,7 ,14 ,15 In a cross-sectional study, the Jewish, Polish, Irish and Italian whites who perceived chronic discrimination had a higher likelihood of an increased waist circumference (WC) than the Jewish, Polish, Irish and Italian whites who did not perceive chronic discrimination. This relationship, however, was not found among other white Americans, Hispanic Americans and African-Americans.16 In a longitudinal study, participants who indicated constantly high levels of discrimination over the 9-year study period showed the largest gains in WC.6 Lewis et al7 found that experiences of discrimination were positively associated with visceral fat in white American and African-American women, although a recent study did not find this longitudinal association in African-American adults.15
The current literature has several gaps, however. First, most research on perceived discrimination and obesity has been conducted on ethnic minority groups in the USA, predominantly African-Americans and Hispanic Americans. These groups differ from those residing in Europe, due to different migration histories and ethnic origins, among other factors.17 Hence, the US findings cannot simply be generalised to the European-based ethnic minority groups. Second, most discrimination studies on obesity have focused on discrimination in general, with relatively few studies investigating the association with a specific type of discrimination. A type of discrimination that is particularly relevant to ethnic minority groups is perceived discrimination based on ethnic background. Little is known whether perceived ethnic discrimination (PED) is associated with obesity in different ethnic groups. Although different from ethnic discrimination, some studies have assessed the association of racism with obesity. For example, Cozier et al18 found that everyday racism is longitudinally associated with obesity among African-American women. Finally, only a few studies have assessed whether discrimination is differentially associated with general or abdominal obesity. It might be possible that the association differs by type of obesity, given potential differences in the underlying biological mechanisms (eg, stress-related pathways, abdominal fat accumulation).7 ,19
In this study, we aimed to investigate the association of PED with general and abdominal obesity in five ethnic minority groups residing in Amsterdam, the Netherlands. The Netherlands is considered to have a multicultural or inclusive model, according to Meuleman's20 country typology using the Migrant Integration Policy Index (MIPEX).21 MIPEX is a systematic evaluation of integration policies in European countries. The inclusive model is characterised by social and political tolerance, respect for cultural differences and opportunities to acquire citizenship.20 ,21 The ethnic minority groups included in this study differ in migration history and country of origin among other factors (see box 1), providing an opportunity to assess whether there are ethnic variations in the associations of PED with general or abdominal obesity.
Sociodemographic information on the ethnic minority groups in the Netherlands.22
Ghanaians: Ghanaian people came to the Netherlands between 1974 and 1983 mainly for economic reasons. A second group of migrants from Ghana arrived in the 1990s due to political instability and drought.
Moroccans and Turks: Moroccan and Turkish immigrants moved to the Netherlands during the 1960s and early 1970s to fill shortages in the Dutch labour market in mainly unskilled occupations. A second wave of immigrants from Turkey and Morocco arrived in the 1970s and 1980s, mainly spouses and children as part of the family reunification.
South-Asian and African Surinamese: Migrants from Suriname, a former Dutch colony, came to the Netherlands between 1975 and 1980 due to the political instability and dire socioeconomic circumstances in their home country. The South-Asian Surinamese people originate from Northern India and African Surinamese people originate from West Africa.
Methods
Study population
Baseline data were used from the Healthy Life in Urban Settings (HELIUS) study, a multiethnic cohort study in Amsterdam, the Netherlands. The study has been described in detail elsewhere.17 Participants aged 18–70 years were randomly sampled from the municipality register, stratified by ethnicity (see below for definition). The overall participation rate was 50% (around 40% for Turkish and Moroccan participants, 50% for Surinamese participants, 60% for Ghanaian participants). Data were collected through questionnaires (either self-administered or by an ethnically matched interviewer) and a physical examination. The physical examination was conducted by trained research staff at the designated research locations of HELIUS. All participants gave written informed consent. All HELIUS protocols were approved by the Ethical Review Board of the Academic Medical Center, University of Amsterdam.
