Increases in Psychological Stress Are Associated With Higher Fasting Glucose in US Chinese Immigrants

Abstract

Background

The majority of Chinese Americans is foreign-born, and it is well-documented that immigration to the United States (US) leads to increased risk for chronic diseases including type 2 diabetes. Increased disease risk has been attributed to changes in lifestyle behaviors following immigration, but few studies have considered the psychosocial impact of immigration upon biomarkers of disease risk.

Purpose

To examine associations of psychological stress and social isolation with markers of type 2 diabetes risk over time among US Chinese immigrants.

Methods

In this longitudinal study of 614 Chinese immigrants, participants completed assessments of perceived stress, acculturative stress, negative life events, and social isolation annually at three time points. Fasting blood samples were obtained at each time point to measure blood glucose, glycated hemoglobin, and insulin resistance. Mean duration between baseline and follow-up assessments was approximately 2 years.

Results

Increases in migration-related stress, perceived stress and social isolation were associated with significant increases in fasting glucose at follow-up independent of age, body mass index, length of US residence, and other potential covariates. Moreover, increases in glucose varied depending on perceived stress levels at baseline, such that those with higher baseline stress had a steeper increase in glucose over time.

Conclusions

Psychological stress and social isolation are associated with increases in fasting glucose in a sample of US Chinese immigrants. Findings suggest that the unique experiences of immigration may be involved in the risk of developing type 2 diabetes, a condition that is prevalent among US Chinese despite relatively low rates of obesity.

Lay summary

Many Chinese Americans are born outside the United States (US), and moving to the US can increase their risk of chronic diseases like type 2 diabetes. This increased risk is often linked to lifestyle changes after immigration, but not much research has looked at how stress and social isolation may affect disease risk. We conducted a longitudinal study that followed 614 Chinese immigrants over a 2-year period to see if stress and social isolation were linked to diabetes risk. Participants filled out surveys about their levels of stress, social isolation, and any negative life events that had occurred, and had their blood tested for markers that are associated with diabetes risk. The study showed that higher stress and social isolation were linked to higher fasting glucose (blood sugar) levels, regardless of other factors like age or how long they had lived in the US. People who started with higher stress levels also had a bigger increase in fasting blood sugar levels over time. In summary, among Chinese immigrants in the US, stress and social isolation leads to increases in fasting blood sugar levels over time, which can be indicative of type 2 diabetes.

Introduction

Type 2 diabetes represents a significant threat to global human health. In the United States (US) alone, annual health spending attributed to treating diabetes and preventing complications was estimated at $237 billion [1]. These costs are projected to continue to rise over the next decade, putting further strain on our healthcare system. The increasing prevalence of diabetes has been attributed primarily to high rates of overweight and obesity resulting from sedentary lifestyles and intake of energy-dense foods. However, an often overlooked fact is that approximately 20% of patients with diabetes are neither overweight nor obese [2]. Although the phenomenon of diabetes in the non-obese is not well understood, data suggest that diabetes is more deadly in non-obese individuals compared to their obese counterparts, even after adjustment for known risk factors [3]. These findings are extremely relevant for certain US populations, such as Asian Americans.

Asian Americans represent the fastest growing racial group in the US [4]. Chinese Americans are the largest subgroup, representing 24% of US Asians. The majority of Chinese Americans are foreign-born [4]; and studies with diverse immigrant populations in the US, Canada, and Australia suggest that with increasing length of residence in host countries comes elevated risk for various chronic conditions, including type 2 diabetes [5]. Despite having healthier profiles upon arrival (e.g., the “healthy immigrant” phenomenon), there is a subsequent deterioration in health that appears to begin in the migrating generation [6]. Indeed, data from US studies report that Asian Americans consistently have higher rates of diabetes than Whites despite being thinner [7–9]. Although there is heterogeneity in diabetes risk across specific Asian American subgroups, a large-scale study reported that Chinese Americans have higher prevalence of both prediabetes and diabetes compared to White adults at each weight category (including healthy weight, overweight, and obesity body mass index [BMI] levels) [10].

