This review concluded that all classes of second-line anti-hyperglycaemic drugs, when added to metformin, achieved clinically meaningful reductions in glycated haemoglobin in patients with type 2 diabetes inadequately controlled with metformin monotherapy. The authors' conclusions reflect the evidence presented and appear likely to be reliable.

To compare safety and efficacy of all available classes of anti-hyperglycaemic therapies in patients with type 2 diabetes who were inadequately controlled with metformin monotherapy.

PubMed, EMBASE, BIOSIS Previews, and the Cochrane Central Register of Controlled Trials (CENTRAL) were searched from 1980 up to October 2009 for studies published in English. Grey literature, conference proceedings and web sites were searched to identify further studies (details were provided in an online appendix, along with the search strategies).

Placebo or active-treatment controlled randomised trials (RCTs) in adults or children with type 2 diabetes were eligible for inclusion. Eligible trials had to compare one or more relevant drugs (see below) either added to metformin (due to inadequate glycaemic control with metformin alone) or replacing metformin (because of intolerance). Definitions of inadequate control were provided. Trials had to be at least four weeks in duration. The following types of drug were eligible: sulfonylureas, meglitinides, thiazolidinediones, dipeptidyl peptidase-4 inhibitors, glucagon-like peptide-1 analogues, insulins and insulin analogues, alpha-glucosidase inhibitors, and weight-loss agents (orlistat and sibutramine). Outcomes of interest included glycated haemoglobin (HbA_1c), hypoglycaemia, body weight, quality of life, long-term complications of diabetes, severe adverse events, and mortality.

In included trials, the mean baseline glycated haemoglobin ranged from 6.6% to 10% and the duration of diabetes ranged from 1.8 to 10.3 years. All trials were in adults (mean age range 49 to 63 years). Treatments, regimens, and baseline metformin doses varied, as did the criteria for defining metformin monotherapy failure and the duration with stable dose at baseline. Around half the trials were placebo-controlled. Treatment durations ranged from one to 66 months (although most used six or 12 months). Most trials were multinational or were conducted in the USA. Detailed trial and patient characteristics were provided in separate online appendices).

Two reviewers independently selected studies for inclusion.

Risk of bias was assessed independently by two reviewers using the SIGN-50 (Scottish Intercollegiate Guidelines Network) instrument. The following criteria were evaluated: focus of question; randomisation; allocation concealment; subject and investigator blinding; baseline similarity of groups; validity and reliability of outcome measures; acceptability of drop-out rates and comparability between groups; use of intention-to-treat analysis; and comparability of site results (for multi-study sites). Studies were then rated as being good or poor.

Two reviewers independently extracted data in order to calculate mean differences or odds ratios.

Bayesian indirect comparison meta-analyses were conducted on drug classes for glycated haemoglobin, body weight, and overall hypoglycaemia; metformin monotherapy was the reference group. Results were reported with 95% credible intervals. The probability of being best and the mean rank of drug classes were calculated. Pair-wise meta-analyses were also performed. All analyses were conducted using random-effects models; fixed-effect models were used in sensitivity analyses.

Tests were also performed to evaluate inconsistency and model convergence. Trial arms with second-line agent doses below World Health Organization Defined Daily Doses were excluded from meta-analyses; the robustness of this approach was tested through an alternate, dose-stratified model of all the evidence. A number of meta-regressions and sensitivity analyses were also performed.

Forty-nine RCTs were included in the review; sample sizes ranged from 13 to 2,789 participants. Twenty-nine out of 49 trials were of poor methodological quality. Inadequate allocation concealment, failure to use an intention-to-treat analysis, lack of blinding, and comparability of site results were frequent limitations.

All classes of second-line agents added to metformin compared with metformin alone achieved clinically-meaningful reductions in glycated haemoglobin (estimates ranged from -0.65% to -0.96%; 40 RCTs; 17,795 participants); there were no significant differences between drug classes. Thirty-four trials (16,704 participants) reported hypoglycaemia. Insulins and insulin secretagogues were associated with significantly more events of overall hypoglycaemia than the other classes.

A statistically significant increase in body weight (reported as an outcome in 30 trials, 15,265 participants) was observed with most drug classes (increases ranged from 1.8 to 2.96kg), although there was no significant weight gain with dipeptidyl peptidase-4 inhibitors, alpha-glucosidase inhibitors and basal insulin. Glucagon-like peptide-1 analogues resulted in weight loss. There were no significant differences between treatments for long-terms complications (although limited data were available). There was also little data available for mortality and quality of life. Severe adverse events, although reported in 23 studies, were rarely defined.

Meta-regression and sensitivity analyses generally produced similar results to the main analyses.

When added to metformin, all classes of second-line anti-hyperglycaemic drugs achieved clinically meaningful reductions in glycated haemoglobin in patients with type 2 diabetes inadequately controlled with metformin monotherapy. Severe hypoglycaemia events were rare for all agents. A modest increase in body weight was found with most second-line therapies, except for dipeptidyl peptidase-4 inhibitors, alpha-glucosidase inhibitors and glucagon-like peptide-1 analogues.

The review addressed a clear question and was supported by reproducible eligibility criteria. Attempts to identify relevant trials were undertaken by searching relevant databases. Various methods were used identify unpublished studies. The restriction to including studies in English meant that some relevant studies may have been missed, and the possibility of language bias could not be ruled out. Suitable methods were employed to reduce the risks of reviewer error and bias throughout the review.

The quality of included trials was assessed, with the results used in sensitivity analyses. Sufficient trial details were provided. Appropriate methods were used to pool data. However, results were not reported on optimal treatment probabilities, mean ranks, and investigations into inconsistency.

The authors' conclusions reflect the evidence presented and appear likely to be reliable.

Practice: The authors did not state any implications for practice.

Research: The authors stated that further research was required to determine whether thiazolidinediones, glucagon-like peptide-1 analogues and dipeptidyl peptidase-4 inhibitors provide greater benefits in patient groups at higher risk of severe hypoglycaemia or its consequences. The authors added that further long-term studies were needed to explore differences in glycaemic durability between agents over time, especially for the newer, more expensive oral anti-diabetes drugs. Studies should also assess long-term complications and mortality.

Canadian Agency for Drugs and Technologies in Health (CADTH) funded by Canadian federal, provincial and territorial governments.

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.