Impact of neighbourhood socio-economic status on cardiovascular risk factors in a French urban population

Cardiovascular diseases (CVDs) remain a major burden worldwide with a positive association between neighbourhood deprivation and CVD,1,2 partially explained by an unequal distribution of cardiovascular risk factors (CVRFs).3 A few studies have focused on the impact of neighbourhood socio-economic status (SES) on CVRF repartition but never all together.4,5 The aim of this study is to offer an overall picture of neighbourhood deprivation impact on the distribution of CVRF. CARVAR 92 is an ongoing prospective observational cohort study of CVRF in the Western Suburbs of Paris (‘Hauts-de-Seine’). The detailed protocol has been previously published.6 The study was approved by the National Commission for Data Protection and Liberties and by the Institutional Data Protection Authority of Paris Saclay University. The target population was subjects without known CVD, aged between 40 and 70 years, included between 2010 and 2018. Diabetes mellitus was defined as a patient treated for diabetes mellitus or with a fasting plasma glucose value >1.26 g/l (7 mmol/L), hypertension was defined as a patient treated for hypertension or blood pressure exceeding 140 over 90 mmHg in nondiabetics and 130 over 80 mmHg in diabetic patients and obesity was defined as a body mass index ≥30 kg/m2. Dyslipidemia was defined as a patient treated for dyslipidemia or LDL-c >1.6 g/L. After factorial analysis for mixed data including numerous socio-economic variables such as education, profession, income, material possessions, we defined three socio-economic clusters of neighbourhoods: Cluster 1 [neighbourhood of high SES (high NSES)], Cluster 2 (medium NSES), and Cluster 3 (low NSES). The association between each CVRF and SES was assessed by a logistic model adjusted for known confounding factors from previous literature. Age was always included in the models, as well as sex when the analysis concerned the whole population. Obesity was also included in the modelling of diabetes. Each interaction was tested, and significant interactions (P < 0.1) were included in the model.