Correlates of Pre-Diabetes and Type 2 Diabetes in US South Asians: Findings from the Mediators of Atherosclerosis in South Asians Living in America (MASALA) Study

Abstract

Purpose

In this study, we aim to elucidate the role of sociodemographic, lifestyle and cultural factors in pre-diabetes and diabetes in South Asian immigrants to the United States (US), a population at high risk of type 2 diabetes.

Methods

We performed a cross-sectional analysis of a community-based cohort of 899 South Asians without known cardiovascular disease from the Mediators of Atherosclerosis in South Asians Living in America (MASALA) study. Glycemic status was determined by fasting glucose, 2-hour post-challenge glucose and use of diabetes medication. We used multinomial logistic regression models to estimate the independent associations of sociodemographic, lifestyle and cultural factors with pre-diabetes and diabetes, adjusting for confounders identified using directed acyclic graphs.

Results

Approximately 33% of participants had pre-diabetes and 25% had diabetes. In multivariate analyses, an independent correlate of pre-diabetes was low exercise. Additional covariates associated with diabetes included: lower family income, less education, high chronic psychological burden score, and greater time spent watching television, and fasting monthly or annually was inversely associated with diabetes prevalence.

Conclusions

We found several modifiable risk factors associated with pre-diabetes and diabetes that may help guide diabetes prevention interventions for South Asian immigrants to the US.

Introduction

The prevalence of type 2 diabetes in individuals of South Asian origin is rising. (1) In India alone, there are over 65 million people with diabetes, making it the country with the second highest number of cases worldwide. (2) South Asians have a higher diabetes prevalence compared to most other racial/ethnic groups (3), as well as more cardiovascular disease (CVD) complications with diabetes (4, 5), and a higher mortality rate mainly due to higher rates of CVD. (6) Understanding the drivers of increased diabetes risk in South Asians is important for improving prevention and treatment options for this high-risk population.

South Asians’ increased cardiometabolic risk is multi-factorial reflecting a mixture of genetic, environmental, and lifestyle factors. (7) Greater visceral adiposity, insulin resistance (IR) and impaired β-cell function are known to contribute to the increased diabetes risk in South Asians. (7) Urbanization and immigration are also contributory factors with an observed gradient of higher diabetes prevalence in urban Indian settings compared to rural areas (8), and even higher diabetes prevalence with immigration to more affluent countries such as the United States (US) and the United Kingdom (UK) which may be attributed to diet and physical activity changes and psychosocial stressors. (7) However, few studies have measured several lifestyle, behavioral, psychosocial and biologic factors concurrently in immigrant South Asians in the US.

Less is known about the association between non-biologic factors and diabetes risk among South Asians. Therefore, we aimed to determine the non-biologic correlates of pre-diabetes and type 2 diabetes in a community-based cohort of middle-aged South Asians in the US, which is representative of the US South Asian population. (9) We hypothesized that sociodemographic, cultural, lifestyle, and psychological factors would be associated with diabetes in South Asians. Gaining a better understanding of modifiable risk factors for diabetes in US South Asians can help guide the delivery of tailored interventions to decrease their diabetes risk.

Materials and Methods

Study Design

We performed a cross-sectional analysis of a community-based cohort of South Asians without known cardiovascular disease from the Mediators of Atherosclerosis in South Asians Living in America (MASALA) study. The MASALA study is modeled on the Multi-Ethnic Study of Atherosclerosis (MESA) study with similar recruitment methods, eligibility criteria, questionnaire and clinical measurements. Detailed MASALA study methods have been published. (9)

Study Subjects

To be eligible for the study, participants had to self-report South Asian ethnicity, be between the ages of 40–84, and be able to speak and/or read English, Hindi or Urdu. (9) Exclusion criteria included a physician diagnosed heart attack, stroke or transient ischemic attack, heart failure, angina, use of nitroglycerin; a history of cardiovascular procedures (coronary artery bypass graft, angioplasty, valve replacement, pacemaker or defibrillator implantation, or any surgery on the heart or arteries); current atrial fibrillation; active treatment for cancer; life expectancy < 5 years due to a serious medical illness; impaired cognitive ability; plans to move out of the study region in the next 5 years; living in a nursing home or on a waiting list; and weight > 300 lbs. (9)

