Abstract
Objectives:
Adults with diabetes are at an increased risk of atherosclerotic cardiovascular disease (ASCVD), and food insecurity may be a major and underappreciated risk compounder in this population. We sought to analyze the prevalence of food insecurity and its association with ASCVD in adults with diabetes.
Methods:
A total of 6424 participants with diabetes were included from the 2019 and 2020 National Health Interview Survey. Food insecurity was determined with a 10-question U.S. Adult Food Security Survey Module, and classified as high, marginal, low, and very low. ASCVD was defined as a self-reported history of coronary artery disease, myocardial infarction, or stroke.
Results:
Of the 6424 included participants (weighted: n = 21 690 217), 5 405 543 (24.4%) reported a history of ASCVD and 2 946 061 (13.3%) were identified as food insecure (low or very low food security). Adults with food insecurity were more likely to have ASCVD than adults who were food secure (28.9% vs 23.7%; P = 0.008). In the multivariate analyses adjusted for traditional cardiovascular risk factors, all levels of food insecurity were associated with ASCVD compared with food-secure adults (marginal security: odds ratio [OR]: 1.60; 95% confidence interval [CI], 1.18 2.18]; P = 0.003; low security: OR: 2.09; 95% CI, 1.58 2.74]; P<0.001; very low security: OR: 1.69; 95% CI, 1.22 2.34]; P = 0.001). The association persisted when adjusted for income, location, education, and insurance status. In adults with diabetes and ASCVD, income was a negative factor for food insecurity (OR: 0.71; 95% CI, 0.62 0.80; P < 0.001), but female sex and smoking were positive factors (OR: 1.90; 95% CI, 1.29 2.80; P = 0.001; and OR: 1.97; 95% CI, 1.23 3.18; P = 0.005; respectively). At younger ages, the prevalence of food insecurity increased, especially in adults with ASCVD.
Conclusions:
We showed that 13% of U.S. adults with diabetes are food insecure, which was associated with ASCVD independent of traditional and socioeconomic risk factors. Our findings emphasize the importance of recognizing food insecurity as a driver of ASCVD in adults with diabetes, and encourage future efforts at reducing this disparity.
Introduction
Food insecurity is a household-level economic and social condition of limited or uncertain access to adequate food. Access to food is a global social determinant of health not limited to third-world countries, but extends to developed nations as well. Approximately 1 in 8 Americans experience a form of food insecurity [1]. Food insecurity has been linked to poor medication self-management, increased risk of hospitalization, mental stress, greater health care expenditure, and mortality [2–5].
Additionally, food insecurity is linked to atherosclerotic cardio-vascular disease (ASCVD), with 1 in 7 U.S. adults with ASCVD experiencing food insecurity [6]. Adults with diabetes make up approximately 10% of the adult U.S. population [7], and are an especially vulnerable population for ASCVD [8]. Lack of reliable access to nutrition could negatively affect the achievement of glycemic targets and accelerate ASCVD [9 11]. The purpose of this study was to characterize the prevalence of food insecurity in adults with diabetes, its association with ASCVD, as well as the effect of other demographic or socioeconomic factors.
Methods
Population:
Participant data was obtained from the 2019 and 2020 adult National Health Interview Survey (NHIS). The NHIS is a cross-sectional survey used to assess the association between nutritional status and health outcomes, and is conducted annually. Data from 2019 and 2020 were pooled together and weighted in accordance with NHIS weighting procedures to extrapolate the data to the U.S. population and account for nonresponse bias and reinterview cases [12]. All data were self-reported. Baseline characteristics were obtained, including demographic information, medical history, social history, ASCVD information, and food insecurity status. Food insecurity was determined with the 10-question U.S. Adult Food Security Survey Module, and classified on a four-item scale (high security, marginal security, low security, very low security), as well as a two-item scale (food secure, food insecure). ASCVD was defined as a self-reported history of coronary artery disease (CAD), myocardial infarction (MI), or stroke. In our study, we included all adult participants with a history of diabetes, and excluded adults in whom food insecurity was unable to be ascertained. A total of 6424 participants met these criteria.
