The effectiveness, safety and epidemiology of the use of acarbose in the treatment of patients with type II diabetes mellitus. A model of medicine-based evidence.

Objective: To assess the efficacy, safety and extent of perceived indications of acarbose, a new antidiabetic agent, under routine clinical practice conditions in an unselected Northern Italian population of type II diabetic patients.

Methods: The study population was assigned to three different groups according to the physician's clinical judgement: group A (acarbose considered as an elective treatment); group B (acarbose considered to be of uncertain benefit): group C (acarbose deemed not to be appropriate). Group B patients were randomized either to continue their standard treatment or to add acarbose to it. Patients with type II diabetes mellitus were recruited from 17 diabetes outpatient clinics from one Italian region (Lombardy). A total of 1027 patients were recruited (group A: 283; group C: 494; group B: 250, of whom 124 were randomly assigned to standard treatment + acarbose and 126 to standard treatment alone). Acarbose was administered for 1 year at a median dose of 100 mg 3 times daily. Drug efficacy was evaluated in terms of mean HbAlc, pre- and post-prandial glycaemic values. Additional endpoints were the proportion of patients with HbA1c levels below 8% at the end of the study period and the proportion of subjects who needed a modification in the standard treatment. The safety and tolerability profiles of the drug were also investigated. Data on HbA1c, fasting and post-prandial blood glucose levels were analysed over time using repeated-measures analysis [Generalized Estimating Equation (GEE) models].

Results: The analysis of Group B showed that, after treatment for 1 year, the mean reduction in HbA1c levels in the acarbose group with respect to the control group was 0.30% (95% confidence limits -0.60 +0.02; P = 0.07), while the mean reduction in post-prandial glycaemia was 17 mg-dl(-1) (95% c.l. -33.5 -0.8; P = 0.04). No difference resulted for fasting blood glucose levels. When looking at the baseline HbA1c levels, it emerged that the mean benefit associated with the use of acarbose was 0.14% (95% c.l. -0.6 +0.28; P = 0.5) in patients with HbAlc levels below 8%, 0.28% (95% c.l. -0.6 +0.05; P = 0.09) in those with values between 8% and 9.9% and 0.65% (95% c.l. -1.36 +0.06; P = 0.07) in those with values > or =10%. Only patients treated with diet+/-oral anti-diabetic agents (OAA) benefited from acarbose treatment (mean benefit = 0.37%, 95% c.l. -0.65 -0.08), while no effect was shown for insulin-treated subjects. The proportion of patients with HbA1c below 8% increased from 31% to 44% in the acarbose group and from 40% to 45% in the control group (absolute difference between baseline and end-of-study values = 8.0% in favour of acarbose-treated patients; P = 0.058). Patients treated with acarbose were significantly more likely to undergo a dose reduction in concomitant diabetic treatments compared with the control group; they were also less likely to require an increase in the dose of standard treatment and to start insulin during the study period. One third of the patients could not assume the drug for the whole study period, mainly due to gastrointestinal side-effects.

Conclusions: The design adopted in this study allowed an integrated evaluation of the overall effectiveness of acarbose in clinical practice. The benefits of the drug in an unselected population of non-insulin-dependent diabetes mellitus (NIDDM) patients are significant but of marginal clinical relevance. Only a better definition of the subgroups of patients who are more likely to benefit from long-term treatment, particularly through possible postponement of secondary OAA failure, will allow a reliable definition of the cost-effectiveness of this complementary component of anti-diabetic strategy.