Introduction

The prevalence of type 2 diabetes among African-American women is double that in non-Hispanic white women [1]. Differences in traditional risk factors, including BMI and socioeconomic status (SES), do not fully explain this racial disparity [2].

Racial discrimination is a critical psychosocial stressor for African-American people, who experience racism more frequently than other ethnic groups [3], and may contribute to observed differences in health between African-Americans and white Americans.

To our knowledge, no prospective studies have been conducted regarding perceived racism and type 2 diabetes incidence, although we have evaluated racism in relation to weight gain and incident obesity [4, 5]. Cross-sectional studies have found associations between perceived racism and diabetes-related outcomes [6,7,8].

In the current study, we evaluated the association of perceived racism with incident type 2 diabetes using data from the Black Women’s Health Study (BWHS). We hypothesised that women with high levels of perceived racism would have higher risk of type 2 diabetes. We assessed whether BMI and diet are mediators of the relationship.

Methods

Study population

The BWHS is a prospective follow-up study that enrolled approximately 59,000 African-American women aged 21–69 years at baseline in 1995 [9]. The cohort is followed every 2 years using postal questionnaires to update information including diagnosis of diabetes. The study protocol is approved by the institutional review board of Boston University. Follow-up has been successful for 87% of potential person-years of follow-up up to 2013.

Diabetes assessment

Follow-up questionnaires up to 2013 asked about new diagnoses of diabetes during the previous 2 year period. Incident cases were defined as self-reported physician-diagnosed type 2 diabetes that occurred at age 30 years or older, excluding type 1 diabetes and gestational diabetes, with no report of a previous diagnosis.

In a validation study among BWHS participants, the diagnosis of diabetes was confirmed in 220 (96%) of the 229 self-reports [10]. We estimated the prevalence of undiagnosed diabetes to be 6.4% using tests for HbA1c in blood samples from 1873 participants who had never reported diabetes [11].

Ascertainment of perceived racism

The 1997 questionnaire contained questions on perceived racism adapted from Williams et al [3]. In a validation study, we confirmed the predetermined two major domains of the racism questions [4]. We created two summary racism variables based on results from principal component analysis. An ‘everyday racism’ summary score averaged responses from five questions regarding the frequency of racism in daily life. Response options were ‘never’, ‘a few times a year’, ‘once a month’, ‘once a week’ and ‘almost every day’ and were scored 1 (never) to 5 (almost every day). Total scores were categorised into quartiles (quartile 1 lowest; quartile 4 highest). A ‘lifetime racism’ summary score summed the number of positive responses to three yes/no questions about experiences of ‘being treated unfairly due to your race’ at work, in housing or by the police. The summary score was 0 (none), 1, 2 or 3 (all) questions answered ‘yes’.

Covariates

Current weight, marital status, medical care use, reproductive and medical history, smoking, alcohol use, medications and physical activity were assessed in 1995 and on each subsequent follow-up questionnaire. Dietary patterns of vegetables/fruit and meat/fried food consumption were calculated from the 1995 and 2001 food frequency questionnaires. Self-reported adult height was recorded in 1995. First-degree family history of diabetes was ascertained in 1995 and 1999. Data on years of education were collected in 1995 and 2003. Marital status was recorded in 1995, 1997, 1999 and 2005. BMI was updated with each questionnaire, based on self-reported height and current weight. In a validation study of anthropometric measures conducted in 115 BWHS participants, Spearman correlations for self-reported vs technician-measured weight and height were 0.97 and 0.93, respectively [12].

Neighbourhood SES was measured by linking participants’ current address to 2000 US Census block groups using geocoding (Mapping Analytics, Rochester, NY, USA) [13]. We used factor analysis to calculate a score for neighbourhood SES, with further division into quintiles (quintile 1 lowest; quintile 5 highest).

Statistical analyses

Baseline for the present analyses was 1997. There were 51,090 women who answered the perceived racism questions. We excluded women with a history of diabetes, cancer, myocardial infarction, stroke and/or coronary artery bypass graft surgery. The final sample analysed included 45,781 women.

We used Cox proportional hazards models to estimate HRs and 95% CIs for incident type 2 diabetes. Person-years were calculated from 1997 until diagnosis of type 2 diabetes, loss to follow-up, death, occurrence of any of the exclusion criteria or end of follow-up in 2013, whichever came first. We modelled time-varying variables using the Andersen–Gill approach. We used SAS statistical software version 9.4 (SAS Institute, Cary, NC, USA).

