On 2015 Jul 24, Jay Kaufman commented:

Etsuji Suzuki of Okayama University, Japan contacted the authors to note that on page 3 of the article, the text states that that conditioning on CVD in Figure 1c opens four paths that bias the effect of obesity on mortality. However, the first of these paths that is listed in the text (obesity <- smoking -> CVD -> mortality) does not in fact contain a collider on the path, since no nodes are entered and exited through arrowheads. Therefore, the text should state that there are three such paths, not four. This error in the exposition of the problem changes neither the quantitative results nor the conclusions of the paper.

On 2015 Jul 24, Jay Kaufman commented:

Etsuji Suzuki of Okayama University, Japan contacted the authors to note that on page 3 of the article, the text states that that conditioning on CVD in Figure 1c opens four paths that bias the effect of obesity on mortality. However, the first of these paths that is listed in the text (obesity <- smoking -> CVD -> mortality) does not in fact contain a collider on the path, since no nodes are entered and exited through arrowheads. Therefore, the text should state that there are three such paths, not four. This error in the exposition of the problem changes neither the quantitative results nor the conclusions of the paper.