Emotional distress in cancer survivors from various ethnic backgrounds: Analysis of the multi‐ethnic HELIUS study

Insight into emotional distress of cancer survivors from ethnic minority groups in Europe is scarce. We aimed to compare distress levels of survivors from ethnic minorities to that of the majority population, determine whether the association between having cancer (yes vs. no) and distress differs among ethnic groups and investigate sociocultural correlates of distress.


| BACKGROUND
Emotional distress is common among cancer patients. 1,2 It can persist long into survivorship and negatively impact various aspects of quality of life. 3 Insights into emotional distress are therefore crucial to understand and subsequently address cancer survivors' supportive care needs. Of note, emotional distress seems to differ between patients of different ethnic groups: patients from ethnic minority groups have been found to be more distressed than patients from the majority group 4,5 ; but also differences in the other direction have been found. 6 Controlling for sociodemographic or clinical variables seems to explain these differences in some 7,8 but not all 4,6 studies. As studies so far have been mainly conducted in the US and Australia, insights in the European context are lacking. This is alarming as it seems likely that also in the European context emotional distress relevantly differs between ethnic groups. Our first aim is to address this knowledge gap by describing the emotional distress of cancer patients from ethnic minority groups living in the Netherlands and comparing it to that of native Dutch patients, both in non-adjusted and adjusted analyses.
Not only having (had) cancer but also ethnicity seems associated with emotional distress. For example, irrespective of having cancer, Turkish and Moroccan immigrants were found to have a higher prevalence of depressive disorder as compared to native Dutch. 9 Similar results have been found in a study using HELIUS data (of which a subset is used for the analyses reported here, see 'Methods'). 10 The authors found that individuals from ethnic minorities, including Surinamese, Turkish, Moroccan and Ghanaian individuals, had a higher risk of mental health problems as compared to Dutch. While providing valuable insights, most studies investigated the association of cancer with distress and the association of ethnicity with distress separately. It is therefore poorly understood whether having cancer is more distressing for some ethnic groups than for others. Our second research aim is therefore to determine whether the association of having (had) cancer (yes vs. no) and distress differs among ethnic groups.
For cancer patients from ethnic minorities, several sociocultural factors are likely related to distress. Previous studies among minority groups, including cancer studies, found that variables indicative of problems navigating the healthcare system 4 and cultural experiences (i.e. poor acculturation and discrimination) 11,12 are related to more distress while individual resources such as religious coping 13 and social support 14 are related to less distress. To be able to address cancer patients' supportive care needs, correlates of distress need to be identified, also for the European context. Our third aim is therefore to explore correlates of emotional distress in cancer patients from ethnic minorities living in the Netherlands. Correlates of interest pertain to health care, cultural adaptation/experience and individual coping resources.

| METHODS
See Textbox S1 for the terminology and a description of the relevant ethnic minority groups.

| Setting
Data were derived from the HELIUS (HEalthy LIfe in an Urban Setting) study, a multi-ethnic study conducted in Amsterdam, the Netherlands. 15 Figure 1 for the flowchart.

| Measures
Participants with Surinamese, Ghanaian and Turkish ethnicity received the questionnaire in their mother tongue, that is, Dutch, English and Turkish, respectively. For participants with Moroccan ethnicity, the questionnaire was not translated as most Moroccans are unable to read the official Moroccan language (Moroccan Arabic) and their spoken language (Berber) consists of many different dialects. Assistance from a trained and ethnically matched interviewer was available for participants unable to complete the questionnaire themselves.

| Ethnicity
Ethnicity was defined by participants' and their parents' country of birth (as derived from the municipality register of Amsterdam).
Participants of Dutch ethnic origin are those who themselves and F I G U R E 1 Flow chart. Of the 90,019 persons invited for participation in HELIUS, 24,789 participants provided baseline data, that is, underwent physical examination and/or completed the baseline questionnaire (28% response rate). Please see Snijder and colleagues 15 for more details on the recruitment and participant flow. NCR = Netherlands Cancer Registry. whose parents were born in the Netherlands. Participants of non-Dutch ethnic origin are those who were born outside the Netherlands and have at least one parent born outside the Netherlands (first generation) or those who were born in the Netherlands but whose both parents were born outside the Netherlands (second generation). The differentiation between African versus South-Asian Surinamese was made based on participant self-report.

| Cancer-related variables and morbidity
Matching with NCR data was performed within a previous project, 17 and permission from NCR was obtained to use these cancer-related variables for the current study. Only tumors diagnosed prior to the baseline assessment and those labeled as malignant were included.
For participants with multiple malignant tumors (n = 24), data were extracted for the most recently diagnosed cancer.
Morbidity was self-reported as presence versus absence of 19 predefined chronic conditions (excluding cancer, as this was operationalized with NCR data).

