Rosiglitazone for type 2 diabetes mellitus

Bernd Richter; Elizabeth Bandeira‐Echtler; Karla Bergerhoff; Christine Clar; Susanne H Ebrahim

doi:10.1002/14651858.CD006063.pub2

. 2007 Jul 18;2007(3):CD006063. doi: 10.1002/14651858.CD006063.pub2

Rosiglitazone for type 2 diabetes mellitus

Bernd Richter ^1,^✉, Elizabeth Bandeira‐Echtler ¹, Karla Bergerhoff ¹, Christine Clar ², Susanne H Ebrahim ³

Editor: Cochrane Metabolic and Endocrine Disorders Group

PMCID: PMC7389529 PMID: 17636824

Abstract

Background

Objectives

To assess the effects of rosiglitazone in the treatment of type 2 diabetes.

Search methods

Studies were obtained from computerised searches of MEDLINE, EMBASE and The Cochrane Library.

Selection criteria

Studies were included if they were randomised controlled trials in adult people with type 2 diabetes mellitus and had a trial duration of at least 24 weeks.

Data collection and analysis

Two authors independently assessed trial quality and extracted data. Pooling of studies by means of fixed‐effects meta‐analysis could be performed for adverse events only.

Main results

Eighteen trials which randomised 3888 people to rosiglitazone treatment were identified. Longest duration of therapy was four years with a median of 26 weeks. Published studies of at least 24 weeks rosiglitazone treatment in people with type 2 diabetes mellitus did not provide evidence that patient‐oriented outcomes like mortality, morbidity, adverse effects, costs and health‐related quality of life are positively influenced by this compound. Metabolic control measured by glycosylated haemoglobin A1c (HbA1c) as a surrogate endpoint did not demonstrate clinically relevant differences to other oral antidiabetic drugs. Occurrence of oedema was significantly raised (OR 2.27, 95% confidence interval (CI) 1.83 to 2.81). The single large RCT (ADOPT ‐ A Diabetes Outcomes Progression Trial) indicated increased cardiovascular risk. New data on raised fracture rates in women reveal extensive action of rosiglitazone in various body tissues.

Authors' conclusions

New studies should focus on patient‐oriented outcomes to clarify the benefit‐risk ratio of rosiglitazone therapy. Safety data and adverse events of all investigations (published and unpublished) should be made available to the public.

Plain language summary

Rosiglitazone for type 2 diabetes mellitus

Diseases of the heart and blood vessels account for approximately 70% of all mortality in people with diabetes. Compared to their non‐diabetic counterparts the relative risk of mortality caused by disorders of the heart and blood vessels is two to three for men and three to four for women with diabetes. Type 2 diabetes is mainly characterised by a reduced ability of the hormone insulin to stimulate glucose uptake in body fat and muscles (insulin resistance) and affects most people suffering from diabetes. Several medications are on the market to treat diabetes, amongst them rosiglitazone as a member of the 'glitazones' reduced risk markers for diseases of the heart and blood vessels. Since the two biggest trials in people with type 2 diabetes showed that improved blood glucose alone is not enough to reduce the risk of the above mentioned diseases we looked for longer‐term studies investigating 24 weeks as a minimum of rosiglitazone treatment on patient‐oriented outcomes. As patient‐oriented outcomes we defined mortality, complications of diabetes, side effects of the medication, health‐related quality of life, costs and metabolic control (lowering of blood glucose to near normal levels).

Eighteen trials randomised 3888 people to rosiglitazone therapy. The longest duration of rosiglitazone treatment was four years, most trials lasted around half a year. Unfortunately, the published studies of at least 24 weeks rosiglitazone treatment in people with type 2 diabetes mellitus did not provide relevant evidence that patient‐oriented outcomes are positively influenced by this agent. The chance of developing oedema was approximately doubled, the risk of cardiovascular diseases increased. The single large randomised controlled trial showed evidence of raised cardiovascular risk after rosiglitazone treatment. Moreover, new safety data show increased numbers of broken bones in women. This finding was published years after approval of this agent by drug regulatory authorities. New ways of exploring drug effects, for example by early long‐term studies in many people, as well as public access to all safety data of published and unpublished investigations have to be established.

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', 'Cochrane Review Group (CRGs)'). For an explanation of methodological terms, see the main glossary in The Cochrane Library.

There are two main types of diabetes mellitus, type 1 (formerly termed insulin‐dependent diabetes mellitus) and type 2 (formerly termed non‐insulin dependent diabetes mellitus):

Type 1 diabetes mellitus

Type 1 diabetes is a chronic disease characterised by hyperglycaemia due to absolute deficiency of insulin secretion which is caused by autoimmune destruction of the pancreatic beta cells. Evidence of autoimmunity is provided by the appearance of autoantibodies prior to the onset of clinical disease. The clinical presentation ranges from mild nonspecific symptoms or no symptoms to coma. Although type 1 diabetes usually develops before 30 years of age, it can occur at any age. At presentation, most patients are thin and have experienced weight loss, polyuria, polydipsia, fatigue, and diabetic ketoacidosis.

