tion is diagrammed in Fig. 1. Two patients were excluded because
of outlier problem. Most of eligible subjects (93 patients, 73.2%)
were female. Mean ages and body mass index (BMI) were 51.8 ± 9.3
years and 25.1 ± 3.4 kg/m2, respectively. Co-morbid diseases were
hypertension, dyslipidemia, asthma and thyroid diseases. All three
patients who had thyroid disease were in metformin group. Two of
them were well controlled withoutmedication and one of them was
well controlled by thyroxine. Baseline characteristics of patients in
each treatment group were similar (Tables 1 and 2). The rate of
treatment compliance was the same in all four groups (99%).