Statistical analysis:
All participants with diabetes were stratified by level of food insecurity, and baseline characteristics, including demographics (age, sex, race), traditional cardiovascular risk factors (hypertension, hyperlipidemia, obesity, active smoking), socioeconomic risk factors (education level, income, insurance status, location of residence), and ASCVD, were calculated. Differences in baseline characteristics by food insecurity were assessed using a 2 analysis for the categorical variables. The association between ASCVD and food insecurity (classified as high security, marginal security, low security, or very low security) was measured using multivariate logistic regression models, adjusting for traditional cardiovascular risk factors and other socioeconomic factors. The association between participant characteristics and food insecurity in all participants with diabetes, as well as participants with diabetes and ASCVD, was also assessed using multivariate logistic regression models. All analyses were performed and reported using data weighting in R, version 4.0.4, using the package survey.
Results
Of a total of 61 065 survey participants, 6424 had diabetes (weighted: n = 21 690 217; 9.2%) and were included in our study. Participants with diabetes were more likely to be food insecure compared with participants without diabetes (13.3% vs 7.6%; P < 0.001). Of all diabetic participants, 776 (weighted: n = 2 946 061; 13.3%) reported low or very low food security, and a total of 1662 participants (weighted: n = 5 405 543; 24.4%) reported a history of ASCVD. Of these participants, 577 (weighted: n = 1 887 796; 8.7%) reported a history of stroke, 1163 (weighted: n = 3 731 094; 17.2%) a history of CAD, and 718 (weighted: n = 2 340 545; 10.8%) a history of MI.
Baseline characteristics stratified by food insecurity are reported in Table 1. Adults with food insecurity were more likely to have ASCVD than adults who were food secure (28.9% vs 23.7%; P = 0.008; Table 2). When divided by individual outcome, adults with food insecurity were more likely to have a history of stroke (12.1% vs 8%; P = 0.001) and MI (13.8% vs 10%; P = 0.007). Difference in CAD did not reach statistical significance (18.6% vs 16.6%; P = 0.229).
Table 1:
Participant characteristics by food insecurity
| Participant characteristics | All, n (%) | Food secure, n (%) | Food insecure, n (%) | P value | |||
|---|---|---|---|---|---|---|---|
| (n = 21 690 217) | (n = 19 243 042) | (n = 2 946 061) | |||||
| Age, y | < 0.001 | ||||||
| <40 | 1 519 725 | (6.8) | 1 200 302 | (6.2) | 319 423 | (10.8) | |
| 40 65 | 10 238 431 | (46.1) | 8 429 522 | (43.8) | 1 808 908 | (61.4) | |
| 65 | 10 430 947 | (47) | 9 613 217 | (50) | 817 730 | (27.8) | |
| Sex | < 0.001 | ||||||
| Male | 11 080 630 | (49.9) | 9 951 929 | (51.7) | 1 128 701 | (38.3) | |
| Female | 11 108 473 | (50.1) | 9 291 113 | (48.3) | 1 817 360 | (61.7) | |
| Race | < 0.001 | ||||||
| Non-Hispanic white | 13 010 389 | (58.6) | 11 716 401 | (60.9) | 1 293 988 | (43.9) | |
| Hispanic | 3 867 192 | (17.4) | 3 159 730 | (16.4) | 707 462 | (24) | |
| Non-Hispanic black | 3 343 305 | (15.1) | 2 668 815 | (13.9) | 674 490 | (22.9) | |
| Other | 1 968 217 | (8.9) | 1 698 096 | (8.8) | 270 121 | (9.2) | |
| Hypertension | 16 020 546 | (72.3) | 13 730 203 | (71.4) | 2 290 343 | (77.9) | 0.003 |
| Hyperlipidemia | 13 643 409 | (61.9) | 11 792 113 | (61.5) | 1 851 297 | (64.2) | 0.277 |
