To conduct a meta-analysis of the efficacy of metformin in the treatment of non-insulin dependent diabetes mellitus, by comparing metformin with placebo and sulfonylurea.

The Current List of Medical Literature was searched from 1957 to 1959, the Cumulated Index Medicus was searched from 1959 to 1965, and MEDLINE and EMBASE were searched from inception to January 1996. The search terms used were 'metformin', 'phenformin' and 'humans'. Reference lists of all relevant articles were scanned.

Study designs of evaluations included in the review

Included studies had to be full publications of randomised controlled trials (RCTs). RCTs could be blind or open, parallel or crossover.

Specific interventions included in the review

The included studies compared metformin (in doses from 500 to 3000mg) with sulfonylurea or placebo. The types of sulfonylurea included were glibenclamide (in doses from 1.75 to 20mg), gliclazide (from 80 to 240mg), tolbutamide (up to 3000mg) and glipizide (up to 30mg).

Participants included in the review

Non-insulin dependent diabetes mellitus (NIDDM) was used as an inclusion criterion. Participants could be obese or non-obese, but it is unclear how many participants were included in the meta-analysis or what their characteristics were.

Outcomes assessed in the review

The outcome measures included in the review were fasting blood glucose (mmol/l), glycosylated haemoglobin (%) and body weight (kg).

How were decisions on the relevance of primary studies made?

The author does not state how the papers were selected for the review, or how many of the reviewers performed the selection.

The author does not state that they assessed validity.

The author does not state how the data were extracted from the review, or how many of the reviewers performed the data extraction. Data were extracted on: study design, study duration, metformin dose, sulfonylurea dose, type of sulfonylurea, patient status prior to randomisation and results.

How were the studies combined?

Weighted mean differences and 95% confidence intervals (CI) for continuous data were calculated for changes in outcomes (fasting blood glucose, glycosylated haemoglobin and body weight) compared to baseline (see Other Publications of Related Interest no.1). A fixed-effect model was used, but in the case of heterogeneity, a random-effects model was also used.

How were differences between studies investigated?

Heterogeneity of the included studies was discussed in terms of study design, metformin dose, duration of treatment and the type and dose of sulfonylurea used. A sensitivity analysis including only studies comparing metformin with glibenclamide was performed.

Nineteen RCTs were included in the meta-analysis. Nine studies compared metformin with placebo and ten studies compared metformin with sulfonylurea. Ten of the studies were double-blind, one was blind and eight were open. Thirteen were parallel studies and six were crossover studies.

The weighted mean difference (WMD) between metformin and placebo after treatment for fasting blood glucose was (-2.0 mmol/l (95% CI: -2.4, -1.7) and for glycosylated haemoglobin (fixed-effect: -0.9% (95% CI: -1.1, -0.7), random effects: -1.0% (95% CI: -1.7, -0.4)). Body weight WMD was not significant after treatment (fixed-effect: 0.8kg (95% CI: -1.0, 2.5), random effects: 1.0kg (95% CI: -1.8, 3.7)). Sulfonylurea and metformin lowered glycosylated haemoglobin (random effects: 0.1% (95% CI: -0.2, 0.4)) and blood glucose equally, while there was a significant WMD in body weight (-2.9 kg (95% CI: -4.4, -1.1)) because of a 1.7 kg mean increase after sulfonylurea and a 1.2 kg decrease after metformin.

Metformin lowers blood glucose and glycosylated hemoglobin significantly, compared with placebo. Metformin and sulfonylurea have an equal effect on fasting blood glucose and glycosylated hemoglobin, but the body weight is significantly lower after metformin, compared with sulfonylurea treatment because of an increase in body weight after sulfonylurea treatment.

The review question was clearly stated and reasonably well supported by the inclusion criteria, although it is not clear how many reviewers were involved in applying these criteria. A number of databases were searched using appropriate key terms, but it is unclear whether there were language restrictions or whether an attempt was made to identify unpublished literature. Although the review was limited to RCTs and the details of individual studies were given, these are not discussed in terms of the validity of included studies. Pooling of studies seemed appropriate and heterogeneity of studies was considered, although this was not adequately addressed. The author's conclusions seem to follow from the results of the meta-analysis but, given the limitations mentioned above, these should be treated with caution.

Practice: The author states that the blood glucose-lowering effect of metformin is equal to sulfonylurea. Body weight, however, is significantly lower after metformin than after sulfonylurea treatment. Because obesity is a major problem in NIDDM, metformin treatment in this respect has an advantage over sulfonylureas.

Research: The author does not state any implications for research.

1. Bracken MB. Statistical methods for analysis of effects of treatment in overviews of randomized trials. In Effective Care of the Newborn Infant. Sinclair JC, Bracken MB, Eds., Oxford, Oxford University Press, 1992, p.13-8.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.