Study subjects were recruited from two clinical sites – the San Francisco Bay Area through the University of California, San Francisco (UCSF) and the greater Chicago area through Northwestern University (NWU). Sampling methods have been reported. (9) Between October 2010 and March 2013, a total of 906 South Asian men and women were enrolled – 496 were enrolled at the UCSF site and 410 were enrolled at the NWU site. (9)

Study Measurements

All participants completed a detailed questionnaire to ascertain sociodemographic information, medical history, family history, medication use, cultural practices, and behaviors. (9) Physical activity was assessed using the Typical Week’s Physical Activity Questionnaire. (10) Dietary intake over the previous year was assessed using the Study of Health Assessment and Risk in Ethnic Groups (SHARE) food frequency questionnaire, which was created and validated among South Asians in Canada. (11) Several psychosocial scales were administered including the Spielberger trait anxiety scale (12), the Center for Epidemiologic Studies depression scale (13) and one to measure chronic psychological burden. (14) Traditional Indian beliefs were examined using a 7-item scale from prior qualitative research and scored using a Likert scale with higher scores representing weaker traditional beliefs. (15)

Clinical Measurements

Participant weight was measured on a standard balance-beam scale or digital weighing scale and height was measured using a stadiometer. Weight and height measurements were used to calculate body mass index (BMI). (9) Waist circumference was measured using a flexible tape measure at the site of maximum circumference midway between the lower ribs and the anterior superior iliac spine. (9) Abdominal computed tomography (CT, Philips Medical Systems, Andover, MA; Toshiba Medical Systems, Tustin, CA; Siemens Medical Solution, Malvern, PA) was used to determine abdominal visceral and subcutaneous fat area. A trained radiology technician used a lateral scout image of the spine to establish the correct position (between the L4 and L5 vertebrae) for the abdominal CT using standardized protocols. (9) Visceral and subcutaneous abdominal fat were measured at the L4–L5 level using the Medical Image Processing, Analysis, and Visualization (MIPAV) software. (16) The subcutaneous compartment is composed of tissue outside the visceral cavity but within the body contour. The muscles in the abdomen were segmented and then omitted from the calculation of subcutaneous fat. Visceral fat was defined as those pixels within the appropriate Hounsfield Unit (HU) range and within the contour of the visceral cavity.

Seated resting blood pressure was measured 3 times using an automated blood pressure monitor (V100 Vital Signs Monitor; GE Healthcare, Fairfield, CT) with the average of the last 2 readings being used for analysis. (9) Hypertension was defined as a systolic blood pressure (SBP) ≥140 mmHg and/or diastolic blood pressure (DBP) ≥90 mmHg or use of anti-hypertensive medication.

Blood samples were obtained after a 12-hour fast. Total cholesterol, triglycerides, and high-density lipoprotein cholesterol (HDL-C) were measured using enzymatic methods, and LDL-cholesterol (LDL-C) was calculated. Alanine aminotransferase (ALT) and gamma glutamyl transferase (GGT) were measured using spectrophotometry (Quest Labs, San Jose, CA). Adiponectin and resistin levels were measured using the Millipore Luminex adipokine panel A (EMD Millipore; Billerica, MA). The inter-assay coefficient of variations (CV) for was 2.34–4.12% for adiponectin and 3.25–5.03% for resistin.

