When you do your study without accounting for treatments that are given to people with high risk of disease and which reduce cholesterol, you will find that lower cholesterol corresponds with disease due to those treatments. In other words, they have performed a large study that is likely to be confounded by indication.