The increased disease risk observed in Asian Americans has been attributed primarily to dietary changes and weight gain following immigration [11]. However, studies suggest that Asian Americans have higher fruit and vegetable consumption and lower soda consumption than non-Hispanic whites [12]. And a longitudinal study reported that changes in diet and acculturation can be quite limited in Chinese immigrants [13], and thus unlikely to wholly explain the observed increase in disease risk.

Models of immigrant health point to potential alternative factors and pathways that may be contributing to poorer health. For example, immigrants report psychological stress and social isolation post-migration—factors that can contribute to poor health overall [14–16] and to diabetes risk specifically [17, 18] via behavioral and/or inflammatory pathways—and these stressors associated with the immigrant experience might play a role in contributing to the excess disease risk observed [19, 20]. Immigration is often associated with significant upheaval across multiple domains of one’s life (family and interpersonal relations, job and professional responsibilities). In addition, immigrant health can be impacted by structural factors and anti-immigration policies that contribute to persistent economic disadvantages, deter immigrants from accessing available health care resources and—in a pernicious negative cycle—lead to greater immigration-related stressors [21, 22]. Racialized stereotypes may also contribute to and promote internalized racism and oppression [23], which predicts poorer mental health [24, 25]. These findings are relevant given data from other populations that internalized racism and perceived discrimination are associated with glucose intolerance [26] and abdominal obesity [27, 28]. Similarly, in a study of 132 Latino immigrants, greater perceived discrimination was associated with higher fasting glucose levels in women [29].

Some Chinese immigrants may also feel socially isolated, due to language barriers and being separated from family members, or experience feelings of “not belonging” in their new environment [30]. Social isolation may increase risk for chronic disease [31, 32] via alterations in immune system regulation and inflammatory pathways [33–35]. Studies suggest that inflammatory processes play a central role in contributing to diabetes in both obese and non-obese individuals [36]. Evidence is also accumulating from animal studies that inflammation can directly interfere with insulin signaling and lead to insulin resistance [37]. Similarly, higher levels of C-reactive protein, an inflammatory marker, were associated with insulin resistance independent of abdominal obesity in a Chinese non-diabetic population [38]. Although few studies of social isolation and diabetes risk exist, one study noted that subjective loneliness and low family support predicted diabetes among US Latinos [39]. Data from a mouse model also are suggestive, by demonstrating that social isolation (i.e., chronic individual housing) accelerated weight gain in mice [40].

Thus, the objectives of the present study were to examine whether psychological stress and social isolation are associated with insulin resistance and markers of type 2 diabetes risk in Chinese immigrants, a relatively non-obese population. It was hypothesized that stress and social isolation would be associated with higher insulin resistance and risk markers over time.

Methods

Study Design and Participants

In this longitudinal study, we enrolled 650 Chinese immigrants between January 2016 and June 2018. Eligibility criteria included: self-reported Chinese heritage; age 35–65 years; and immigration from Asia as an adult (age 18+ years). Exclusion criteria included: self-reported history of diabetes, cancer, auto-immune disorders, or HIV infection; current use of anti-inflammatory medications; current pregnancy or breastfeeding; or inability to provide informed consent. The study was approved by the Fox Chase Cancer Center Institutional Review Board, and all participants provided written informed consent prior to engaging in any study procedures.

Procedures

Participants were recruited from Chinese-serving community organizations, community events and activities, churches and Chinese language schools, and through community contacts and leaders within the Philadelphia Chinese community. Research staff screened interested participants for eligibility, obtained written informed consent, and then scheduled an appointment for the data collection visit. All study materials were available in English and Chinese. Study interviews were conducted in either English or Chinese in the participant’s preferred dialect (Mandarin or Cantonese) by multilingual study staff.

