The Role of Ethnicity in Expression, Prevalence, and Severity of Generalized Anxiety Disorder and their Overlap with Depressive Symptoms: Epidemiological Evidence from a Rapidly Developing Middle Eastern Country

Qatar small but high-income country a large population structure and A total of 2,640 participants interviewed by phone using the Generalized Anxiety Disorder (GAD-7) and the Patient Health Questionnaire (PHQ-2). Bivariate and multinomial logistic regression analyses were conducted to explore the associations of ethnicity (Arab versus Non-Arab) with severity and symptom profile of GAD and overlap with depressive symptoms. Results The point prevalence of mild, moderate, and severe symptoms were 13.4% (95% CI: 11.9 – 15.0), 2.7% (95% CI: 2.0 – 3.5), and 0.9% (95% CI: 0.6– 1.4), respectively. Arab ethnicity was associated with mild GAD (OR=1.78, p=0.009) and moderate-to-severe GAD (OR=2.36, p=0.044). Relative to non-migrants, migrant types were not significantly associated with mild or with moderate-to- severe GAD. Interactions between depressive symptoms and ethnicity were evident and statistically significant for the association with mild GAD versus no GAD (OR=0.34, p=0.003). Ethnicity moderated the association of depression symptoms and GAD severity, with potential implications for early screening and community intervention.


Introduction
Anxiety disorders constitute the largest group of mental disorders in most western societies; it is estimated that one in four individuals have experienced an anxiety disorder in their lifetime including generalized anxiety disorder, panic disorder, phobias, social anxiety disorder, obsessivecompulsive disorder, and post-traumatic stress disorder (1)(2)(3)(4)(5).
In contrast, epidemiological data on anxiety disorders remain scarce in the Middle East. Only a handful of studies reported the prevalence of anxiety disorders in war-afflicted countries (6,7); there is also paucity of epidemiological studies from war-free countries Arab countries.
Additionally, significant differences in the prevalence rates of these disorders have been reported among different Arabic speaking populations (8).
Comparison in prevalence estimates between cultures is difficult because of differences in language, assessment instruments, and geopolitical or sociodemographic contexts. However, direct comparison in prevalence rates between different cultural groups may be possible within the same country (9). For example, studies in the United States have shown that White Americans were more likely than minority groups to be diagnosed with most anxiety disorders, except post-traumatic stress disorder, which was most prevalent in African Americans (10,11). However, similar studies in the Arab world are lacking.
A small country in the Arabian Peninsula, Qatar has a unique sociocultural setting with migrants that have very distinct migratory pathways from immigrants to Europe and North America. Qatar's population comprises mostly of expatriate migrant workers from diverse ethnicities and different socio-cultural backgrounds (12). Migrants of Qatar are predominantly from the Indian subcontinent and various Arab nationalities including Egyptians, Syrians, Jordanians, Lebanese, Palestinians, Tunisians, and people from other Arabian Gulf countries (13). Non-migrants or Qataris, who are also of Arab ethnicity, are a minority in the population of Qatar (14).
Anxiety disorders are the most common among mental illnesses in Qatar's primary care centers (15)(16)(17). The lifetime prevalence of generalized anxiety disorder (GAD) reported in this setting ranged between 10.4% (15) and 20.4% (17) and was significantly associated with female and younger age (18 to 34 years). However, generalizability from these studies is limited primarily because only Qataris were included (16). Thus, the role of ethnicity in the epidemiology of GAD symptoms in Qatar including its potential moderating effects on severity and overlap of these symptoms with other psychiatric symptomology such as depression remains largely unstudied.
The role of ethnicity in the epidemiology of GAD is of interest not only for Qatar, but worldwide, because symptoms of GAD that occur with symptoms of other mood disorders such as depression tend to more debilitating and resistant to treatment compared to GAD symptoms that occur on their own (18). Understanding the role of ethnicity in these associations may largely uncover cognitive/ behavioral culture-specific mechanisms (e.g. patterned interpretations or reactions to stressors) that may underlie the expression (types or profile of symptoms most commonly presented) or the development of generalized anxiety symptoms (19), and their overlap with other moodrelated symptoms.
By using a relatively brief measure, the current study aimed to estimate the point prevalence and severity of GAD symptoms in the general population of Qatar and compare its distribution across ethnicity, socio-demographics, migration-, work-, and health-related characteristics in this population. To explore the potential role of ethnicity as a moderator of the associations between severity of GAD and depression symptoms was the secondary aim of this study.

