Question

In an occupational epidemiology study, CHEMICAL A was shown to be associated with liver cancer with an Odds Ratio (OR) of 5.0. Subsequently, in the same occupational setting, suspicions arose that CHEMICAL B was associated with liver cancer. To further investigate this, a second study was conducted in which exposures to both CHEMICAL A and CHEMICAL B were measured, and assessed in relation to liver cancer. If CHEMICAL B was a confounder of the association between CHEMICAL A and liver cancer, then (1 mark):
a) The ORs for CHEMICAL A and liver cancer would be 1.0 within each stratum of CHEMICAL B in the second study (Exposed to CHEMICAL B versus Unexposed to CHEMICAL B)
b) CHEMICAL B would be associated with liver cancer, but not with exposure to CHEMICAL A;
c) CHEMICAL A would be a mediator of the effect of CHEMICAL B on liver cancer;
d) The ORs for CHEMICAL A and liver cancer would be still be 5.0 within each stratum of CHEMICAL B in the second study;
e) None of the above.

If, on the other hand, CHEMICAL B was an effect modifier of the relationship between CHEMICAL A and liver cancer, then (1 mark):
a) CHEMICAL B would be associated with liver cancer, and also with exposure to CHEMICAL A;
b) CHEMICAL B would be a mediator of the effect of CHEMICAL A on liver cancer;
c) The ORs for CHEMICAL A and liver cancer would be different within each stratum of CHEMICAL B in the second study (Exposed to CHEMICAL B versus Unexposed to CHEMCIAL B)
d) The ORs for CHEMICAL A and liver cancer would be 1.0 within each stratum of CHEMICAL B in the second study (Exposed to CHEMICAL B versus Unexposed to CHEMICAL B)
e) The ORs for CHEMICAL A and liver cancer would be 5.0 within each stratum of CHEMICAL B in the second study (Exposed to CHEMICAL B versus Unexposed to CHEMICAL B)

Answer 1

1.

If in a sub-group analysis, the original association between the exposure and outcome doesn’t hold up in BOTH sub-groups the factor is a confounder. But if the original association holds up in one group but not in another, then the factor is an effect modifier.

If CHEMICAL B was a confounder of the association between CHEMICAL A and liver cancer, then

a) The ORs for CHEMICAL A and liver cancer would be 1.0 within each stratum of CHEMICAL B in the second study (Exposed to CHEMICAL B versus Unexposed to CHEMICAL B)

If, CHEMICAL B was an effect modifier of the relationship between CHEMICAL A and liver cancer, then

c) The ORs for CHEMICAL A and liver cancer would be different within each stratum of CHEMICAL B in the second study (Exposed to CHEMICAL B versus Unexposed to CHEMCIAL B)