Question about HW2 Question 10d

Question about HW2 Question 10d

by Kristen Krysko -

Sorry I wasn't able to make it to section today due to clinic. I have a question about the answer key for question 10d. 

It is stated: 

It is concerning that they included prevalent users of AC because any baseline covariate included in the analysis should be carefully excluded as a potential mediator. Treatment modality could be an example of such a mediator. Indication for anticoagulants (Aspirin, warfarin, clopidogrel, Enoxaparin) includes heart disease, venous thrombosis, or immobility. These comorbidities could have influenced the selection of treatment modality, and treatment modality is related with the probability of PCSM or all cause mortality. 


I'm having trouble understanding how this is an example of mediation. I clinically agree that it is the comorbidites that influence treatment modality (rather than anticoagulants themselves), but I conceptualized the DAG as below. If comorbidities cause use of anticoagulants and also affect treatment modality, wouldn't this place treatment modality on a confounding path as below? 

I thought PSA, gleason score and tumor stage may be mediators if anticoagulants are truly protective as including prevalent users means that these baseline measures of cancer severity may be affected by anticoagulant use.  


Thanks,

Kristen

Re: Question about HW2 Question 10d

by Monica Ospina Romero -

Kirsten,

Thank you for reviewing the answer key and participating in this forum.

The DAG and your assumption about the association between anticoagulant use and treatment modality seem correct. In the answer key, we are explaining that we should be cautious when including prevalent user into the sample and selecting confounders to adjust for in the model. Some of the covariates can be mediators instead of confounders. In contrast, when we exclude prevalent users and adjust for baseline covariates, we are less concerned that these covariates are mediators because they were measured before the study participants were exposed. We use treatment modality as an example of a potential mediator, assuming anticoagulant use might be a contraindication for one of those two treatments options (RT vs RP). This scenario might be possible for warfarin where a doctor should balance between the risk of bleeding or a thrombotic event, but maybe not the case for aspirin or clopidogrel that can be stopped without a major concern before any surgery.


Re: Question about HW2 Question 10d

by Kristen Krysko -

Thank you Monica. 

That explanation makes sense, and I did mention that if anticoagulant use directly affected the treatment modality then it could be a mediator. There are peri-operative strategies that can be used for drugs like warfarin such as bridge to heparin or enoxaparin, so I didn't think it would be the anticoagulant that directly affects treatment modality choice, but thought it was more the comorbidities that affect treatment modality due to surgical risks as shown in the DAG I posted. 


Thanks!

Kristen