Study Sample
Briefly, a representative sample was drawn from a comprehensive frame of cellphone numbers in Qatar using a list-based probability dialing technique (20). A systematic stratified sample was drawn separately for migrants and non-migrants; weights were constructed to adjust for nonresponse and sampling disproportionality (as non-migrants were the minority and had to be oversampled). The total target sample of completed interviews was 2,252 (see Appendix I) or approximately 750 interviews per group, which was calculated using standard sample size formula for complex survey design (21,22) allowing for statistically adequate comparisons at the 5% significance level between the main three population groups in Qatar: Low-Income Migrants (LIMs), High-Income Migrants (HIMs), and non-migrants or Qatari Nationals (QNs).

Data Collection
After obtaining approval of Qatar University's Institutional Review Board (Reference number: 264-E/13), the interviews were conducted over the phone by trained interviewers at the Social and Economic Survey Research Institute of Qatar University in February of 2018. Verbal consent was obtained from all respondents before starting the interviews.

Translation
The following website (www.phqscreeners.com) provided officially translated versions of the GAD-7 and PHQ-2 free of charge in the following languages: Arabic, Hindi, Malayalam, Tagalog, Tamil, and Urdu. Using internationally accepted translation methodology, the MAPI research institute developed these translated versions (23). The Nepali and Sinhalese versions of the GAD-7 were not available from the same source and were translated by two independent researchers who were fluent in each of these languages and English. The two independent researchers discussed disagreements in translation and resolved these disagreements by consensus for each language.

GAD Symptoms
The seven-item Generalized Anxiety Disorder (GAD-7) is a relatively brief and well-validated screening measure for severity of GAD symptoms used in both clinical and community samples to date (24). Although not a diagnostic instrument, the GAD-7 assesses the frequency over the past two weeks of the seven symptoms criteria for GAD in the Diagnostic and Statistical Manual of Mental Disorders (DSM) version 5 (24) using the following 4-point response options for each symptom: 0 = "not at all", 1 = "several days", 2 = "more than half the days", and 3 = "nearly every day". A total GAD-7 score can be generated for each participant with score range from 0 to 21. Cut-off scores of 5, 10, and 15 denote mild, moderate and severe levels of GAD symptoms, respectively (25,26).Previous studies have shown that GAD-7 exhibits good psychometric properties such as good test-retest reliability of 0.82, internal consistency or alpha of 0.89 (27)(28)(29), and an acceptable sensitivity, but lessthan-ideal specificity (24,27).
In this study, participants were categorized based on these previously established cut-off scores (24) and the distribution of all seven GAD items (dependent variable) in the sample as follows: normal levels = a total score between 0 and 4, mild levels = a total score between 5 and 9, or moderate-tosevere levels of GAD = a total score between 10 and 21. Previous research has shown that the latter levels of severity indexed by a total score of 10 or greater may suggest the need for further evaluation by a health professional (24). For each of the seven items, mild levels of the symptom occurred "several days" in the past two weeks, while moderate-to-severe levels of each symptom occurred "half-of-the-days" or "every day" (24).

Depressive Symptoms
The two-item Physician Health Questionnaire or PHQ-2 was used to measure the frequency of depressed mood and anhedonia in the past two weeks with the following 4-point response options for each symptom: 0 = "not at all", 1 = "several days", 2 = "more than half the days", and 3 = "nearly every day".
The total score range for the two items was 0 to 6. In our study, depressive symptoms were modelled as a continuous composite score and as a dichotomous independent variable using a cut-off score ≥ 3 to denote potentially clinically significant depressive symptoms (29,30). Previous studies support that the PHQ-2 has test-retest reliability of 0.79 and internal consistency or alpha of 0.81 (28).

