Meglitinide analogues for type 2 diabetes mellitus

Abstract

Background

In type 2 diabetes mellitus, impairment of insulin secretion is an important component of the disease. Meglitinide analogues are a class of oral hypoglycaemic agents that increase insulin secretion, in particular, during the early phase of insulin release.

Objectives

The aim of this review was to assess the effects of meglitinide analogues in patients with type 2 diabetes mellitus.

Search methods

We searched several databases including The Cochrane Library, MEDLINE and EMBASE. We also contacted manufacturers and searched ongoing trials databases, and the American Diabetes Association (ADA) and European Association for the Study of Diabetes (EASD) web sites.

Selection criteria

We included randomised controlled, parallel or cross‐over trials comparing at least 10 weeks of treatment with meglitinide analogues to placebo, head‐to‐head, metformin or in combination with insulin.

Data collection and analysis

Two authors independently extracted data and assessed trial quality.

Main results

Fifteen trials involving 3781 participants were included. No studies reported the effect of meglitinides on mortality or morbidity. In the eleven studies comparing meglitinides to placebo, both repaglinide and nateglinide resulted in a reductions in glycosylated haemoglobin (0.1% to 2.1% reduction in HbA1c for repaglinide; 0.2% to 0.6% for nateglinide). Only two trials compared repaglinide to nateglinide (342 participants), with greater reduction in glycosylated haemoglobin in those receiving repaglinide. Repaglinide (248 participants in three trials) had a similar degree of effect in reducing glycosylated haemoglobin as metformin. Nateglinide had a similar or slightly less marked effect on glycosylated haemoglobin than metformin (one study, 355 participants). Weight gain was generally greater in those treated with meglitinides compared with metformin (up to three kg in three months). Diarrhoea occurred less frequently and hypoglycaemia occurred more frequently but rarely severely enough as to require assistance.

Authors' conclusions

Meglitinides may offer an alternative oral hypoglycaemic agent of similar potency to metformin, and may be indicated where side effects of metformin are intolerable or where metformin is contraindicated. However, there is no evidence available to indicate what effect meglitinides will have on important long‐term outcomes, particularly mortality.

Plain language summary

Meglitinide analogues for type 2 diabetes mellitus

In type 2 diabetes mellitus, impairment of insulin secretion is an important component of the disease. The meglitinide analogues ("meglitinides") are a class of oral antidiabetic agents that increase insulin secretion in the pancreas. The properties of this class of drug suggest that they have the potential to produce a rapid, short‐lived insulin output. Two analogues are currently available for clinical use: repaglinide and nateglinide. Altogether 15 studies were included in this review, eight of the studies used repaglinide as the study drug and five used nateglinide. Two compared the two meglitinides to one another. In total, 3781 people participated in the 15 studies which mainly lasted between 10 and 24 weeks (one study had a duration of 52 weeks). 
 No studies reported the effect of meglitinides on mortality or diabetes related complications. In the eleven studies comparing meglitinides to placebo, both repaglinide and nateglinide resulted in an improved blood sugar control. Weight gain was generally greater in those treated with meglitinides compared with metformin (up to three kg in three months), another oral antidiabetic drug. Here, diarrhoea occurred less frequently and hypoglycaemia occurred more frequently but rarely severely enough as to require assistance. 
 Meglitinides may offer an alternative oral hypoglycaemic agent of similar potency to metformin, and may be indicated where side effects of metformin are intolerable (in particular persistent diarrhoea) or where metformin is contraindicated. However, there is no evidence available yet to indicate what effect meglitinides will have on important long‐term outcomes, in particular, on mortality. As yet, the experience with meglitinides in terms of side effects is limited. Results from other Cochrane review groups may provide additional information about the potential role of meglitinides in the management of type 2 diabetes mellitus.

Background

Description of the condition

Diabetes mellitus is a metabolic disorder resulting from a defect in insulin secretion, insulin action, or both. A consequence of this is chronic hyperglycaemia (that is elevated levels of plasma glucose) with 
 disturbances of carbohydrate, fat and protein metabolism. Long‐term complications of diabetes mellitus include retinopathy, nephropathy and neuropathy. The risk of cardiovascular disease is increased. For a detailed overview of diabetes mellitus, please see under 'Additional information' in the information on the Metabolic and Endocrine Disorders Group in The Cochrane Library (see 'About the Cochrane Collaboration', 'Collaborative Review Groups ‐ CRGs'). For an explanation of methodological terms, see the main Glossary in The Cochrane Library.

In type 2 diabetes mellitus, compromise of insulin secreting ß‐islet cell function in the pancreas is an important component of the disease, along with increased insulin resistance (De Fronzo 1999). The UK Prospective Diabetes Study demonstrated that continued loss of ß cell function determines the progressive deterioration in blood glucose control (UKPDS 1995). The loss of ability to generate a rapid, early phase, release of insulin from ß cells, in response to a glucose load (that is a meal), is recognised among type 2 diabetes patients and among those at high risk of developing the disease (that is first degree relatives of type 2 diabetes patients, women with a history of gestational diabetes) (Dornhurst 2001). There is evidence to support the importance of the early phase of insulin release, normally occurring within minutes after a glucose load, in the control of postprandial blood glucose (Gerich 2003). It has been suggested that there may be a specific benefit in lowering post prandial glucose (Bastyr 2000; Breuer 2000). The evidence for this comes from epidemiological studies, which were used for decisions on classification of diabetes, such as DECODE (DECODE 1999). These studies showed that the plasma glucose level, after a glucose load of 75 g, was a better predictor of heart disease than fasting glucose levels. However, the studies mainly involved oral glucose tolerance tests (OGTT) in people without diabetes mellitus (de Vegt 1999). During an OGTT, glucose levels are measured at intervals after the consumption of an oral glucose load. The DECODE study reported that glucose levels, measured at two hours after an oral glucose load, were independently related to mortality. Even if fasting plasma glucose was normal (under 6.1 mmol/L), rising glucose levels at two hours were a predictor of an increase in ischaemic heart disease. The DECODE study also noted that 31% of people with abnormal two hour OGTT glucose levels that were diagnostic of diabetes mellitus had normal fasting plasma glucose levels. 
 Bonora and colleagues (Bonora 2001) showed that, in people with type 2 diabetes, elevated glucose levels after meals were seen in patients whose glycosylated haemoglobin was normal and that glycosylated haemoglobin was more closely related to preprandial than postprandial glucose levels. This is presumed to be an effect of duration, in that preprandial levels correlate more closely with mean 24 hour glucose levels. Conversely, Bastyr and colleagues (Bastyr 2000) found that a strategy of focusing on lowering postprandial glucose (using insulin lispro) had more effect on HbA1c than reduction of fasting plasma glucose using bedtime NPH insulin. In a recent review of the topic, Gerich and colleagues (Gerich 2003) concluded that postprandial glucose was a greater risk factor for cardiovascular disease than fasting plasma glucose. Hanefeld 1996 found that the incidence of myocardial infarction, in an 11‐year follow‐up study in patients with type 2 diabetes mellitus, was associated with increased postprandial glucose but not fasting plasma glucose. 
 Increased two‐hour OGTT levels may reflect insulin resistance, whereas high fasting plasma glucose reflects beta‐cell dysfunction (Yudkin 1994). Sulphonylureas mainly improve late insulin release, whereas the meglitinides affect early insulin release. Metformin works by reducing hepatic over‐production of glucose, and the glitazones by increasing peripheral glucose utilisation. Gerich 2003 argued that the sulphonylureas, glitazones, and metformin will act mainly to reduce fasting and other non‐prandial levels, and that the most effective drugs for reducing postprandial glucose are short‐acting insulins and the meglitinides. The meglitinides may be of particular value in patients with insulin resistance (Shiba 2003). 
 However, for reduction of postprandial glucose to be accepted as a therapeutic goal in addition to glycosylated haemoglobin, it is not enough to show a correlation between postprandial glucose and ischaemic heart disease. Evidence that reducing postprandial glucose (independently of glycosylated haemoglobin) will reduce ischaemic heart disease events is also needed. Postprandial glucose may reflect a more general metabolic problem, and may be a correlate, not a cause, of ischaemic heart disease.

Description of the intervention

The meglitinide analogues are a class of oral hypoglycaemic agents that increase insulin secretion by binding to specific sites on the sulphonylurea receptor found on insulin secreting ß islet cells in the pancreas. The pharmacokinetic and pharmacodynamic properties of this class of drug suggest they have the potential to produce a rapid, short‐lived insulin secretary response (Landgraf 2000). These properties, along with clinical trial evidence, supports the potential role of meglitinide analogues in augmenting the early phase insulin release and, therefore, postprandial glucose control (Dornhurst 2001). Two analogues are currently available for clinical use: repaglinide and nateglinide. While the chemical structures and mechanisms differ between these agents, the effect on early phase insulin release is similar, with a rapid rise in insulin concentrations after dosing and a short half‐life. Early trial evidence supports their effect in reduction of postprandial glucose and reduction in hypoglycaemic episodes (Landgraf 2000).

Objectives

To assess the effects of meglitinide analogues in patients with type 2 diabetes mellitus compared to placebo, metformin and head to head.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled clinical trials were considered eligible for this review. We also planned to include cross‐over trial designs, where each treatment period exceeded 10 weeks and a "wash‐out" period of four weeks was allowed for between treatments, but none were found. 
 The minimum accepted duration was 10 weeks of treatment, based on the time taken for glycosylated haemoglobin to reliably reflect changes in glycaemic control.

Types of participants

Patients with type 2 diabetes mellitus.

Types of interventions

The potential therapeutic combinations for meglitinide analogues are outlined below. As a priority, we focused on the following comparisons, based on their perceived clinical relevance. Other comparisons are the subject of reviews being undertaken by other review teams within the Cochrane Metabolic and Endocrine Disorder Review Group. Their results, when available, may be quoted and referenced in this review for completeness, in order to give readers a comprehensive picture of the meglitinides.

• meglitinide analogues (repaglinide, nateglinide) versus placebo;

• repaglinide versus nateglinide;

• metformin versus meglitinides.

Within each comparison, we considered the evidence for monotherapy, combination therapy with oral agents included in this review (metformin), and combination therapy with insulin.

Types of outcome measures

Primary outcomes

The following outcome measures were sought:

• mortality: diabetes related deaths (including death from vascular disease, renal disease, hyper‐ and hypoglycaemia or sudden death), and total mortality;

• diabetes related complications: non‐fatal myocardial infarction, angina, heart failure, stroke, renal failure, amputation (of at least one digit), vitreous haemorrhage, retinal photocoagulation, blindness in one eye, or cataract extraction.

Secondary outcomes

Preliminary searches indicated that it was likely that no studies of sufficient duration to measure long‐term mortality or complication rates would be available. We therefore also assess the following outcome measures:

• glycaemic control measured by glycosylated haemoglobin ( a clinically significant reduction in glycosylated haemoglobin was taken, a priori, to be a fall in HbA1c of 0.5%);

• weight or body‐mass index (BMI) change;

• hypoglycaemia, graded as mild (symptoms easily controlled by the person); moderate (normal activities interrupted but help not required); severe (ideally, the Diabetes Control and Complication Trial definition ‐ patient unable to deal with hypoglycaemia without assistance, and associated with a blood glucose level under 50 mg/dl or 4 mmol/L, or with prompt recovery after oral carbohydrate or glucagon or intravenous glucose) (DCCTRG 1993);

• fasting blood glucose and postprandial glucose (only if no glycosylated haemoglobin levels were reported);

• plasma lipids (triglycerides, total cholesterol, HDL‐ and LDL‐cholesterol);

• adverse drug effects (short and long‐term);

• quality of life (ideally using a validated instrument) and patient satisfaction.

Search methods for identification of studies

Electronic searches

The following databases were searched:

• The Cochrane Library (2006, Issue 3);

• MEDLINE (1993 until October 2006);

• EMBASE (1993 until October 2006);

• Science Citation Index (1981 until October 2006);

• ISI Proceedings (1990 until October 2006).

Ongoing trials database:

• www.controlled‐trials.com

For detailed search strategies please see under Appendix 1. 
 There were no language restrictions on searching.

