Efficacy of combined treatments in NIDDM patients with secondary failure to sulphonylureas. Is it predictable?

The treatment of NIDDM patients with secondary failure to sulphonylurea is a common problem. We performed a crossover study in 50 NIDDM patients with secondary failure to glibenclamide by comparing the addition to sulphonylurea of either a low-dose bedtime NPH insulin or a t.i.d. oral metformin and by analyzing treatment efficacy in relation to patient and disease characteristics. Both combined therapies clearly improved glycaemic control. HbA1 c were similarly reduced by the addition of either bedtime NPH insulin (7.6+/-0.34 vs 8.7+/-0.35, p<0.01) or metformin (7.6+/-0.22 vs 8.6+/-0.31, p<0.01). Also fasting plasma glucose (FPG) and post-prandial plasma glucose (PPPG) significantly decreased (p<0.01) with both treatments. Bed-time NPH insulin was more effective on FPG reduction than metformin (-36+/-2% vs -25+/-2%, p<0.01); in contrast, metformin addition was more effective on PPPG reduction than bedtime NPH insulin addition (-30+/-2% vs 20+/-3%, p<0.01). Serum cholesterol was marginally but significantly decreased after metformin (5.49+/-0.19 vs 5.91 +/-0.18 mM, p<0.05) but not after NPH insulin. Body weight increase was significantly greater after insulin addition than after metformin (1.47+/-0.25 Kg vs 0.64+/-0.17 p=0.02). All patients preferred the addition of metformin rather than NPH insulin. None of the measured clinical and metabolic variables (before treatment FPG and PPPG, HbA1 c, post-glucagon C-peptide levels, insulin sensitivity, patient age, BMI and diabetes duration) significantly correlated to the efficacy of the two combined treatments studied. In conclusion, in NIDDM patients with secondary failure to sulphonylureas the addition of either low-dose bedtime NPH insulin or t.i.d. metformin is similarly effective in improving glycaemic control. Metformin is better accepted by patients and provides a modest advantage in terms of body weight and cholesterol levels. The most common clinical and metabolic variables are not useful for predicting the efficacy of these two combined treatments.