This review assessed drug treatments for the prevention of type 2 diabetes. The authors concluded that there was insufficient evidence to draw firm conclusions about the best agent for preventing the development of diabetes, and that further long-term studies are required. Despite limitations of the review, overall, the authors' conclusions are likely to be reliable.

To evaluate pharmacological treatments for the prevention of type 2 diabetes.

The Cochrane Controlled Trials Register, MEDLINE and EMBASE were searched to June 2004. The reference lists of original studies and narrative reviews were screened. Some details of the search strategy were reported. Studies that were only published as abstracts were excluded. No language restrictions were applied.

Study designs of evaluations included in the review

Randomised controlled trials (RCTs) and cohort studies with more than 50 patients were eligible for inclusion. The authors stated that reviews were to be excluded. However, they did include 6 RCTs from one review in the paper. The duration of follow-up, where reported, ranged from 1 to 12 years.

Specific interventions included in the review

Studies of drug treatments were eligible for inclusion. In studies with multiple interventions, only the results of drug intervention arms compared with a placebo or control arm were included. The included studies evaluated:

oral hypoglycaemic agents, i.e. metformin (a biguanide), acarbose, tolbutamide (a sulphonylurea) and troglitazone (a thiazolidinedione);

antiobesity agents, i.e. orlistat;

antihypertensive agents, i.e. thiazide diuretics, angiotensin- converting enzyme (ACE) inhibitors, angiotensin-receptor blockers, verapamil and valsartan;

statins, i.e. pravastatin, simvastatin and atorovastatin;

fibrates, i.e. bezafibrate; and

oestrogen replacement, including combined oestrogen plus progesterone.

The control treatments included placebo, no treatment, atenolol and amlodipine-based treatments.

Participants included in the review

The search was limited to adults (aged more than 18 years). Studies of patients with pre-existing diabetes were excluded. The participants in the primary studies were patients with impaired glucose tolerance, women with gestational diabetes, postmenopausal women, obese individuals, and patients with hypertension, coronary artery disease or dyslipidaemia.

Outcomes assessed in the review

Studies that reported (or provided sufficient data to permit the calculation of) the incidence of type 2 diabetes using an intention-to-treat (ITT) analysis were eligible for inclusion. For some studies, the review also assessed weight change, cardiovascular events and adverse effects.

How were decisions on the relevance of primary studies made?

Two reviewers independently selected studies. Any disagreements were resolved by consensus and inter-rater agreement was assessed using Cohen's kappa coefficient.

The authors did not state that they assessed validity. However, they did comment on the power of some studies, the use of analyses adjusted for potential confounding factors in other studies, and the drop-out rates.

The authors did not state how the validity assessment was performed.

Two reviewers independently extracted the data. Any disagreements were resolved by consensus and inter-rater agreement was assessed using Cohen's kappa coefficient. For each study, the incidence of diabetes was extracted for each treatment arm, together with the relative risk (RR) or hazard ratio (HR) and the associated 95% confidence intervals (CIs).

How were the studies combined?

The studies were grouped by drug treatment and combined in a narrative.

How were differences between studies investigated?

Some differences between the studies were mentioned in the text, while others were apparent from the tables.

The authors stated that 25 articles were included; some articles included more than one study. Oral hypoglycaemic agents were evaluated in 8 RCTs (n=4,946) and 2 cohort studies (n=433). Antiobesity agents were evaluated in 1 RCT plus one pooled analysis of 3 RCTs (n=3,947). Antihypertensive agents were evaluated in 6 RCTs included in an existing systematic review (n not reported) plus 2 subsequent RCTs (n=16,595). Statins were evaluated in 4 post hoc analyses of RCTs (n=35,790). Fibrates were evaluated in 1 post hoc analysis of an RCT (n=303). Oestrogen replacement was evaluated in 1 RCT (n=2,029) and 5 cohort studies (n=24,146).

Inter-rater agreement was 1.0 for study inclusion and 0.91 for the data extraction.

Oral hypoglycaemic agents. Biguanides (5 studies): 1 large RCT reported that metformin significantly reduced the risk of diabetes at a mean 2.8 years compared with placebo (RR 0.69, 95% CI: 0.57, 0.83). The reduction in risk of diabetes remained statistically significant after including a 1- to 2-week period after stopping metformin. The other 4 studies found no significant reduction in the risk of diabetes with biguanides compared with placebo when using ITT analysis. The authors considered that these studies might have been underpowered.

Acarbose (1 RCT plus 1 cohort study): the RCT reported that acarbose significantly reduced the risk of diabetes at 29 months compared with placebo (RR 0.75, 95% CI: 0.63, 0.90), but drop-out rates due to acarbose-associated gastrointestinal toxicity were high (25%). After a 3-month washout period, an analysis of 60% of eligible patients showed the development of diabetes in 15% of acarbose-treated patients and 10.5% of placebo-treated patients.