Searching other resources

• handsearches: reference lists of the relevant trials and reviews identified were searched for additional studies;

• information regarding unpublished trials was sought from the pharmaceutical companies, Novo Nordisk and Novartis, that produce repaglinide and nateglinide respectively. We did not contact authors directly for additional information;

• the web sites of the American Diabetes Association (ADA) (http://www.diabetes.org/) and the European Association for the Study of Diabetes (EASD) (http://www.easd.org/) were searched for recent meeting abstracts.

Data collection and analysis

Selection of studies

Two authors (PR, CB) independently reviewed the titles, abstracts and key words of all the records retrieved in the search. Full articles were retrieved for further assessment if the information given suggested that the study 1. included patients with type 2 diabetes mellitus, 2. compared meglitinide analogues with one of the defined therapy comparisons, 3. assessed one or more of the defined primary outcome measures and 4. met the defined criteria for trial design. Full articles were retrieved for clarification when there was doubt about eligibility. Had differences of opinion existed, they would have been resolved by discussion. However, there was complete agreement between the authors on the studies for inclusion (Cohen's kappa =1, Cohen 1960).

Data extraction and management

Two authors, using a predefined data extraction form, undertook the data extraction independently. Again, differences were resolved by discussion, with reference back to the original paper. It was not necessary to seek a third opinion.

Extraction included the following information:

• general information ‐ author and year, country, setting, published/unpublished, language of publication, source of funding, abstract / full article, duplicate publications;

• trial characteristics ‐ design (parallel or cross‐over); quality by method and security of randomisation, duration, blinding, allocation concealment;

• participants ‐ age, gender, race, selection method, exclusions, duration of diabetes, initial glycosylated haemoglobin, body mass index (BMI), baseline similarity of groups (including co‐morbidity), assessment of compliance;

• interventions ‐ type of meglitinide, dose of meglitinide, combination therapy with meglitinide, comparator regimen, placebos;

• results ‐ comparability at baseline, losses/drop‐outs, mortality and morbidity outcomes, glycosylated haemoglobin, postprandial glucose, hypoglycaemia, adverse effects, weight change, patient preference, quality of life, study duration, as observed and by intention‐to‐treat.

Assessment of risk of bias in included studies

The quality of reporting of each trial was assessed, based largely on the quality criteria specified by Schulz and Jadad (Jadad 1996; Schulz 1995), and as described in the manual of the Centre for Reviews and Dissemination (CRD) for randomised controlled trials (Khan 2000).

In particular the following factors were considered: 
 (1) Minimisation of selection bias 
 (a) was the randomisation procedure adequate? 
 (b) was the allocation concealment adequate?

(2) Minimisation of attrition bias 
 (a) were withdrawals and dropouts completely described? 
 (b) was analysis by intention‐to‐treat?

(3) Minimisation of detection bias 
 (a) were outcome assessors blind to the intervention?

Based on these criteria, studies were broadly subdivided into the following three categories (see Cochrane Handbook for Systematic Reviews of Interventions ‐ Higgins 2005):

A ‐ all quality criteria met: low risk of bias; 
 B ‐ one or more of the quality criteria only partly met: moderate risk of bias; 
 C ‐ one or more criteria not met: high risk of bias.

Two authors assessed each trial independently. Differences were resolved by discussion, with reference back to the original paper. Had it been necessary, a third opinion would have been sought to resolve any remaining differences. There were no differences that could not be resolved by discussion and reference to the original paper (Cohen's Kappa score = 1).

Measures of treatment effect

Data were summarised statistically where available, and sufficiently similar. We report both dichotomous and continuous data.

Dichotomous data

Dichotomous data were expressed as relative risks (RR).

Continuous data

Continuous data were expressed as weighted mean differences (WMD) and an overall WMD was calculated. Overall results were calculated based on the fixed effect model.

Assessment of heterogeneity

Heterogeneity was tested for using the Z score and the χ² statistic with significance being set at α < 0.1. The I² statistic was used to estimate the proportion of the total variation in study estimates that could be explained by heterogeneity (Higgins 2002; Higgins 2003). Where significant heterogeneity was found, it was considered to be unreasonable to assume that there was one 'true' effect underlying the data, that was constant across different populations, and therefore no summary statistic was calculated.

There were insufficient studies reporting data in a way that could be summarised to enable meaningful exploration of causes of heterogeneity. In the future, if more studies are published, we will assess possible sources of heterogeneity by subgroup and sensitivity analyses as described below. Had there been sufficient studies reporting data consistently, a funnel plot would have been undertaken to assess the effect of small studies on the estimates.

Subgroup analysis and investigation of heterogeneity

The following subgroup analyses were planned for the main outcome measures where significant effects had been found:

• different types of meglitinide analogues (repaglinide, nateglinide);

• dose;

• baseline glycosylated haemoglobin levels (low, medium, high);

• age;

• gender;

• race;

• weight, body mass index.

Sensitivity analysis

We planned to do sensitivity analyses in order to explore the influence of the following factors on effect size:

• repeating the analysis excluding studies published in abstract form only;

• repeating the analysis taking into account the study quality, as specified;

• repeating the analysis excluding any very long or large studies to establish how much they dominate the results;

• repeating the analysis excluding studies using the following filters: language of publication, source of funding (industry versus other), and country.

Results

Description of studies

Results of the search

The initial searches of the electronic databases, using the search strategies outlined above, yielded 596 references. After removing database duplicates and comparisons not being undertaken by this review, the remaining titles and abstracts were reviewed by two researchers (CB/PR) independently, with a resulting 78 full journal articles and 169 abstracts for potential inclusion. Following review of the full articles a final total of 15 studies were included and 41 excluded. Six studies, only published in abstract form, are awaiting the response of the authors to a request for further details (under 'Studies awaiting assessment'). Wherever a study initially published in abstract form was subsequently superseded by a full journal article, the latter was used in preference.

Novartis and Novo Nordisk were contacted for additional information regarding unpublished data and provided a list of publications, all of which had been identified by our search.

Interrater agreement

There was no discordance between the two researchers in terms of studies that should be included for review (Cohen's kappa =1.0).

Included studies

A summary of the characteristics of the included studies is given in the Table Characteristics of included studies.

Study design

All 15 of the included studies were parallel, randomised controlled trials. Eleven studies were double blind, with neither the patients nor the clinician knowing the treatment allocation (Bengel 2005; Chuang 1999; Goldberg 1998; Hanefeld 2000; Horton 2000; Jovanovic 2000; Marre 2002; Moses 1999; Moses 2001; Saloranta 2002; Van Gaal 2001). Four were open label studies ( Derosa 2003; Furlong 2002; Raskin 2003; Rosenstock 2004), one of which included participants who received insulin ( Furlong 2002). Three studies (Chuang 1999; Goldberg 1998; Jovanovic 2000) had a run‐in period of two weeks, two a run‐in period of 6 to 8 weeks (Bengel 2005; Derosa 2003) and one a run in period of three months (Rosenstock 2004), prior to randomisation and beginning the study intervention to enable wash out of other oral hypoglycaemic agents. Two studies did not include a run‐in period but, as all participants were sulphonylurea naive, the lack of run‐in was unlikely to be a major source of error (Moses 2001; Van Gaal 2001). The remaining seven studies had a run in of four weeks. Derosa 2003 reported the longest follow up at 60 weeks. Five of the trials ( Chuang 1999; Horton 2000; Jovanovic 2000; Marre 2002; Saloranta 2002) reported follow up of 24 weeks and the remaining nine a period between 10 and 20 weeks (Bengel 2005; Furlong 2002; Goldberg 1998; Hanefeld 2000; Moses 1999; Moses 2001; Raskin 2003; Rosenstock 2004; Van Gaal 2001).

Interventions*

Appendix 3 summarises which comparisons of treatment interventions were undertaken in each study. Eleven studies assessed meglitinides versus placebo (Bengel 2005; Chuang 1999; Goldberg 1998; Hanefeld 2000; Horton 2000; Jovanovic 2000; Marre 2002; Moses 1999; Moses 2001; Saloranta 2002; Van Gaal 2001). Three of these were in the setting of combination therapy with metformin (Horton 2000; Marre 2002; Moses 1999), while the rest used monotherapy. Four studies compared meglitinides to metformin; three as monotherapy (Derosa 2003; Horton 2000; Moses 1999) and one in combination with insulin (Furlong 2002). 
 Eight of the studies used repaglinide as the study drug and five used nateglinide. Two compared the two meglitinides to one another (Raskin 2003; Rosenstock 2004). The treatment regimen and dose varied (detailed in table Characteristics of included studies).

(*two studies included more than two comparator arms in their design and therefore contribute to greater than one comparison as described above (Moses 1999; Horton 2000))

Participants

In total, 3781 people participated in the 15 studies included in this review. Appendix 3 outlines how many people received the various combinations of therapy being considered in this review. Details of race, sex, age and body mass index are included in the table Characteristics of included studies. All participants had received some sort of treatment prior to the study but not all had received oral therapy, that is had been on diet only regimens (Moses 2001; Van Gaal 2001; Derosa 2003). Eight of the studies included a mixture of pharmacologically naive and previously treated participants (Bengel 2005; Chuang 1999; Goldberg 1998; Jovanovic 2000; Hanefeld 2000; Horton 2000; Rosenstock 2004; Saloranta 2002). Only Goldberg 1998 reported results for subgroups (and here reporting was incomplete) with the other five studies not separating treatment naive participants from the other patients in the results. The remaining four studies only included people who had previously received pharmacological treatment for their diabetes mellitus (Moses 1999; Raskin 2003; Furlong 2002; Marre 2002).

Outcome measures

None of the identified studies reported outcomes related to mortality or long‐term complications of diabetes. 
 Appendix 4 and Appendix 5 summarise the outcomes glycosylated haemoglobin and weight: Weight change was reported in nine of the studies (Chuang 1999; Derosa 2003; Furlong 2002; Goldberg 1998; Marre 2002; Moses 1999; Raskin 2003; Rosenstock 2004; Van Gaal 2001). Adverse events and hypoglycaemia were not reported consistently throughout the studies but were reported to some extent in all but two of the studies (Bengel 2005; Derosa 2003). 
 Data about the effect of treatment on lipids were presented for four of the studies (Bengel 2005; Derosa 2003; Goldberg 1998; Marre 2002). Quality of life was reported by two trials (Furlong 2002; Moses 2001). Both used the validated World Health Organisation questionnaires (Well‐Being Questionnaire, and Diabetes Treatment Satisfaction Questionnaire).

Excluded studies

The 41 excluded studies and reasons for exclusion are summarised in the table of Characteristics of excluded studies. The most common reasons for exclusion were that the study duration was less than 10 weeks or that the study design was not appropriate for inclusion.

Risk of bias in included studies

The main quality issues of the studies are summarised in the Appendix 6 . Apart from poor reporting of allocation concealment, the main problem with the studies was reporting of results in a format that could be utilised for meta‐analysis. This was a particular problem with the placebo comparison studies, where several only reported glycosylated haemoglobin A1c (HbA1c) results in graphical form (Chuang 1999; Derosa 2003; Furlong 2002; Horton 2000; Jovanovic 2000; Marre 2002; Moses 2001; Saloranta 2002). For these studies estimates have been included in the main outcomes tables only.

Effects of interventions

The results are presented by comparator, examining the evidence of efficacy by first reviewing the studies comparing meglitinide monotherapy to a placebo. Secondly, the comparison of the two meglitinides currently licensed for use provided us with information as to whether the therapies could be considered together in future comparisons or whether their properties were sufficiently different as to require separate consideration. Finally, comparisons with metformin were undertaken. Three potential treatment regimens were anticipated: use of meglitinides as monotherapy, use of meglitinides in combination with metformin, use of meglitinides in combination with insulin. Each of these combinations was considered for the three comparator options noted above.

The results for the outcomes glycosylated haemoglobin A1c (HbA1c), and weight change are summarised in Appendix 4 and Appendix 5, respectively. Reporting of hypoglycaemia was inconsistent with different definitions being used throughout. Hypoglycaemia has been described in the text but it was not possible to summarise results in tables in any meaningful way.

Comparisons of meglitinides with placebo

Five studies compared monotherapy with repaglinide to placebo (Chuang 1999;Goldberg 1998;Jovanovic 2000;Moses 2001;Van Gaal 2001). One study compared repaglinide to placebo when used in combination with metformin (Moses 1999). No study compared repaglinide to placebo when used in combination with insulin.