At baseline and two subsequent follow-up time points (approximately one year apart), data were obtained on sociodemographic characteristics, length of US residence, psychological stress (including perceived stress, acculturative stress and non-migration-specific stress) and social isolation. Data collection visits were scheduled to take place in the morning, and participants were asked to fast for 8–12 hours prior to their visits. At each study visit, staff used standard methods to conduct an anthropometric examination for weight, standing height, and waist circumference [41]. Similar to prior studies [42, 43], height was measured using a stadiometer, weight by an electronic digital scale, and waist circumference with an anthropometric measuring tape. All measurements were recorded in duplicate; mean weight and height were used to compute BMI.

At each study visit (baseline and follow-up time points), we collected blood samples in one EDTA (lavender-top) tube and 2 serum-separator tubes. Blood samples were labeled with the participant’s study ID and date and transported to the Fox Chase Cancer Center Biosample Repository Facility where the blood samples were spun down, separated, and aliquoted for clinical testing or storage at −80°C until analysis.

The annual follow-up assessments followed similar procedures as the baseline visit. Research staff contacted individual participants by mail and telephone to confirm their continued participation and to schedule follow-up appointments. Psychological stress and social isolation were assessed at each follow-up time point. An exam was conducted to measure weight, height and waist circumference, and fasting blood samples were collected.

Measures

Demographic and health characteristics

Demographic and health characteristics including participants’ age, sex, highest education level, occupation, length of US residence, and marital status was assessed at baseline. Education was categorized into the following four groups: (1) 0–12 years without completing high school or vocational/technical school; (2) high school graduate or GED (General Educational Development); (3) some college or technical school; and (4) college graduate or post-graduate education. Occupational categories were classified as: (1) not employed, farmer/farm worker, machine or vehicle operator, craftsworker, or service worker; (2) clerical or sales worker; and (3) manager/administrator or professional, similar to prior work [44]. Current smoking status and family history of diabetes were also assessed.

Psychological stress

Psychological stress was assessed using three measures. To capture global stress, we used the 10-item Perceived Stress Scale [45, 46], a well-established instrument that has been extensively used in studies of stress and health [45, 47] and has been translated into Chinese [48]. For each of the 10 items, participants are asked to respond how often they felt or thought a certain way over the past month. Response options range from “0 = never” to “4 = very often” and are summed across the 10 items for a potential total score ranging from 0 to 40. The Perceived Stress Scale demonstrated strong internal reliability in the present sample (Cronbach’s α = .85).

The Migration-Acculturation Stressor Scale (MASS) [49, 50] was used to assess levels of acculturative stress. This scale includes 22 items describing potential acculturative stressors (e.g., cultural or social difficulties). For each item, participants were asked if they had encountered that difficulty in the past year, with answers ranging from “1 = no” to “5 = very much.” A total acculturative stress score was calculated by summing across all items. The MASS has been validated among Chinese immigrant populations [51] and has high internal reliability (α = .85) [52].

Finally, we utilized a modified version of the Life Experiences Survey (LES) to assess non-migration specific life stressors [53]. The modified LES was developed for, and demonstrated good convergent validity in, an ethnically and socioeconomically diverse urban population [54]. In brief, the LES assesses the occurrence of various life events over the past year, as well as the respondent’s perception of the impact of that event. Participants rated each life event on a 7-point scale ranging from −3 (extremely negative) to +3 (extremely positive). If an event did not occur, the item was scored as 0. Standard procedures [53, 55] were used to calculate scores based on the events reported. Specifically, positive and negative life events were differentiated based on whether respondents reported a positive (1 to 3) or negative (−1 to −3) impact for that event; then, the impact scores for all negatively rated life events were summed and the absolute value was used to represent the total negative life events impact score. This measure has demonstrated good reliability and validity [53] including in a prior study of Chinese immigrants [42].

