Individual/Neighborhood Social Factors and Blood Pressure in the RECORD Cohort Study
Which Risk Factors Explain the Associations?
Recent studies have started to suggest that, beyond effects of individual socioeconomic profiles, socioeconomic characteristics of residential neighborhoods are independently associated with blood pressure. However, mechanisms involved in these associations remain unknown. To distinguish between different mechanisms, we investigated whether specific risk factors of hypertension (physical inactivity, alcohol consumption, smoking, body mass index, waist circumference, and resting heart rate) intervene as mediators in the associations between individual or neighborhood socioeconomic characteristics and systolic blood pressure. We relied on data from the RECORD Cohort Study (Residential Environment and CORonary heart Disease) on 5941 participants recruited in 2007–2008, aged 30 to 79 years, residing in 1824 neighborhoods in the Paris metropolitan area. Systolic blood pressure increased independently and regularly with both decreasing individual education and decreasing residential neighborhood education. Body mass index/waist circumference and resting heart rate mediated an appreciable share of the associations between education and blood pressure and, adding validity to the finding, were the 2 most significant mediators for the effects of both individual education and neighborhood education. We found that 52% (95% CI: 25% to 79%) of the association between neighborhood education and blood pressure was mediated by body mass index/waist circumference and 20% (95% CI: 5% to 36%) by resting heart rate. Future research will have to clarify the exact mechanisms through which body weight and shape and resting heart rate intervene as mediators in the associations between individual/neighborhood education and blood pressure.
Considering socioeconomic characteristics is useful both in the clinical setting to improve risk stratification of patients at risk of hypertension and, from a public health perspective, to identify population-level determinants of blood pressure when defining interventions.1 Regardless of the aim, recent studies suggest that a better assessment of socioeconomic differences in blood pressure may be obtained by considering social circumstances both at the individual level and at the residential neighborhood level.2–5 To date, fewer studies have quantified neighborhood socioeconomic influences on blood pressure than on behavioral risk factors of cardiovascular diseases (smoking and physical inactivity) or obesity.6
As recently emphasized,7 knowledge useful for public health action is identifying the different mechanisms underlying associations between individual/neighborhood socioeconomic characteristics and blood pressure (eg, through known risk factors of hypertension) on which it would be possible to intervene to address social disparities in blood pressure.1,8 However, on the one hand, the very few studies that investigated intermediate mechanisms through which individual socioeconomic variables relate to blood pressure9 have generally included all of the mediating risk factors simultaneously in the models, not permitting us to disentangle the independent mediating role of different risk factors of hypertension.1 On the other hand, the only study that investigated mediating processes between neighborhood socioeconomic characteristics and blood pressure has only considered weight status as a mediator,2 not allowing us to compare different risk factors according to their importance in explaining social environment effects on blood pressure. Therefore, our aims were to assess whether individual and neighborhood socioeconomic characteristics influence blood pressure and to investigate the extent to which health behavior, body weight and shape, and resting heart rate may contribute to these relationships.
The RECORD Cohort Study (Residential Environment and CORonary heart Disease) includes 7293 participants who were recruited between March 2007 and February 2008. The participants benefitted from a free medical checkup, offered every 5 years by the French National Health Insurance System for Salaried Workers to all working and retired employees and their families. Participants were recruited without a priori sampling during these 2-hour–long preventive checkups conducted by the Centre d’Investigations Préventives et Cliniques10,11 in 4 of its health centers, located in the Paris metropolitan area (Paris, Argenteuil, Trappes, and Mantes-la-Jolie). Eligibility criteria were age 30 to 79 years, ability to fill out the study questionnaires, and residence in 1 of 10 (of 20) administrative divisions of Paris or 111 other municipalities of the metropolitan area selected a priori. Among those who came to the health centers and who were eligible on the basis of age and residence, 10.9% were not selected for participation because of linguistic or cognitive difficulties in filling out questionnaires. Individuals selected for participation were informed about the study by trained survey technicians. Of these, 83.6% accepted to participate and completed the data collection protocol.
Participants were accurately geocoded on the basis of their residential address in 2007–2008. Research assistants corrected all of the incorrect or incomplete addresses with the participants by telephone. Extensive investigations with local departments of urban planning were conducted to complete the geocoding. Precise spatial coordinates and block group codes were identified for 100% of the participants. The study protocol was approved by the French Data Protection Authority.