| Obesity (body mass index >30 kg/m2) | 11 796 555 | (54.4) | 10 041 989 | (53.4) | 1 754 567 | (60.7) | 0.003 |
| Current smoker | 2 717 754 | (12.3) | 2 017 403 | (10.5) | 700 351 | (23.9) | < 0.001 |
| Education level | < 0.001 | ||||||
| Less than high school | 5 441 792 | (24.5) | 4 331 717 | (22.5) | 1 110 075 | (37.7) | |
| High school graduate | 10 439 824 | (47) | 9 080 109 | (47.2) | 1 359 715 | (46.2) | |
| College degree or above | 6 307 487 | (28.4) | 5 831 216 | (30.3) | 476 271 | (16.2) | |
| Income, U.S. $ | < 0.001 | ||||||
| 0 – 35 000 | 8 162 034 | (36.8) | 6 074 872 | (31.6) | 2 087 162 | (70.8) | |
| 35 000 – 50 000 | 3 590 414 | (16.2) | 3 166 511 | (16.5) | 423 903 | (14.4) | |
| 50 000 – 75 000 | 3 953 831 | (17.8) | 3 700 309 | (19.2) | 253 522 | (8.6) | |
| 75 000 – 100 000 | 2 423 239 | (10.9) | 2 317 261 | (12) | 105 978 | (3.6) | |
| >100 000 | 4 059 585 | (18.3) | 3 984 089 | (20.7) | 75 496 | (2.6) | |
| Medical insurance | < 0.001 | ||||||
| Private insurance | 11 608 214 | (52.3) | 10 628 504 | (55.2) | 979 710 | (33.3) | |
| Public insurance | 9 154 151 | (41.3) | 7 524 491 | (39.1) | 1 629 660 | (55.3) | |
| Uninsured | 1 426 738 | (6.4) | 1 090 048 | (5.7) | 336 690 | (11.4) | |
| Location | < 0.001 | ||||||
| Central urban | 6 402 845 | (28.9) | 5 381 649 | (28) | 1 021 196 | (34.7) | |
| Fringe urban | 4 728 310 | (21.3) | 4 320 225 | (22.5) | 408 086 | (13.9) | |
| Medium/small urban | 6 908 938 | (31.1) | 5 976 160 | (31.1) | 932 779 | (31.7) | |
| Nonmetropolitan area | 4 149 008 | (18.7) | 3 565 008 | (18.5) | 584 000 | (19.8) | |
Table 2:
Atherosclerotic cardiovascular disease (stroke, myocardial infarction, coronary artery disease) by food insecurity
| All, n (%) | Food secure, n (%) | Food insecure, n (%) | P value | |
|---|---|---|---|---|
| (n = 21 690 217) | (n = 19 243 042) | (n = 2 946 061) | ||
| All atherosclerotic cardiovascular disease | 5 405 543 (24.4) | 4 55 607 (23.7) | 851 935 (28.9) | 0.008 |
| Stroke | 1 887 796 (8.5) | 1 531 691 (8) | 356 105 (12.1) | 0.001 |
| Myocardial infarction | 2 340 545 (10.5) | 1 933 695 (10) | 406 850 (13.8) | 0.007 |
| Coronary artery disease | 3 731 094 (16.9) | 3 184 288 (16.6) | 546 806 (18.6) | 0.229 |
In the multivariate analyses adjusted for traditional cardiovascular risk factors (Fig. 1), all levels of food insecurity were associated with ASCVD compared with food-secure adults (marginal security: odds ratio [OR]: 1.60; 95% confidence interval [CI], 1.18 2.18; P = 0.003; low security: OR: 2.09; 95% CI, 1.58 2.74]; P<0.001; very low security: OR: 1.69; 95% CI, 1.22 2.34; P = 0.001). The association persisted when adjusted for additional socioeconomic factors, such as income, location, education, and insurance status (marginal security: OR: 1.46; 95% CI, 1.07 1.98; P = 0.016; low security: OR: 1.84; 95% CI, 1.38 2.44; P < 0.001; very low security: OR: 1.48; 95% CI, 1.05 2.09; P = 0.025).
Figure 1.
Logistic regression models for atherosclerotic cardiovascular disease by food insecurity.
*Versus high food security, adjusted for traditional cardiovascular risk factors (age, sex, race, hypertension, hyperlipidemia, smoking status)
yVersus high food security, adjusted for traditional cardiovascular risk factors (age, sex, race, hypertension, hyperlipidemia, smoking status) and socioeconomic risk fac-tors (income, location of residence, education level, insurance status). Reference is high food security.
