Evaluation of the association between glucose variability,as measured by the difference between maximum and minimum, intermittent, random blood glucose observations, and long term clinical outcome in subjects with type 2 diabetes treated with insulin