In the article by Nguyen et al. (1), a coding error led to an error in the direct and indirect effect bias sensitivity analysis coefficients and in Web Figure 1. This error does not substantively change the conclusions.
In the original article, the authors reported that “[t]he direct effect plot (Web Figure 1) indicates very little bias in the direct effect; the direct effect coefficient remains consistent (ranging from 0.195 to 0.219) and nonsignificant. The indirect effect coefficient ranges from −0.930 to −1.306; however the significance level is more sensitive to bias. Specifically, the indirect effect coefficient has P < 0.1 when λ ranges from −0.3 to 1.5, but the indirect effect has P > 0.1 when λ ranges from −1.5 to −0.4. This suggests that the indirect effect may be overestimated due to confounding bias for values of λ below −0.4—that is, if the potential BMI (under the active intervention) of persons more likely to adhere to the intervention was in actuality at least 0.4 units lower than that of persons less likely to adhere” (1, pp. 354–355).
The corrected description of the bias sensitivity analyses results, with the corrected coefficients, is as follows: “The direct effect plot (Web Figure 1) indicates very little bias in the direct effect; the direct effect coefficient remains significant across all λ (ranging from −0.522 to −1.126). The indirect effect coefficient ranges from −0.213 to 0.039. The indirect effect coefficient has P < 0.1 when λ ranges from −1.5 to −1.3, and the indirect effect has P > 0.1 when λ ranges from −1.2 to 1.5. This suggests that the indirect effect may be overestimated due to confounding bias for values of λ below −1.2—that is, if the potential BMI (under the active intervention) of persons more likely to adhere to the intervention was in actuality at least 1.2 units lower than that of persons less likely to adhere.”
The online version of the article and the Web material have been updated. The authors regret these errors.
Supplementary Material
REFERENCE
- 1. Nguyen QC, Osypuk TL, Schmidt NM, et al. Practical guidance for conducting mediation analysis with multiple mediators using inverse odds ratio weighting. Am J Epidemiol. 2015;181(5):349–356. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