Social isolation

Social isolation was assessed using a 9-item scale derived from the National Social Life, Health, and Aging Project. The items measure social disconnectedness and perceived isolation and have been predictive of health outcomes [56]. The scale had high internal reliability (α = .84) in the present study. Respondents are asked to rate how often they have someone (family member or friend) they can turn to or rely on for specific situations or how often they feel isolated or lack companionship. Response options range from “1 = hardly ever or never” to “3 = often.” Items are reverse-scored and then averaged across all items, so that a higher mean score reflects greater social isolation.

Outcome variables

The primary outcomes included insulin resistance, fasting glucose, and glycated hemoglobin (HbA1c). Blood samples were tested using standard laboratory methods. Plasma insulin was measured by ultra-sensitive insulin ELISA assay kit (ALPCO, Salem, NH, cat. # 80-INSHUU-10), and plasma glucose was measured by the glucose oxidase method with a glucose analyzer (Yellow Springs Instruments, Yellow Springs, OH). The intra-assay % CV (Coefficient of Variability) was 4.4% for insulin and 0.73% for glucose; and the inter-assay % CV was 2.5% for insulin and 0.64% for glucose. Insulin resistance was calculated according to the homeostasis model assessment of insulin resistance (HOMA-IR): Fasting insulin (μU per mL) × Fasting glucose (mmol per L)/22.5 [57, 58].

Fasting glucose and HbA1c were selected due to their clinical relevance in diagnosing type 2 diabetes and assessed using standard clinical laboratory assays [59]. To measure HbA1c, whole blood was transported locally to Quest Diagnostics and tested using a direct whole blood enzymatic quantification of HbA1c by fructosyl dipeptide oxidase [60]. Fasting blood glucose was measured at the Fox Chase Cancer Center Clinical Laboratory using a glucose uptake colorimetric assay with the VITROS^® XT 3400 Chemistry Analyzer. Fasting glucose and HbA1c tests were performed in CLIA-approved and CAP-accredited clinical testing facilities.

Analytic Plan

The variables were first examined for outliers, skewness, and kurtosis, and, if necessary, normalizing transformations were applied. The relationship between baseline demographic variables and outcome measures was assessed via plots and, where appropriate, Spearman’s correlation tests (or Kruskal–Wallis tests) and regression analyses. To examine the association between psychological factors and markers of diabetes risk and insulin resistance in Chinese immigrants over time, we used a series of linear mixed-effects models with random intercepts for participants. The outcomes for these models were change in each diabetes risk factor, measured as the difference between the risk markers at the 2 follow-up time points from baseline. We aimed to assess both the effect of baseline psychosocial measures on the risk markers obtained at follow-up assessments (i.e., do psychosocial factors at baseline predict changes in diabetes risk markers), and associations between changes in psychosocial measures and changes in the risk markers obtained at follow-up assessments (i.e., are changes in psychosocial factors associated with changes in diabetes risk markers at follow-up?). Therefore, we separated the psychosocial measures into their baseline value and change from baseline. These were included in the model, along with (categorical) time and relevant covariates. Due to the clinical relevance of selected markers (glucose, A1c), we also ran models in which the outcomes were the risk marker values measured at each of the 3 study timepoints, rather than change in risk marker values. These analyses allowed us to examine interactions between the baseline psychosocial measure and the indicator value at each follow-up time point. Separate models were fit for each psychosocial/risk marker pair. In these models, we included potential confounders (e.g., education) with an anticipated association with risk markers. These models accounted for correlation within participants using Generalized Estimating Equations with robust standard errors. Statistical analyses were conducted using R software (version 4.2).

