The Study of Women’s Health Across the Nation is a multi-ethnic longitudinal study of midlife women from seven sites in the United States. Each of the clinical sites selected White women and women of one designated racial/ethnic minority group to participate. African-American women were recruited in Boston, MA, Chicago, IL, Detroit-area, MI, and Pittsburgh, PA; Chinese women in Oakland, CA; Hispanic women in Newark, NJ; and Japanese women in Los Angeles, CA. Women were eligible for the baseline assessment in 1996 if they were between 42 and 52 years of age, had an intact uterus and at least one ovary, had no reported current hormone use, had at least one menstrual period in the past 3 months, and self-identified as one of the site’s designated racial/ethnic groups. There have been an additional 15 approximately annual visits following baseline. Visits include both self-reported questionnaires and clinic visits. Further detailed information about SWAN study procedures and methods have been published elsewhere (26). Of the initial 3,302 enrolled women, 2,091 (63%) participated in study visit 15 (2016). The women who participated in visit 15 but were missing data for neuropathy status, diabetes status, body mass index, or baseline perceived discrimination were excluded from the present analysis, leaving 1,718 women (82%) in the analytic sample.

Perceived discrimination was assessed during the baseline visit using the Detroit Area Study Everyday Discrimination Scale (27,28); responses on this scale have been shown to be consistent over time among women in the SWAN cohort (19). This scale measures ten experiences involving discrimination over the past 12 months such as being treated with less respect or courtesy than other people, receiving poorer service than others, and being ignored. Each question was assessed using a response scale of never, rarely, sometimes, and often. Responses were summed and averaged across the 10 questions to yield final scores between 1 (low perceived discrimination) and 4 (high perceived discrimination) (29). The questions above were not specific to reasons for discrimination, however there was a follow up question that asked participants to attribute their experience of discrimination to one major reason (race/ethnicity, sex, etc.). Previous studies in SWAN show that African-American and Chinese women report slightly higher levels of perceived discrimination than other racial/ethnic groups (29).

PN was assessed during visit 15 using both a symptom questionnaire and monofilament testing. The Michigan Neuropathy Screening Instrument (MNSI) symptom questionnaire acquired information about 15 neuropathy symptoms including burning pain, sensitivity, prickling, lack of hot/cold differentiation, severe dryness of the foot, numbness, open sores, and amputations (30). Participants who identified having 4 or more of these 15 symptoms were categorized as having PN from the screening instrument. This questionnaire and cut-point have been previously validated with clinical diagnoses of PN (31).

The monofilament test was first administered using a 10-gram filament placed on the big toe for less than one second for four repetitions. If the participant correctly identified all 4 repetitions, this procedure was repeated with a 1.4-gram monofilament in order to measure more minute levels of insensitivity. Participants who responded correctly to 2 or fewer of 4 attempts on either the 10-gram or the 1.4-gram test were classified as having PN from the monofilament test. For this analysis, women who screened positive on either the questionnaire or the monofilament test were classified as having PN in order to capture a broad range of neuropathy symptoms and sub-types.

Diabetes status was ascertained using a combined measure that included self-reported use of a diabetes medication including insulin, self-reported doctor-diagnosed diabetes, or a fasting glucose ≥ to 126 mg/dL at any follow-up visit. Race/ethnicity was self-identified at baseline as African-American, Chinese, Hispanic, Japanese, or White. To ensure temporality between perceived discrimination and body mass index (BMI), and because BMI has been hypothesized to change through the menopausal transition independently of discrimination (32–34), BMI was ascertained concurrently (visit 15) with PN when most women in the sample were post-menopausal. BMI was calculated as weight in kilograms divided by measured height in meters squared. Smoking status was based on self-report at visit 15 of smoking regularly since the last study visit. Financial strain at baseline was assessed by self-report of how hard it is for the respondent to pay for basics.

Frequencies, means, and 95% confidence intervals (CI) were calculated to define the characteristics of the eligible study population. Chi-square tests and t-tests were used to test for significant differences in characteristics between women with and without diabetes and with and without PN. Logistic regression adjusted for race/ethnicity, financial strain, smoking status, and diabetes was performed to investigate the relationship between perceived discrimination reported at baseline and PN at visit 15. Additionally, an interaction between race/ethnicity and perceived discrimination was tested. BMI was added to the models in order to assess whether BMI could potentially mediate the relationship between perceived discrimination and PN. A formal mediation analysis to determine whether BMI mediated the effect of discrimination on PN was conducted using the STATA package Paramed (StataCorp LP, College Station, Texas), which decomposes the total effect of discrimination on PN into direct (effect of discrimination alone), indirect (effect of discrimination via BMI) and interactive (discrimination and BMI interaction) effects using the methods outlined by Vanderweele, 2013 (35–37). Because diabetes is a strong risk factor for PN we also tested for interactions between discrimination and diabetes status in all models.

