Observational studies are useless for proving anything, but they do provide interesting starting points for speculation. They can also give some interesting insights in to human behaviour. I particularly enjoyed the American Paradox study.
On the raw data a high intake of saturated fat didn't quite reverse the women's arteriosclerosis, but after modeling it did. Modeling appears to involve adjusting your results to estimate what the values would have been if your patients didn't smoke, didn't have diabetes, didn't eat cholesterol (gasp) etc.
Glancing through the patient characteristics in table 1 you can see why adjustment might be needed.
Look at smoking. The high saturated fat group had the highest percentage of smokers and a hugely higher pack-years value (34 pack-years) than any other group (all below 18). They boozed the most, ate most trans fats and pooped the least fibre. They may even, wait for it, have eaten an egg a day! On the plus side they obviously had the highest HDL cholesterol and lowest triglycerides, because they also ate the least carbohydrate.
If you had to describe this group in one word it would have to be "naughty". These naughty people did all the WRONG things (according to the AHA) and, after adjustments, began the reversal process of their IHD. Thumbing your nose at the AHA, even in this mild manner (by my standards), looks to be good for you.
The lowest saturated fat group had the highest carbohydrate intake, let's call them the Goody Goodies. They made excellent progress as coronary bypass fodder.
Just observational. But what an observation of disobedience!