In the present study, only participants recruited in the Paris health center were considered. After excluding individuals with missing values for selected variables (see below), 5941 participants, living in 110 different municipalities or 1824 neighborhoods, were included in the analyses.
Systolic Blood Pressure
During the health checkup, supine brachial blood pressure was measured by trained nurses 3 times in the right arm after a 10-minute rest period, using a manual mercury sphygmomanometer.10 A standard cuff size was used, but a large cuff was utilized if necessary. The first Korotkoff phase was used to define systolic blood pressure (SBP; please see the online Data Supplement at http://hyper.ahajournals.org for comparable analyses conducted for diastolic blood pressure). The mean of the last 2 measurements was taken as the outcome.11
Individual Sociodemographic Variables
Various sociodemographic characteristics of participants were considered: age, sex, marital status, individual education, parental education, occupation, employment status, household income, self-reported financial strain, dwelling ownership, and Human Development Index of each participant’s country of birth. Age was considered as a continuous variable. Marital status was coded in 2 classes: living alone or as a couple. Education was divided in 4 classes: (1) no education; (2) primary education and lower secondary education; (3) higher secondary education and lower tertiary education; and (4) upper tertiary education. For parental education level, we created an education variable by adding the mother’s and father’s education level (1: primary school or less; 2: secondary school; and 3: tertiary school) and divided the resulting score into 3 classes (2, 3 to 4, and 5 to 6). Mother’s and father’s education were also considered separately.12 Regarding occupation, in accordance with the French National Institute of Statistics and Economic Studies, 4 categories were distinguished: (1) blue-collar workers; (2) low white-collar workers; (3) intermediate occupations; and (4) high white-collar workers. Employment status was coded in 3 categories: employed, unemployed, and retired. Household income adjusted for household size was divided into 4 categories. A binary variable for self-reported financial strain and a binary variable for dwelling ownership were determined. Each individual’s self-reported country of birth was also taken into account. We followed an approach by Merlo13 in attributing to each individual the 2004 Human Development Index of his/her country of birth, as a proxy of the country’s social development level. Following the United Nations Development Programme,14 a binary variable was used to distinguish people born in low-development countries (Human Development Index <0.5) from others.
Antihypertensive Medication Use
By merging Système d’Information Inter Régimes de l’Assurance Maladie or Federative Information System of Sick Insurance data for all of the reimbursed healthcare consumption in participants in 2006–2008 to the RECORD database, we created a binary variable indicating whether individuals had been reimbursed for any antihypertensive medication over the previous year.
Risk Factors of High Blood Pressure
The following risk factors of high blood pressure were considered as possible mediators in the associations between socioeconomic variables and blood pressure: physical inactivity, alcohol consumption, smoking, body mass index, waist circumference, and resting heart rate. Family history of hypertension was also taken into account for adjustment.
Family history of hypertension was assessed from the questionnaire. Participants were asked whether they were physically active (at work, during transportation, or during leisure time) for an equivalent of >1 hour of walking per day. Alcohol consumption was coded in 4 categories: never drinker, former drinker, light drinker, and drinker (>2 glasses per day for women and 3 glasses per day for men). For smoking, we distinguished between nonsmoker, former smoker, and current smoker.
Height (using a wall-mounted stadiometer) and weight (using calibrated scales) were recorded by a nurse.10 Body mass index was divided into 3 categories (normal: <25 kg/m2, overweight: 25 to <30 kg/m2, and obese: ≥30 kg/m2). Waist circumference was measured in centimeters using an inelastic tape placed midway between the lower ribs and iliac crests on the midaxillary line.11 It was divided into 3 categories (<94 cm, 94 to ≤102 cm, and >102 cm among men; <80 cm, 80 to ≤88 cm, and >88 cm among women). Resting heart rate was measured by ECG after a 5- to 7-minute rest period10 and was subsequently divided into 3 classes: <60 bpm, 60 to <70 bpm, and ≥70 bpm (70 rather than 80 bpm was used as a cutoff because only 4.8% of the participants had a resting heart rate ≥80 bpm).
Neighborhood Socioeconomic Variables
Neighborhoods were assessed as census block groups.2 These 1824 local units were defined from the 1999 census so as to be relatively homogeneous in terms of sociodemographic and housing characteristics. The median number of residents per neighborhood was 2393 in 1999 (interquartile range: 2084 to 2903).
