Excluding all confounders in a study like this is statistical error. You cannot do that stuff mathematically with such noisy data.

Similar mistakes happen when someone tries to show meat is cancerous or increases all cause mortality.