Glucose Metabolism Measurements

After obtaining fasting blood samples, a 75-g oral glucose load was administered to participants who were not taking diabetes medications and blood samples were drawn 30-minutes and two hours following the glucose challenge. Fasting plasma glucose and 2-hour post-challenge glucose were measured using a hexokinase method (Quest Labs, San Jose, CA). Fasting serum samples were batched for insulin measured by the sandwich immunoassay method (Roche Elecsys 2010, Roche Diagnostics, Indianapolis, IN). The homeostasis model assessment (HOMA)-IR was used to measure IR and calculated as [Insulin₀(μIU/ml) × Glucose₀(mmol/l)/22.5] and HOMA-β was used to measure β-cell function and was calculated as [20 × Insulin₀(μIU/ml)/ Glucose₀(mmol/l)−3.5]. (17) We excluded participants taking exogenous insulin from the analyses of HOMA-IR and HOMA-β.

Individuals were categorized with normal glucose tolerance (NGT), pre-diabetes or diabetes based on the American Diabetes Association (ADA) criteria. (18) NGT was defined as having a fasting plasma glucose (FPG) of < 100 mg/dL and 2-hour glucose of < 140 mg/dL; pre-diabetes was defined as having a FPG of 101–125 mg/dL or 2-hour glucose of 140–199 mg/dL and type 2 diabetes was defined as having a FPG of ≥126 mg/dL, or 2-hour glucose ≥ 200 mg/dL or use of medications for diabetes. We had FPG, 2-hour post-challenge glucose, and diabetes medication use information on 899 of the study participants.

Statistical Analysis

Characteristics of the MASALA participants were compared by glycemic category using chi-square, analysis of variance (ANOVA) or Kruskall-Wallis tests as appropriate. Variance inflation factors were used to assess collinearity.

We used a directed acyclic graph (DAG) to summarize our prior understanding of the causal relationships between exposures of interest and pre-diabetes and diabetes. We then analyzed the DAG using an on-line tool (19) to identify a minimal sufficient adjustment set (MSAS) of confounders for each exposure of interest. Under the causal assumptions encoded in the DAG and provided that the regression model for the outcome is correctly specified, adjusting for an MSAS is sufficient to obtain an unconfounded estimate of the overall effect of each exposure of interest on the outcome, without adjusting away indirect effects via mediators. (19, 20) Predictors included a traditional cultural beliefs scale (the base question was “How much would you wish these traditions from South Asia would be practiced in America?” The seven items included: performing religious ceremonies; serving sweets at ceremonies; fasting on specific occasions; living in a joint family; having an arranged marriage; eating a staple diet of chapattis, rice, daal, vegetable, and yogurt; using spices for health and healing), socioeconomic status (education and family income), fasting and dietary pattern, chronic burden and psychological disorders (depression and anxiety), sedentary behavior (time spent watching television) and physical activity (total exercise in MET-min/week). Multinomial logistic regression was used to estimate the causal effects of each exposure of interest on pre-diabetes and type 2 diabetes, adjusting for the exposure-specific MSAS detailed in the Appendix. All models were further adjusted for study site, a strong independent correlate of these outcomes, and a potential marker for unmeasured confounding (Appendix). If non-linearity was detected in the effects of continuous variables, they were spline- or log-transformed. For categorical predictors, tests of heterogeneity were performed. The relative risk ratio (RRR) with 95% confidence intervals was used to summarize model results. We used STATA (version 13.1, College Station, TX) for our analyses.

Results

Among the 899 participants with categorized glycemic status, the mean age was 55±9 years and 54% were men. Forty-two percent of participants had NGT, 33% had pre-diabetes, and 25% had type 2 diabetes. Of those with type 2 diabetes, 65% had known diabetes and were taking diabetes medications while 35% were newly diagnosed. Baseline characteristics of the study participants by glycemic category are shown in Table 1. Participants with pre-diabetes and diabetes were older, more likely to be male, be from the NWU site, and have lower socioeconomic status than those with NGT. Those with pre-diabetes and diabetes reported less total exercise and a greater number of minutes of watching television per week than those with NGT.

Table 1.

Baseline characteristics^* of MASALA study participants, 2010–2013.

