Originally Posted by
Subsistence
Wish I could edit, but to clarify.
Removing studies focusing on specific conditions prevents those conditions from being overrepresented in the study population. But if these conditions are more prevalent in some BMI ranges then others, then the aforementioned removal of studies does nothing to control for this (nor does it intent to). In fact, detecting such things would be part of the purpose of the study. Unfortunately unlike obesity contributing to a heart attack, the relationship is reversed with conditions like cancer contributing to weight loss (opposed to vice versa).