Baseline data were collected from January 2011 to December 2015. The total number of participants who filled in the questionnaire and attended the physical examination was 22 165. We excluded the Javanese Surinamese (N=233), Surinamese of different or unknown origin (N=267), as well as participants of unknown ethnic background (N=48) because of the relatively small sample size. In addition, we excluded the Dutch origin group (N=4564) as their median PED score was close to 1 (ie, no discrimination perceived), making it difficult to investigate the association with the necessary statistical precision. Furthermore, participants with missing PED scores (further N=195) were excluded. Subsequently, a total sample of 16 858 participants was included in our analysis: 3021 South-Asian Surinamese, 4110 African Surinamese, 2297 Ghanaian, 3562 Turkish and 3868 of Moroccan origin.
Ethnicity
Individuals were considered of non-Dutch ethnic origin if they were born outside the Netherlands and at least one parent was born outside the Netherlands (first generation), or if they were born in the Netherlands, and at least one parent was born outside the Netherlands (second generation).17 The Surinamese subgroups were determined by self-reported ethnic origin. Participants were considered of Dutch origin if the participant and both parents were born in the Netherlands.
Perceived ethnic discrimination
PED is conceptualised as the subtle forms of interpersonal discrimination on grounds of ethnic background as experienced in daily life.8 PED was assessed with a nine-item questionnaire based on the Everyday Discrimination Scale (EDS).23 EDS was developed in the USA,24 yet its conceptual work was based on a qualitative study conducted among African Surinamese women in the Netherlands and African-American women in the USA.25 The EDS has not been formally validated for use in the ethnic groups studied in HELIUS, but there is evidence for its suitability in multiethnic samples in the USA.23 Participants respond to nine statements using a five-point Likert scale (1=never to 5=very often): ‘You're treated with less courtesy’; ‘You're treated with less respect’; ‘You receive poorer service than other people (in restaurants or shops)’; ‘People act like they think you're not smart’; ‘People act like they're afraid of you’; ‘People act like they think you're dishonest’; ‘People act like they're better than you are’; ‘You're threatened or harassed’ and ‘You are called names or insulted’. We adjusted the scale by asking participants specifically about discrimination because of their ethnic background. We calculated the mean score of the nine EDS items. In case one of the nine questions was missing, the average of the other eight items was calculated. In case more than one item was missing, the variable was deemed missing.
General obesity
The body mass index (BMI) reflects general obesity and is calculated by dividing someone's weight (kg) by the square of their height (m).26 Both height and body weight of the participants were measured in duplicate in barefoot participants wearing light clothes only.17
Abdominal obesity
WC (cm) is a measure of abdominal fat mass, and is used as a measure for abdominal obesity. WC was measured using a non-elastic, flexible tape measure at the level midway between the lower rib margin and the iliac crest. WC was measured in duplicate, and a third measurement was taken if the difference between the first two measurements was >1 cm.
Covariates
We considered the following covariates: demographics (age, sex, marital status), socioeconomic status (educational level, employment status), other psychosocial stressors (domestic stress, any negative life events), health behaviours (smoking, alcohol consumption, physical activity) and depressive symptoms. Categories for marital status were: married/living with partner, divorced/separated/widowed and never married. There were four categories for highest education attained: no education or elementary schooling only, lower vocational or lower secondary education, intermediate vocational or intermediate/higher secondary education, and higher vocational education or university. Current employment status was categorised into: currently employed, unemployed or not in the labour force. Domestic stress was defined as perceived psychosocial stress (feeling irritable, anxiety or trouble sleeping) arising from one's living circumstances at home (never, some periods, several periods, constantly). Any negative life events were operationalised as any type of self-reported threatening experience (eg, the death of a close relative or friend, major financial crisis) during the previous 12 months (yes/no).
Physical activity was self-reported and included different activities (eg, household and occupational activities, sports), measured with the Short QUestionnaire to ASsess Health-enhancing physical activity (SQUASH).27 We subdivided the total activity scores into quartiles. Alcohol use was determined with the following categories: no consumption (abstinence from alcohol during the previous 12 months), low consumption (1–2 glasses), moderate consumption (3–4 glasses) and high consumption (5+ glasses), on a typical day when alcohol is consumed. For smoking, participants were classified as current smokers, ex-smokers or never-smokers.