The initial model included age and questionnaire cycle. The primary multivariate model included age, questionnaire cycle, marital status (married/living as married, separated/divorced/widowed, single/never married), quintiles of neighbourhood SES, number of years of education (≤ 12, 13–15, ≥ 16), family history of diabetes (yes/no), vigorous physical activity (h/week) (< 1, 1–4, ≥ 5), walking for exercise (h/week) (< 1, 2–3, 3–4, ≥ 5), alcohol use (never, past, current no. of drinks/week 1–3, 4–6, or ≥ 7) and smoking status (never, past, current).

We evaluated BMI and diet as potential mediators of the relationship between perceived racism and diabetes. We included time-varying categorical BMI (in kg/m2: < 25, 25–29, 30–34, 35–39, ≥ 40) and vegetable/fruit and meat/fried food dietary pattern scores (continuous) to our multivariate model. We estimated mediation proportion and 95% CI using the partial likelihood of the Cox models.

Results

Table 1 shows participant characteristics by everyday racism quartiles and by lifetime racism categories.

Table 1 Baseline (1997) characteristics of participants in the analysis sample from the BWHS, by perceived everyday and lifetime racism
Full size table

Table 2 shows association and mediation analysis results. A total of 5344 incident cases of diabetes occurred during 576,577 person-years of follow-up. In the multivariate model, women with the highest exposure (fourth quartile) to everyday racism had a 31% increased risk of diabetes. For lifetime racism, women with the highest exposure (yes to all) had a 16% increased risk of diabetes compared with women who answered no to all questions. Adjustment for dietary patterns did not change the estimates in any appreciable way. Adjustment for BMI greatly attenuated racism associations with diabetes. For everyday racism, the HR controlling for BMI was 1.13 (95% CI 1.04, 1.22). BMI explained 55% of the association. For lifetime racism, the addition of BMI to the model resulted in HR 1.07 (95% CI 0.98, 1.18). BMI explained 52% of the association.

Table 2 Association of perceived racism and incidence of type 2 diabetes in the BWHS, 1997–2013
Full size table

Discussion

In this large prospective study of African-American women, we found that high scores of perceived racism were associated with higher incidence of type 2 diabetes.

To our knowledge, ours is the first study to prospectively examine self-reported perceptions of racism in relation to incident diabetes. Cross-sectional studies have explored the relationship between perceived racism and diabetes-related outcomes. Internalised racism has been found to be associated with diabetes diagnosis [6] and elevated fasting glucose [7]. Perceived discrimination was associated with increased complications of diabetes [8].

Chronic stress induced by experiences of racism may affect risk of diabetes through several pathways. Animal and human data suggest that stressors can lead to weight gain [14]. In addition, stress tends to alter the pattern of food consumption, promoting cravings for nutrient-dense ‘comfort foods’ [14]. Our results show that about half of the association between greater experiences of racism and diabetes is mediated by BMI, and previous BWHS studies have also shown that more frequent experiences of racism were associated with higher weight gain and incident obesity. We did not find evidence of mediation by diet, at least as measured by the vegetable/fruit and meat/fried food dietary patterns. Because we assessed diet at only two points in time (1995 and 2001), in contrast to BMI, which was updated every 2 years, it is possible that we did not capture small changes in dietary behaviour resulting from exposure to racism. Independent of behavioural mechanisms, chronic stress may lead to long-term activation of the hypothalamic–pituitary–adrenal axis, resulting in alterations of innate immune activity, which may lead to insulin resistance and type 2 diabetes [15].

Strengths of the current study include a large, geographically diverse sample, a prospective study design and a long follow-up. We were also able to control for important confounding factors such as age and SES. Successful cohort follow-up lessened the potential for bias from selective losses. The measures of perceived racism are highly reproducible [3] and have been used previously in the BWHS [4, 5]. Validation studies were used to assess the accuracy of self-reported diabetes and self-reported weight and to estimate prevalence of undiagnosed diabetes in the cohort.

In summary, this large prospective study of African-American women suggests that perceived racism is associated with a higher risk of incident type 2 diabetes. BMI appears to play an important mediating role. Racial discrimination is an important psychosocial stressor for African-American women [16] and may explain in part the higher burden of obesity and type 2 diabetes among these women. As both type 2 diabetes and obesity are important risk factors for cardiovascular disease, these results underscore the public health importance of continuing antidiscrimination efforts globally and domestically.