The following socioeconomic variables were defined at the neighborhood level: (1) the proportion of residents aged ≥15 years with an upper tertiary education (1999 census); (2) median income in 2005 (Tax Registry of Direction Générale des Impôts, General Directorate of Taxation); and (3) mean value of dwellings sold in 2003–2007 (Paris-Notaries). All of the neighborhood sociodemographic variables were divided into 4 categories composed of a similar number of individuals.
We excluded 176 participants with missing information for SBP and 50 individuals with missing data for individual education. Individuals with missing information for any of the mediating risk factors were also excluded, resulting in a final sample of 5941 individuals.
Initial Multilevel Analyses
To account for the strong within-neighborhood correlation in SBP, individual and neighborhood predictors of SBP were analyzed with multilevel linear regression models.15 To derive parsimonious models, only the individual and neighborhood variables that were associated with SBP were retained in the model.
The individual risk factors of hypertension were then further entered into the model. Our aim was to assess the extent to which risk factors of hypertension mediated, that is, explained, the associations between individual/neighborhood socioeconomic variables and SBP. Family history of hypertension was considered not to be a plausible mediator between the participant’s individual and neighborhood education level and his/her blood pressure, but was entered into the model for adjustment. Our aim was to rank risk factors of hypertension according to their importance in mediating the associations between individual/neighborhood socioeconomic variables and SBP.
We used the path analysis model described in the Figure to approximately quantify the share of the associations between individual/neighborhood socioeconomic variables and SBP that was statistically explained by each of the mediating risk factors. When required assumptions are met,16 this approach allows one to decompose an association (between individual/neighborhood socioeconomic factors and SBP) into direct and indirect effects (through each of the mediating risk factors).17
In these mediation analyses, individual and neighborhood socioeconomic variables were incorporated as ordinal variables. As explained in the online Appendix (please see online Data Supplement), all of the potential mediating risk factors were introduced as binary or ordinal variables to ensure homogeneity in the definition of mediators (see coding in footnotes to Table 2). Because body mass index and waist circumference were too correlated (Pearson correlation among the 3-category ordinal variables: 0.69; P<0.0001) to be introduced into the path analysis model simultaneously, we constructed a 9-category ordinal variable combining these factors (see the online Data Supplement).
All of the regression equations involved in the path analysis model (for SBP and all of the mediating risk factors) were adjusted for individual sociodemographic variables, antihypertensive medication use, and family history of hypertension. Technical details pertaining to the model, interpretation of the coefficients, and assumptions required for a valid decomposition into direct and indirect effects16 are described in the online Appendix.
In our sample, mean SBP was 127.6 mm Hg (95% CI: 127.2 to 128.0 mm Hg), mean body mass index was 25.3 kg/m2 (95% CI: 25.2 to 25.5 kg/m2), and mean resting heart rate was 62.2 bpm (95% CI: 62.0 to 62.5 bpm). Overall, 11.1% of the participants were on antihypertensive medication and 3.7% were on β-blockers. Details of the distribution of individual and neighborhood variables and SBP levels according to these variables are reported in the online Appendix. For example, mean age-adjusted SBP was 126.1, 127.3, 129.9, and 131.8 mm Hg and 126.0, 126.6, 128.0, and 130.4 mm Hg in the 4 categories of decreasing individual and neighborhood education, respectively.
The individual and neighborhood variables that were associated with SBP after mutual adjustment are shown in Table 1. A higher SBP was observed among individuals who did not own their dwelling and among people born in a low human development country. SBP was lower among the unemployed. However, individual socioeconomic influences were dominated by the strong dose-response increase in SBP with decreasing education level of participants. The other individual socioeconomic variables were not associated with SBP (SBP tended to increase with decreasing mother’s education, but the association was not statistically significant after adjustment).
Regarding neighborhood influences, SBP showed a much stronger pattern of association with neighborhood education than with the other neighborhood variables. After controlling for individual covariates, SBP regularly increased with decreasing neighborhood education (Table 1). Once neighborhood education was introduced into the model, the other neighborhood variables were not associated with SBP.
