Inquiry - E-value interpretation in mediation analysis

[Johnathan T]

Dear Lavaan team & members,

I am reaching out to request your expertise on interpreting E-values in light of a previous publication's recommendations to compare E-values with reported confounder risk/odds ratios for sensitivity analysis (DOI: https://doi.org/10.1016/j.jclinepi.2023.09.014). In short, the paper promotes the practice of interpreting E-values in the context of the odds ratios (ORs) measured from known confounding variables.

Hypothetically, if an E-value is reported in a study and all the included confounders in the model but 1 had ORs below the LL of this E-value, is there validity in reporting these results? Particularly, if the only confounder that fell into the E-value's range was a known very significant effector of the mediation model, could I interpret my model as moderately robust?

Below is an theoretical example that is applicable to my inquiry:
-Suppose a mediation model was examining the potential mediators between smoking and peripheral neuropathy. Confounders of the model included: age, gender, BMI, and diabetes status. After running the mediation model, a mediator "x" was shown to significantly mediate the relationship between smoking and peripheral neuropathy. However, its reported E-value was 1.21 [1.13-1.27], and all of the confounders had the following ORs: age [1.08]. gender [1.03], BMI [1.03], and diabetes [1.14]. Considering the strong link already known between diabetes and peripheral neuropathy, would the results of this particular study be warranted for publication at all?

My logic is that an unmeasured confounder must have quite a strong relationship to negate the effects of the mediation model if it matches the influence of diabetes on peripheral neuropathy. Therefore, the results of the mediation model may have usefulness because an unmeasured confounder of this magnitude would be unlikely in most patient populations.

I have yet to find any studies that interpret mediation results in this context, and I would love to hear your opinions.

Sincerely,
Johnathan T

[Mugo Muiruri]

good 

--
You received this message because you are subscribed to the Google Groups "lavaan" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lavaan+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/lavaan/0b055a3f-e9ae-4aaf-9c8e-485bdabf408en%40googlegroups.com.