The median baseline age of participants was 46.2 years and the median age at visit 15 was 65.4 years. Approximately one-third of the sample (33.9%) reported financial strain at baseline (Table 1). The average discrimination score was 1.7 (standard deviation (SD) = 0.5) with a range of 1.0 to 3.6, 8.91% of the sample reported no instances of discrimination at baseline. The prevalence of diabetes was 3.6% at baseline and 18.6% by visit 15. At visit 15, 68.6% of the sample was overweight or obese, and the average BMI was 29.2 kg/m² (SD = 7.1).

By visit 15, 26.1% of women had PN. The prevalence of PN was higher among women with diabetes (40.9%) compared to women without diabetes (22.8%). Of those with PN, 62.8% were classified as having the condition based on the MNSI questionnaire, 51.7% based on the monofilament test and 14.5% based on both measures (Table 2).

In unadjusted analyses, financial strain and higher perceived discrimination at baseline as well as higher BMI, smoking, and diabetes at visit 15 were positively associated with PN at visit 15 (Table 3). Each unit increase in level of perceived discrimination was associated with 43% higher odds of PN (odds ratio (OR) = 1.43, 95% CI: 1.14, 1.79). Japanese women had 72% lower odds of PN compared to white women (OR = 0.28, 95% CI: 0.16, 0.48), but there were no significant differences in the odds between African-American, Chinese, or Hispanic women and White women.

After adjustment for race/ethnicity, financial strain, current smoking, and diabetes, perceived discrimination remained positively associated with PN (Model 1). Each unit increase in perceived discrimination was associated with 29% higher odds of PN (OR = 1.29, 95% CI: 1.01, 1.66). Next we further adjusted for BMI (Model 2). Doing so attenuated the estimate for perceived discrimination, and it was no longer significantly associated with PN (OR = 1.21, 95% CI: 0.94, 1.56). In the fully adjusted model, Japanese women had 67% lower odds of PN as compared to White women (OR=0.33, 95% CI: 0.19, 0.58), and financial strain, smoking status, diabetes and BMI were positively associated with PN. There was no significant interaction between discrimination and racial/ethnic group or diabetes.

Sensitivity analysis indicated no significant interactive effect between discrimination and BMI in association with PN (likelihood ratio chi-squared = 0.29, two-sided P-value = 0.59), therefore controlled (non-interactive) direct and indirect effects are presented in the mediation analysis (Table 4). The total effect of discrimination on PN is estimated to be 28.3% (95% CI: 17.1, 36.5%) mediated by BMI (indirect effect); the direct effect of discrimination is non-significant and is slightly higher (OR = 1.21, 95% CI: 0.94, 1.56) than the indirect effect of discrimination via BMI (OR = 1.07, 95% CI: 1.03, 1.12).

This is the largest known study of PN in women without and with diabetes. Among women aged 60 to 73 years old, the prevalence of neuropathy was 22.8% in women without diabetes and 40.9% in women with diabetes. Perceived discrimination in midlife was a predictor of having PN 20 years later, and this association was mediated significantly by BMI. We also found that financial strain, higher BMI, current smoking, and diabetes were all associated with an increased odds of PN, confirming prior observations (20–23).

This is the first analysis of the association of discrimination and PN, and we found that each increasing level of perceived discrimination in midlife was associated with 43% higher odds of PN. Discrimination has been found to be a strong indicator for risk of many serious health conditions, including diabetes, obesity and hypertension (10,24–27). Discrimination, along with other types of stressors, increases allostatic load, or the overstimulated or abnormal neurological, immunological, and inflammatory physiological responses induced by chronic stressors (18,19,45–56). Discrimination and other chronic stressors are associated with heightened responsiveness of the cardiovascular system including increased blood pressure reactivity (57,58), red blood cell oxidative stress (59) and increased inflammation (54). Cardiometabolic risk factors including diabetes are associated with PN among individuals with and without diabetes (60) and more recently, diabetic PN was associated with endothelial dysfunction (61). Thus, chronic stressors such as discrimination may be associated with PN through the dysregulation of cardiovascular and metabolic function and pro-inflammatory responses that lead to damage of the peripheral nerves (16, 17). Furthermore, previous work has found an association between discrimination and chronic pain (62,63), potentially through psychological distress (62). Because the symptoms of PN, particularly in the early stages, are characterized by chronic pain it could be that discrimination is at play through the same neurological pathways as chronic pain among both diabetic and non-diabetic PN cases.

