Epidemiology of Diabetes Among Arab Americans

The study was reviewed and approved by the Wayne State University Institutional Review Board, and all subjects provided written informed consent. The study was conducted in two areas of Dearborn, Michigan, totaling two square miles, that are primarily inhabited by Arab Americans. The study population included randomly selected nonpregnant adults 20–75 years of age with native Arab ancestry. An initial sampling frame of households was constructed using city tax rolls. This list captured addresses of all residential and commercial units. Study staff then enumerated all 2,813 housing units in the areas, including both single and multiple residency housing units. Each household was assigned a sequential number, and a randomized list of households was generated using random numbers. Phone number and family name of head of household for these units were obtained from a variety of sources. Letters were mailed to the identified households, and within a week of these mailings, each household was contacted in person by a trained bilingual interviewer to determine ancestry and to identify the members of the household. If there were no members of Arab ancestry, the household was replaced by the next randomized household. An adult member of the eligible household was then interviewed, and all members of the household meeting the inclusion criteria at the time of screening were enumerated by age.

To select subjects, the interviewer opened a sealed envelope that contained a list of possible household sizes and the index numbers of the members who should be chosen for each household size. If there were one or two eligible members in the household, only one individual was invited to participate. If there were more than two, one or two were invited to participate according to a prespecified algorithm. If a member of a multi-person household refused, the interviewer noted this and chose the next number on the list; no further replacements were permitted. All individuals selected and willing to participate were then scheduled to be seen at a local ACCESS health clinic for testing.

On the assigned study day, subjects reported to the health clinic in the morning after a 12-h overnight fast. Standardized questionnaires translated into Arabic were used by interviewers to assess demographic and socioeconomic factors including ethnic background (country of origin), employment history, and educational status. Arab Americans were defined by self-report of Arab ancestry of the individual, a parent, or a grandparent. Individuals who were born in an Arab country and immigrated to the U.S. were considered immigrants. The age of immigration and length of residence in the U.S. were recorded. A standardized medical history was obtained. Height, body weight, and hip and waist circumferences were measured in light clothing and without shoes. Waist-to-hip ratio (WHR) (the ratio of waist to hip circumferences) and BMI (body weight in kilograms divided by the square of height in meters) were calculated. Obesity was defined as a BMI ≥30 kg/m². Central body fat distribution was defined as a WHR ≥0.8 for women and ≥0.9 for men.

Fasting blood samples were collected for measurement of plasma glucose and HbA_1c. Subjects who were not fasting were rescheduled on another day. All subjects without a documented history of diabetes received a 75-g oral glucose load over 10 min; blood samples were collected at 120 min for measurement of plasma glucose. All blood samples were centrifuged for 10 min, and plasma was stored at −70°C. Whole-blood samples for HbA_1c were stored at 4°C. Plasma glucose concentrations were measured by an automated glucose oxidase method using Behring Diagnostics Reagents (SVR Glucose Test). HbA_1c was measured in a Cobas Mira Plus analyzer with a Unimate HbA_1c kit calibrated to the Diabetes Control and Complications Trial methodology standards (Roche Diagnostics, Indianapolis, IN). The mean normal HbA_1c was 5.1% (95% CI 4.5–5.7%).

Individuals were considered to have diabetes if they reported a previous medical diagnosis of diabetes and/or were using insulin or oral antidiabetic agents. Fasting plasma glucose (FPG) ≥126 mg/dl (7.0 mmol/l) and/or 2-h plasma glucose ≥200 mg/dl (11.1 mmol/l) were considered diagnostic of diabetes. An FPG ≥110 and <126 mg/dl (6.1–7.0 mmol/l) was considered diagnostic of impaired fasting glucose (IFG). A FPG <126 mg/dl (7.0 mmol/l) and a 2-h plasma glucose ≥140 mg/dl and <200 mg/dl (7.8–11.0 mmol/l) indicated the presence of IGT. Subjects were designated as having normal glucose tolerance (NGT) if FPG was <110 mg/dl (6.1 mmol/l) and the 2-h plasma glucose was <140 mg/dl (7.8 mmol/l). A family history of diabetes was considered present if any of the subjects’ grandparents, parents, siblings, or children had diabetes.

The primary aim of this study was to determine the prevalence of diabetes and glucose intolerance among Arab Americans. With a sample size of 400 subjects, the maximal standard error for the estimate of prevalence would be 2.5% and the CI for the estimate would be ±5%. Because there are factors, such as genetic predisposition and diet, that are common to members of a family unit, family members from 400 households were to be recruited.

We sampled two members from multi-person households for the following reasons: 1) It increased the likelihood that at least one member from that household would participate. We projected a 75% recruitment rate. 2) It increased the sample size and hence reduced the variability. 3) The demographics of multi-person households differ from single person households. We did not want to underrepresent the characteristics of subjects in multi-person households.

Data entry and verification were performed by Automated Resource Management, Inc., in Ann Arbor, Michigan. Statistical analyses were performed using SAS by the Biostatistics Core of the Michigan Diabetes Research and Training Center.