Characteristic	Normal glucose tolerance, n=375	Pre-diabetes n=295	Diabetes n=229	P-value
Sociodemographic measures
Age (years)	53.4 ± 9.2	55.8 ± 9.6	57.7 ± 8.7	< 0.001
Male sex	174 (46)	170 (58)	139 (61)	0.001
Clinical site:				< 0.001
Northwestern University	130 (32)	168 (41)	111 (27)
UCSF	245 (50)	127 (26)	118 (24)
Country of birth:				0.68
India	310 (41)	252 (34)	190 (25)
Other South Asian countries	23 (39)	20 (34)	16 (27)
US and other Diaspora countries	42 (48)	23 (26)	23 (26)
Years in the US	27 ± 10	27 ± 11	28 ± 11	0.32
Religious affiliation:				0.64
Hindu & Jain	282 (43)	211 (32)	167 (25)
Sikh	30 (44)	21 (31)	17 (25)
Islam	19 (30)	26 (41)	19 (30)
Other	17 (35)	19 (39)	13 (27)
None	27 (47)	18 (31)	13 (22)
Level of education:
≥Bachelor’s degree	339 (90)	261 (88)	190 (83)	0.02
Incomes, < $100,000	106 (29)	111 (38)	105 (47)	< 0.001
Behavioral factors and Dietary Intake
Total calories (kcal/day)	1695 ± 517	1707 ± 515	1630 ± 472	0.19
Carbohydrate, % energy intake	57 ± 5	56 ± 6	56 ± 6	0.25
Total protein, % energy intake	15 ± 2	15 ± 2	15 ± 2	0.60
Total fat, % energy intake	29 ± 5	29 ± 5	30 ± 6	0.30
Dietary pattern:				0.86
Western diet	123 (33)	99 (34)	72 (32)
Sweets and Refined Grains	122 (33)	99 (34)	83 (37)
Fruits and Vegetables	125 (34)	94 (32)	68 (30)
Current smoking	13 (3)	9 (3)	9 (4)	0.40
Alcohol consumption, ≥ 1 drink/week	121 (32)	105 (36)	72 (31)	0.54
Total exercise (MET-min/week)	1102.5 (390,	900 (315, 1575)	810 (315, 1680)	0.006
TV watching (minutes/week)	420 (210, 840)	420 (210, 840)	420 (300, 840)	< 0.001
Psychological measures
CES-D score	6 (3, 10)	6 (2, 11)	7 (3, 11)	0.21
Spielberger anxiety score	16.3 ± 4.3	15.8 ± 4.3	16.2 ± 4.6	0.33
Presence of chronic stress	188 (50)	145 (49)	119 (52)	0.81
Presence of chronic stress in the last 6 months	175 (47)	135 (46)	109 (48)	0.92
Cultural measures
Frequency of fasting, ≥1×/month	64 (17)	62 (21)	36 (16)	0.24
Type of food eaten at home:				0.79
Only and mostly South Asian	195 (52)	159 (54)	128 (56)
Equally South Asian and other food	151 (40)	118 (40)	83 (36)
Only and mostly other food	29 (8)	18 (6)	17 (7)
Cultural Traditions scale score	14.4 ± 6.2	14.0 ± 6.1	13.6 ± 6.5	0.33
Health and Family History
History of gestational diabetes	15 (8)	11 (9)	14 (16)	0.07
Family history of diabetes	158 (43)	155 (53)	155 (69)	< 0.001
Clinical measures
Body mass index, kg/m2	25.3 ± 4.4	26.4 ± 4.1	26.8 ± 4.2	< 0.001
Waist circumference, cm	90.0 ± 9.7	94.0 ± 10.2	95.8 ± 10.4	< 0.001
Subcutaneous fat area, cm2	233 ± 91	232 ± 83	248 ± 110	0.12
Visceral fat area, cm2	117 ± 50	142 ± 55	153 ± 59	< 0.001
Systolic blood pressure, mmHg	121 ± 15	126 ± 17	129 ± 15	< 0.001
Diastolic blood pressure, mmHg	72 ± 10	74 ± 10	74 ± 9	0.006
Hypertension	100 (27)	121 (41)	147 (64)	< 0.001
Total cholesterol, mg/dL	192 ± 35	190 ± 35	176 ± 40	< 0.001
HDL-cholesterol, mg/dL	53 ± 15	49 ± 12	47 ± 12	< 0.001
Triglycerides, mg/dL	108 (77, 143)	124 (95, 159)	131 (95, 179)	< 0.001
LDL-cholesterol, mg/dL	116 ± 31	114 ± 31	99 ± 33	< 0.001
ALT, mg/dL	17 (14, 24)	19 (15, 25)	19 (15, 27)	0.004
GGT, mg/dL	18 (14, 25)	21 (15, 28)	22 (17, 32)	< 0.001
eGFR, ml/min/1.73m2	95.2 ± 19.4	92.3 ± 17.2	93.6 ± 21.5	0.15
Adiponectin, ng/mL	11,578 (7496; 16,220)	10,368 (7230; 15,182)	9449 (6013; 13,574)	< 0.001
Resistin, ng/mL	19.5 (16.1, 24.5)	20.7 (16.6, 25.2)	19.8 (16.2, 25.8)	0.35
Glucose metabolism measures
Fasting glucose, mg/dL	91 (86, 94)	101 (94, 106)	120 (107, 145)	< 0.001
2-hour post-challenge glucose, mg/dL	104 (89, 120)	147 (122, 162)	223 (208, 253)	< 0.001
Fasting insulin, pmol/L	50 (36, 70)	70 (48.4, 96)	69.4 (47, 112.9)	< 0.001
2-hour post-challenge insulin, pmol/	394 (253, 684.9)	699 (440.3, 1116)	682.7 (426, 1021.2)	< 0.001
HOMA-IR, (pmol/L*mg/dL)¶	1.81 (1.33, 2.63)	2.91 (1.99, 3.98)	3.43 (2.30, 6.00)	< 0.001
HOMA-β, ((pmol/L)/(mg/dL))¶	116.2 (79.4, 161.2)	112.4 (76.9, 154.9)	75.8 (52.6, 117.9)	< 0.001