Results

Participant Characteristics

Participants included 377 women and 273 men. One male participant was subsequently withdrawn due to having a current diagnosis of diabetes, and one female participant was withdrawn due to not completing any of the study assessments. At baseline (Timepoint 1 or T1), 28 participants were missing data on relevant covariates (e.g., age, BMI, length of US residence, education, and occupation); an additional 6 participants were missing data on fasting glucose, resulting in a final sample of 614 participants. Participants were, on average, 51.0 years of age (Standard Deviation (SD) = 7.6 years; range = 35–65 years). Over half of the study sample (56.7%) was female. Due to the focus of the present study, all participants were of Asian race (100%) and nonHispanic ethnicity (100%). With respect to education, 33.3% of participants reported receiving less than a high school education, 24.6% completed high school or its equivalent, 25.4% attended some college or technical school, and 16.7% had received a college or post-graduate degree. Over one-third of participants (34.1%) reported either being not employed or working in the service industry (e.g., restaurant worker) or as a laborer (e.g., farm worker, machine operator), 38.8% were employed in a clerical or sales worker position, and 27.0% reported managerial or professional occupations. Other participant characteristics are presented in Table 1.

Table 1.

Participant Characteristics at Baseline (n = 614)

Variable	Mean (SD) or %
Age (years)	51.0 (7.6)
	Range: 35–65 years
Sex
Female	56.7%
Male	43.3%
Education
<High school	33.3%
High school grad/GED	24.6%
Some college/technical school	25.4%
College grad/post-grad	16.7%
Occupation
Not employed/laborer/service worker	34.1%
Clerical or sales worker	38.8%
Manager/administrator or professional	27.0%
Married	88.9%
Length of US residence (years)	17.2 (8.9)
	Range: 0.06–43.4 years
BMI (kg per m2)	23.9 (3.1)
Current smoker	10.1%
Have a family history of diabetes	22.4%

BMI Body Mass Index.

At the 1-year post-baseline assessment (Timepoint 2 [T2]), 516 of the 614 participants returned and completed study procedures; at Timepoint 3 (T3, approximately 2-year post-baseline), 435 participants completed study procedures. This represents an overall 70.8% retention rate. Age, sex, BMI, education, occupation, and length of US residence were not associated with dropout. Participants who dropped out of the study reported higher levels of social isolation (p < .05) and had slightly higher levels of fasting glucose (p = .065), but lower levels of HbA1c (p < .05) at baseline.

Sociodemographic Correlates of Risk Markers

Correlational analyses of baseline data indicated that length of US residence was positively associated with HOMA-IR (p < .05) and HbA1c (p < .05). Among categorical covariates, ANOVA tests revealed that education level was significantly associated with each risk marker, such that higher levels of education were associated with lower HOMA-IR (p < .05), HbA1c (p < .05), and fasting glucose (p < .05). Higher occupational class was associated with greater HbA1c (p < .01) and fasting glucose (p < .01), as well as HOMA-IR although this association did not reach statistical significance (p < .08). Increased BMI was associated with higher HOMA-IR, HbA1c, and fasting glucose (p < .01 for each). Higher age was associated with higher HbA1c and fasting glucose (p < .01), but not HOMA-IR. Smoking status, family history of diabetes, and marital status were not significantly associated with any of the risk markers.

Across study time points, acculturative stress, perceived stress and negative life events decreased over time (all p-values < .01; see Table 2). Social isolation significantly increased between T1 and T2 (p < .01) and decreased between T2 and T3 (p < .01). In addition, HbA1c and fasting glucose levels increased significantly over time, but HOMA-IR did not.

Table 2.

