Thread by @WhitneyEpi

@MaxJordan_N @EpiEllie @AmakaEMD @DrDeidraCrews That sounds reasonable. Admittedly I am not trained in genetic epi, so I may be totally off-base and welcome correction. Let’s say there are a bunch of genetic variants that increase risk of hypertension. Holding all else equal (waving hands), 2 increase risk by same degree 1/

@MaxJordan_N @EpiEllie @AmakaEMD @DrDeidraCrews Variant A is represented at 4% among a US Black same, 8% among an insular European sample that’s often studied. Variant B is present in 4% of US Black sample and 0% in European samples (my understanding is that people of African descent in US more genetically varied than...

@MaxJordan_N @EpiEllie @AmakaEMD @DrDeidraCrews ...most “White” populations as conceptualized in these US sense so I think this is plausible. So here there is some relationship between the “race” node and genetic variants bc of all the processes that have grouped people and structured society. But it’s not a 1-1 overlap.

@MaxJordan_N @EpiEllie @AmakaEMD @DrDeidraCrews 5/ so then there’s ALL the other things and environments that produce risk. In the US these are often patterned around racial categories. In fact sometimes they are SO racially patterned that it’s hard to tease each one apart from other “exposures” that are racially patterned...

@MaxJordan_N @EpiEllie @AmakaEMD @DrDeidraCrews 6/ ...(see #epitwitter literature on “structural confounding”, #causalinf literature on positivity, and tons of good stuff in Sociology segregation health literature)

@MaxJordan_N @EpiEllie @AmakaEMD @DrDeidraCrews 7/ In both popns combined, in simple bivariable analyses, variant B will be associated with higher absolute level of risk of outcome than variant A. But if you restricted to Black popn, you might see that effects actually equal.

@MaxJordan_N @EpiEllie @AmakaEMD @DrDeidraCrews 8/ And Black race will be statistically associated with higher risk than White race but not bc of any increased genetic risk, just bc of the racial patterning of other non-genetic risk factors.

@MaxJordan_N @EpiEllie @AmakaEMD @DrDeidraCrews 9/ and each non-genetic factor in example actually increase risk a lot more than variant A or variant B. Absolute risks associated with these factors will be high, but relative risks may vary a lot depending on whether research is in Black & White popns combined, Black or White

@MaxJordan_N @EpiEllie @AmakaEMD @DrDeidraCrews In this example, (1) the variants, (2) race, and (3) all the non-genetic exposures that are driven by racialization are all statistically associated with the outcome but for different reasons. And different analyses will give different estimates of effects.

@MaxJordan_N @EpiEllie @AmakaEMD @DrDeidraCrews 11/ This example just scratches at the surface but I hope starts to illustrate how considering these 3 sets of factors as statistically connected through racialization processes can guide better study design and analyses and make us attentive to sources of bias.