Individual risk factors for high blood pressure were then introduced into the model (Table 1, second column). Factors associated with a higher SBP included family history of hypertension, alcohol consumption, overweight or obesity, high waist circumference, and a medium or high heart rate. As in other studies where blood pressure was also measured after a tobacco-free interval,18 smoking was associated with a lower SBP. Regular physical activity, which included activity at work, was associated with a slightly higher SBP. Interestingly, associations between individual or neighborhood education and SBP were drastically reduced after adjustment for risk factors. The association with individual education persisted after adjustment, whereas that with neighborhood education was reduced to nonsignificance.
The path analysis model depicted in the Figure was used to assess the role played by possible mediating risk factors in explaining associations between individual/neighborhood education and SBP (please see the online Data Supplement for interpretation of the coefficients). As shown in the left part of Figure, a high individual education was associated with excessive alcohol use, lower odds of being physically active (including activity at work), lower odds of smoking and obesity, and with a lower resting heart rate. High neighborhood education was associated with alcohol use, smoking, lower odds of obesity, and lower resting heart rate. Indirect effects of individual/neighborhood education on SBP through each of these mediators, expressed in millimeters of mercury, are shown in the online Appendix. Only the main indirect effects (“explaining” ≥10% of the associations of interest as reported in Table 2) are commented on below.
As shown in the Figure, alcohol consumption increased (rather than decreased) with both individual and neighborhood education. Therefore, alcohol consumption tended to mask rather than explain the associations between individual/neighborhood education and SBP (Table 2).
The main indirect effects of education level on SBP through mediating risk factors involved body mass index/waist circumference and heart rate (Table 2). Adding validity to the finding, these 2 factors were the main mediators for the effects of both individual education and neighborhood education. Because decreased individual and neighborhood education were associated with a higher body mass index/waist circumference and resting heart rate, the indirect effects of individual/neighborhood education on SBP through these 2 risk factors significantly contributed to the higher SBP observed in low-education individuals and low-education neighborhoods.
According to the proportion of the associations explained by each of the risk factors (Table 2), body mass index/waist circumference was, by far, the most significant contributor to the relationships between education and SBP: the corresponding indirect effects represented 28.0% (95% CI: 16.3% to 39.7%) of the individual education-SBP association and 51.6% (95% CI: 24.5% to 78.6%) of the neighborhood education-SBP association. Resting heart rate was the second contributor, accounting for ≈14.7% (95% CI: 6.1% to 23.3%) of the association of SBP with individual education and 20.4% (95% CI: 4.6% to 36.2%) of its association with neighborhood education.
As shown in the Figure, after accounting for all of the indirect effects, the direct (residual) association between neighborhood education and SBP was no longer statistically significant, but the association persisted between individual education and SBP.
We found that a decrease in both individual education and neighborhood education was independently associated with a nonnegligible and regular increase in SBP. However, the most innovative aspect of our study was our analyses, which identified body mass index/waist circumference and resting heart rate as the most important intermediate variables contributing to these associations.
Strengths of the present study include meticulous geocoding of the participants, the large study territory that allowed comparison of diverse neighborhoods, the fact that both individual and neighborhood socioeconomic variables were taken into consideration, and the multilevel path analysis framework2 implemented to investigate mediating pathways in the education-SBP associations. The primary study limitation is the use of cross-sectional data, which did not enable us to confirm that the temporal sequence of phenomena was coherent with the mediation hypothesis (individual/neighborhood education → risk factors of hypertension → SBP).
Associations Between Neighborhood Education and Blood Pressure
Compared to physical inactivity or obesity,6 few studies have investigated relationships between neighborhood socioeconomic characteristics and blood pressure.2–5 In the present study, comparing 3 socioeconomic variables (education, income, and dwelling values), only neighborhood education was independently related to SBP, which is consistent with previous findings established at the individual level.19 It is tempting to speculate that a reason why this particular socioeconomic variable emerged as a predictor of blood pressure is that a high neighborhood education level is associated with health norms and knowledge that promote healthy behavior in terms of diet, physical activity, and healthcare use. However, because of the strong correlation between neighborhood education and alternative neighborhood socioeconomic factors or other environmental variables, and in the absence of other supporting data, this statement may be excessive. At least it is an interesting hypothesis for future research.
Mechanisms in the Association Between Education and Blood Pressure
An original aspect of our study was the attempt to investigate which, of a number of risk factors of hypertension, intervene as mediating mechanisms in the associations between individual/neighborhood education and SBP (under the hypothesis of the causal diagram in the Figure and the assumptions16 discussed in the online Appendix). The only other study that examined intermediate processes between neighborhood socioeconomic variables and blood pressure only considered weight status as a mediator,2 not providing a distinction between the different underlying mediating mechanisms, as we did here.