Patient Health Questionaire-9 (PHQ-9) was used to assess the presence of depressive symptoms over the past 2 weeks,28 with sum scores varying between 0 and 27. A cut-off value of ≥10 was used to determine depressive symptoms. In receiver operating characteristic curves in previous studies, the cut-off point of ≥10 tended to have the most optimal sensitivity and specificity for screening for depression.29
Statistical analyses
We first calculated the mean BMI and WC by ethnicity. We used linear regression to assess the association of PED with BMI and WC. The associations differed by ethnicity (WC: p value for interaction (PED×ethnicity) p=0.019; BMI: p=0.058), but not by sex (p value for interaction (PED×sex) for BMI and WC p>0.54), so the main associations were presented by ethnic group. An additional three-way interaction analysis (PED×ethnicity×sex) was statistically significant for BMI and WC (p<0.001). We therefore also presented the sex-specific results by ethnicity. If data were missing for a particular covariate, the participant was excluded in the analyses that included that covariate (eg, participants with missing data on smoking were excluded only from the models that included the smoking variable). The number of missing covariates was very low (<2%) and is therefore unlikely to influence the results.
We used four consecutive models for assessing the association of PED with BMI and WC. For model 1, we adjusted for age and sex. In model 2, we additionally adjusted for potential confounders: marital status, educational level, employment status, domestic stress and any negative life events. We further adjusted for health behaviours in model 3 and for depressive symptoms in model 4 as potential intermediate factors. In order to determine whether PED was differentially related to BMI and WC, the linear regression analyses were repeated using z-score transformations of BMI and WC, ensuring comparability of the regression coefficients of the associations for BMI and WC.
Results
Table 1 shows the characteristics of the study population by ethnicity (see online supplementary material). Turkish and Moroccan participants tended to be younger than the other groups, and were more likely to be married/living with partner and to have no or only elementary education. There were more women than men in each of the five ethnic groups. The prevalence of depressive symptoms was highest in Turkish and Moroccan participants and lowest in the Ghanaians. The median PED score was highest among the African and South-Asian Surinamese, and lowest in Ghanaian and Turkish participants. The mean BMI of all groups was above 25 (ie, overweight), with the highest means in Ghanaian and Turkish participants (around 28.5). The Turkish group also had the highest mean WC. Compared with the other groups, South-Asian Surinamese participants tended to have a lower BMI and WC.
Characteristics of the participants by ethnic minority group*
Table 2 presents the associations of PED with BMI and WC. PED was significantly associated with BMI in the South-Asian Surinamese, African Surinamese and Turkish participants after adjusting for potential confounders (model 2). After further adjustment for health behaviours and depressive symptoms, the associations attenuated slightly but remained statistically significant (models 3 and 4). No significant associations were observed in Ghanaian and Moroccan participants. A somewhat similar pattern emerged for WC, with statistically significant associations in the South-Asian Surinamese, African Surinamese and Turkish participants (model 2). Similar to BMI, the associations between PED and WC changed little after adjusting for health behaviours and depressive symptoms (models 3 and 4). Again, we found no significant associations with WC in Moroccan and Ghanaian participants.
The association of perceived ethnic discrimination (mean score varying from 1 to 5) with BMI and WC by ethnicity
Table 3 presents the associations of PED with BMI and WC using the z-scores. The standardised regression coefficients of both measures tended to be similar in all ethnic groups, with the CIs largely overlapping. This indicates that the strength of the association of PED with BMI and WC is similar across the ethnic groups.
The associations of perceived ethnic discrimination (mean score varying from 1 to 5) with z-scores of BMI and z-scores of WC by ethnicity
Table 4 presents the fully adjusted model with BMI and WC with both β coefficients and z-scores, stratified by sex and ethnicity. There were no important differences across the different models. In men, we found significant associations in Turkish and Moroccan participants. In women, the associations were significant in the African and South-Asian Surinamese participants. As in non-sex-stratified z-score models in table 3, the associations were similar for BMI and WC for men and women.
The associations of perceived ethnic discrimination (mean score varying from 1 to 5) with BMI and WC by ethnicity stratified by sex in the fully adjusted model,† both with and without z-scores
Discussion
This study aimed at investigating whether PED is associated with general and abdominal obesity in different ethnic minority groups in Amsterdam, the Netherlands. We found that the associations varied by ethnicity. Overall, we observed consistently positive associations in the African Surinamese, South-Asian Surinamese and Turkish participants for general obesity (BMI) and abdominal obesity (WC). We further found that positive associations were only observed in Surinamese women and Turkish and Moroccan men. No associations were seen in Surinamese men and Turkish and Moroccan women. In Ghanaian participants, we did not find any associations. The associations of PED with abdominal and general obesity were of similar strength in the ethnic groups.
