To the Editor The Viewpoint by Dr Chokshi and colleagues1 discussed apparent nonlinear associations often reported in studies. However, the examples used to highlight the inherently paradoxical nature of J-shaped curves were misleading, especially in the context of a risk factor such as blood pressure.
Studies have shown an erosion of the J-shaped curve when a more informed analytical approach is taken and the risk of confounding is rigorously mitigated.2 For example, in a large cohort of approximately 5000 men with a history of myocardial infarction,3 a J-shaped curve was observed between blood pressure and risk of all-cause and CHD mortality. But after accounting for reverse causality (2-year wash out), the same study reported linear associations.
Kiran A, Conrad N, Rahimi K. Nonlinear Exposure-Outcome Associations and Public Health Policy. JAMA. 2016;315(12):1287-1288. doi:10.1001/jama.2015.18026