^*Values represent n(%) for chi-square analyses, mean±SD for ANOVA and median (25^th percentile, 75^th percentile) for the Kruskall-Wallis test. P-values resulted using the chi-square test, ANOVA or Kruskall-Wallis test as appropriate.

^¶These analyses exclude participants who were using insulin.

Abbreviations: MASALA (Mediators of Atherosclerosis in South Asians Living in America), MET (metabolic equivalent), CES-D (Centers for Epidemiologic Studies Depression Scale); HDL (high-density lipoprotein), LDL (low-density lipoprotein), ALT (alanine aminotransferase), GGT (gamma glutamyl transferase), eGFR (estimated glomerular filtration rate), HOMA (homeostasis model assessment), IR (insulin resistance)

Table 2 shows the independent associations of social, behavioral/lifestyle, and cultural factors with pre-diabetes and diabetes estimated using multinomial models adjusting for the MSAS for each factor. Living in the greater Chicago area (Northwestern study site) was associated with a greater prevalence of pre-diabetes and diabetes after adjusting for age and sex. Compared to never fasting, fasting monthly or annually was associated with a lower prevalence of diabetes (p=0.005 for test of heterogeneity). Of the socioeconomic variables, lower income (< $40,000 annually) and having less than a Bachelor’s degree were associated with a greater prevalence of diabetes although the tests of heterogeneity were not statistically significant. Higher chronic psychological burden score was associated with a greater prevalence of diabetes. Of the lifestyle behaviors, we found that exercise was associated with a lower prevalence of prediabetes while greater time spent watching television was associated with a greater prevalence of diabetes. These associations were not modified by country or region of birth.

Table 2.