Strikingly, it was found that the 2 risk factors (of the 5 variables considered) that emerged as the main statistical mediators were the same for the effects of individual education and neighborhood education. Body mass index/waist circumference was by far the main mediating variable,1 but resting heart rate also had a nonnegligible mediating role.
The fact that a significant part of the association between individual education and blood pressure was mediated by the lower body weight of high educated individuals may be partly because of their more accurate knowledge of health risks associated with obesity and greater motivation to control weight. Also, in additional path analyses reported in the online Appendix, we found that low individual education was associated with perceived stress, which was associated with a higher body mass index/waist circumference, which was, in turn, associated with a higher blood pressure. However, not to mention uncertainties in the causal model involving stress (see Discussion in the online Data Supplement), perceived stress only explained 5% of the mediating role that body weight had in the association between individual education and blood pressure. Similarly, the pathway from “individual education” to “symptoms of depression” to “body mass index/waist circumference” to “SBP” was statistically significant but only accounted for 2.4% of the mediating effect of body weight (see the online Data Supplement).
The large share of the relationship between neighborhood education and SBP that was mediated by body mass index/waist circumference may be interpreted in light of the growing literature showing effects of a number of environmental dimensions on physical activity and dietary behavior.20,21 A number of mechanisms related to the neighborhood environment, including specific behavioral habits, a more unfavorable food environment, and a weaker potential for active living in disadvantaged neighborhoods, may increase blood pressure through body weight and shape modification.
Resting heart rate statistically explained a nonnegligible share of the individual/neighborhood education-SBP associations. We may be tempted to hypothesize that part of the mediating effect of heart rate is attributable to the effects of stress on heart rate. Certain authors5 have interpreted associations between individual or neighborhood socioeconomic characteristics and heart rate in relation to the chronic stressors associated with poverty or life conditions of disadvantaged neighborhoods (eg, crowding, noise, unemployment, crime, and violence). Therefore, given reported effects of heart rate on hypertension incidence,22 it is possible that part of the effect of individual/neighborhood education on SBP through heart rate reflects the greater exposure to economic and environmental stressors in high-deprivation neighborhoods. However, additional analyses reported in the online Appendix indicated that perceived stress was not associated with SBP and that perceived stress played no part in the nonnegligible mediating effect that resting heart rate had in the associations between individual/neighborhood education and blood pressure. Longitudinal studies using other measures of psychological stress are needed to explore this issue further.
Although heart rate was assessed after a sufficient resting period, we cannot exclude the possibility that the higher heart rate associated with low education reflected a temporary increase in heart rate when visiting the health center (white coat syndrome). Moreover, we could not verify whether heart rate really intervened as a mediator between individual/neighborhood education and blood pressure or whether 1 or more factors, including sympathetic activity, were in fact influencing both heart rate and blood pressure. Thus, whereas our study allows one to generate innovative hypotheses on the mediating role of resting heart rate in socioeconomic effects on blood pressure, its findings will have to be replicated with a longitudinal design in the next stages of the RECORD Cohort Study.
Finally, additional analyses reported at the end of the online Appendix showed that, among hypertensive participants, a high individual or neighborhood education was not associated with awareness of hypertension, antihypertensive medication use, and control of hypertension, making it rather unlikely that these factors contribute to the observed education-blood pressure associations.
From a clinical viewpoint, our study suggests that, after accounting for individual risk factors, socioeconomic characteristics of residential environments were not especially useful for identifying individuals with high blood pressure. However, implications are different from a public health perspective, which would conclude that neighborhood environments affect blood pressure through specific risk factors. In this latter approach, it is critical to disentangle the mechanisms through which neighborhood physical and social environments influence blood pressure (eg, through weight status or heart rate modification) to effectively intervene in reducing sociospatial disparities in blood pressure.1,7
We thank Alfred Spira, head of the French Institute for Public Health Research, for his advice and support. We are also grateful to Insee, the French National Institute of Statistics and Economic Studies, which provided support for the geocoding of the RECORD participants and allowed us access to relevant geographical data (with particular thanks to Aline Désesquelles, Pascale Breuil, and Jean-Luc Lipatz). We thank Geoconcept for giving us access to the Universal Geocoder software. We are grateful to Paris-Notaries for providing geographical data used in the present analysis. We also thank the Caisse Nationale d’Assurance Maladie des Travailleurs Salariés (France) and the Caisse Primaire d’Assurance Maladie de Paris (France) for helping make this study possible. We are grateful to Alain Weill from Caisse Nationale d’Assurance Maladie des Travailleurs Salariés for his support in merging the healthcare consumption data from the Système d’Information Inter Régimes de l’Assurance Maladie or Federative Information System of Sick Insurance to the RECORD database.