Ethnicity
Respondents were categorized into: Arab, South Asian, South-East Asian and other (including participants from East Asia, Asia other, Africa, Latin America, Europe, UK, Russia, US, Canada, Australia) based on country of origin and the preferred language for completing the interview (Arabic, Hindi, Malayalam, Tagalog, Tamil, Urdu, Nepali, Sinhalese, and English). These categories were further collapsed into "Arab" and "Non-Arab" as these were the largest two cultural groups that provided sufficient statistical power to conduct the analysis.

Migrant Types
Respondents were classified into one of three main population groups based on self-reported nationality and income. Qatari participants or QNs were classified as non-migrants. All non-Qatari respondents were considered migrants and were either classified as LIMs or HIMs based on their reported income. Participants who had a combined household income of less than $1,100 per month were categorized as LIMs, while those with combined household income greater than $1,100 per month were classified as HIMs.
This income cut-off was previously shown to reliably delineate between LIMs and HIMs in Qatar (31).

Other variables
We also collected standard sociodemographic information (age, sex, education, marital status, employment status) during the phone interview.
Overall health status as measured on a scale ranging from zero to a hundred (32). Quartiles of health rating scores (including poor, fair, good, and excellent) were generated based on the distribution of these scores in the sample. Participants were also asked if they ever been diagnosed with or told by a health professional that they have any of the following health conditions:

1) Diabetes 2) High Blood Pressure or high cholesterol; 3) Asthma; 4) Heart
Disease; 5) Mental or Psychological Problems such as Anxiety, Sleep Problems or Depression; 6) Cancer or Cancerous Tumors; 7) Disability (physical, visual, hearing) or any other condition (please specify); and 8) Never been diagnosed with any health condition.

Statistical Analyses
We estimated the point prevalence of GAD symptoms using weighted proportions and corresponding 95% Confidence Intervals (CIs). We fitted multinomial logistic regression models to the data and estimated odds ratios (OR) and standard errors (SE) for mild and moderate-to-severe symptoms of GAD relative to no GAD symptoms as a function of each independent variable entered into the model. We ran model diagnostics to assess effects of outliers, influential observations, and violations of model assumptions, including the Suest-based Huasmen test for independence of irrelevant alternative assumption (33). Two-tailed p-values < 0.05 were used to indicate statistically significant results.
Only one potential explanatory variable was entered into each univariable model and the corresponding unadjusted OR was estimated for each level of the dependent variable relative to the baseline level (i.e. normal levels of GAD symptoms or score of 0 -4). We also fitted a fully adjusted model by simultaneously entering all of the following explanatory variables into the model: sex, age (continuous), current marital status, migrant status, migrant type (LIMs, HIMs), ethnicity, current employment status, education, net household income, current health rating, and ever diagnosed with a chronic health condition. The contribution of each variable to the final or reduced model was assessed using a variety of fit statistics, including the F-adjusted Wald test and the F-adjusted mean residual goodness of fit test.
We tested statistical interactions between ethnicity and depression symptoms for different levels of GAD severity by adding a two-way interaction terms between ethnicity and depression symptoms to the final reduced model. We evaluated the goodness of fit of the reduced final model with and without these interaction terms using the design-adjusted Wald test. To improve the process of interpreting this interaction between ethnicity and depression on GAD-7 severity, the average marginal effects generated from these models were plotted (34).
All statistical analyses were carried out in STATA version 13 and results were weighted to adjust for non-response and unequal probability of selection. Table 1  Only 7.2% of the respondents reported having poor health, and 20.8%

Sample characteristics
reported diagnosis with a severe health condition. As representative of the general population of Qatar, males outnumbered females by 4 to 1 , while migrants outnumbered non-migrants by 11 to 1 (https://www.psa.gov.qa/en/Pages/default.aspx). Only 24 respondents did not answer all seven questions of the GAD scale. With those who did not answer one or more of the seven questions (n=66), these respondents also were excluded from the rest of the analysis.