A particular strength of this study is that it has a large sample size. Another strength of the study is the composition of the multiethnic sample, including groups with a similar migration history and demographics but different ethnic identity (Turkish and Moroccan); groups with similar ethnic identity but different ancestry (African Surinamese and South-Asian Surinamese); and groups with similar ancestry but different ethnic identity and migration history (Ghanaian and African Surinamese; UZ Ikram, MB Snijder, C Agyemang, et al. Ethnic discrimination and metabolic syndrome. Psychosom Med. Forthcoming 2016).
This study also has some limitations. First, the data were cross-sectional, which makes it difficult to assess temporality. However, the few longitudinal studies (from the USA) indicate that racism or perceived discrimination precedes increase in WC6 and obesity.18 It is unknown whether that applies to our findings and Europe-based ethnic minority groups as well. Second, since the EDS, as used in this study, is conceptually based on the discrimination experiences of the African Surinamese and African-Americans, the EDS might possibly not adequately capture the discrimination experiences of other ethnic groups. However, the EDS was shown to be suitable in several ethnic groups in an American sample.23 Similarly, the validity of other variables such as depressive symptoms, domestic stress and adverse life events is currently unknown in our ethnic minority groups. Third, residual confounding might have occurred, as some constructs were not measured comprehensively. For example, the socioeconomic status variable could have included income, which was not assessed in the HELIUS study. Finally, since the participation rate was 50%, the representativeness of our sample may be hampered, thereby potentially affecting the external validity of our findings.
Our results suggest that PED is positively associated with general and abdominal obesity in some ethnic groups, even after partial adjustment for health behaviours and depression. As mentioned in the introduction, this points towards direct or other indirect mechanisms that we have not assessed, such as hypercortisolism or differences in food intake. It could be that among some ethnic groups (eg, Surinamese subgroups) PED is associated with increased allostatic load and hypothalamic–pituitary–adrenal axis dysregulation.11–13 Even small increases in serum cortisol levels contribute to abnormalities in glucose metabolism, similar to metabolic syndrome patients.30 Chronic stress may also lead to changes in food intake. Eating high sugar foods may be a coping mechanism to reduce stress, as it is shown that cortisol responses to a stress test may be attenuated on drinking sucrose-sweetened beverages.31 In addition, our findings imply that PED is similarly related to general and abdominal obesity, and that the strength of the associations is similar for BMI and WC. This suggests that perceived discrimination might affect adiposity (eg, through hypercortisolism) across the whole body, rather than specific areas (eg, abdominal region). Increased fat and sugar intake due to chronic stress could also be equally related to increased BMI and WC.10 ,31
We found ethnic and sex variations in the associations of PED with general and abdominal obesity. This corroborates previous evidence that the association of perceived discrimination with obesity differed by race/ethnicity and sex.16 Reasons for these variations are not clear. These variations can possibly be understood by differences in three areas: perception of discrimination, available psychosocial buffers and biological responses to PED. First, it is possible that some members of ethnic minority groups may perceive ethnic discrimination as more stressful or threatening than others due to differences in migration history and role and stereotypes within the ethnic community and larger society. For example, qualitative evidence from Lithuania suggests that members of the Polish minority perceive discrimination as a hazard to ‘social or material welfare’,32 while in Jews discrimination tends to evoke associations with danger to ‘health and life’ due to the experience of the holocaust.32 Similarly, discrimination in the Dutch context might invoke memories of past injustices associated with slavery and indenture labour during the Dutch Colonial Empire among the Surinamese subgroups.33 In addition, in the USA, gender roles tend to be more egalitarian among African-Americans than Hispanic Americans, who tend to espouse more traditional notions of femininity and masculinity.34 Such differences in cultural notions of gender could potentially make Turkish and Moroccan men more vulnerable to PED than their female counterparts or Surinamese men. Furthermore, ethnic discrimination and sexism may “narrowly intertwine and combine under certain conditions into one, hybrid phenomenon” (ref. 25, p. 31), which Essed25 describes as gendered racism. Negative gendered stereotypes may differ across ethnic groups, and may differently impact the perception of discrimination. For example, while white American women are seen as dependent and monogamous, African-American women tend to be viewed as sexually promiscuous and dominant.25 In the Netherlands, similar stereotypes could disproportionally affect Surinamese women. Negative stereotypes may exist for (mainly Muslim) Turkish and Moroccan men, who are frequently confronted with images of ‘archaic migrant masculinity’ and are framed as dangerous in media rhetoric.35 ,36