Sources of Funding
As part of the RECORD project, the present work was funded by the National Research Agency (ANR; Health-Environment Program); the Institute for Public Health Research (IReSP); the National Institute for Prevention and Health Education (INPES; Prevention Program 2007); the National Institute of Public Health Surveillance (InVS; Territory and Health Program); the French Ministries of Research and Health (Epidemiologic Cohorts grant 2008); the National Health Insurance Office for Salaried Workers (CNAM-TS); the Ile-de-France Health and Social Affairs Regional Direction (DRASSIF); the Ile-de-France Public Health Regional Group (GRSP); and the Ile-de-France Youth and Sports Regional Direction (DRDJS).
- Received September 28, 2009.
- Revision received October 20, 2009.
- Accepted December 23, 2009.
Chaix B, Ducimetiere P, Lang T, Haas B, Montaye M, Ruidavets JB, Arveiler D, Amouyel P, Ferrieres J, Bingham A, Chauvin P. Residential environment and blood pressure in the PRIME Study: is the association mediated by body mass index and waist circumference? J Hypertens. 2008; 26: 1078–1084.
Diez-Roux AV, Nieto FJ, Muntaner C, Tyroler HA, Comstock GW, Shahar E, Cooper LS, Watson RL, Szklo M. Neighborhood environments and coronary heart disease: a multilevel analysis. Am J Epidemiol. 1997; 146: 48–63.
Riva M, Gauvin L, Barnett TA. Toward the next generation of research into small area effects on health: a synthesis of multilevel investigations. J Epidemiol Community Health. 2007; 61: 853–861.
Nilsson PM. Adverse social factors can predict hypertension–but how? Eur Heart J. 2009; 30: 1305–1306.
Kivimaki M, Lawlor DA, Smith GD, Keltikangas-Jarvinen L, Elovainio M, Vahtera J, Pulkki-Raback L, Taittonen L, Viikari JS, Raitakari OT. Early socioeconomic position and blood pressure in childhood and adulthood: the Cardiovascular Risk in Young Finns Study. Hypertension. 2006; 47: 39–44.
Thomas F, Bean K, Pannier B, Oppert JM, Guize L, Benetos A. Cardiovascular mortality in overweight subjects: the key role of associated risk factors. Hypertension. 2005; 46: 654–659.
Williams RB, Marchuk DA, Siegler IC, Barefoot JC, Helms MJ, Brummett BH, Surwit RS, Lane JD, Kuhn CM, Gadde KM, Ashley-Koch A, Svenson IK, Schanberg SM. Childhood socioeconomic status and serotonin transporter gene polymorphism enhance cardiovascular reactivity to mental stress. Psychosom Med. 2008; 70: 32–39.
United Nations Development Programme. Human Development Report 2007/2008. New York, NY: United Nations Development Programme; 2008. Available at: http://hdr.undp.org/. Accessed August 1, 2009.
Chaix B, Rosvall M, Merlo J. Recent increase of neighborhood socioeconomic effects on ischemic heart disease mortality: a multilevel survival analysis of two large Swedish cohorts. Am J Epidemiol. 2007; 165: 22–26.
Conen D, Glynn RJ, Ridker PM, Buring JE, Albert MA. Socioeconomic status, blood pressure progression, and incident hypertension in a prospective cohort of female health professionals. Eur Heart J. 2009; 30: 1378–1384.
Glanz K, Sallis JF, Saelens BE, Frank LD. Healthy nutrition environments: concepts and measures. Am J Health Promot. 2005; 19: 330–333, ii.
Reed D, McGee D, Yano K. Biological and social correlates of blood pressure among Japanese men in Hawaii. Hypertension. 1982; 4: 406–414.