Psychological Stress, Social Isolation, and Risk Markers Across Time Points

Variables	T1 (n = 614)	T2 (n = 516)	T3 (n = 435)
	Mean (SD) or %	Mean (SD) or %	Mean (SD) or %
Psychological stress
Acculturative stress	40.28 (12.88)	37.76 (11.87)	36.06 (11.22)
Perceived stress	11.53 (7.07)	11.54 (6.55)	9.55 (6.44)
Number of negative life events	1.29 (2.37)	0.99 (1.83)	0.86 (1.72)
Social isolation	1.50 (0.38)	1.54 (0.36)	1.49 (0.37)
Outcome variables
HOMA-IR	1.26 (0.89)	1.27 (0.98)	1.29 (0.99)
Fasting glucose (mg per dL)	89.95 (19.79)	91.54 (15.35)	92.45 (16.76)
<100 mg per dLa	85.9%	84.1%	82.3%
100–125 mg per dLb	10.6%	12.8%	14.1%
≥126 mg per dLc	3.5%	3.1%	3.6%
HbA1c (%)	5.62% (0.49)	5.69% (0.55)	5.73% (0.49)
<5.7%a	58.4%	57.0%	51.5%
5.7% to 6.4%b	37.3%	36.4%	41.4%
≥6.5%c	4.3%	6.6%	7.1%

^aNormal.

^bBorderline or indicating possible prediabetes.

^cConsistent with diabetes.

HOMA-IR homeostatic model assessment of insulin resistance; HbA1c glycated hemoglobin; Timepoint 1 T1; Timepoint 2 T2; Timepoint 3 T3.

Psychological Stress and Changes in Risk Markers Over Time

Analyses revealed that increases in migration-related stress predicted significant increases in fasting glucose at follow-up time points (β = .11, p < .05; see Table 3). Increases in perceived stress over time were also associated with significant changes in fasting glucose at follow-up (β = .24, p < .05). Increased social isolation predicted significant increases in fasting glucose as well (β = 2.27, p < .01). The associations of migration-related stress, perceived stress, and social isolation with increases in fasting glucose were independent of the effects of age, BMI, occupation, length of residence in the US, education, and gender, which were included as covariates in the models. However, no significant associations were observed with the other risk markers (HbA1c and HOMA-IR). Finally, changes in negative life events did not lead to significant changes in fasting glucose, HbA1c or HOMA-IR.

Table 3.

Effects of Change in Stress, Social Isolation, and Negative Life Events From Baseline to Follow-up on Change in Fasting Glucose Levels

	Fasting glucose
	β (LCL, UCL)
Change in migration-related stress	0.11 (0.02, 0.20)*
BMI	0.57 (0.10, 1.05)*
Age (per 10 years)	1.55 (−0.36, 3.44)
Length of US residence	−0.01 (−0.18, 0.15)
Sex
Female	Ref
Male	−1.24 (−4.02, 1.55)
Change in perceived stress	0.24 (0.09, 0.39)**
BMI	0.56 (0.08, 1.03)*
Age (per 10 years)	1.78 (−0.13, 3.69)
Length of US residence	−0.02 (−0.19, 0.14)
Sex
Female	Ref
Male	−1.07 (−3.86, 1.72)
Change in social isolation	2.27 (0.56, 3.97)**
BMI	0.57 (0.09, 1.04)**
Age (per 10 years)	1.57 (−0.33, 3.47)
Length of US residence	−0.04 (−0.20, 0.13)
Sex
Female	Ref
Male	−1.20 (−3.98, 1.58)
Change in negative life events	0.14 (−0.45, 0.72)
BMI	0.59 (0.11, 1.06)*
Age (per 10 years)	1.58 (−0.33, 3.49)
Length of US residence	−0.04 (−0.21, 0.12)
Sex
Female	Ref
Male	−1.15 (−3.95, 1.65)

Note: Change over time was calculated for each independent variable from T2 to T1 and from T3 to T1, if available. The results presented reflect the effect of change in the independent variable from T1 to T2 predicting change in fasting glucose at T2, and the effect of change in the independent variable from T1 to T3 predicting change in fasting glucose at T3, averaged across T2 and T3. All models are also adjusted for education level (<high school [referent category], completed high school or GED, some college or technical schooling, or college graduate or post-graduate education), and occupational category (not employed/laborer/service worker [referent category], clerical or sales, or managerial/professional occupation), data not shown.

