Response to invited editorial ‘What can be learnt from an atypical population?’

Abstract

In their editorial, Jelinek and Hammond highlight the nearly twofold variations in the prevalence of hypertension that we found across participating centres of the French CONSTANCES cohort. Based on our findings that socioeconomic status (SES), a known determinant of hypertension, did little to explain variations between centres, they raised the two following questions: (a) is this the consequence of an absence of association between SES and hypertension, which would be contrary to a large body of evidence; (b) could it be a consequence of the selected nature of our sample. Regarding the first question (association between SES and hypertension in CONSTANCES), we would like to underline that, as shown in Table 2 and Table 3, socioeconomic indicators were strongly and significantly associated with the prevalence of hypertension in age-adjusted models as well as in multivariate models. To illustrate, odds ratios for hypertension prevalence comparing the highest versus lowest level of education were 0.57 (95% confidence interval (CI) 0.53–0.62) in men and 0.45 (95% CI 0.41–0.49) in women in age-adjusted models (Table 2), meaning that higher education was strongly associated with a lower prevalence of hypertension in both men and women. Corresponding odds ratios in multivariate full model were 0.79 (95% CI 0.70–0.89) in men and 0.72 (95% CI 0.62–0.83) in women (model 4 from Table 3). Moreover, similar protective associations were found for higher versus lower occupation and less deprived versus more deprived area of residence. We would like to emphasise the linearity of the associations between socioeconomic indicators and hypertension (see Supplementary material), which means that, for example, having a high school diploma compared to a professional qualification was as protective as having a lower tertiary education compared to a high school diploma. Thus, despite selection, our results show a pronounced socioeconomic gradient and strengthen the role of socioeconomic indicators as determinants of hypertension prevalence. As mentioned in the discussion section, we could hypothesise that those associations would have been even stronger in a more representative sample. We reported a relatively low contribution of socioeconomic indicators to the geographical variations observed in hypertension prevalence. This means that in this sample, distributions of education, occupation and FDep scores did not really differ from one region to another. As stated in the methodological section, we chose to add the individual and neighbourhood-level factors sequentially in a pre-established order following a life course approach in our models. We first included body mass index as a proximal indicator reflecting the consequences of diet and physical activity from childhood up to inclusion. We postulate that modulating this order and introducing socioeconomic variables first would have modified the role of the latter indicators, as body mass index, which is associated with SES, may have absorbed a part of the geographical variations in socioeconomic disparities. To illustrate, introducing the socioeconomic variables before body mass index corresponds to a diminution of 20.1% in centre-level variance among men and 24.9% among women. Thus, more than selection bias, the order of introduction of determinants of hypertension prevalence in the models might have led to observing a low contribution of socioeconomic factors to geographical disparities in hypertension. Nevertheless, we agree that having a more varied distribution of education (or occupation or areas of residence) across regions might have increased the contribution of socioeconomic factors to the territorial variations in the prevalence of hypertension in our sample.