| Demographics
Age and sex were derived from the municipality register. Relationship status, education, employment and occupational level were selfreported.

| Putative sociocultural correlates
Correlates related to navigating the health care system were operationalized as health literacy and satisfaction with one's general practitioner (GP). The former was assessed with the Set of Brief Screening Questions. 23 Higher mean scores represent better health literacy (range [1][2][3][4][5]. Satisfaction with the GP was measured with 1 item ('How satisfied are you with your GP?'). Higher scores indicate more satisfaction (range [1][2][3][4][5]. Predictors related to adaptation to and experiences with the host culture were operationalized as acculturation, perceived discrimination and several proxy measures. Acculturation was assessed with regard to a participant's ethnic identity, cultural orientation and social network. Scores were dichotomized into being adapted to the Dutch culture (Berry's acculturation strategies of integration and assimilation 24 ) versus not being adapted to the Dutch culture (separation and marginalization). Perceived discrimination was assessed with the Everyday Discrimination Scale. 25 Higher sum scores indicate greater perception of discrimination (range 9-45).
Proxy measures for acculturation include age of migration, residence duration (both in years) and self-reported difficulty with the Dutch language (yes vs. no).
Individual resources were operationalized in terms of support and religiosity. Emotional support was measured with the daily emotional support subscale of the interconnected Social Support Questionnaire for Transactions (SSQT) and the Social Support Scale for Satisfaction with the supportive transactions (SSQS). 26 Answers were scored on a 4-point Likert scale ranging from 1 (seldom or never/much less than I like) to 4 (often/more then I like). Answer categories 3 (just as much as I like) and 4 (more than I like) were combined to represent sufficient support. 26 A higher sum score indicates more supportive emotional transactions (SSQT, range [5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20] and more satisfaction with the support (SSQS, range 5-15). Religiosity was assessed in terms of whether one was raised according to a specific religion, one's current religion and the frequency of practicing ones religion (How often have you attended a religious service in the past 6 months (in a Church, mosque, synagogue, etc.)?), dichotomized into less versus at least once a month.

| Statistical analyses
First, possible violations of assumptions for regression analyses were tested. Next, differences in participant characteristics between Dutch versus each non-Dutch ethnic group were tested with independent samples t-tests (continuous variables) and chi-square tests for independence (categorical variables). In the same way, we tested the difference between participants who were included in versus excluded from analyses of the current paper.
To address research aim 1, participants with cancer were selected. General linear models were conducted with PHQ-9 or MCS-12 as dependent variable and ethnicity as independent variable.
The models were run as unadjusted (analysis of variance) and adjusted analyses (analysis of covariance). For the latter, models were run for each covariate (i.e. morbidity, demographics and cancerrelated variables as presented in Table 1) that also included ethnicity.
Covariates with p < 0.2 were entered simultaneously into one model, Given the small size of some groups, a fixed and limited set of covariates was selected.
To increase the statistical power for the analyses pertaining to research aim 1 and 2, we also conducted the analyses with data of the non-Dutch ethnic groups pooled into a single group (Dutch vs. non-Dutch). Also for these analyses, covariates were selected based on p < 0.20 and backward deletion.
To address research aim 3, General linear models were conducted with data of participants with cancer and those of non-

| RESULTS
See Table 1 for a sample description and supplementary Table S1 for descriptives of key variables. Participants excluded from analyses (for any reason, see flowchart) were younger, more often Moroccan and Turkish and less often Dutch and Ghanaian than those included (differences with at least small effect size; see Table S2). As the assumption of homoscedasticity appeared violated, p-values based on robust standard errors are reported for the main analyses.

| Research aim 1
Analyses of (co)variance revealed that ethnicity was a significant predictor of depressive symptoms and mental health, see

| Research aim 2
Models were adjusted for sex, age, morbidity and education. For depressive symptoms, the interaction term between ethnicity and cancer diagnosis was only significant for the comparisons between

F I G U R E 2 Significant interaction between cancer and ethnicity
for the outcome depressive symptoms. The model was adjusted for sex, age, (co-)morbidity and education. The PHQ-9 ranges from 0 to 27. The plots of the remaining five significant interactions are displayed in the supplement, see Figure S1.
T A B L E 2 Comparison of emotional distress across ethnic groups (research aim 1). For analyses conducted with non-pooled data, the model testing the effect on PHQ-9 controls for sex, employment, and co-morbidity; the adjusted model testing the effect on MCS-12 controls for age, sex and co-morbidity. For analyses conducted with the pooled data, the model testing the effect on PHQ-9 controls for age, sex, employment and co-morbidity and the model testing the effect on MCS-12 controls for age, education, employment and co-morbidity.