Mild levels of GAD and bivariate associations with explanatory variables
Moderate-to-severe levels of GAD and bivariate associations with explanatory variables Like mild GAD, moderate-to-severe levels of GAD ( Unlike mild GAD, moderate-to-severe GAD was not significantly associated with post-secondary education (OR=1.74, p=0.533) and only significantly associated with fair (OR=2.75, p=0.003) and poor health (OR=5.71, p<0.001) relative to excellent health rating, respectively. Multivariable regression models with mild and moderate-to-severe relative to normal levels of GAD are shown in Table 3. Results from fully adjusted and final (reduced) adjusted models were largely consistent for mild and moderate-to-severe GAD levels. Therefore, unless otherwise stated, the results described below are from the final models (see Table 3).  (Table 3). Only poor health relative to excellent health rating remained significantly associated with moderate-tosevere GAD (OR=3.50, p=0.009). Similar to mild GAD, depressive symptoms were significantly associated with moderate-to-severe GAD (OR=3.35, p<0.001).

Interactive Effects of Ethnicity on Associations between Depression and GAD
We found evidence that the association between depressive symptoms and severity levels of GAD differed by ethnicity. The interactions between depressive symptoms that may be clinically relevant and ethnicity were statistically significant for the association with mild levels of GAD (OR=0.34, p=0.003), but not for moderate-to-severe levels of GAD (OR=0.50, p=0.277).
Similar results were observed when depression was treated as a continuous score.
To aid in interpreting these two-way interactions, Figure 1 presents plots of the population-averaged predicted probability for mild and moderate-tosevere GAD as a function of depressive symptoms among Arabs and non-Arabs. Although the predicted probability of mild GAD for Arabs, on average, was significantly lower than non-Arabs for those who were depressed compared to those who were non-depressed, the opposite effects were observed for moderate-to-severe GAD. The average marginal effect of depressive symptoms for Arabs compared to non-Arabs was a statistically significant decrease in the predicted probability of mild GAD by 15.1% (p=0.029). In contrast, the average marginal effect of depressive symptoms for Arabs compared to non-Arabs increased the predicted probability of moderate-to-severe GAD by 4.1% (p=0.569).