For nateglinide, monotherapy was compared to placebo in three studies (Hanefeld 2000;Horton 2000; Saloranta 2002). Two studies compared nateglinide to placebo in combination with metformin (Horton 2000; Marre 2002). No study compared nateglinide to placebo when used in combination with insulin.

Repaglinide versus placebo: monotherapy

(Chuang 1999; Goldberg 1998; Jovanovic 2000; Moses 2001; Van Gaal 2001)

Glycosylated haemoglobin

Four of the five studies comparing repaglinide to placebo as monotherapy reported clinically significant reductions in HbA1c in the repaglinide treatment groups (0.5% to 2.1%) (Goldberg 1998; Jovanovic 2000; Moses 2001; Van Gaal 2001). The fifth study reported a non clinically significant reduction in glycosylated haemoglobin (Chuang 1999). One study (Jovanovic 2000) compared two different fixed doses of repaglinide (1.0 mg and 4.0 mg) taken three times per day. There was little difference in the reported reduction in HbA1c from baseline for the two doses (0.7% and 0.5% respectively) despite the 4.0 mg treatment group having a higher baseline HbA1c (8.2% versus 8.7%). Chaung et al (Chuang 1999) compared two fixed doses (0.5 mg and 2.0 mg) taken three times per day but did not demonstrate a clinically important reduction in glycosylated haemoglobin with either dose (HbA1c reduction of 0.1% and 0.3% respectively). 
 In four studies glycosylated haemoglobin in the placebo group rose during the study period by between 0.1 to 1.4% (Chuang 1999; Goldberg 1998; Jovanovic 2000; Van Gaal 2001). In Moses et al (Moses 2001) the placebo group achieved a small reduction in glycosylated haemoglobin of 0.2%. 
 One study only reported placebo data in graphical form (Chuang 1999) and three only reported standard error of the mean (or standard deviation) in graphical form (Jovanovic 2000; Moses 2001; Van Gaal 2001). Jovanovic 2000 only reported standard error of the mean (or standard deviation) for a post‐hoc subgroup. Data for these studies are reported in Appendix 4. There was insufficient information to allow meta‐analysis.

Lipids

Only one study reported findings for fasting lipids (Goldberg 1998). The authors found no significant difference between those receiving repaglinide and placebo for HDL cholesterol, LDL cholesterol and triglycerides, but the numbers of patients analysed differed for each lipid outcome (ranging from 56 to 62 out of the 66 patients randomised in the repaglinide group and from 23 to 30 out of the 33 patients randomised to placebo). One other study only stated that triglyceride levels decreased for both active treatment doses of repaglinide and cholesterol was reduced in the 2.0 mg repaglinide group (Chuang 1999). No levels were reported. There was insufficient information to allow meta‐analysis.

Quality of life

In one study, the quality of life was assessed using a variety of validated tools (Moses 2001). The data for quality of life assessment were published in a separate report including only some participants of the total study population (n = 253). The authors reported that repaglinide improved patient satisfaction: 9% (SD 20) improvement in the repaglinide group versus 2% (SD 18) improvement in the placebo group (difference between groups P < 0.01), using the WHO Diabetes Satisfaction Questionnaire, which consists of a scale from 0 to 100. They found no significant difference between repaglinide and placebo using the WHO Wellbeing Questionnaire, or the EuroQual Health Status Measure‐5D.

Weight

Three studies reported weight gain in the repaglinide treated groups. One small study found that repaglinide increased weight by 2.1 kg (P < 0.01) while reporting no significant change in weight in the placebo group (weight loss 0.8 kg) (Van Gaal 2001). The second study reported a weight gain of 0.8 kg in the 2.0 mg repaglinide group; in the lower dose repaglinide group the mean weight was unchanged and for placebo a mean fall of 1.5 kg was achieved (Chuang 1999). Goldberg 1998 reported an increase in weight in the repaglinide group of 0.7 kg with a fall of 2 kg in the placebo group. For the comparison of 2.0 mg repaglinide versus placebo the mean difference in weight gain compared to placebo was 2.3 kg (95% confidence interval (CI) ‐1.8 to 6.4) with repaglinide (Chuang 1999). The lower dose of repaglinide was reported not to have caused a change in weight but no other details were provided. As only one trial is available at the current time, we have not reported any summary totals but will do so in the future as the data allow.

Hypoglycaemia

Symptomatic hypoglycaemic episodes, reported by the participant based on typical symptoms of hypoglycaemia but not necessarily confirmed by blood glucose, were reported by all five studies. Three studies found rates of symptomatic hypoglycaemia ranging from 17% to 44% in the treated groups (Goldberg 1998; Moses 2001;Van Gaal 2001). Two studies compared two different doses of repaglinide, and reported higher rates of symptomatic hypoglycaemia with 4.0 mg compared with 1.0 mg (35% versus 27%, respectively) (Jovanovic 2000) and 2.0 mg compared with 0.5 mg (17% versus 11%, respectively) (Chuang 1999). One study (Moses 2001) reported that three patients (1%) receiving repaglinide experienced major hypoglycaemic episodes requiring third party help. The four other studies reported no major hypoglycaemic episodes.

Severe adverse events (or sufficient to cause termination of therapy)

Serious adverse events and adverse events leading to withdrawal were inconsistently and generally incompletely reported. 
 Four studies reported withdrawals due to adverse events (Chuang 1999; Goldberg 1998; Jovanovic 2000; Moses 2001). Jovanovic 2000 found withdrawal rates due to adverse events of 6% to 8% with repaglinide 1.0 mg and 4.0 mg, respectively, compared with 16% for placebo (Jovanovic 2000). Specific reasons included: hypoglycaemic events (one with repaglinide 1.0 mg versus two with repaglinide 4.0 mg versus three with placebo) and myocardial infarction (one in each repaglinide group). No details were given of other adverse events leading to withdrawal. Moses 2001 found a total of twelve serious adverse events, with similar rates for repaglinide and placebo (2.6% with repaglinide versus 2.9% with placebo). The authors reported that adverse events leading to withdrawals were mostly minor, including: elevated liver enzymes (two with placebo versus one with repaglinide); abdominal pain, dyspepsia, constipation (four with repaglinide), and facial oedema (one with repaglinide). Three patients on repaglinide were withdrawn due to more serious adverse events: one each due to grand mal convulsions, paraesthesia, and pancreatitis. Goldberg 1998 reported withdrawal due to adverse events in one patient with repaglinide (myocardial infarction) and one patient with placebo (foot ulcer). They reported five serious adverse events in patients taking repaglinide: myocardial infarction (mentioned above), hospitalisation for epistaxis, congestive cardiac failure plus probable atrioventricular block in a patient on beta‐blockers, a patient with a malignant cutaneous neoplasm and a patient with diverticulitis diagnosed during follow‐up. The researchers did not consider these events to be related to the study drug. Chuang 1999 reported two severe adverse events: jaundice and ketoacidosis. Four patients withdrew from the trial (one from jaundice in the placebo arm, one from erectile dysfunction in the 0.5 mg repaglinide arm and two from dizziness and fever in the 2.0 mg repaglinide arm).

Other adverse events

The proportion of patients reporting any adverse event ranged from 29% to 80% with repaglinide and from 18% to 71% with placebo (Chuang 1999; Goldberg 1998; Jovanovic 2000; Moses 2001). One study reported that the most common adverse events found for 1.0 mg and 4.0 mg repaglinide were: headache (11% and 14%); upper respiratory tract infection (both 13%); and dizziness (2% and 10%) (Jovanovic 2000). The same study noted that cardiovascular events were reported by 9% and 14% of patients taking repaglinide 1.0 mg and 4.0 mg compared with 8% taking placebo (the cardiovascular events included chest pain (3% in repaglinide groups versus 1% in placebo group), hypertension, ECG abnormalities). A second study reported that the most common adverse events, possibly or probably related to study drug, were: rash (3% with repaglinide versus 3% with placebo), nausea (2% with repaglinide versus 3% with placebo) and fatigue (2% with repaglinide versus 3% with placebo) (Goldberg 1998). The third study reported that adverse events were similar for repaglinide and placebo but gave few details (Moses 2001). Chuang 1999 indicated that 55%, 58% and 79% of participants in the placebo, 0.5 mg and 2.0 mg repaglinide groups respectively, reported minor adverse events. Seventy‐eight percent of these were upper respiratory tract infections and believed to be unrelated to the treatments. The final study only reported hypoglycaemic events, but commented that no serious adverse events occurred (Van Gaal 2001).

Diarrhoea

The two studies that reported the proportion of participants experiencing diarrhoea, found that 2% of participants receiving an incremental regimen of repaglinide experienced diarrhoea (Goldberg 1998), and in the fixed dose repaglinide regimens 5% to 6% reported diarrhoea, with no substantial difference between different doses (Jovanovic 2000). The proportion of patients receiving placebo that reported diarrhoea was 0% to 1% (Goldberg 1998; Jovanovic 2000).

Subgroup analyses

Two studies that included a mixture of previously treated, and previously untreated, patients with diabetes mellitus reported subgroup analyses according to previous treatment status, but analyses did not appear to be on an intention‐to‐treat basis and were not stated a priori (Goldberg 1998; Jovanovic 2000). These subgroups are reported in more detail below. Chuang 1999 also included a mix of treated and untreated patients, but no information was given about the numbers of each and results were not reported separately. Moses 2001 and Van Gaal 2001 included only those treated by diet alone but did not report primary outcomes in such a way as to allow meta‐analysis. 
 Jovanovic 2000 reported a statistically significant HbA1c reduction after repaglinide in a subgroup of patients who had been previously treated with an oral hypoglycaemic agent, compared with placebo (change in HbA1c from baseline: ‐0.1% with repaglinide 1.0 mg versus ‐0.1% with repaglinide 4.0 mg versus +1.8% with placebo, P < 0.001 for repaglinide versus placebo). While statistically significant, the fall from baseline HbA1c would not be considered to be clinically significant. Also, the numbers analysed were not reported and it was not clear whether analysis was on an intention‐to‐treat basis. The same study depicted a larger change in HbA1c from baseline in the treatment naive patient group (figures not reported, read from graph: ‐1.3% with repaglinide 1.0 mg versus ‐1.8% with repaglinide 4.0 mg versus +1% with placebo, P value not reported). This analysis was not on an intention‐to‐treat basis (95 randomised, 64 analysed). 
 The other study (Goldberg 1998) conducted a post‐hoc, subgroup analysis after finding that the proportion of sulphonylurea‐naive patients differed between treatment groups, at baseline (15/67 [23%] in repaglinide group versus 3/33 [9%] with placebo, P = 0.097). They reported that repaglinide significantly reduced HbA1c, compared with placebo, in patients previously treated with sulphonylurea (change from baseline: ‐0.1% with repaglinide versus +1.5% with placebo, P < 0.001). This analysis was not on an intention‐to‐treat basis (81 randomised, 74 analysed). They found that repaglinide significantly reduced HbA1c, from baseline to 20 weeks, in sulphonylurea‐naive patients (51 patients randomised, number analysed was not reported: change from baseline ‐2.7%, P not reported). There were only three sulphonylurea‐treated patients in the placebo group.

Summary

Repaglinide monotherapy compared to placebo produced a clinically significant reduction in glycosylated haemoglobin, without evidence of major adverse events or severe hypoglycaemia.

Repaglinide versus placebo: combination therapy with metformin

(Moses 1999)

Glycosylated haemoglobin

Both treatment arms resulted in a reduction in glycosylated haemoglobin. The larger reduction was obtained with combination therapy (1.4% versus 0.3%; P < 0.05); both groups had a similar baseline HbA1c (8.6% versus 8.3%). The reduction from baseline was clinically and statistically significant in the combination therapy arm (95% confidence interval 1.0 to 1.9; P < 0.05). Sixty percent of those treated with combination therapy had "adequate" control (HbA1c less than 7%) by the end of the study period, compared to 20% in the metformin monotherapy group. The mean difference in glycosylated haemoglobin reduction for repaglinide compared to placebo was 1.1% (95% CI ‐1.7 to ‐0.4) with repaglinide.