Multivariate Analysis – Results of Multinomial Logistic Regression

Primary Predictor	Pre-Diabetes		Diabetes
	RRR (95% CI)	p-value	RRR (95% CI)	p-value	Test of heterogeneity (p-value
Northwestern Univ. sitea	2.50 (1.82–3.43)	< 0.001	1.79 (1.27–2.53)	0.001
Cultural practices*§b	1.04 (0.93–1.15)	0.50	0.99 (0.89–1.10)	0.84
Dietary Pattern (reference: Western)c					0.97
Sweets and Refined Grains	1.02 (0.65–1.60)	0.93	1.01 (0.62–1.64)	0.98
Fruits and Vegetables	0.93 (0.60–1.44)	0.75	0.87 (0.53–1.40)	0.56
Fasting (reference: never/almost never)c					0.005
Monthly/annually	0.73 (0.49–1.08)	0.12	0.47 (0.30–0.74)	0.001
Weekly	1.40 (0.80–2.48)	0.24	0.78 (0.41–1.47)	0.44
Family income (reference >$100K annually)d					0.11
< 40K	1.51 (0.87–2.64)	0.14	2.34 (1.33–4.11)	0.003
40–75K	1.22 (0.74–2.03)	0.44	1.55 (0.92–2.64)	0.10
75–100K	0.82 (0.48–1.43)	0.49	1.04 (0.58–1.86)	0.90
Education (reference >Bachelor’s)d					0.26
< Bachelor’s	1.11 (0.64–1.90)	0.72	1.79 (1.04–3.07)	0.04
Bachelor’s	0.92 (0.64–1.33)	0.65	1.10 (0.74–1.64)	0.63
Chronic Burdene					0.46
Burden score 1	0.97 (0.67–1.42)	0.89	1.01 (0.67–1.54)	0.95
Burden score 2	1.22 (0.74–2.04)	0.44	1.42 (0.81–2.46)	0.22
Burden score 3–5	1.47 (0.74–2.91)	0.27	2.10 (1.04–4.21)	0.04
Depression*§e	1.00 (0.93–1.08)	0.98	1.04 (0.96–1.12)	0.34
Anxiety*§e	0.78 (0.60–1.02)	0.07	0.83 (0.62–1.10)	0.20
Exercise (MET-min/week)*f	0.81 (0.68–0.97)	0.02	0.87 (0.72–1.06)	0.17
TV watching (min/week)*f	0.93 (0.77–1.13)	0.46	1.23 (1.02–1.49)	0.03

^*Per standard deviation

^§RRR reported for 50% increase in log-transformed predictors

^aAdjusts for: age, sex

^bAdjusts for: age, sex, site, country of birth, religion, years in the United States

^cAdjusts for: age, sex, site, country of birth, religion, cultural practices, chronic burden, anxiety, depression, income, education, years in the United States

^dAdjusts for: age, sex, site, cultural practices

^eAdjusts for: age, sex, site, income, education, years in the United States, religion, cultural practices

^fAdjusts for: age, sex, site, country of birth, religion, cultural practices, chronic burden, anxiety, depression, income, education, years in the United States, dietary pattern, fasting

Discussion

In this cross-sectional analysis of a large community-based cohort of US South Asians without known cardiovascular disease, there was a high prevalence of pre-diabetes (33%) and diabetes (25%). In multinomial logistic regression models, lower income, less education, increased chronic psychological burden and greater time spent watching television were associated with a greater prevalence of pre-diabetes and diabetes. On the other hand, more exercise and fasting monthly or annually were associated with a lower prevalence of pre-diabetes and diabetes.

While the MASALA cohort was recruited from two urban areas in the US, it is grossly representative of the middle- to older-age South Asian population in the US when compared to the 2010 US Census. The MASALA cohort was 84% Asian Indian, which is comparable to the 83% reported in the 2010 US Census. (9) There was a somewhat lower proportion of Pakistanis (5% vs 10.6% in the 2010 Census) and a higher proportion of Bangladeshis and Sri Lankans. (9)

The prevalence of diabetes in the MASALA study was 25% (in Asian Indians – 25%; in non-Asian Indians – 26.5%), and diabetes prevalence has ranged from 17.4% to 35.4% in other studies of US South Asians (21, 22), considerably higher than the estimated diabetes prevalence of 9.1% in India, 10.3–10.4% in South India, and 9.1% in North India. (2, 23, 24) Compared to other ethnic groups, South Asians have a higher prevalence of pre-diabetes and diabetes. (22, 25–27) Since the MASALA study excluded individuals with cardiovascular disease, our prevalence estimates may be lower than in the overall population of middle-older aged US South Asians. Therefore, understanding the risk factors for diabetes in US South Asians is critical.