Discussion
The latest reported worldwide lifetime, 12-month, and 30-day prevalence estimates of GAD were 3.7%, 1.8% and 0.8%, respectively (5). Of note, these estimates were derived from community surveys that used the WHO Composite International Diagnostic Interview or CIDI (4) and not the GAD-7. Most of the GAD-7 data come from studies conducted in primary care or outpatient clinics in Europe (35)(36)(37). Only three studies validated the use of GAD-7 as a screening tool in general population samples in Lebanon (38).
Currently, no population-based data is available for the GAD-7 from the Middle East. The present study is the first to examine the prevalence and severity of GAD symptoms in a relatively large sample of the general population in Qatar, a rapidly developing and war-free country in the Middle East.
Overall, our findings in relation to significant associations between GAD and female gender are in line with findings from around the world (1,5,8,(39)(40)(41).
Depression symptoms were also highly comorbid with GAD symptoms in our study, which is also consistent with what has been reported in the literature (2,5).
Although both types of migrants (HIMs and LIMs) compared to nonmigrants were positively associated with mild and moderate-to-severe GAD symptoms in our study, these associations were not statistically significant. In contrast, negative associations between migrant status and GAD were previously reported in Canada and Australia (41,42). In our final multivariable model, Arab ethnicity rather than migrant status was significantly associated with higher prevalence estimates of mild GAD (OR=1.78) and moderate-to-severe (OR=2.36) versus normal levels of GAD.
In our study, differences between Arabs and non-Arabs in GAD symptoms were mainly due to temperament. Irrespective of severity, Arabs had slightly higher endorsement of temperament-based (e.g., feeling nervous, anxious, or on edge, difficulty relaxing, or being easily upset or irritable) rather than cognitive-based (excessive anxiety or fear that something terrible would happen) symptoms of GAD than non-Arabs.
This finding may further highlight the role of culture in the etiology of anxiety disorders; specific cultural settings may predispose individuals with high levels of certain temperament traits to GAD symptoms. Existing evidence suggests that temperament plays an important role in the development of GAD (43,44). Specifically, "anxious temperaments" (fearfulness, neuroticism, and negative affectivity) may have genetic ties to anxiety symptoms (45,46) and the development of anxiety disorders in general (47,48). Furthermore, a previous epidemiologic study that incorporated temperament assessment alongside the Arabic version of CIDI 3.0 (49) found a significant association among anxious temperament and DSM-IV mood, anxiety, and impulse control disorders in the Arab context (50).
Our results also provide evidence for ethnicity as a potential moderator of the association between depressive symptoms and severity of GAD symptoms.
For those who were depressed compared to those who were non-depressed, the predicted probability of mild GAD for Arabs, on average, was significantly lower than non-Arabs. Though not statistically significant, but may be of clinical relevance, we observed the opposite effects in the likelihood of experiencing moderate-to-severe GAD symptoms. Compared to non-Arabs, depressive symptoms were more closely associated with moderate-to-severe than mild GAD symptoms in Arabs.
These differences in severity levels of GAD symptoms between Arabs and non-Arabs in our study and their differential associations with depressive symptoms may reflect underlying GAD subtypes and require further future investigation to delineate potential clinical significance of these findings.
The present study offered unprecedented epidemiological data on GAD in a relatively large sample of the general population in Qatar using the GAD-7.
Nevertheless, the study had many limitations. The study was cross-sectional and the data was based on self-report. Previous history of GAD symptoms was not assessed and limited information on functional impairment was collected. Except for the current status of depressive symptoms, the study did not assess the present and past history of other mental illness, including different types of anxiety. This is a major limitation as GAD is known to be commonly comorbid with other mental disorders. The accuracy of GAD-7 as a screener may be largely compromised in the absence of comorbidity information and in the context of a cross-sectional study design. Furthermore, clinical diagnostic interviews for GAD were not included. Given the very high false-positive rate previously reported for using the GAD-7 as a screener in Arabic speaking outpatient sample (38), future population-based studies should further delineate the specificity of this instrument in follow-up clinical interviews. Although the interview questionnaire was thoroughly translated and administered in the most common languages in Qatar, we could not rule out other sources of response bias for job security reasons in Arab and non-Arab migrants if any underlying mental disorder is identified. In addition, whether the GAD-7 questions measured the same concept and were associated with the same level of functional impairment in ethnically diverse groups were not verified in our sample. It is also important to highlight that the DSM-5 criteria for GAD diagnosis require the presence of anxiety symptoms for at least six months, whereas the GAD-7 queries only symptoms present within the last two weeks.
Despite these limitations, our study is the first in the Middle East providing population-based data using the GAD-7. Our preliminary findings support the existence of differences between Arab and non-Arab ethnicities on temperament, but not cognitive items of the GAD-7; better instruments that capture these dimensions of GAD, or the inclusion of comprehensive temperament-assessment scales comprehensive assessment of GAD are needed to replicate these findings in future studies.

Conclusions
Qatar has experienced rapid economic growth, migration, and urbanization in the past three decades; the prevalence estimates of mild and moderate-tosevere GAD symptoms in the general population of Qatar are comparable to estimates previously reported in the Western literature. Findings from this study supported independent associations between ethnicity and GAD.
Prevalence of mild and moderate-to-severe GAD symptoms was higher for Arabs than non-Arabs irrespective of migrant status or migrant type. Arab ethnicity was associated with a higher prevalence of temperament-based symptoms of GAD irrespective of symptom severity. Depression symptoms were more closely associated with moderate-to-severe than mild GAD symptoms in Arabs compared to non-Arabs. Future research is needed to provide additional information on subtype of GAD symptoms in general population samples, which can help tailor screening and interventions for ethnically diverse populations.  Note. All percentages are based on weighted proportions and therefore differ from the raw percentages    is reference category. CI is 95% confidence intervals. All estimates were weighted. The following GAD-7 cut-offs were used to define Normal (total score between 0 and 4), Mild (total score between 5 and 9), or Moderate-to-Severe levels of GAD (total score between 10 and 21). PHQ-2 § Model was based on adjustment for all variables and n=2,071. ¶ Model adjusted was based on n=2,202 and adjustment for all variables with exception of employment status, education and income.