In adults with diabetes, age >65 y and a higher income (per $100) were strong protective factors for food insecurity (OR: 0.29; 95% CI, 0.18 −0.47; P < 0.001; OR: 0.75; 95% CI, 0.70–0.80; P <0.001; respectively). Female sex (OR: 1.38; 95% CI, 1.12–1.71; P = 0.003), actively smoking (OR: 1.82; 95% CI, 1.42–2.33; P < 0.001), and public health insurance (OR: 1.37; 95% CI, 1.09–1.72; P = 0.008) were negative risk factors for food insecurity. In the subset of adults with diabetes and ASCVD, income remained a strong protective factor (OR: 0.71; 95% CI, 0.62–0.80; P < 0.001), female sex (OR: 1.90; 95% CI, 1.29–2.80; P = 0.001) and actively smoking (OR: 1.97; 95% CI, 1.23–3.18; P = 0.005) remained strong negative risk factors. (Table 3). At younger ages, participants’ prevalence of food insecurity increased, especially in adults with ASCVD (Fig. 2). Figure 3 shows the prevalence of food insecurity in adults with and without ASCVD, stratified by sex.
Table 3.
Logistic regression models for food insecurity in all adults and adults with atherosclerotic cardiovascular disease
| All adults | Adults with atherosclerotic cardiovascular disease | |||||
|---|---|---|---|---|---|---|
| Odds ratio | 95% confidence interval | P value | Odds ratio | 95% confidence interval | P value | |
| Age 40–65 y * | 0.96 | 0.61–1.52 | 0.857 | 1.28 | 0.24–6.76 | 0.771 |
| Age ≥65 y * | 0.33 | 0.20–0.54 | < 0.001 | 0.39 | 0.07–2.05 | 0.266 |
| Female sex | 1.43 | 1.15–1.76 | < 0.001 | 1.9 | 1.29–2.8 | 0.001 |
| Current smoker | 1.77 | 1.38–2.27 | < 0.001 | 1.97 | 1.23–3.18 | 0.005 |
| Hispanic race † | 1.12 | 0.81–1.56 | 0.482 | 1.64 | 0.85–3.15 | 0.140 |
| Non-Hispanic black race † | 1.34 | 1.01–1.78 | 0.041 | 1.23 | 0.74–2.02 | 0.425 |
| Other race † | 1.27 | 0.85–1.91 | 0.240 | 1.21 | 0.57–2.58 | 0.612 |
| Income (per $10 000) | 0.74 | 0.69–0.79 | < 0.001 | 0.71 | 0.62–0.8 | < 0.001 |
| Fringe urban area ‡ | 0.80 | 0.57–1.12 | 0.193 | 1.25 | 0.66–2.38 | 0.493 |
| Medium/small urban area ‡ | 0.90 | 0.68–1.18 | 0.443 | 1.28 | 0.77–2.13 | 0.337 |
| Nonmetropolitan area ‡ | 0.81 | 0.60–1.10 | 0.173 | 1.17 | 0.66–2.07 | 0.588 |
| High school diploma § | 0.83 | 0.64–1.07 | 0.145 | 0.79 | 0.51–1.22 | 0.284 |
| College degree or above § | 0.72 | 0.52–0.99 | 0.042 | 0.99 | 0.57–1.7 | 0.961 |
| Public insurance ║ | 1.38 | 1.10–1.73 | 0.006 | 0.89 | 0.60–1.33 | 0.583 |
| Uninsured ║ | 1.23 | 0.8–1.88 | 0.350 | 1.48 | 0.52–4.23 | 0.466 |
Versus age <40 y
Versus non-Hispanic white race
Versus central urban area
Versus less than high school education
Versus private insurance
Figure 2.
Prevalence of food insecurity stratified by age category (<30 y, 30–39 y, 40–49 y, 50–59 y, 60–69 y, 70–79 y, and >80 y) and atherosclerotic cardiovascular disease.
Figure 3.
Prevalence of food insecurity stratified by atherosclerotic cardiovascular disease in male and female participants.