Quality of life

Not reported

Lipids

The authors reported no substantial change in lipids during the study period.

Weight

A mean weight loss of 0.9 kg was achieved in the metformin plus placebo group (not statistically significant; P not reported), but the combination therapy group gained a mean 2.4 kg, during the study (P < 0.05). The mean difference was a 3.3 kg (95% CI 1.9 to 4.7) gain for those treated with combination therapy over the metformin‐placebo treated group

Hypoglycaemia

Nine patients (33%) on combined therapy reported a total of 30 hypoglycaemic episodes (one patient in the combined therapy group accounted for 12 of 30 hypoglycaemic episodes). 
 Those on metformin monotherapy reported no hypoglycaemia.

Severe adverse events (or sufficient to cause termination of therapy)

Five severe adverse events were reported, but none were considered to be related to the treatments and the authors did not report to which arm of the study these five participants belonged.

Diarrhoea

More diarrhoea was reported in the arm treated with metformin monotherapy than metformin‐repaglinide combination therapy (29% versus 19%; P = 0.5 estimated from percentages). The difference was not statistically significant.

Subgroup analysis

Observed changes were reported to be similar for men versus women, and for those aged less than or equal to 65 years versus those greater than 65 years, but no data were presented. All participants had previously received oral treatment for their diabetes mellitus.

Summary

Combination of metformin and repaglinide produced a clinically (and statistically) significant reduction in glycosylated haemoglobin over three months compared with metformin alone without any reported severe hypoglycaemia or other adverse events, but at the expense of a statistically significant weight gain.

Nateglinide versus placebo : monotherapy

(Bengel 2005; Hanefeld 2000; Horton 2000; Saloranta 2002)

Glycosylated haemoglobin

Three studies reported that nateglinide 120 mg produced a reduction in HbA1c of around 0.5% (0.4% in Bengel 2005, 0.6% in Hanefeld 2000 and 0.5% in Horton 2000). In all three studies HbA1c in the placebo group increased (0.5% in Bengel 2005, 0.1% in Hanefeld 2000 and 0.5% in Horton 2000). In the fourth study (Saloranta 2002) nateglinide 120 mg resulted in a small reduction in HbA1c of 0.2% (estimated from graph), while the HbA1c rose by 0.2% in the placebo arm. However, the mean baseline HbA1c in this study was less than 7%; lower than the other three studies, and this may account for the lower effect seen in this study. 
 Two studies compared more than one dose of nateglinide. Hanefeld 2000 reported the mean HbA1c change by dose to be: ‐0.2% with 30 mg, ‐0.4% with 60 mg, ‐0.6% with 120 mg and ‐0.6% with 180 mg. The other study reported the mean HbA1c change by dose to be: ‐1.0% with 30 mg, ‐0.2% with 60 mg, and ‐0.2% with 120 mg (Saloranta 2002) (estimated from graph). 
 Hanefeld 2000 provided results as mean difference in HbA1c with 95% confidence intervals. For Bengel 2005, Horton 2000 and Saloranta 2002, estimates of HbA1c change had to be made from graphical representations. Therefore, we did not have sufficient information in the appropriate format to undertake meta‐analysis.

Quality of life

Not reported

Lipids

Two of the studies commented that there were no substantial changes in lipids from baseline, but findings were not reported (Bengel 2005; Hanefeld 2000).

Weight

Horton 2000 reported no substantial change in weight from baseline (less than one kg) but no other details were provided. In Saloranta 2002, it is not clear if mean weight change or maximum weight change is being reported but the authors describe no statistically significant differences between treatment groups. In other two studies, neither final weight, body mass index (BMI) nor weight change were reported (Bengel 2005; Hanefeld 2000).

Hypoglycaemia

Only one study recorded the proportion of participants reporting symptoms of hypoglycaemia by treatment group. They reported symptomatic hypoglycaemia in 12% of those receiving nateglinide 30 mg, 11% in those receiving 60 mg and 23% in those receiving 120 mg of nateglinide compared with 5% with placebo (Saloranta 2002). "Confirmed" cases were lower (2% on 30 mg, 4% on 60 mg, 5% on 120 mg and 1% on placebo). Horton 2000 reported symptomatic hypoglycaemia in 13% (three events confirmed by plasma glucose of less than 3.3 mmol/L) in those receiving 30 mg nateglinide but did not report the frequency in other treatment arms. The authors did note that 14% of the adverse events were due to hypoglycaemia. Hanefeld 2000 did not report hypoglycaemia by treatment group but did note that only three episodes were confirmed by plasma glucose, and all three occurred in the nateglinide 120 mg group. One participant, receiving nateglinide 30 mg, experienced a fall with a short period of loss of consciousness and spontaneous recovery that was attributed to hypoglycaemia by the authors. Bengel 2005 did not report any information about hypoglycaemia.

Severe adverse events (or sufficient to cause termination of therapy)

Adverse events led to withdrawal in 2% to 8% patients treated with nateglinide and from 3% to 5% of patients treated with placebo.

Other adverse events

None of the studies adequately described specific details of adverse events by treatment group. One study found that adverse events were more common with nateglinide compared with placebo (49% versus 35%, P not reported) but reported no details (Hanefeld 2000). One study found that central nervous system disorders (primarily tremor) were more common with nateglinide but, again, provided no further details of the types of disorders or the frequency (Saloranta 2002). The third study reported adverse events in 77.7% with nateglinide compared with 68.6% with placebo (Horton 2000). The study reported overall adverse events for all four treatment groups combined (nateglinide alone, metformin alone, nateglinide plus metformin and placebo). They stated that specific causes of adverse events were similar among treatment groups but presented no data by treatment group. Bengel 2005 only reported that three participants withdrew form the trial due to non serious adverse events but gave no information about adverse events in those who completed the trial.

Diarrhoea

Only one study specifically reported the number of participants with diarrhoea who received nateglinide compared with placebo (Horton 2000). Five (of 179) patients reported diarrhoea in the nateglinide group and nine (of 172) in the placebo group. One participant withdrew from Bengel 2005 with diarrhoea.

Subgroup analyses

One study reported subgroup analyses exploring the influence of sex, BMI, age and baseline HbA1c on change in HbA1c for nateglinide compared with placebo (Saloranta 2002). They found that neither sex nor age (less than 65 years versus greater than or equal to 65 years) influenced results (data were not presented). They also reported that nateglinide 120 mg reduced HbA1c more in patients with a baseline HbA1c less than or equal to 6.5% compared with baseline HbA1c greater than 6.5% (difference from placebo in two HbA1c groups: ‐0.5% versus ‐0.3%, P not reported). Furthermore, they reported that nateglinide 120 mg reduced HbA1c more in obese patients with BMI less than or equal to 30 kg/m² , compared with BMI more than 30 kg/m2 (nateglinide minus placebo: ‐0.6% versus ‐0.3%, P not reported). 
 Two of the studies (Hanefeld 2000; Saloranta 2002) included only treatment naive participants while Horton 2000 and Bengel 2005 included a mixture of treated and untreated participants. Horton 2000 and Bengel 2005 did not present the results separately for these two groups. Data for the primary outcomes of HbA1c and weight were insufficient to allow meta‐analysis.

Summary

From review of the placebo studies, nateglinide 120 mg produced a reduction in HbA1c over 12 to 24 weeks that was of borderline clinical significance (taken as 0.5%). Reporting of adverse events was limited in the included studies, but no major adverse events or hypoglycaemic episodes were reported. 
 While any comparison of the effect on HbA1c for repaglinide and nateglinide must be made with caution, as no direct comparison has been made in these studies, there is evidence that the clinical effects of the two drugs may differ. The next but one section of this review concentrates on the evidence obtained from comparing the two drugs in head to head trials.

Nateglinide versus placebo: combination therapy with metformin

(Horton 2000; Marre 2002)

Glycosylated haemoglobin

A clinically important reduction in HbA1c, from baseline, was produced in the combination therapy arms of both studies. The HbA1c results were estimated from graphs for both studies, as no figures for change from baseline were presented. In one study, where all participants had previously received metformin (Marre 2002), the HbA1c in the metformin monotherapy arm was stable. In the two nateglinide treatment arms HbA1c fell by 0.5% and 0.7%, with the larger drop in the group receiving the higher dose of nateglinide; baseline HbA1c was 8.0% (60 mg nateglinide) and 8.2% (120 mg nateglinide). In the other study (Horton 2000), after a four week run‐in period of diet only therapy, both the metformin monotherapy and the combination therapy arms experienced a drop in HbA1c (0.8% versus 1.4% respectively). Participants in Horton 2000 were a mixture of those who had previously received oral treatment and those who were treatment naive. Data were not presented separately for these groups. 
 As mean change had to be estimated from a graph, no estimate of variance was available and we were not able to present the results from this comparison in a meta‐analysis.

Quality of life

Not reported

Lipids

Only one study reported blood lipids and did not identify any substantial change from baseline in any of the treatment arms but not details were reported (Marre 2002).

Weight

The greatest weight gain was reported in Marre 2002 among those receiving combination therapy with 120 mg nateglinide (one kg), with only small increases in weight in the other two groups. In the other study, the authors state that there was no statistically significant weight change, but figures were not presented (Horton 2000). 
 Marre 2002 reported a mean difference in weight gain comparing the 120 mg nateglinide treatment group and the placebo treatment group, when all participants were receiving metformin, of +0.9 kg (95%CI 0.4 to 1.5) additional gain in the nateglinide group. As Marre 2002 was the only study providing data that could be included, no summary totals could be calculated.

Hypoglycaemia

Hypoglycaemic episodes were low in both studies, with no reports of major hypoglycaemic events requiring help. In the Marre 2002 study there were 13 and 25 reported episodes in the low and high dose nateglinide combination therapy groups respectively (with one and five confirmed by blood glucose, respectively) compared with six (one confirmed) in the metformin monotherapy group. In the other study (Horton 2000) there were 45 reports (five confirmed) for combination therapy versus 18 reports (one confirmed) in the metformin monotherapy group.

Severe adverse events (or sufficient to cause termination of therapy)

In both studies, 50% of participants reported adverse events. In the Horton 2000 study, withdrawals due to treatment adverse events occurred in 12 (six thought to be due to therapy) in the monotherapy arm with metformin, and 16 (six thought to be due to therapy) in the combination therapy. 
 In the Marre 2002 study there were two deaths, both in the nateglinide treatment groups and both from cardiac causes. Neither was thought to be related to the drug treatment.

Diarrhoea

In Horton 2000, reporting of diarrhoea was higher in the monotherapy arm: 20% with metformin alone versus 15% with metformin plus nateglinide. Diarrhoea reporting was lower in the other study (Marre 2002), where all participants had previously been treated with metformin: 8% in the monotherapy arm versus 5.6 to 5.8% in the combination therapy arms.

Subgroup analyses

Marre 2002 reported that combination treatment with nateglinide produced the greatest reduction in HbA1c among those with a baseline HbA1c of greater than 9.5% (reduction of 1.4%, metformin alone data not reported). Similarly Horton 2000 identified the largest reductions in glycosylated haemoglobin, compared with placebo, among those with baseline HbA1c of greater than 9.5% (reduction of 2.5% with combination therapy versus 1.5% with metformin alone; relative to placebo). 
 Marre 2002 only included participants previously treated with oral hypoglycaemic agents but Horton 2000 did not separate subgroups by prior treatment.

Summary

Both studies, comparing nateglinide in combination with metformin, to metformin monotherapy reported reductions in glycosylated haemoglobin that are of clinical significance. However, since no point estimates were provided (with measures of variance), it is not possible to combine the data in the form of a meta‐analysis. While there was evidence that the combination therapy resulted in more hypoglycaemia, there were no severe hypoglycaemic episodes. Adverse events were similar for combination therapy and metformin alone.

Comparison of repaglinide with nateglinide

Two studies compared repaglinide to nateglinide, head to head (Raskin 2003; Rosenstock 2004). The meglitinides were used as monotherapy in Rosenstock 2004 and in combination with metformin in Raskin 2003.

