Yep. The epitome of this is the infamous Finnish researcher who conducts ostensibly reasonable statistical analysis in order to prove his desired contrarian conclusion that it's beneficial to drink up to six pints of regular strength beer every day, and harmful to have days off drinking...
Even if a model actually succeeds in properly balancing the controls so that as well as controlling for basic demographics it also ensures its non coffee drinkers don't drink a commensurately larger portion of other caffeinated drinks or unhealthy alternative drinks or substitute for the caffeine hit with stronger drugs, and don't have a health reason for cutting caffeine or a cultural/social reason linked to other behaviours or diets, you've then got the problem that people with different genes and physical health likely metabolise coffee in very different ways, and to recommend more or less coffee to a particular individual, you've probably got to untangle all that too...
Even if a model actually succeeds in properly balancing the controls so that as well as controlling for basic demographics it also ensures its non coffee drinkers don't drink a commensurately larger portion of other caffeinated drinks or unhealthy alternative drinks or substitute for the caffeine hit with stronger drugs, and don't have a health reason for cutting caffeine or a cultural/social reason linked to other behaviours or diets, you've then got the problem that people with different genes and physical health likely metabolise coffee in very different ways, and to recommend more or less coffee to a particular individual, you've probably got to untangle all that too...