Discussion
In this study, we demonstrated that between 2019 and 2020, 13% of U.S. adults with diabetes were food insecure compared with 7% of adults without diabetes. Very low, low, and marginal food insecurity were all associated with a greater prevalence of ASCVD, which persisted independent of socioeconomic and traditional cardiovascular risk factors. This relationship was unexpected in younger adults, where the overall prevalence of food insecurity was higher. More than half of adults with diabetes and ASCVD under age 30 y were food insecure. In addition, female sex and tobacco use nearly doubled the odds of food insecurity in adults with diabetes and ASCVD. Women with ASCVD had the highest rate of food insecurity of nearly 1 in 4 when compared with adults by sex and ASCVD. Higher income was a protective factor against food insecurity in this population, reinforcing the effect of financial insecurity and ASCVD, especially in adults with diabetes.
Although an emerging body of literature has documented associations between food insecurity and chronic noncommunicable disease, the mechanisms remain unclear. Those who are food insecure have a higher mortality risk and shorter life span compared with their secure counterparts. The replacement of healthy diets with inexpensive high-calorie diets may potentiate hyperglycemia, insulin resistance, ectopic adipose deposition, and target organ damage [5,9,13]. Recently chronic allostatic load is a common denominator for several chronic diseases, including diabetes and ASCVD [13,14].
Indeed, studies have highlighted a bidirectional relationship between food insecurity and ASCVD [15], where food insecurity may lead to ASCVD, which may, in turn, perpetuate additional insecurity, through economic hardship. Although the effect of food insecurity on ASCVD has been previously examined, to our knowledge, no studies have described this relationship in U.S. adults with diabetes on a national level. Food insecurity has been linked with not taking prescribed medications because of their cost in this population [16], which is further compounded with the baseline increased risk for ASCVD in adults with diabetes independent of food insecurity, with diabetes often treated as a CAD risk equivalent [17].
Our analysis shows that food insecurity further increased the risk of ASCVD in adults with diabetes. We also highlight certain adult populations at a greater risk of food insecurity, including those of young age, female sex, and using tobacco. Although the reasons for this are currently unclear, some factors may be relevant given the population being examined, which may include type 1 diabetes, the use of recreational tobacco products, and those caring for children, which may exacerbate economic hardships. However, these reasons are merely hypothesis generating.
Our study emphasizes the negative effect of food insecurity on ASCVD in adults with diabetes, a group that may require special attention and intervention on a population level. From a public health standpoint, there is growing evidence that screening for food insecurity may be important [8,18,19]. All health care providers should familiarize themselves with available social services in their community that can assist with addressing food insecurity, such as food banks or governmental support (e.g., Supplemental Nutrition Assistance Program; Women, Infants, and Children program). Health care efforts to address food insecurity on a community level are also important, such as hospital-associated food pantries, nutrition prescriptions, and public aid.
Our study has multiple limitations that must be acknowledged. This cross-sectional survey across the U.S. population reflects only a limited snapshot in time of a carefully chosen sample, where all data were extracted from self-reported interviews and therefore do not reflect causal relationships between ASCVD and food insecurity. The study design necessarily implies a potential risk for sample and reporter bias. For instance, the NHIS survey does not include people without access to phones or those who are unhoused, a group with likely the lowest food security that is not captured. This may underestimate the true prevalence of food insecurity, and result in bias as well. We were also unable to differentiate between adults with types 1 and 2 diabetes. Lastly, this was an observational study; thus, we were unable to determine causality or directionality between food insecurity and ASCVD. A strength is that this study encompasses a large representative sample of the adult U.S. population.
Conclusions
We showed that 13% of adults with diabetes are food insecure compared with 7% of adults without diabetes. All levels of food insecurity were associated with ASCVD, which persisted independent of socioeconomic and traditional cardiovascular risk factors. Food insecurity was especially prevalent in younger adults, women, and smokers. Our findings emphasize the importance of recognizing and addressing food insecurity as a driver of ASCVD in adults with diabetes.
References
- [1].Coleman-Jensen A, Rabbitt MP, Gregory C, Singh A. Household food security in the United States in 2018. Available at https://www.ers.usda.gov/webdocs/publications/94849/err-270.pdf. Accessed November 15, 2021.