Repaglinide versus nateglinide (monotherapy)

(Rosenstock 2004)

Glycosylated haemoglobin

A clinically significant reduction in glycosylated haemoglobin was reported in both arms of the study (mean reduction of 1.6% in those randomised to repaglinide versus 1% in those randomised to nateglinide). Baseline HbA1c values were similar in both arms (8.9%). Only 9 (12%) of those receiving repaglinide had been titrated to the maximum dose whereas 57 (77%) reached maximum dose. 
 Rosenstock 2004 reported a mean difference in reduction of HbA1c of 0.5% (95% CI ‐0.9 to ‐0.1) more in the repaglinide than the nateglinide treated group.

Quality of life

Not reported

Lipids

Not reported

Weight

A weight gain of 1.8 kg was reported in the repaglinide arm, versus 0.7 kg gain in the nateglinide arm. The data reported did not permit inclusion in any meta‐analysis.

Hypoglycaemia

No major hypoglycaemia was reported in either treatment group.

Severe adverse events (or sufficient to cause termination of therapy)

Two participants withdrew due to adverse events (no details reported) in the repaglinide group. There were no withdrawals due to adverse events in the nateglinide group.

Other adverse events

Little information was presented regarding adverse events, with the authors reporting no significant differences between treatment groups.

Subgroup analysis

The inclusion criteria for Rosenstock 2004 may have included a mixture of treated and untreated people with diabetes mellitus but data were not presented separately. No other subgroups were reported.

Repaglinide versus nateglinide (in combination with metformin)

(Raskin 2003)

Glycosylated haemoglobin

The single study reported a clinically significant reduction in glycosylated haemoglobin in both arms of the study, with the greatest reduction in the repaglinide arm (HbA1c reduction of 1.3% versus 0.7%). Baseline HbA1c values were similar in both arms (8.4% versus 8.2%). Fifty‐nine percent of patients achieved an HbA1c of less than 7% in the repaglinide plus metformin arm, compared with 46% in the nateglinide plus metformin arm. This was despite only 7% of the repaglinide arm requiring, or reaching, maximal dose. Eighty percent of the nateglinide arm reached maximal dosage. All participants had been treated with oral hypoglycaemic agents previously. 
 Raskin 2003 presented a mean difference in reduction of HbA1c which was of 0.6% (95% CI ‐0.9 to ‐0.3) more in the repaglinide than in the nateglinide treated group. As it is the only study of this comparison, no totals could be calculated.

Quality of life

Not reported

Lipids

Not reported

Weight

A gain of 0.6 kg was reported in the repaglinide arm of the study, compared to a loss of 0.5 kg in the nateglinide arm. The data reported did not permit inclusion in any meta‐analysis.

Hypoglycaemia

Seven percent of the repaglinide arm experienced symptomatic hypoglycaemia versus 2% in the nateglinide arm (no P value reported). No major hypoglycaemic events occurred in either treatment group (that is none with symptoms so severe that the participant could not treat themselves).

Severe adverse events (or sufficient to cause termination of therapy)

No severe adverse events were reported. One patient, in the nateglinide arm, had to withdraw due to adverse events but the event was not specified.

Other adverse events

The most commonly reported adverse event was upper respiratory tract infection (21% (repaglinide) versus 12% (nateglinide); no P value reported).

Diarrhoea

A similar frequency of diarrhoea was recorded for both the treatment arms, with less than 8% of participants reporting diarrhoea.

Summary

Both Rosenstock 2004 and Raskin 2003 reported a greater drop in HbA1c after 16 weeks of therapy with repaglinide, compared with nateglinide. These two studies provide evidence that it is not appropriate to assume that the two meglitinides are sufficiently similar for us to combine them for the purpose of comparison with other diabetes therapies. Repaglinide and nateglinide will, therefore, be considered separately in the remaining comparisons.

Comparisons of meglitinides with metformin

Repaglinide was compared to metformin monotherapy in two studies (Derosa 2003; Moses 1999). Moses 1999 included a placebo comparison arm and has been described above. Derosa 2003 included only participants who were treatment naive. One study (Furlong 2002) compared repaglinide to metformin in patients also receiving insulin therapy.

One study compared nateglinide to metformin with both medicines given as monotherapy (Horton 2000) to a group of participants who were a mixture of treatment naive and those who had already received oral therapy of some description. Horton 2000 also included a treatment arm comparing nateglinide to placebo (described above). No studies compared nateglinide to metformin in patients who were also prescribed insulin.

Repaglinide versus metformin (monotherapy)

( Derosa 2003; Moses 1999)

Glycosylated haemoglobin

In the 12 month study (Derosa 2003) both treatments were associated with a clinically important reduction in HbA1c (0.8% for repaglinide, and 0.9% for metformin). The baseline HbA1c was similar in both treatment arms (7.6% repaglinide versus 7.4% metformin). In the shorter study (Moses 1999), HbA1c was also reduced in both treatment groups. Again, the reduction was similar in both arms (0.4% repaglinide versus 0.3% metformin) but not clinically significant. HbA1c at baseline was 8.6% in both treatment arms. 
 Moses 1999 showed that the mean difference in HbA1c was 0.1% (95% CI ‐0.6 to 0.7) lower in those treated with repaglinide versus metformin.

Quality of life

Not reported

Lipids

There were no substantial changes reported in lipid profiles in either study.

Weight

In the 12 month study (Derosa 2003), both treatments were associated with reductions in weight (0.8 kg in repaglinide versus 2.0 kg in metformin) and BMI, but these were not statistically significant in terms of the baseline to follow‐up change. In Moses 1999, a 3.0 kg weight gain occurred in the repaglinide group (versus a 0.9 kg reduction in weight in the metformin group). 
 Moses 1999 demonstrated a mean difference in weight gain comparing the repaglinide treatment group and the metformin treatment group of a +3.8 kg (95% CI 2.5 to 5.2) additional gain in the repaglinide group. As Marre 2002 was the only study providing data that could be included, no summary totals were reported.

Hypoglycaemia

In Moses 1999, three (11%) patients reported nine events in the repaglinide group, while none of the metformin group reported symptoms of hypoglycaemia. No severe hypoglycaemia was reported. No information about hypoglycaemia was reported in Derosa 2003.

Severe adverse events (or sufficient to cause termination of therapy)

No severe adverse events were reported in the Moses 1999 study. Six participants withdrew from the metformin group and three from the repaglinide group. Reasons for withdrawal were not reported. No information was reported regarding adverse events in Derosa 2003.

Other adverse events

Diarrhoea

In the Moses 1999 study, diarrhoea was reported in 7% of the repaglinide group and 30% of the metformin group.

Subgroups

Derosa 2003 included only treatment naive participants while Moses 1999 was restricted to those who had previously received oral treatments.

Summary

While both studies had limitations in terms of reporting of methodological details, these findings support that repaglinide had an effect on HbA1c, in both treatment naive and previously treated patients, which was similar to the effect of metformin. In treatment naive patients, the fall in HbA1c was clinically significant (more than 0.5%). In previously treated patients, the treatment effect on HbA1c was less substantial despite higher baseline glycosylated haemoglobin levels. Where recorded (Moses 1999), few adverse effects were reported with repaglinide; diarrhoea was reported less than with metformin treatment. No major hypoglycaemia was reported, although slightly more symptoms of hypoglycaemia were experienced among the patients in the repaglinide group. Weight gain was greater in those treated with repaglinide.

Repaglinide versus metformin (in combination with insulin)

(Furlong 2002)

Glycosylated haemoglobin

Only participants receiving metformin plus insulin achieved a reduction in HbA1c over the 13 week trial period (0.4% reduction in mean HbA1c). In the repaglinide plus insulin group, the mean HbA1c increased by 0.4%. Baseline mean HbA1c levels were comparable (8.4% in the metformin‐insulin versus 8.1% in the repaglinide‐insulin group). Data were only presented in graphical form and therefore could not be included in a meta‐analysis.

Quality of life

Quality of life was measured using two validated questionnaires (Well‐Being Questionnaire, and Diabetes Treatment Satisfaction Questionnaire). Small improvements in Well‐Being Questionnaire were reported in each treatment group. In the Treatment Satisfaction Questionnaire, those receiving metformin‐insulin combination reported an improvement in satisfaction (32 to 34) but the repaglinide‐insulin group reported a deterioration in satisfaction (33 to 29). The difference between the two groups is statistically significant but the authors do not report what, a priori, they would consider to be a clinically important change. Insufficient information was reported to enable inclusion in meta‐analysis.

Lipids

Not Reported

Weight

Both treatment groups experienced an increase in mean weight (0.9 kg ‐ metformin‐insulin versus 2.7 kg repaglinide‐insulin). Furlong 2002 reported a mean difference in weight gain comparing the repaglinide treatment group and the metformin treatment group of +1.8 kg (95% CI 0.7 to 2.9) additional gain in the repaglinide group. As Furlong 2002 was the only study providing data that could be included, no summary totals were calculated.

Hypoglycaemia

No episodes of serious hypoglycaemia were reported although five episodes of nocturnal hypoglycaemia were reported in the repaglinide‐insulin group. It was not reported how many individuals this involved.

Serious adverse events

Five serious adverse events were recorded throughout the study; all occurred in the repaglinide‐insulin group. Three occurred in one patient (chest pain), one myocardial infarction and one adrenal tumour.

Other adverse events

The most commonly reported non‐severe adverse events were gastrointestinal side effects, respiratory and urinary tract infections, headache and itch but no details were reported.

Subgroups

All participants were injecting insulin at recruitment. Data were not presented for other subgroups.

Summary

In summary, the evidence from the one study suggests that metformin is superior to repaglinide in combination with insulin, with a greater reduction in HbA1c and less weight gain.

Nateglinide versus metformin (monotherapy)

(Horton 2000)

Glycosylated haemoglobin

A clinically significant drop in HbA1c was achieved for both relevant arms of the study (0.5% for nateglinide versus 0.8% for metformin monotherapy (estimated from graph ‐ no data provided)), having started from a baseline HbA1c of 8.3% and 8.4% respectively. This result was adjusted for treatment centre and baseline HbA1c, but no unadjusted values were provided to assess the impact of the adjustment factors. The study, because of drop‐outs and design, was not powered to detect differences between the two treatment arms. Insufficient information was reported to enable inclusion in a meta‐analysis, with the mean HbA1c change estimated from a graph.

Quality of life

Not reported

Lipids

Not reported

Weight

No significant body weight changes were reported for either group, but absolute values were not reported.

Hypoglycaemia

Similar numbers of hypoglycaemic episodes were reported by participants in each treatment group (23 with nateglinide versus 18 with metformin) with very few confirmed by low blood glucose (three with nateglinide versus one with metformin). No severe hypoglycaemic episodes were reported.

Severe adverse events (or sufficient to cause termination of therapy)

Withdrawals due to adverse events were more common with metformin (12 versus 5), although only six versus one were considered to be related to treatment, by the clinicians. One death due to cardiac disease occurred in the metformin group but was reported to be unrelated to treatment.

Other adverse events

Diarrhoea

Twenty percent of metformin treated patients experienced diarrhoea; reportedly three to four times more than those in the nateglinide arm, but no data were provided.

Subgroups

Horton 2000 included a mixture of participants previously treated and untreated with oral hypoglycaemic agents. No subgroup data were presented.

Summary

Horton 2000 suggested that nateglinide produced clinically significant reductions in HbA1c. Reductions appear to be of a level similar to, or slightly less than, metformin. Fewer gastrointestinal adverse events were reported with nateglinide, and no serious hypoglycaemia was noted. The proportion of treatment naive patients was not reported and therefore it is difficult to draw any comparisons between nateglinide and metformin effects on HbA1c.

Discussion

In order to evaluate the role of meglitinide analogues in the treatment of type 2 diabetes mellitus, we present a systematic review of the evidence for a number of comparators thought to provide the most relevant information to clinical practice.