Sulphonylureas (2 studies): neither study reported a statistically significant decrease in diabetes with tolbutamide compared with placebo, but the authors stated that both studies were small and potentially underpowered. Thiazolidinediones (1 RCT and 1 small cohort study): the RCT reported that troglitazone significantly reduced the risk of diabetes compared with placebo (RR 0.45, 95% CI: 0.25, 0.83), but the drop-out rate was 33%. The cohort study also reported a significant reduction in diabetes with thiazolidinediones.

Antiobesity agents (1 RCT and one pooled analysis of 3 RCTs).

The RCT reported that orlistat significantly reduced the risk of diabetes compared with placebo (RR 0.63, 95% CI: 0.46, 0.86), but the drop-out rate was 57% and 91% of orlistat-treated patients reported gastrointestinal adverse effects. The pooled analysis reported a non statistically significant reduction in the risk of diabetes with orlistat compared with placebo (RR 0.25, 95% CI: 0.05, 1.2); the mean drop-out rate was greater than 30%.

Antihypertensive agents (8 RCTs).

Two RCTs reported no statistically significant effect of thiazides on the risk of diabetes compared with placebo.

Two RCTs reported that ACE inhibitors significantly reduced the risk of diabetes compared with placebo; the RRs were 0.66 (95% CI: 0.51, 0.85) and 0.26 (95% CI: 0.13, 0.53), respectively.

Two RCTs reported no statistically significant effect of angiotensin-receptor blockers on the risk of diabetes compared with placebo.

One RCT reported that a verapamil-based treatment significantly reduced the risk of diabetes compared with an atenolol-based treatment (RR 0.85, 95% CI: 0.77, 0.95).

One RCT reported that a valsartan-based treatment significantly reduced the risk of diabetes compared with an amlodipine-based treatment (RR 0.77, 95% CI: 0.69, 0.86).

Statins (4 RCTs).

One of 4 post hoc analyses of RCTs reported that a statin (pravastatin) significantly reduced the risk of diabetes compared with placebo (RR 0.70, 95% CI: 0.50, 0.99). The other 3 analyses reported no statistically significant difference in the risk of diabetes between statins (pravastatin, simvastatin and atorovastatin) and placebo.

Fibrates (1 RCT).

One post hoc analysis of an RCT reported that bezafibrate significantly reduced the risk of diabetes compared with placebo (HR 0.70, 95% CI: 0.49, 0.99).

Oestrogen replacement (1 post hoc analysis of an RCT and 5 cohort studies).

The post hoc analysis of the RCT reported that oestrogen plus progesterone significantly reduced the risk of diabetes compared with no treatment (RR 0.65, 95% CI: 0.48, 0.89). One cohort study reported a significantly reduced risk of diabetes for current users of oestrogen compared with never users (RR 0.82, 95% CI: 0.7, 0.96), but no significant difference between former users and never users. One cohort study reported a significantly reduced risk of diabetes for oestrogen users compared with nonusers (RR 0.29, 95% CI: 0.15, 0.58). Two cohort studies reported no statistically significant difference in the risk of diabetes for current oestrogen users compared with never or past users (1 study), or for current users compared with never users or for past users compared with never users (1 study).

There was insufficient evidence to draw firm conclusions about the best agent for preventing the development of diabetes. Further long-term studies that use the development of diabetes as the primary outcome are required.

The review addressed a clear question that was defined in terms of the intervention, outcomes and study design; the inclusion criteria for the interventions were broad, which was appropriate given the nature of the review. Three relevant databases were searched and attempts were made to reduce publication and language bias. The authors acknowledged the difficulty of identifying studies that reported the incidence of diabetes as a secondary or post hoc end point. Two reviewers independently selected studies and extracted the data, thus reducing the potential for reviewer bias and errors. Validity was not formally assessed and study quality issues were only mentioned briefly; this means that results from these studies and any synthesis might not be reliable. There were few details on the 6 RCTs included in an existing systematic review. Given the diversity of the studies, a narrative synthesis was appropriate. Overall, although there were limitations to the review, the authors' conclusions about the lack of adequate evidence to draw firm conclusions are likely to be reliable.

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

Research: The authors stated that there is a need for further long-term studies that assess the development of new-onset diabetes as the primary outcome and differentiate prevention of diabetes from delayed development of diabetes. They also stated that further information on adverse effects, compliance with treatment and the cost-effectiveness of drug treatments is required.

This additional published commentary may also be of interest. Kenealy T, Arroll B. Review: Mixed signals from trials concerning pharmacologic prevention of type 2 diabetes mellitus. ACP J Club 2005;143:44.

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.