While perceived discrimination was a strong predictor for PN in the unadjusted model, this association was attenuated to non-significance when adjusted for BMI – BMI was the most highly significant predicator of PN in the present analysis. Formal mediation results suggest that the association between discrimination and PN is approximately 28.3% (95% CI: 17.1‒36.5%) mediated by BMI. In SWAN and elsewhere, discrimination has been shown to be an independent predictor of BMI (64,65), and discrimination has been found to be a predictor of inflammation with significant moderation by BMI (66). Furthermore, increased body mass index has been shown to contribute to decreased nerve conduction among individuals without diabetes who do not yet have any symptoms of PN (39). Even when no neuropathy symptoms were reported by those participants, they had significantly lower thermal and pain thresholds. This lends support to the mediation results and suggests that the underlying mechanism for discrimination’s effect on risk of PN may be related to the physiological effects of obesity and inflammation.

Other factors such as low socioeconomic status and smoking were found to be associated with prevalent PN. Women with lower socioeconomic status have been shown to have higher allostatic load stress markers, which according to the proposed biological mechanism above, could account for some of the increased odds of developing PN for women reporting financial strain (17,68). People of lower socioeconomic status also have poorer nutrition, exercise less, have potentially more exposure to environmental toxins, and have less access to health care than women of higher socioeconomic status (69–71). Even women who seek care for PN symptoms may be less likely to be properly tested and diagnosed based on their socioeconomic status (72). All of these factors may lead to poorer management and greater exacerbations of diseases such as PN among women of lower socioeconomic status. Furthermore, a recent meta-analysis found that smoking was significantly associated with an increased odds of diabetic PN (40). Our results are consistent with these findings. Smoking is hypothesized to contribute to neuropathies by constricting the blood vessels that supply nerves with blood and nutrients (73).

The observed prevalence of PN in our present analysis is higher than reported in previous studies among women with and without diabetes (74,75). In this analysis of the full SWAN cohort, the prevalence of PN is similar to that observed eight years earlier at the Michigan SWAN site (5). Notably, this current analysis found a high prevalence (22.8%) of PN in women without diabetes, suggesting that increased clinical focus on the problem of PN in the general population of women is need, especially given evidence that PN is a significant determinant of multiple dimensions of disability (76,77). Furthermore, the increased prevalence of PN may be due to the method of measurement of PN. Many previous studies have used only monofilament testing to characterize PN, which in the present analysis would lead to a reduction in the number of PN cases by approximately 50% (4,23). Early work describing the use of two instruments (in this case the MNSI and monofilament testing) proved to provide a more sensitive and reliable measure of neuropathies among individuals with diabetes than one instrument alone as each may be capturing differing severities of PN (31).

There are several limitations with this study. First, PN was not measured at baseline or at other early study visits, so we could not separate existing PN from neuropathy that developed over the course of the study period. This may have introduced issues of temporality into our study, however increasing age contributes to neuropathy development, and measures were chosen to establish temporality between perceived discrimination (baseline) and BMI (the mediator, measured concurrently with PN). Perceived discrimination was assessed at baseline, but these experiences can accumulate over time, and can be a regular and pervasive experience throughout the life course. Perceived discrimination has been shown to be stable over time in the SWAN cohort (19). Because 68 women (10.6%) reported that they experienced discrimination mainly because of their physical appearance, we ran a sensitivity analysis with these women removed in order to reduce the possibility of reverse causation – i.e. that for some, higher BMI (physical appearance) could be leading to discrimination. The estimates of the association of perceived discrimination with PN did not change significantly in this sensitivity analysis (unadjusted OR = 1.39, 95% CI: 1.10, 1.75; adjusted model OR = 1.18, 95% CI: 0.90, 1.53). Despite these limitations, this analysis has many strengths overall. Using two standardized measures of PN gives more robust estimates of disease (19,40). This large, community-based cohort design allowed us to investigate PN among women with and without diabetes, as well as women of diverse racial/ethnic backgrounds.

Nearly all previous PN research has focused primarily on populations with diabetes, however our analysis found that almost one-fourth of women without diabetes have PN. More research is needed to determine the contributing factors to non-diabetic PN. Clinicians should consider this high prevalence when diagnosing and treating both patients with and without diabetes for neuropathy. This study also found that midlife perceived discrimination was a significant predictor of PN and that this association was significantly mediated by BMI. Our findings re-affirm the impact of financial strain, BMI and diabetes as significant correlates of PN measured by symptom questionnaire and monofilament testing and show the impact of discrimination as an important risk factor for PN in women with and without diabetes.