Summary of main results

As yet, no data are available about the impact of meglitinides on mortality or diabetes complications. For this review, a clinically significant reduction in glycosylated haemoglobin was taken to be a 0.5% reduction in HbA1c. None of the studies included in the review reported what level of glycosylated haemoglobin reduction would have been considered to be of clinical significance by the trialists. It is possible that this information was included in their power calculations but it was not reported separately. From a public health perspective, an epidemiological investigation of the UKPDS reported that a 1% reduction in HbA1c was estimated to result in a 21% reduction in diabetes related complications or deaths (Stratton 2000). 
 The placebo versus meglitinides studies demonstrated that both repaglinide and nateglinide are efficacious, producing both statistically and clinically significant reductions in glycosylated haemoglobin, at least in the short term (10 to 24 weeks). Two studies compared repaglinide to nateglinide, again demonstrating a clinically significant reduction in glycosylated haemoglobin for both treatments (Raskin 2003; Rosenstock 2004). There was sufficient difference in effect on glycosylated haemoglobin between the two drugs that consideration of the two meglitinides together, when comparing to other hypoglycaemic agents, was considered to be inappropriate. This is a finding that is in keeping with evidence from reviews of the pharmacokinetics and pharmacodynamics that suggest that nateglinide binds more rapidly and dissociates more quickly with the receptor, than repaglinide (Dornhurst 2001). This lower binding affinity may explain the lower potency, reflected in the higher doses of nateglinide required for therapy, and the clinical effect on glycosylated haemoglobin. In theory, it is possible that the pharmacokinetic properties of nateglinide could be an advantage clinically if the shorter action than repaglinide led to fewer hypoglycaemic episodes but only if the lesser potency was acceptable, as it may be in combination therapies.

When compared to metformin monotherapy, both repaglinide and nateglinide produce a similar (or slightly smaller) reduction in glycosylated haemoglobin than metformin. The combination of metformin with a meglitinide produced a clinically significant additional reduction in glycosylated haemoglobin when compared to metformin monotherapy. Only one trial considered the addition of meglitinides to insulin therapy (Furlong 2002). However, from this there was evidence that metformin in combination with insulin was more effective in reducing glycosylated haemoglobin than repaglinide in combination with insulin. The maximum follow up in any of the trials was one year, with a single study (Derosa 2003) providing some evidence to support an ongoing reduction in glycosylated haemoglobin with meglitinide therapy, under trial conditions, at one year.

Only two studies presented information about quality of life comparing repaglinide to placebo and repaglinide plus insulin to metformin plus insulin (Furlong 2002; Moses 2001). The authors of both studies reported no substantial change in quality of life using a variety of validated diseases specific and non specific tools. Treatment satisfaction, measured using the WHO Diabetes Treatment Satisfaction Questionnaire, improved statistically significantly in those treated with repaglinide alone, and when compared with those treated with placebo. When combined with insulin, patient satisfaction improved with the metformin combination but deteriorated with the repaglinide combination. The clinical significance of the changes observed is not clear and may relate to the previous therapy experiences and "expectations" of new treatments where the study was an open label trial (Furlong 2002). From the limited evidence available, it does not appear that the introduction of repaglinide causes any substantial change in quality of life.

Increasing weight in those with type 2 diabetes mellitus is recognised to lead to increased insulin resistance and, therefore, deteriorating diabetic control. Weight change is, therefore, an important factor in therapy choice. Metformin is favoured for the treatment of type 2 diabetes mellitus, particularly in obese patients, because of its beneficial effect on weight (Kirpichnikov 2002). For both repaglinide and nateglinide, in almost all studies where weight was reported, weight gains occurred. Where meglitinides were compared directly to metformin, those treated with metformin experienced the greater weight losses.

Given the evidence to support the similar degree of effect on glycosylated haemoglobin by the meglitinides and metformin, a major issue in choosing therapy will be the side effect profile of the drugs. A well recognised side effect of metformin therapy is diarrhoea (Kirpichnikov 2002). In the trial setting, 20% to 30% of participants receiving metformin reported diarrhoea. This was lower in studies recruiting only patients who were already long‐term metformin users, presumably because those severely affected by diarrhoea have been prevented from using metformin on a long‐term basis. In the clinical setting, persistent diarrhoea is reported to be an important cause of poor therapy compliance and would be an indication for changing to an alternative therapy. Evidence from the meglitinide trials with metformin comparison arms suggests that both repaglinide and nateglinide produce fewer gastrointestinal adverse events including less diarrhoea. There was no evidence of serious adverse events associated with meglitinides. However, the longest reported follow up was 12 months and the total number of patients exposed to meglitinides in trials included in this review, was relatively small (repaglinide n = 1029; nateglinide n = 1610). A number of studies reported cardiovascular deaths in the meglitinide treated groups. At this stage there is no clear evidence of an increased risk of cardiovascular complications among meglitinide treated patients with diabetes, and as the risk of cardiovascular disease is substantially increased in type 2 diabetes mellitus compared with the general population (Watkins 2003), it is not unexpected that such studies report deaths.

While obtaining improvements in glycosylated haemoglobin is a priority in terms of improving long‐term survival, it cannot be at the expense of unacceptable hypoglycaemia. There are four components to hypoglycaemia that provide useful comparison: frequency, severity, lack of awareness and timing. None of the studies reported any severe hypoglycaemic events (where severe was defined as requiring assistance). All the trials reported hypoglycaemia based on patient symptoms. None discussed lack of hypoglycaemia awareness. Only a few of the studies indicated the timing of hypoglycaemic episodes, and the authors stated that no hypoglycaemia occurred at night (in keeping with the relatively short half life of the meglitinides) (Dornhurst 2001). As might be expected, more hypoglycaemia was reported in those treated with meglitinides than those receiving a placebo. In the two 'head to head' trials of repaglinide versus nateglinide, fewer participants treated with nateglinide reported hypoglycaemia symptoms (2% versus 7%). Compared to metformin, participants receiving meglitinides reported more hypoglycaemia symptoms.

At the present time only one trial reported follow up for longer than six months. No studies have reported long‐term outcomes such as mortality or morbidity.

Overall completeness and applicability of evidence

It is difficult to comment on the representativeness of the participants in the trials identified. Little detailed information about the participants was presented. Apart from one study, most of the trials have largely had Caucasian participants. Age, sex and BMI of participants were in keeping with what might be expected in clinical practice. However, it should be noted that patients with existing complications of diabetes ‐ cardiovascular disease or renal disease, or with hepatic disease were, in general, excluded from the studies. It should also be noted that it is difficult to say how typical patients with type 2 diabetes mellitus, who participate in clinical trials, may be of the wider population with type 2 diabetes mellitus.

Potential biases in the review process

The review currently only includes comparisons of meglitinides with metformin (plus or minus insulin). Comparisons with other oral hypoglycaemic agents are currently being undertaken by other Cochrane review teams and may be referenced by this review once available. A wide search of electronic reference databases was conducted, and the drug companies contacted for additional information. No unpublished study data were identified. The main limitations for interpreting the results of this review relate to the relatively poor reporting of the factors important for assessing quality of the trials such as allocation concealment, randomisation methodology and blinding methods. As a result it was difficult to exclude the possibility of biases; a particular issue because all available trials have been funded by the pharmaceutical companies producing meglitinides.

Agreements and disagreements with other studies or reviews

A number of reviews (of variable quality Appendix 2) have been published, describing the evidence of the efficacy of meglitinides for the treatment of type 2 diabetes mellitus. We restricted our review of their findings to those published after 2000. The main findings of these reviews were in keeping with this systematic review. They reported clinically significant reductions in glycosylated haemoglobin compared with placebo and approximate equivalency to metformin and sulphonylureas. The potential benefit from greater flexibility in dosing regimens was highlighted in one review (Inzucchi 2002). We found few studies that made any attempt to assess quality of life or treatment satisfaction but the limited evidence available did not support the increased potential for flexibility in dosing resulting in improved treatment satisfaction or quality of life. Several authors highlighted the need to consider the place of meglitinides in the existing treatment of type 2 diabetes mellitus, and where cost was considered, they recognised that meglitinides role in therapy was not clear.

Authors' conclusions

Implications for practice.

Meglitinides may offer an alternative oral hypoglycaemic agent of similar potency to metformin, and may be indicated where side effects of metformin are intolerable (in particular persistent diarrhoea) or where metformin is contraindicated. However, there is no evidence available yet to indicate what effect meglitinides will have on important long‐term outcomes, in particular, on mortality. Meglitinides do appear to result in more hypoglycaemia and weight gain than metformin, and are substantially more expensive. For metformin, there is long experience of use in many patients with type 2 diabetes mellitus and serious side effects are rare. As yet, the experience with meglitinides in terms of adverse events is limited. Results from other Cochrane review groups may provide additional information about the potential role of meglitinides in the management of type 2 diabetes mellitus.

Implications for research.

Research needs can be divided into three areas: safety, efficacy and economics:

• Safety: Long‐term follow up (years) of large numbers of patients using meglitinides are required to assess safety. Large observational studies could provide this information. A review of the evidence of safety from currently available trials and observation cohorts would provide some more information regarding potential safety while long‐term follow up is undertaken.

• Efficacy: Long‐term outcomes such as mortality and morbidity from diabetes associated complications require to be assessed. Independent studies of the comparison of repaglinide with nateglinide would afford better assessment of the comparability of the two meglitinides.

• Economics: Inclusion of cost, quality of life and long‐term outcomes in future RCTs are required assess the cost‐effectiveness of meglitinides. Flexibility of treatment regimens has been used for marketing the meglitinides. It is not clear, from the current evidence, how much of gain in terms of quality of life or treatment satisfaction, this potential flexibility actually offers.

What's new

Date	Event	Description
28 October 2008	Amended	Converted to new review format.

Appendices

Appendix 1. Search strategy

Search terms

Unless otherwise stated, search terms are free text terms; MesH = Medical subject heading (Medline medical index term); exp = exploded MeSH; the dollar sign ($) stands for any character(s); the question mark (?) = to substitute for one or no characters; tw = text word; pt = publication type; sh = MeSH; adj = adjacent. MEDLINE (OVID interface) 1. (repaglinide$ or nateglinide$ or meglitinide$).mp. 2. randomized controlled trial.pt. 3. random$.tw. 4. exp Randomized Controlled Trials/ 5. meta‐analysis.pt. 6. (meta‐analysis or metaanalysis).tw. 7. exp Meta‐Analysis/ 8. systematic review.tw. 9. 2 or 3 or 4 or 5 or 6 or 7 or 8 10. 1 and 9 Embase (OVID interface) 1. (repaglinide$ or nateglinide$).mp. 2. random$.tw. 3. Randomized Controlled Trial/ 4. exp "Systematic Review"/ 5. Meta Analysis/ 6. 2 or 3 or 4 or 5 7. 1 and 6 Science Citation Index and ISI Proceedings (repaglinide or nateglinide (limited to meeting abstracts))

Appendix 2. Summary of reviews of meglitinides

Author	Search meth. stated	Comprehensive search	Inc. criteria stated	Quality ass. stated	Meta‐analy. adequate	Overall review qual.	Notes
Prescrire2000	partially	yes	no	no	N/A	adequate
DTB2003	no	no	no	no	N/A	not assessable
CCOHTA2001 (nateglinide only)	no	no	no	no	N/A	not assessable	2 page summary, probably of a larger review
Inzucchi2002	yes	yes, but Medline and bibliographies of retrievd studies only	yes	no details	N/A	good
Moses2002	no	no	no	no	N/A	poor	does not really aim to be a systematic review; it selected 3 important studies
Dornhorst2001	no detail	no detail	no detail	no details	N/A	not assessable	appears good but methods not reported
EMEA2001 repaglinide only	no	no	no	no	N/A	not assessable	this is a review for licensing decisions