Because sampling of households was random, the enumeration of all subjects in each household was a random sample of the population in these areas. All estimates reported in this article use weights determined by the age/sex distribution of the enumeration. In this manner, the estimates of prevalence are appropriate for the population of interest. The weights for men and women were as follows: 20–29 years of age, 0.145 for men and 0.154 for women; 30–39 years of age, 0.161 for men and 0.112 for women; 40–49 years for age, 0.094 for men and 0.097 for women; 50–59 years of age, 0.058 for men and 0.059 for women; and ≥60 years of age, 0.056 for men and 0.064 for women. When analyses were by sex, the weights for each sex were normalized to a sum of 1.

An enumeration sampling list of 2,813 residential households was constructed from tax rolls and screened for eligibility. A random sample of 942 households was selected from the sampling list. Of these, 415 households were excluded because 1) they did not have household members of Arab ancestry (84% of those excluded), 2) they were unoccupied or nonexistent (13%), 3) all members of the household did not meet the age criteria (1%), or 4) the interviewer was unable to establish contact despite multiple visits on different days and times (1%).

Of the 527 remaining households, 459 were willing to participate (87% household response rate). There were 1,096 eligible subjects in these households (562 men, 534 women). A total of 626 subjects (274 men, 352 women) were randomly selected and 542 (214 men, 328 women) gave informed consent and agreed to be tested. The subject response rate was lower for men (78%) than for women (93%). Among the 542 individuals tested, 46 refused the oral glucose tolerance test and only a fasting blood sample was obtained.

Demographic characteristics of the study population are presented in Table 1. The mean age of subjects was 38 years. Most (95%) were immigrants with a mean length of stay in the U.S. of 11 years. The majority were Lebanese (65%), but there were also large numbers of Yemenis (15%) and Iraqis (9%). More men had completed high school than women (69 vs. 49%). Eighty percent of men and 19% of women were employed; 73% of the women described themselves as homemakers.

The prevalences of diagnosed and undiagnosed diabetes, IFG, IGT, and combined glucose intolerance are shown in Table 2 by age and sex. The age-adjusted rate of diabetes was 15.5% (95% CI 12.2–18.7%) in women and 20.1% (15.0–25.2%) in men. Among those with diabetes, 47.8% of the women and 57.2% of the men were undiagnosed. The prevalence of undiagnosed diabetes (9.5% [95% CI 7.0–12.1%]) was similar to that of diagnosed diabetes (8.3% [6.2–10.4%]). Only two subjects met epidemiological criteria for type 1 diabetes, with age at onset <30 years and current insulin therapy. Only one subjects, a 29-year-old man, had a fasting C-peptide level consistent with type 1 diabetes. In general, the prevalence of diabetes increased with age in both sexes and was consistently higher in men than in women 20–49 years of age. The age-related increase in prevalence was most prominent in 40- to 49-year-old men and in women ≥50 years of age. The age-standardized prevalence of diabetes (either previously diagnosed or newly diagnosed) was 14.5% (95% CI 11.1–17.9%) in Lebanese, 25.5% (17.4–33.7%) in Yemenis, and 26.2% (15.3–37.1%) in Iraqis.

The prevalence of IGT/IFG was much higher in men (29.7% [23.4–35.9%]) than in women (16.8% [12.8–20.8%]). The overall prevalence of IGT/IFG was significantly greater in men than in women regardless of age. The age-adjusted rates of combined glucose intolerance were 32.3% (27.8–36.7%) for women and 49.8% (43.1–56.4%) for men (P < 0.0001).

Selected demographic and biomedical indicators of the populations with NGT, IGT/IFG, and diabetes are presented in Table 3. Subjects with diabetes and IGT/IFG were older and had greater BMI and WHR values than subjects with NGT. Mean BMI and WHR levels were highest in subjects with diabetes, intermediate in subjects with IGT/IFG, and lowest in the NGT group. HbA_1c levels were highest in subjects with diagnosed diabetes (8.0 ± 1.8%), intermediate in subjects with undiagnosed diabetes (5.9 ± 1.1%), and lowest in subjects with IGT, IFG, and NGT (5.3 ± 0.4, 5.3 ± 0.4, and 5.0 ± 0.4%, respectively). FPG levels were 196 ± 63 mg/dl in subjects with diagnosed diabetes, 133 ± 40 mg/dl in subjects with undiagnosed diabetes, and 102 ± 11, 115 ± 4, and 96 ± 8 mg/dl in subjects with IGT, IFG, and NGT, respectively.

This study provides the first representative, population-based estimates of the prevalence of diabetes and other categories of glucose intolerance in an Arab population outside the Middle East. Abnormal glucose tolerance is common in Arab Americans, affecting 41% of the population 20–75 years of age and >70% of those over 60 years of age. Diabetes (either diagnosed or undiagnosed) affects 18% of subjects aged 20–75 years and 36% of men and 54% of women over the age of 60 years. These rates are considerably higher than those reported for the white, African-American, and Hispanic populations in the U.S. and for rural Arab populations (2,8). They are, however, consistent with rates reported in urban Arab populations (2,9–15).

