Positive association between job decision authority and systolic blood pressure: a statistical artifact?

We read with great interest the paper by Juvanhol et al (1) on the association between demand‒control model (DCM) components and blood pressure (BP) using the baseline data of the Brazilian Longitudinal Study of Adult Health (ELSA-Brasil). The authors reported a marginal, but positive cross-sectional association between job decision authority and systolic blood pressure (SBP) (0.59 mmHg, 95% CI 0.00-1.18) among non-users of anti-hypertensives from five universities and one research institute (2). And they claimed that (in the contemporary world of work), "the health benefits of increasing decision authority may not be as evident as in the historical context where the DCM was developed, which was characteristically one of industrialization and repetitive tasks." We agree with the authors that, in principle, the associations between psychosocial working conditions and health outcomes should be examined and interpreted with a socioecological perspective with consideration of societal and organizational structures and contexts (3). However, we could not find any convincing structure- and context-based explanation of the authors for the positive association between decision authority and SBP, for example, one grounded in their in-depth understanding of detailed working conditions and contexts of the Brazilian workers at higher education institutes during 2008-2010. Instead, we have found serious weaknesses in their analyses and interpretations, which undermine their finding and claim. First, the authors should have tested effect modification by race/color or socioeconomic status (SES) in their analysis. As expected, high decision authority was initially negatively associated with SBP among both users and non-users of anti-hypertensives (model 1, table 3) and also after controlling for gender, age, and recruitment site (model 2). However, only when race/color, schooling and mother's schooling were additionally included in their analysis (model 3), the direction of the association between decision authority and SBP changed from negative to positive among non-users of anti-hypertensives, while it substantially attenuated but remained negative among users of anti-hypertensives. Although the authors discussed these changes as a confounding effect of SES, it does not explain why the direction of the association was different between non-users and users of anti-hypertensives in model 3. In addition, there is a possibility of effect modification by race/color, schoolin