Appendix 3. Summary of included studies: Comparisons of therapies

Comparison	Therapies	Study	[n] intervention	[n] control	Duration	Treatment naive pat.
meglitinide vs. placebo	repaglinide vs. placebo: monotherapy	Chaung 1999	1.5 mg per day = 36; 6mg per day = 34	22	24 weeks	Mixture (no details)
		Goldberg 1998	67	33	18 weeks	All Treated
		Jovanovic 2000	3 mg per day = 140; 12mg per day = 146	75	24 weeks	Mixture (23% treatment Naive in placebo group; 27‐28% treatment naive in treatment groups)
		Moses 2001	270	138	16 weeks	All Treatment Naive
		van Gaal 2001	13	13	10 weeks	All Treatment Naive
	repaglinide vs. placebo: combination therapy with metformin	Moses 1999	29	27	16‐20 weeks	All Treated
	nateglinide vs. placebo: monotherapy	Bengel 2004	33	14	16 weeks	Mixture (no details)
		Hanefeld 2000	90 mg per day = 51; 180 mg per day = 58; 360 mg per day =63; 540 mg per day = 57	60	12 weeks	All Treatment Naive
		Horton 2000	179	172	24 weeks	Mixture (no details)
		Saloranta 2002	90 mg per day =166; 180 mg per day = 175; 360 mg per day = 171	163	24 weeks	All Treatment Naive
	nateglinide vs. placebo: combination therapy with metformin	Horton 2000	172	176	24 weeks	Mixture (no details)
		Marre 2002	180 mg per day = 155; 360 mg per day = 160	152	24 weeks	All Treated
repaglinide vs. nateglinide	Monotherapy	Rosenstock 2004	76 (repaglinide)	74 (nateglinide)	16 weeks	Mixture (no details)
	combination therapy with metformin	Raskin 2003	96 (repaglinide)	96 (nateglinide)	16 weeks	All Treated
meglitinides vs. metformin	repaglinide vs. metfomin: monotherapy	Derosa 2003	56	56	60 weeks	All Treatment Naive
		Moses 1999	29	27	16‐20 weeks	All Treated
	repaglinide vs. metfomin: combination therapy with insulin	Furlong 2002	39	41	13 weeks	All Treated
	nateglinide vs. metformin: monotherapy	Horton 2000	179	176	24 weeks	Mixture (no details)

Appendix 4. Summary of outcomes: Glycosylated haemoglobin

Comparison	Study	HbA1c [%]	Variation	Variation	Comments
meglitinides vs. placebo	repaglinide vs. placebo: monotherapy	placebo	repaglinide (lowest dose)	repaglinide (highest dose)
	Chaung 1999	+0.28 (est)	‐0.14	‐0.27	no measure of variance
	Goldberg 1998 all	+1.2	‐0.7		no measure of variance
	Goldberg 1998 (previously treated subgroup)	+1.5 (SEM 0.2)	‐0.1 (SEM 0.1)		post hoc analysis
	Jovanovic 2000 all	+1.4 (est)	‐0.7 (est)	‐0.5 (est)	estimated from graph
	Jovanovic 2000 (previously treated)	+1.8	‐0.07	‐0.05	no measure of variance
	Jovanovic 2000 (treatment naive)	+0.1	‐1.3	‐1.8	estimated from graph
	Moses 2001	‐0.15	‐1.4		no measure of variance
	van Gaal 2001	+0.1	‐2.1		no measure of variance
	repaglinide vs. placebo: combination	metformin plus placebo	metformin plus repaglinide
	Moses 1999	‐0.3	‐1.4		no measure of variance
	nateglinide vs. placebo: monotherapy	placebo	nateglinide (120mg)
	Bengel 2004	+0.5	‐0.4		no measure of variance
	Hanefield 2000	+0.07 (95%CI 0.18 to 0.32)	‐0.55 (95%CI ‐0.79 to ‐0.3)		(also reported 30, 60, 180mg)
	Horton 2000	+0.5 (est)	‐0.5 (est)		estimated from graph
	Saloranta 2002	+0.16 (SD 0.05)	‐0.23		no measure of variance (also reported 30, 60mg doses in grpahical form only)
	nateglinide vs. placebo: combination	metformin plus placebo	metformin plus nateglinide (lowset dose)	metformin plus nateglinide (highet dose)
	Horton 2000	‐0.8 (est)	‐1.4 (est)		estimated from graph
	Marre 2002	‐0.05 (est)	‐0.49 (est)	‐0.73 (est)	estimated from graph
meglitinide head to head	repaglinide vs. nateglinide: monotherapy	repaglinide	nateglinide
	Rosenstock 2004	‐1.57 (SEM 0.15)	‐1.04 (SEM 0.14)
	repaglinide vs. nateglinide: combination	metformin plus repaglinide	metformin plus nateglinide
	Raskin 2003	‐1.28 (SEM 0.1)	‐0.67 (SEM 0.1)
meglitinides vs. metformin	repaglinide vs. metformin: monotherapy	metformin	repaglinide
	Derosa 2003	‐0.9 (est)	‐0.8 (est)		estimated from graph
	Moses 1999	‐0.33 (SEM 0.24)	‐0.38 (SEM 0.23)
	repaglinide vs. metformin: combination with insulin	insulin plus metformin	repaglinide plus metformin
	Furlong 2002	‐0.4 (est)	+0.4 (est)		estimated from graph
	nateglinide vs metformin: monotherapy	metformin	nateglinide
	Horton 2000	‐0.8 (est)	‐0.5 (est)		estimated from graph

Appendix 5. Summary of outcomes: Weight

Comparison	Study	Mean change [kg]	Variation	Variation	Comments
meglitinide vs. placebo	repaglinide vs. placebo: monotherapy	placebo	repaglinide (lowest dose)	repaglinide (higest dose)
	Chaung 1999	‐1.5 (SEM 3.0)	"unchanged"	+0.8 (SEM 1.9)
	Goldberg 1998	‐2.0	+0.72		no measure of variance
	van Gaal 2001	‐0.8	+2.1 (SD 0.5)		incomplete measure of variance (for placebo)
	repaglinide versus placebo: combination with metformin	metformin plus placebo	metformin plus repaglinide
	Moses 1999	‐0.86 (SEM 0.51)	+2.41 (SEM 0.05)
	nateglinide vs placebo: combination
	Marre 2002	+0.1 (SEM 0.2)	+0.4 (SEM 0.2)	+1.0 (SEM 0.2)
repaglinide vs. nateglinide	monotherapy	repaglinide	nateglinide
	Rosenstock 2004	+1.8	+0.7		no measure of variance
	combination with metformin	repaglinide	nateglinide
	Raskin 2003	+0.6	‐0.5		no measure of variance
meglitinides vs. metformin	repaglinide vs metformin: monotherapy	metformin	repaglinide
	Derosa 2003	‐2.0 (95%CI ‐0.6 to +4.0)	‐0.8 (95%CI ‐0.43 to +0.28)
	Moses 1999	‐0.86 (SEM 0.51)	+2.98 (SEM 0.49)
	repaglinide vs metformin: combination with insulin	metformin + insulin	repaglinide + insulin
	Furlong 2002	+0.9 (SEM 0.4)	+2.7 (SEM 0.4)

Appendix 6. Risk of bias

Study	Randomisation	Allocation concealm.	Similar at baseline	Eligibility criteria	Point est. reported	ITT analysis	Drop‐outs described	Funding source
Bengel 2004	Yes	Yes	Yes	Yes	No	No	Yes	Industry
Chaung 1999	Unclear	Unclear	No	Yes	No	Yes	Yes	Industry
Derosa 2003	Unclear	Unclear	Yes	Yes	Yes	Yes	Yes	Industry
Furlong 2002	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Industry
Goldberg 1998	Unclear	Unclear	No	Yes	No	Unclear	Yes	Industry
Hanefield 2000	Unclear	Unclear	Yes	Yes	No	Yes	Yes	Industry
Horton 2000	Yes	Unclear	Yes	Yes	No	Yes	No	Industry
Jovanovic 2000	Unclear	Unclear	Yes	Yes	No	Yes	No	Industry
Marre 2002	Yes	Unclear	Yes	Yes	No	Yes	Yes	Industry
Moses 1999	Unclear	Unclear	Yes	Yes	No	Yes	No	Industry
Moses 2001	Unclear	Unclear	Yes	Yes	Yes	Yes	Partially	Industry
Raskin 2003	Unclear	Unclear	Yes	Yes	Yes	No	No	Industry
Rosenstock 2004	Unclear	Unclear	Yes	Yes	Yes	No	Yes	Industry
Saloranta 2002	Unclear	Unclear	Yes	Yes	No	Yes	No	Industry
Van Gaal 2001	Unclear	Unclear	No	Yes	No	Yes	Yes	Industry

Data and analyses

Comparison 1. repaglinide vs. placebo.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 glycosylated haemoglobin (%)	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
1.1 combination therapy (oral)	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2 weight change (kg)	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
2.1 monotherapy (2.0 mg repaglinide)	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.2 combination therapy with metformin	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]

1.1. Analysis.

Comparison 1 repaglinide vs. placebo, Outcome 1 glycosylated haemoglobin (%).

1.2. Analysis.

Comparison 1 repaglinide vs. placebo, Outcome 2 weight change (kg).

Comparison 2. nateglinide vs. placebo.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 weight change (kg)	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
1.1 combination therapy (oral)	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]

2.1. Analysis.

Comparison 2 nateglinide vs. placebo, Outcome 1 weight change (kg).

Comparison 3. repaglinide vs. nateglinide.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 glycosylated haemoglobin (%)	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
1.1 monotherapy	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.2 combination therapy (oral)	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]

3.1. Analysis.

Comparison 3 repaglinide vs. nateglinide, Outcome 1 glycosylated haemoglobin (%).

Comparison 4. repaglinide vs metformin.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 glycosylated haemoglobin (%)	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
1.1 monotherapy	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2 weight change (kg)	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
2.1 monotherapy	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.2 combination therapy (insulin)	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]

4.1. Analysis.

Comparison 4 repaglinide vs metformin, Outcome 1 glycosylated haemoglobin (%).

4.2. Analysis.

Comparison 4 repaglinide vs metformin, Outcome 2 weight change (kg).

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Bengel 2005.

Methods	Double Blind RCT 11 centres 16 weeks treatment
Participants	47 participants Inclusion criteria: Duration of DM: > 3months; Treatment: diet or oral monotherapy HbA1c: 6.5‐9.5% Exclusion criteria: renal disease, severe hyperlipidaemia, ketonuria, diabetic complications, cardiac disease, uncontrolled hypertension Treatment:
Interventions	Treatment 1 placebo (n=14) Treatment 2 nateglinide 120mg (three times per day) (n=33)
Outcomes	HbA1c ‐Yes (no measure of variance) Weight ‐ No Hypoglycaemia ‐ No QoL ‐ No Lipids ‐ Yes (baseline details only) Adverse events ‐No
Notes	Industry funding: Yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	D ‐ Not used

Chuang 1999.

Methods	Double Blind RCT 2 centres 24 weeks treatment
Participants	92 participants Inclusion criteria: Duration of DM: > 3months; Treatment: diet or sulphonyl urea; HbA1c: >7% Exclusion criteria: renal or liver disease, ketonuria, cardiac disease, uncontrolled hypertension
Interventions	Treatment 1 placebo (n=22) Treatment 2 repaglinide 0.5mg (three times per day) (n=36) Treatment 3 repaglinide 2mg (three times per day) (n=34)
Outcomes	HbA1c ‐ graph only Weight ‐ Yes Hypoglycaemia ‐ yes QoL ‐ No Lipids ‐ No Adverse events ‐Yes
Notes	Industry funding: Yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Derosa 2003.

Methods	Open label RCT single centre 52 weeks
Participants	112 participants Inclusion criteria: Duration DM >6 months Treated none HbA1c > 7% Exclusion criteria: cardiovascular disease, abnormal renal function
Interventions	Treatment 1 Repaglinide = 2 x 0.5 mg per day (titration period) increased to 2‐4mg/day (n=56) Treatment 2 Metformin = 2 x 500mg per day (titration period) increased to 1500‐2500 mg/day. (n=56)
Outcomes	HbA1c ‐ graph only Weight ‐ Yes (no measure of variance) Hypoglycaemia ‐ no QoL ‐ No Lipids ‐ Yes (no measure of variance) Adverse events ‐No
Notes	Industry funding: Yes Study would have benefitted from a control arm to isolate whether certain effects were treatment induced
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Furlong 2002.

Methods	Open Label RCT Single centre 13 weeks treatment
Participants	80 participants Inclusion criteria AGE >18 yr old Treated Metformin + bedtime NPH Exclusion criteria history of hypoglycaemia, hepatic or renal failure, cardiovascular disease
Interventions	Treatment 1 4mg (three times per day) repaglinide + NPH insulin (n=39) Treatment 2 850 – 1,000mg metformin + NPH insulin (n=41)
Outcomes	HbA1c ‐ graph only Weight ‐ Yes Hypoglycaemia ‐ Yes QoL ‐ Yes Lipids ‐ No Adverse events ‐ severe only
Notes	Industry funding: Yes Quality of life (well‐being) [mean ± SE] 47.5 (± 1.8) to 48.5 (± 1.8) (NS) (repaglinide); 50.4 (± 2.0) to 50.9 (± 2.1) (metformin) (NS) Patient satisfaction [mean ± SE] 32.5 (± 0.9) to 29.1 (± 1.3) p<0.002 (repaglinide); 32.4 (± 0.8) to 34.1 (± 0.5) p<0.01 (metformin), (p<0.0003 between groups)
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Goldberg 1998.