The prevalence of diabetes in Arab countries varies from 3% in Sudan to 35% in Bahrain (1,2,9–15). Rapid economic development and increasing longevity have been associated with the increasing prevalence of diabetes witnessed over the past 2–3 decades in many Arab countries (1). Higher rates of glucose intolerance are consistently observed in urban areas and have been associated with sociodemographic transformation involving changes in nutritional patterns, physical activity, and obesity (1,2).

The ethnic distribution of our study population reflects the heterogeneity of Arab Americans. Intragroup differences in diabetes reflect the diversity of the genetic composition, cultural backgrounds, and socioeconomic status of the population. The prevalence of diabetes of 14.5% among Lebanese Americans is similar to the rate of 13% reported in a study of 2,518 subjects aged ≥30 years in Lebanon (15).

Factors responsible for the high prevalence of diabetes in Arab Americans are not yet clear. Obesity, defined as BMI ≥30 kg/m², was seen in 34% of the Arab Americans studied, a rate higher than the 26% reported for the U.S. population. Both obesity and central obesity increased with age (from means of 26.6 kg/m² and 0.86, respectively, for younger men to 27.7 kg/m² and 0.93 for older men, and from means of 26.4 kg/m² and 0.79 for younger women to 27.7 kg/m² and 0.88 for older women). In addition, both obesity and central obesity were associated with abnormal glucose tolerance, especially in women. Although diabetes was associated with age in both sexes, age- and sex-related differences in diabetes prevalence and risk factors were evident. In men, diabetes prevalence reached a plateau in the group 40–49 years of age and was relatively unchanged thereafter. In contrast, in women, the prevalence of diabetes was low below the age of 40 years and rose steadily thereafter. The mean age difference between NGT and diabetes was about 10 years in men and 20 years in women. It is plausible that menopause-associated changes in adiposity and insulin sensitivity may be a risk factor for diabetes in women. Finally, diabetes was associated with low employment rates in men and women and lack of a high school education in women. We are currently studying the relevance of these associations and their relation to the process of acculturation.

Our data demonstrate an association between diabetes in men and reported maternal history of diabetes. Maternal history of diabetes was present in 32% of the men with diabetes but in only 15% of those with NGT (P = 0.020). Paternal history was reported in 13 and 10%, respectively. No differences were seen in women with diabetes. Similar observations of excess prevalence of diabetes among siblings and offspring of women with diabetes compared with men with diabetes have been reported (16–18). This maternal transmission of diabetes may reflect sex differences in reporting and recall biases of family history of diabetes (19). However, evaluation of sex-specific parental history of diabetes in nondiabetic subjects revealed reduced insulin responses to oral glucose loads in male offspring of mothers with diabetes compared with those with paternal history (20). Whether the greater maternal history of diabetes in men observed here represents a genetic predisposition in this population is beyond the scope of this study. Certainly, a prospective evaluation of excess maternal transmission of diabetes in this population is warranted. The associations between diabetes prevalence and acculturation, diet, physical activity, education, socioeconomic factors, and perceived stress in this migrant community are currently being investigated.

The prevalence of undiagnosed diabetes among Arab Americans was 10% and was consistent with the low detection rates of diabetes in Arab countries. Multiple factors may have contributed, including lack of access to and use of health care, physician practices, and culturally related health care practices and beliefs, including fear of uncovering medical problems.

The high prevalence of IGT and IFG in Arab Americans is of prognostic importance. IGT and IFG are major risk factors for developing diabetes. Indeed, 15–30% of such individuals are likely to progress to type 2 diabetes over 5 years (21). Considering the relatively young age of the population with IGT and IFG, programs targeting prevention are imperative. Interestingly, the age-standardized prevalence of IGT and IFG was about twice as common in men (29.7%) as in women (16.8%).

In conclusion, this study shows that diabetes is a major health problem in Arab Americans. Relatively high rates of undiagnosed diabetes and abnormal glucose tolerance have substantial clinical relevance. Given the high prevalence of diabetes and impaired glucose intolerance together with low detection rates and the relative youthfulness of the population, it seems reasonable to expect that the future prevalence of diabetes will continue to increase as the population ages. Further increases in prevalence of diabetes will impose a substantial public health burden and present a challenge to health care providers. There is a need for increased public awareness and regular surveillance for diabetes and its complications in this community. It is imperative to develop culturally sensitive community-based intervention strategies aimed at prevention and management.

This study was funded by the American Diabetes Association. Support for the University of Michigan collaborators was provided by the Michigan Diabetes Research and Training Center, National Institutes of Health Grant P60 DK20572.

We acknowledge the dedication of Zahera Zahreldein, Zeinab Ajrouch, and Jawaher Adouh for assistance in fieldwork. We are grateful for the cooperation and support of the ACCESS staff and all the Arab Americans who participated in the study. We also acknowledge the valuable contributions of Drs. Donna Leonetti and Wilfred Fujimoto of the University of Washington for assistance with developing the surveys.