LCL Lower Confidence Limit; UCL Upper Confidence Limit; Timepoint 1 T1; Timepoint 2 T2; Timepoint 3 T3.

^*p < .05

^**p ≤ .01.

Our analyses also indicated that the changes in fasting glucose values over time varied depending on levels of perceived stress at baseline. Specifically, a significant interaction was observed such that those with higher baseline perceived stress had higher mean fasting glucose levels at follow-up (p < .05). For ease of interpretation, the interaction is depicted in Fig. 1 by tertile of baseline perceived stress scores (low, medium, and high). Unlike participants with low perceived stress at baseline, those with medium or high perceived stress showed increasing levels of fasting glucose over time. Those in the highest tertile of perceived stress at baseline ended with the highest mean fasting glucose levels at final follow-up. No other significant interactions were observed.

Fig. 1.

Fasting glucose levels at follow up are greater among participants reporting higher levels of perceived stress at baseline.

Discussion

In this study of foreign-born Chinese Americans, significant increases in fasting glucose and HbA1c were observed over a relatively short period of time. Greater increases in migration-related stress, perceived stress, and social isolation predicted significant increases in fasting glucose, but not HbA1c or HOMA-IR. Analyses also revealed that those immigrants reporting higher baseline levels of perceived stress experienced a greater increase in fasting glucose over time, such that those in the highest tertile of baseline perceived stress had the highest mean fasting glucose level at the end of the study.

These findings are consistent with studies conducted with other populations [17] that suggest a role for psychological stress in increasing risk of type 2 diabetes [61–63]. However, none to date have investigated these associations in immigrant populations. Most studies of immigrant health have focused on behavioral and lifestyle changes or healthcare access barriers as key determinants of health outcomes; few studies have considered the psychological impact of immigration—including any corresponding psychological stressors and potential experiences of social isolation—on biomarkers of health and disease risk. Yet, there is growing evidence that social isolation and stress might lead to medical issues if these conditions persist and are not adequately addressed [64, 65]. Given that immigration to the US can confer significant stress, particularly in the current highly politicized landscape, it is worthwhile to consider how these social and psychological experiences may relate to health conditions experienced within immigrant populations.

One biobehavioral mechanism that may link psychological stress to diabetes risk is proposed to occur via dysregulation of the hypothalamic–pituitary–adrenal axis. Studies have shown that stress-related cortisol secretion is associated with higher fasting blood glucose levels [66], with stress hormones facilitating inflammation, which has been demonstrated to contribute to the development of type 2 diabetes in both obese and nonobese adults [36]. The central role of inflammatory processes in diabetes is highly relevant in light of the abundant and well-established studies of psychological stress and inflammation [67, 68] and the development of metabolic disorders [17, 69, 70].

In contrast, however, no associations were observed between psychological stress and HbA1c. Unlike fasting glucose—which reflects the blood sugar level at a single point in time and can be impacted by acute events such as illness, stress, and recent physical activity—HbA1c provides a measure of blood glucose levels over the prior two to three months. As a result, HbA1c results may be less likely to show associations with recent or current perceived stress. In studies of individuals with type 1 or type 2 diabetes, findings have been variable. Some studies have reported associations between major life events and increases in HbA1c [71], whereas other studies have not observed any significant associations between stressful life events and HbA1c [72].

Of note, the present study is among the first to document increases in social isolation predicting changes in blood glucose levels. Other studies have similarly reported that social isolation is associated with poorer health outcomes [31, 32]. In the Health and Retirement Study cohort, increases in social isolation were associated with greater risk of mortality, physical disability and dementia [73]. A recent systematic review identified both biological and behavioral pathways via which social isolation can negatively impacting health [74]. Specifically, social isolation may lead to pro-inflammatory processes that are known to contribute to diabetes and other chronic diseases [35, 75–77]. And among adults aged 50 or older, socially isolated individuals reported lower levels of moderate to vigorous physical activity and had lower intake of fruits and vegetables compared with those who were not socially isolated [74, 78]. Therefore, it is not surprising that increasing social isolation leads to detrimental changes in blood glucose.