Unadjusted analyses Adjusted analyses a
South-Asian Surinamese (b = −1.76, t (6891) = −1.29, p = 0.198) and Turkish participants (b = −4.11, t (7227) = −1.84, p = 0.65), having a cancer diagnosis, as opposed to being cancer-free, was associated with worse mental health, albeit not significantly. In the analysis in which all non-Dutch groups were pooled, the interaction term was significant (p = 0.001). Simple slope analysis revealed that in Dutch, those having (had) cancer, as opposed to those who were cancer-free, had better mental health (b = 2.236, p = 0.001) whereas there was no significant difference in mental health between non-Dutch individuals with and without cancer (b = −0.850, p = 0.208).

| Research aim 3
Most interactions between ethnicity and sociocultural correlates of  Table 3 for details.

| Sensitivity analyses
Results of analyses performed with bootstrapping and those with another strategy to select covariates (aim 1) had the same pattern as those presented here.

| DISCUSSION
Cancer survivors from most ethnic minority groups in the Netherlands reported more emotional distress than survivors of Dutch origin. After adjusting for demographics and morbidity, most between-group differences remained, yet decreased in size. Similar results have been found among cancer survivors in the US 6 and, in the European context, among migrants. 27 These studies suggest that worse outcomes of ethnic minority groups are not solely a function of their adverse socioeconomic or clinical status and warrant further investigation.
The association between having (had) cancer (yes vs. no) and emotional distress differed by ethnicity, as has been found in a US study. 28 Dutch survivors had similar or even lower levels of emotional distress than their cancer-free peers. Good outcomes of (long-term) cancer survivors have been found previously. 29 Lower levels of health literacy and social support were associated with more distress, as has been found previously. 14,[35][36][37] In order to tailor supportive care, future research should assess specific types and sources of social support. For example, in some cultures, family members might be inclined to engage in overprotection (i.e. providing an excessive amount of support to the patient) or protective buffering (i.e. hiding one's worries and concerns from the patient), which both have been associated with poor outcomes. 38 A higher self-reported frequency of attending a religious service was associated with worse mental health. This finding seems counterintuitive, as prior studies show that spirituality/religious coping among cancer survivors from ethnic (minority) groups is associated with better outcomes. 13  -1419 An unexpected null-finding pertains to perceived discrimination.
The literature consistently shows that discrimination is a predictor of distress and related outcomes, 12 including a previous analysis of HELIUS data. 41 Also in bivariate models, the associations between discrimination and our outcomes were not significant, albeit almost for mental health (data not shown). The selected, small sample might explain this.

| Limitations
This study's findings should be interpreted in light of its strengths and Note: Reference categories: For religiosity: less than once a month or never; for co-morbidity: no co-morbidities; for both acculturation style variables: integration/assimilation. a For the outcome PHQ-9, there were significant interaction effects between ethnicity and two acculturation variables. For these variables, we ran models for each ethnic group separately and report those that provided a significant effect. b Barring acculturation style (Cultural orientation and Social Network, see a) there were no significant interaction effects between ethnicity and the putative correlates. Therefore, we pooled the non-Dutch groups for the remainder analyses.
cancer recurrence) at clinically relevant time points is expected to shed a more nuanced light on between-group differences in outcomes and their developments. Further, as the data are cross-sectional, no conclusions about causality or temporal order can be drawn. Results of the MCS-12 for the Ghanaian group should be interpreted with caution: earlier research found a violation in measurement invariance for mental health in the Ghanaian group which might have led to an overestimation of the differences found. 22 Given the scarcity of research in this field in the European context and the limitations noted above, the authors encourage corroborating and extending studies.

| Clinical implications
Cancer patients from ethnic minorities are at risk to experience emotional distress, more so than patients from the majority group.
Health care providers need to be aware of this vulnerability and consider referral to supportive care where needed. Supportive care should be provided in a culturally sensitive way as has been shown effective in mental health care. 42,43 Such interventions could be targeted at correlates of emotional distress that are amenable to change, such as emotional support.

| CONCLUSION
This study is among the first in the European context to investigate disparities in emotional distress between cancer patients of various ethnic groups. Patients from most ethnic minority groups were found to experience more emotional distress than native Dutch patients.
These differences could not fully be explained by co-morbidity and demographic variables. In individuals from ethnic minority groups, those with cancer tended to be more distressed than their cancerfree peers. In Dutch, those with cancer tended to be less distressed than their cancer-free peers. Within the group of non-Dutch cancer patients, correlates of emotional distress were identified. Some of these correlates are amenable to change and are therefore potential targets for culturally sensitive care.