Second, differences in available psychosocial buffers may help understand the variations in the associations between PED and obesity. A recent study, for example, showed that strong ethnic identity, large ethnic social network and religiosity weaken the association of PED with depression.37 Brody et al11 found that high levels of social support attenuated the association between perceived racial discrimination and allostatic load in African-American youths. In the Netherlands, Ghanaians have high intragroup cohesion, social support and strong religious involvement within their ethnic group, which may protect them against PED.38 Differences in psychosocial buffers may also potentially explain the sex variations within the Surinamese, Turks and Moroccans, although the evidence base is poor for these groups. For example, single motherhood is quite common in the Surinamese,39 and since single motherhood may be linked with poverty,40 this may weaken the ability of Surinamese women to develop effective psychosocial resources in dealing with PED.
Third, the variations in the association could be explained by differences in biological responses to PED across ethnic groups. Recent findings from the HELIUS study suggest that, for example, while PED is related to elevated WC in South-Asian Surinamese participants, Moroccans tend to respond to PED with reduced high-density lipoprotein cholesterol, rather than increased adiposity (UZ Ikram, MB Snijder, C Agyemang, et al. Forthcoming 2016). In an American sample, perceived discrimination was associated with weaker diurnal cortisol rhythms in white Americans, but stronger diurnal rhythms among African-Americans.41 Such differences in physiological responses might potentially be attributable to epigenetic regulations (eg, fetal programming),42 and life course stress, as a response to environmental stressors in the countries of origin and the host country.43 Variations by sex might be explained by differences in physiological responses to ethnic discrimination in males and females. For example, neuroscientific evidence suggests that central stress responses involve asymmetric prefrontal activity in men, but mainly limbic activation in women.44
In conclusion, our study found that PED was associated with both general and abdominal obesity in some ethnic minority groups, with the strength of association with general and abdominal obesity being similar. The associations varied by ethnicity and sex. Our findings may have public health relevance for the many multicultural European cities. More research is needed to understand these variations. It should be investigated if different ethnic and gender groups perceive ethnic discrimination differently, what different psychosocial buffers they employ in coping with discrimination, and which underlying biological mechanisms are involved in these groups. Further understanding of the association of PED with obesity might potentially help in addressing the ethnic differences in obesity.
What is already known on this subject
There are differences in obesity between ethnic groups in the USA and Europe. According to some but inconsistent evidence, perceived ethnic discrimination is associated with obesity in the USA, with variation across ethnic groups. Evidence on European migrant groups is lacking.
What this study adds
In our European sample, we showed important ethnic and sex variations in the association between perceived ethnic discrimination and obesity. When associations were found, they were similar for general and abdominal obesity.
Acknowledgments
The authors gratefully acknowledge the AMC Biobank for their support in biobank management and high-quality storage of collected samples. The authors are most grateful to the participants of the HELIUS study and the management team, research nurses, interviewers, research assistants and other staff who have taken part in gathering the data for this study.
References
Footnotes
Contributors HS and UZI developed the idea of this study and conducted the data analysis. All authors interpreted the data. HS and UZI wrote the manuscript. All authors commented on the drafts of this manuscript.
Funding The HELIUS study is conducted by the Academic Medical Center Amsterdam and the Public Health Service of Amsterdam. Both organisations provided core support for HELIUS. The HELIUS study is also funded by the Dutch Heart Foundation, the Netherlands Organization for Health Research and Development (ZonMw), and the European Union (FP-7).
Competing interests None declared.
Ethics approval This study was approved by the Ethical Review Board of the Academic Medical Center, University of Amsterdam, Amsterdam, the Netherlands.
Provenance and peer review Not commissioned; externally peer reviewed.