Our study suggests geographic differences in diabetes prevalence amongst US South Asians. Participants from the Chicago area had higher rates of pre-diabetes and diabetes than those in the San Francisco Bay Area. While there were differences in country of birth, religion, socioeconomic status, and cultural practices between the two study sites (9), this association remained significant after adjusting for these factors. Lifestyle/behavioral and biologic factors did not differ by study site. Further investigations to elucidate the factors contributing to this difference are important.

Lower family income and lower educational attainment were strongly associated with diabetes. This inverse association has been demonstrated in the US and other developed countries including the UK, Germany, France and Canada. (28–32) South Asians in the UK Whitehall Study had higher odds of diabetes (4.2, 95% CI 3.0–5.8) compared to Caucasians and socioeconomic status (SES) was found to be an important confounding factor for this difference. (33) On the other hand, studies from South Asia show a positive relationship between SES and diabetes prevalence; a Pakistani study showed that diabetes prevalence was 4.5% in an affluent population compared to 1.8% in a less affluent population, with urbanization and obesity playing mediating roles. (34) A nationally representative survey of residents of urban and rural areas in India demonstrated that those in the richest quintile of household wealth had 2.58 times the odds of self-reported diabetes compared to those in the poorest quintile. (35) The pathways through which lower SES influences diabetes prevalence in US South Asians are likely multifactorial. A conceptual framework developed by Brown et al (36) found that healthcare, behavioral, psychological and contextual (neighborhood) factors linked SES with diabetes risk.

Greater chronic psychological burden was associated with diabetes independent of SES. Higher levels of stress are associated with abnormal glucose metabolism and diabetes in other ethnic groups, though the association appears stronger in women than in men in these studies. (37, 38)

While we did not find an association between cultural beliefs and diabetes risk, we did find that fasting monthly or annually was associated with a lower prevalence of diabetes compared to never fasting. Studies looking at the effect of fasting on glucose and diabetes risk have been contradictory. In our pilot study, fasting on specific occasions was associated with higher odds of diabetes. (15) And one prospective study of patients with diabetes showed that there was a deterioration in glycemic control when fasting during Ramadan. (39) However, other studies have shown that insulin sensitivity improved in men with the metabolic syndrome who fast during Ramadan. (40, 41) Routine periodic fasting (fasting for 24 hours once a month) in a primarily Caucasian population was associated with a lower prevalence of diabetes. (42) It is possible that differences in duration and frequency of fasting differentially impact diabetes risk and understanding what aspects of fasting impact diabetes risk will be important.

Sedentary behavior, measured by greater time spent watching television, was associated with diabetes while exercise was inversely associated with pre-diabetes. Sedentary lifestyle is a stronger and more important predictor of the high prevalence of diabetes. (43) While the impact of television watching on diabetes risk in South Asians has not been investigated before, a large prospective study conducted in Europe found that the amount of time spent watching television was an independent predictor of incident diabetes. (44) In addition, South Asian migrants to the US and UK often fall below minimum physical activity recommendations, and results from our pilot study showed that US South Asians participated in less exercise compared to Caucasians and African Americans. (15, 45–47) Level of physical activity is inversely correlated with blood glucose and insulin. (45, 46) Therefore, culturally tailored interventions to reduce sedentary behavior will likely impact diabetes prevalence in US South Asians.