Methods	Double Blind RCT Multi centre 18 weeks treatment
Participants	100 participants (99 analysed) Inclusion criteria AGE 40‐70yr old Treated OHAs Exclusion criteria abnormal ECG
Interventions	Treatment 1 repaglinide 0.25mg before 3 meals (titrated to 8mg three times per day) (n=67) Treatment 2 placebo (n=33)
Outcomes	HbA1c ‐ incomplete Weight ‐ Yes (no measure of variance) Hypoglycaemia ‐ Yes QoL ‐ No Lipids ‐ Yes Adverse events ‐ Yes
Notes	Industry funding: Yes HbA1c change was adjusted for center
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Hanefeld 2000.

Methods	Double blind RCT Multi centre 12 weeks treatment
Participants	289 participants Inclusion criteria Age 30‐75yr old HbA1c 6.8‐10.5% Treated diet alone Exclusion critieria complications of diabetes
Interventions	Treatment 1 Nateglinide 30 mg (three times per day) (n=51) Treatment 2 Nateglinide 60 mg (three times per day) (n=58) Treatment 3 Nateglinide 120 mg tds (n=63) Treatment 4 Nateglinide 180 mg (three times per day) (n=57) Treatment 5 Placebo (n=60)
Outcomes	HbA1c ‐ Yes (no measure of variance) Weight ‐No Hypoglycaemia ‐ Yes QoL ‐ No Lipids ‐No Adverse events ‐ Yes
Notes	Industry funding: Yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Horton 2000.

Methods	Double Blind RCT Multi centre 24 weeks treatment
Participants	699 participants Inclusion criteria AGE >29yr old BMI 20‐35 kg/m2 Duration of DM >=3monts HbA1c 6.8‐11% Exclusion criteria significant diabetes complications, renal disease
Interventions	Treatment 1 nateglinide 120mg (three times per day) + placebo (n=179) Treatment 2 metformin 500mg (three times per day) + placebo (three times per day)(n=176) Treatment 3 metformin 500mg (three times per day) + nateglinide 120mg (three times per day) (n=172) Treatment 4 placebo 1(three times per day) + placebo 2 (three times per day)(n=172)
Outcomes	HbA1c ‐ graph only Weight ‐ No Hypoglycaemia ‐ Yes QoL ‐ No Lipids ‐ No Adverse events ‐ Yes
Notes	Industry funding: Yes HbA1c change was adjusted for center, treatment by center interaction, treatment by baseline interaction and placebo‐subtracted reduction at end point
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Jovanovic 2000.

Methods	Double Blind RCT Multi centre 24 weeks treatment
Participants	361 participants Inclusion criteria AGE 40‐75yr old HbA1c >6.5% Treated OHAs or diet alone Exclusion criteria Cardiac, hepatic or renal disease
Interventions	Treatment 1 repaglinide 1.0mg before 3 meals (n=140) Treatment 2 repaglinide 4.0mg before 3 meals (n=146) Treatment 3 placebo (three times per day)(n=75)
Outcomes	HbA1c ‐ graph only Weight ‐ No Hypoglycaemia ‐ Yes QoL ‐ No Lipids ‐ No Adverse events ‐ Yes
Notes	Industry funding: Yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Marre 2002.

Methods	Double Blind RCT Multi centre 24 weeks treatment
Participants	467 participants Inclusion criteria: Duration of DM: > 3 months; Treatment: diet or oral monotherapy HbA1c: 6.5‐9.5% Exclusion criteria: renal disease, severe hyprlipidaemia, ketonuria, diabetic complications, cardiac disease, uncontrolled hypertension
Interventions	Treatment 1 metformin 1000mg twice daily + placebo (n=152) Treatment 2 metformin 1000mg twice daily + nateglinide 60mg (three times per day) (n=155) Treatment 3 metformin 1000mg twice daily + nateglinide 120mg (three times per day)(n=160)
Outcomes	HbA1c ‐ graph only Weight ‐ Yes Hypoglycaemia ‐ Yes QoL ‐ No Lipids ‐ Yes Adverse events ‐ Yes
Notes	Industry funding: Yes 2 patients died in the combination therapy group, both from cardiac events but neither considered to be related to the drug treament
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Moses 1999.

Methods	Double Blind RCT Multi centre 3 months maintainance treatment
Participants	83 participants (82 analysed) Inclusion criteria AGE <75yr old BMI >=21 kg/m2 Treated metformin alone for >6 months HbA1c > 7.1% Exclusion criteria renal or hepatic failure, cardiac disease
Interventions	Treatment 1 metformin + placebo (n=27) Treatment 2 repaglinide (initiated at 0.5mg (three times per day) and increased stepwise upto 4mg as required) + placebo (n=29) Treatment 3 metformin + repaglinide (initiated at 0.5mg (three times per day) and increased stepwise upto 4mg as required) (n=27)
Outcomes	HbA1c ‐ Yes Weight ‐ Yes Hypoglycaemia ‐ Yes QoL ‐ No Lipids ‐ No Adverse events ‐ incomplete
Notes	Industry funding: Yes For primary eficacy endpoints changes from baseline measured by ANOVA with treatment as a fixed effect. In addtion associations were adjusted for centre and treatment‐by‐centre effects. Bonferroni’s method used to adjust for multiple comparisons.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Moses 2001.

Methods	Double Blind RCT Multi centre 16 weeks treatment
Participants	408 participants (394 analysed) Inclusion criteria AGE >=40yr old Treated suboptimal treatment on diet alone Exclusion criteria renal or hepatic failure, cardiac disease
Interventions	Treatment 1 repaglinide 0.5mg before meals (upto 4 per day) (n=270) Treatment 2 placebo (n=138)
Outcomes	HbA1c ‐ Yes (no measure of variance) Weight ‐ No Hypoglycaemia ‐ Yes QoL ‐ Yes Lipids ‐ No Adverse events ‐ Yes
Notes	Industry funding: Yes no measure of varibility for HbA1c point estimates
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Raskin 2003.

Methods	Open Labelled RCT Multi centre 16 weeks treatment
Participants	192 participants Inclusion criteria AGE <75yr old BMI >=21 kg/m2, Treated metformin alone for >6 months HbA1c > 7.1%
Interventions	Treatment 1 repaglinide (titrated upto 4mg (three times per day) + metformin 1000mg twice or three times daily (n=96) Treatment 2 nateglinide 120mg (three times per day) + metformin 1000mg twice or three times daily(n=96)
Outcomes	HbA1c ‐ Yes Weight ‐ Yes (no measure of variance) Hypoglycaemia ‐ Yes QoL ‐ No Lipids‐ No Adverse events ‐ incomplete
Notes	Industry funding: Yes Doses: at end of 16 weeks, median final doses of mealtime secretagagues were 5.0 mg/day for repaglinide and 360 mg/day for nateglinide % of patients receiving maximal daily dosage: Repag/met group = 7% vs Nat/met group = 82% In both groups, median dose of metformin was 2,000 mg/day.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Rosenstock 2004.

Methods	Open Label RCT Mulitcentre 16 weeks treatment
Participants	150 participants Inclusion criteria AGE >=18yr old Type 2 DM for >=3 months BMI 24‐42 Treatment with diet and exercise in previous 3 months (could have had oral therapy previously)
Interventions	Treatment 1 Repaglinide 0.5mg three times per day (titrated to 4mg as required) Treatment 2 Nateglinide 60mg three times per day (increased to 120mg as required)
Outcomes	HbA1c ‐ Yes Weight ‐ Yes Hypoglycaemia ‐ Yes QoL ‐ No Lipids‐ No Adverse events ‐ incomplete
Notes	Industry funding: Yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	D ‐ Not used

Saloranta 2002.

Methods	Double blind RCT Multi centre 24 weeks treatment
Participants	675 participants Inclusion criteria Duration DM >6 months Treated diet alone
Interventions	Treatment 1 Nateglinide 30 mg (three times per day) (n=166) Treatment 2 Nateglinide 60 mg (three times per day) (n=175) Treatment 3 Nateglinide 120 mg (three times per day) (n=171) Treatment 4 Placebo (n=163)
Outcomes	HbA1c ‐ graph only Weight ‐ No Hypoglycaemia ‐ Yes QoL ‐ No Lipids ‐ No Adverse events ‐ incomplete
Notes	Industry funding: Yes Extensive exclusion criteria.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Van Gaal 2001.

Methods	Double blind RCT Single centre 10 weeks treatment
Participants	26 participants Inclusion criteria HbA1c <10% Druation DM >1yr Treated diet alone
Interventions	Treatment 1 repaglinide 0.5mg twice daily (titrated to 4mg) (n=13) Treatment 2 placebo twice daily(n=13)
Outcomes	HbA1c ‐ Yes (no measure of variance) Weight ‐ Yes (no measure of variance) Hypoglycaemia ‐ Yes QoL ‐ No Lipids ‐ No Adverse events ‐ incomplete
Notes	Industry funding: Yes Values for HbA1c differ slightly in text and in abstract.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Baron 2002	not an RCT
Barone 2004	only variable was metformin ‐ all received repaglinide
Chen 2002	phase II study ‐ no control group
Cozma 2002	phase 1 study
Damsbo 1999	study duration only 20 days, and compared meal regimens, not drugs
De Luis 2001	case series
Foley 2001
Fujita 2001	no control group
Hasslacher 2001	not an RCT
Hatorp 2002	does not report outcomes relevant to review
Hirschberg 2000	study duration less than 10 weeks
Hollander 2001	study only 8 weeks duration
Horton 2001	not an RCT
Horton 2003	re‐analysis of a trial that is already included
Hoyer 2000	pools data from two previous studies ‐ insufficient details reported
Ishii 2001	study duration only two days
Juhl 2000	phase I study
Kahn 2001	study duration less than 10 weeks
Keilson 2000	study duration less than 10 weeks
Khan 2000
Lawrence 2005	a three year follow‐up ‐ not an RCT
Ligueros‐Saylan 2000	duration of trial less than 10 weeks
Mallows 2000	study duration not stated
Merz 2000	study duration less than 10 weeks
Mori 2001	study only three weeks duration
Nishimura 2002	study duration only four weeks
Owens 2000a	study duration only one week
Owens 2000b	studies only one week duration
Rudovich 2002	study only one week duration
Rudovich 2003	no relevant outcomes measured
Saad 2002	results only presented for eight weeks study duration
Schmitz 2002	study duration four weeks
Schumacher 2001	phase I study
Schwartz 1998	duration of trial only four weeks
Srinivasan 2005	a follow‐up study ‐ not an RCT
Strange 1999	study only four weeks duration
Tankova 2003	duration of trial only two months
Torella 2002	insulin and oral treatment varied between arms, so not possible to make any comparisons regarding the effect of repaglinide.
Walter 2000	duration of trial less than 10 weeks
Wang 2004	restricted participation ‐ only enrolled patients whose fasting plasma glucose fluctuated less than 2 mmol/L during the two‐week washout
Whitelaw 2000	insufficient duration ‐ examined effects of a single dose of drug

Contributions of authors

CORRI BLACK: protocol development, quality assessment of trials, data extraction, data analysis, review development, drafting review

PETER DONNELLY: protocol development, quality assessment of trials, data extraction, data analysis, review development

JONATHAN SHEPHERD: protocol development, quality assessment of trials, data extraction, data analysis, review development

LINDA MCINTYRE: protocol development, quality assessment of trials, data extraction, data analysis, review development,

PAM ROYLE: protocol development, searching for trials, quality assessment of trials, data extraction, data analysis, review development.

SIAN THOMAS: protocol development, quality assessment of trials, data extraction, data analysis, review development.

Sources of support

Internal sources

• Department of Public Health, University of Aberdeen, UK.

External sources

• No sources of support supplied

Declarations of interest

None known.

Edited (no change to conclusions)