- [2].Banerjee S, Radak T, Khubchandani J, Dunn P. Food insecurity and mortality in American adults: results from the NHANES-linked mortality study. Health Promot Pract 2021;22:204–14. [DOI] [PubMed] [Google Scholar]
- [3].Johnson KT, Palakshappa D, Basu S, Seligman H, Berkowitz SA. Examining the bidirectional relationship between food insecurity and healthcare spending. Health Serv Res 2021;56:864–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Men F, Gundersen C, Urquia ML, Tarasuk V. Prescription medication nonadherence associated with food insecurity: a population-based cross-sectional study. CMAJ Open 2019;7:E590–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [5].Seligman HK, Bindman AB, Vittinghoff E, Kanaya AM, Kushel MB. Food insecurity is associated with diabetes mellitus: results from the National Health Examination and Nutrition Examination Survey (NHANES) 1999 2002. J Gen Intern Med 2007;22:1018–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Mahajan S, Grandhi GR, Valero-Elizondo J, Mszar R, Khera R, Acquah I, et al. Scope and social determinants of food insecurity among adults with atherosclerotic cardiovascular disease in the United States. J Am Heart Assoc 2021;10:e020028. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [7].U.S. Department of Health and Human Services. National diabetes statistics report, 2020: estimates of diabetes and its burden in the United States. Available at https://www.cdc.gov/diabetes/pdfs/data/statistics/national-diabetes-statistics-report.pdf. Accessed November 15, 2021.
- [8].Seligman HK, Laraia BA, Kushel MB. Food insecurity is associated with chronic disease among low-income NHANES participants. J Nutrition 2010;140:304–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Seligman HK, Jacobs EA, Lopez A, Tschann J, Fernandez A. Food insecurity and glycemic control among low-income patients with type 2 diabetes. Diabetes Care 2012;35:233–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [10].Anand SS, Dagenais GR, Mohan V, Diaz R, Probstfield J, Freeman R, et al. Glucose levels are associated with cardiovascular disease and death in an international cohort of normal glycaemic and dysglycaemic men and women: the EpiDREAM cohort study. Eur J Prev Cardiol 2012;19:755–64. [DOI] [PubMed] [Google Scholar]
- [11].Berkowitz SA, Baggett TP, Wexler DJ, Huskey KW, Wee CC. Food insecurity and metabolic control among U.S. adults with diabetes. Diabetes Care 2013;36: 3093–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [12].Bramlett MD, Dahlhamer JM, Bose J. Weighting procedures and bias assessment for the 2020 National Health Interview Survey. Available at: https://ftp.cdc.gov/pub/Health_Statistics/NCHS/Dataset_Documentation/NHIS/2020/non-response-report-508.pdf. Accessed November 15, 2021.
- [13].Pak TY, Kim G. Association of food insecurity with allostatic load among older adults in the U.S. JAMA Netw Open 2021;4:e2137503. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [14].McClain AC, Xiao RS, Gao X, Tucker KL, Falcon LM, Mattei J. Food insecurity and odds of high allostatic load in Puerto Rican adults: the role of participation in the supplemental nutrition assistance program during 5 years of follow-up. Psychosom Med 2018;80:733–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Liu Y, Eicher-Miller HA. Food insecurity and cardiovascular disease risk. Curr Atheroscler Rep 2021;23:24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [16].Knight CK, Probst JC, Liese AD, Sercy E, Jones SJ. Household food insecurity and medication “scrimping” among U.S. adults with diabetes. Prev Med 2016; 83:41–5. [DOI] [PubMed] [Google Scholar]
- [17].Hajar R Diabetes as “coronary artery disease risk equivalent”: a historical perspective. Heart Views 2017;18:34–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [18].Sun Y, Liu B, Rong S, Du Y, Xu G, Snetselaar LG, et al. Food insecurity is associated with cardiovascular and all-cause mortality among adults in the United States. J Am Heart Assoc 2020;9:e014629. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [19].Walker RJ, Chawla A, Garacci E, Williams JS, Mendez C, Ozieh MN, et al. Assessing the relationship between food insecurity and mortality among U.S. adults. Ann Epidemiol 2019;32:43–8. [DOI] [PMC free article] [PubMed] [Google Scholar]