Indian migrants to the US have also increased their consumption of processed foods(48), and compared to a European meal, a South Asian meal has more calories and a higher percentage of carbohydrates. (49) Furthermore, total carbohydrate intake, glycemic load and glycemic index are associated with an increased diabetes risk in South Indians. (50) Interestingly, results from our pilot study found that higher levels of protein intake (51) were associated with higher odds of diabetes while another study found that lower vegetarianism and greater westernization of the diet was associated with a greater prevalence of the metabolic syndrome in South Asian migrants to the UK. (52) However, neither macronutrient intake nor dietary pattern (“sweets and refined grains” and “fruits and vegetables” pattern being primarily vegetarian) was associated with diabetes in this current study and therefore, further investigation is needed to determine the impact of diet on diabetes prevalence in US South Asians.

This study has several strengths as it is the largest, deeply phenotyped cohort of migrant South Asians in the US with standardized clinical and behavioral/lifestyle measures. In addition, robust epidemiologic and statistical methods were used to derive our multivariate models. There are also several notable limitations. As this is a cross-sectional study, causal inferences cannot be determined. The sociodemographic, cultural, and dietary data were obtained through an interviewer administered questionnaire, which may be limited by recall or social desirability bias. In addition, MASALA was not a nationally representative sample although it was grossly representative of the US South Asian population. There was little variability by SES and few participants from each of the South Asian countries which limited subgroup analyses by SES and nativity.

Conclusion

In conclusion, in a large cohort of US South Asians, we found that lower income, less education, higher chronic psychological burden, and more time spent watching television were associated with a higher prevalence of pre-diabetes and diabetes while exercise and fasting monthly or annually were associated with a lower prevalence. We have identified several modifiable risk factors which can serve as targets for intervention in this high-risk group.

Highlights.

• Cross-sectional analysis of large cohort of US South Asians without cardiac disease

• There was a high prevalence of pre-diabetes (33%) and diabetes (25%)

• Significant difference in diabetes prevalence between the two study sites

• Lower income, less education, and more time watching TV associated with diabetes

• More exercise; fasting monthly / annually associated with lower diabetes prevalence

Abbreviations: (in order of appearance)

MASALA

Mediators of Atherosclerosis in South Asians Living in America

US

United States

CVD

cardiovascular disease

IR

insulin resistance

UK

United Kingdom

MESA

Multi-Ethnic Study of Atherosclerosis

UCSF

University of California, San Francisco

NWU

Northwestern University

SHARE

Study of Health Assessment and Risk in Ethnic Groups

BMI

body mass index

CT

computed tomography

MIPAV

Medical Image Processing, Analysis, and Visualization

HU

Hounsfield Unit

SBP

systolic blood pressure

DBP

diastolic blood pressure

HDL-C

high-density lipoprotein-cholesterol

LDL-C

low-density lipoprotein-cholesterol

ALT

alanine aminotransferase

GGT

gamma glutamyl transferase

CV

coefficient of variation

HOMA

homeostasis model assessment

NGT

normal glucose tolerance

ADA

American Diabetes Association

FPG

fasting plasma glucose

ANOVA

analysis of variance

DAG

directed acyclic graph

MET

metabolic equivalent

MSAS

minimal sufficient adjustment set

RRR

relative risk ratio

SES

socioeconomic status

CES-D

Centers for Epidemiologic Studies Depression Scale

eGFR

estimated glomerular filtration rate

Appendix: MSAS for each exposure

• -Socioeconomic status (income, education): age, sex, site and cultural practices
• -Psychological disorders (depression, anxiety) and chronic psychological burden: age, sex, site, income, education, years in the United States, religion, cultural practices
• -Cultural practices and beliefs: age, sex, site, country of birth, religion, years in the United States
• -Fasting frequency and dietary pattern: age, sex, site, country of birth, religion, cultural practices, chronic burden, anxiety, depression, income, education, years in the United States
• -Sedentary behavior (time spent watching television, decreased physical activity): age, sex, site, country of birth, religion, cultural practices, chronic burden, anxiety, depression, income, education, years in the United States, dietary pattern, fasting