In the present study, we did not observe any associations between the psychosocial factors and HOMA-IR, an indicator of insulin resistance as opposed to hyperglycemia. This is in contrast to other studies of Chinese individuals that have noted associations between chronic stress [79] and occupational stress [80] with the development of insulin resistance. It is possible that the lower levels of HOMA-IR observed in the present study attenuated our ability to detect potential associations. Other published studies in Chinese samples have noted generally higher levels of HOMA-IR than those reported here [79, 80]. The eligibility criteria used in the present study were established in order to enroll study participants who were unaffected by type 2 diabetes and other chronic conditions; however, this may have resulted in a study sample with a more favorable health profile.

There are several limitations to the present study. First, the primary outcome variables were insulin resistance and clinical markers of risk rather than diabetes incidence due to the relatively short follow-up time period. In the present study, 13 participants developed type 2 diabetes over the 2-year period, a number that is insufficient for analytic purposes. However, the selected outcome markers were chosen based upon their well-established predictive associations with diabetes risk [59]. Given that prospective, longitudinal studies among immigrant populations are extremely rare, the present findings are useful for guiding and focusing the selection of variables and risk markers in subsequent studies. Second, because this is not a population-based sample, study findings may not be generalizable across US Chinese immigrants. Unlike many prior studies of Chinese individuals residing in the US that have included predominantly college-educated or higher income individuals [81, 82], the present study sample was comprised of participants with lower educational attainment (58% with a high school education or lower). Thus, our sample is notable given that US Chinese with lower levels of education tend to be underrepresented in research [83], which is relevant in light of the increased diabetes risk associated with lower education [84–86]. A third limitation pertains to the focus on stress with limited attention paid to measures of psychological well-being, which may also be related to glycemic control [87, 88]. In studies of individuals with type 2 diabetes, positive psychological factors such as optimism and positive affect have been associated with better outcomes [89, 90] and lower mortality [91]. Thus, such pathways warrant investigation in future studies. Despite these limitations, the present findings are meaningful in highlighting potential alternate pathways contributing to increased disease risk in this immigrant population.

In summary, this study found that migration-related stress, perceived stress and social isolation are associated with significant increases in fasting glucose in a sample of US Chinese immigrants. These findings suggest that unique experiences resulting from immigration (e.g., acculturative stress and negative impacts on one’s social network) might be involved in the risk of type 2 diabetes, a condition that is more prevalent among US Chinese relative to several other US populations [7, 10] particularly when accounting for body mass [92]. Furthermore, this may have implications for the screening of individuals at risk for diabetes who have recently migrated, as HbA1c has become a primary method for screening for diabetes [93], but this may fail to identify a subgroup of individuals who would be considered as affected using glucose-based tests [94]. Taken together, these findings suggest that a broader focus on supporting psychological functioning, as well as health-promoting behaviors, may be warranted to optimize health outcomes and reduce chronic disease risk within immigrant populations.

Transparency Statement.

Study registration: This study was pre-registered on ClinicalTrials.gov (#NCT02449213) at https://classic.clinicaltrials.gov/ct2/show/NCT02449213. Analytic plan pre-registration: The analysis plan was registered prior to beginning data collection at ClinicalTrials.gov. Analytic code availability: Analytic code used to conduct the analyses presented in this study are not available in a public archive. They may be available by emailing the corresponding author. Materials availability: Materials used to conduct the study are not publically available. To obtain copies of study materials, please contact the corresponding author.

Data Availability

The data used in this manuscript are not available in a public archive. Deidentified data from this study will be made available (as allowable according to institutional IRB standards) by emailing the corresponding author. All uses of the data would be subject to confidentiality and data-use agreements.