Type 2 diabetes mellitus

In type 2 diabetes mellitus, the actions and secretion of insulin are impaired, as opposed to the absolute deficiency of insulin that occurs with type 1 diabetes mellitus. Type 2 diabetes is characterised by two major pathophysiologic defects: (1) insulin resistance, which results in increased hepatic glucose production and decreased peripheral glucose disposal, (2) impaired β‐cell secretory function (Kahn 1997). Insulin resistance is an impaired biological response to the effects of exogenous or endogenous insulin. Insulin resistance in the hepatic and peripheral tissues, particularly skeletal muscle, leads to unrestrained hepatic glucose production and diminished insulin‐stimulated peripheral glucose uptake and utilization (DeFronzo 1992). Insulin secretion by the pancreatic beta cell is initially sufficient to compensate for insulin resistance, thereby maintaining normal blood glucose levels. Hyperinsulinaemia, which accompanies insulin resistance, can maintain sufficiently normal glucose metabolism as long as pancreatic β‐cell function remains normal. However, in patients who may develop type 2 diabetes, insulin secretion eventually fails, leading to hyperglycaemia and clinical diabetes (Warram 1990). Individuals with type 2 diabetes may have few or no classic clinical symptoms (see above) of hyperglycaemia (Ruige 1997). The difficulty in maintaining metabolic control, for example measured by haemoglobin A_1c (HbA_1c) over time may be related to several behavioral factors (for example difficulties with healthy eating, exercise, medication regimens) but primarily reflects the underlying progressive decline in β‐cell function (UKPDS‐16 1995).   Type 2 diabetes has traditionally been treated in a stepwise manner, starting with lifestyle modifications (Armour 2004; Gimenez‐Perez 2001; Moore 2005), exercise (Thomas 2001) and later on pharmacotherapy with oral agents. Several classes of oral agents are available for clinical use. These mainly include insulin secretagogues, drugs that delay the absorption of carbohydrates from the gastrointestinal tract, and insulin sensitizers. Over time, many patients with type 2 diabetes will require insulin therapy (Burt 2005; Misso 2005; Richter 2005; Roberts 2005; Royle 2003; Siebenhofer 2004).   Insulin secretagogues: Currently, the sulphonylureas used are mainly glibenclamide (glyburide), glipizide, chlorpropamide, tolbutamide, and glimepiride. These drugs stimulate pancreatic β‐cell insulin secretion by binding to a sulphonylurea receptor (Lindberg 2002). The short‐acting non‐sulphonylurea insulin secretagogues are repaglinide and nateglinide (Black 2003). These are newer agents that also stimulate insulin secretion by binding to the sulphonylurea receptor.   Alpha‐glucosidase inhibitors: Acarbose and miglitol are α‐glucosidase inhibitors. These drugs slow the absorption of carbohydrates, reducing especially postprandial elevations in plasma glucose levels. They do not significantly lower fasting plasma glucose levels but cause a modest reduction in HbA_1c (Van de Laar 2005).   Insulin sensitizers: Metformin belongs to the biguanides class (Saenz 2005; Salpeter 2003). It might increase insulin sensitivity in the liver by inhibiting hepatic gluconeogenesis and thereby reducing hepatic glucose production. Metformin also seems to increase peripheral insulin sensitivity by enhancing glucose uptake in the muscle. The thiazolidinediones consist of rosiglitazone and pioglitazone. These substances decrease insulin resistance in muscle and adipose tissue by activating the peroxisome proliferator‐activated receptor γ (PPAR−γ) which increases production of proteins involved in glucose uptake. They also decrease hepatic glucose production by improving hepatic insulin sensitivity (Meriden 2004).

Description of the intervention

Type 2 diabetes mellitus can be treated by non‐pharmacological (diet, exercise) and pharmacological means. Insulin, as the natural hormone of the body, might be given as animal (mainly pork or beef) insulin (Richter 2005), genetically constructed 'human' insulin or as insulin‐'analogues' with a modified molecular structure compared to human insulin (Roberts 2005; Siebenhofer 2004). Insulin is currently administered by diabetic people in various ways: Subcutaneous injections, insulin pumps (Misso 2005), and maybe in future by inhalation (Burt 2005; Royle 2003). Oral antidiabetic agents are most often used to treat type 2 diabetes mellitus in its initial stages if lifestyle modifications have failed. The thiazolidinediones rosiglitazone and pioglitazone offer new oral treatment options and affect many tissues and parts of the body. In order to evaluate their effects not only on metabolic control in type 2 diabetes mellitus but also on patient‐oriented outcomes like cardiovascular disease, longer‐term studies of at least 24 weeks continuous intake will be critically appraised in this review.

Adverse effects of the intervention

An increase in bodyweight has been associated with rosiglitazone. Oedema, anaemia and congestive heart failure have been reported in patients receiving rosiglitazone. The patients who appear to be at greatest risk of peripheral oedema, fluid retention and weight gain, congestive heart failure and pulmonary oedema related to rosiglitazone are probably those who use insulin or have New York Heart Association class II, III or IV cardiac status, left‐ventricular dysfunction or renal insufficiency. Some reports of visual impairment in patients taking rosiglitazone were described (Colucciello 2005). Case reports of liver function abnormalities associated with rosiglitazone were documented (Marcy 2004; Menees 2005; Su 2006).

How the intervention might work

Because traditional agents have a limited impact on insulin resistance and β‐cell function, thiazolidinediones may be an appropriate choice especially for combination therapy in patients achieving poor glycaemic control with initial monotherapy. By improving insulin sensitivity, thiazolidinediones may exert beneficial effects on cardiovascular risk factors. The excess cardiovascular risk in type 2 diabetes cannot be attributed to classic risk factors alone (mainly hypertension, hypercholesterolaemia and smoking), but if present, these risk factors are at least as important as in patients without diabetes (Stamler 1993). One explanation for the beneficial effects of thiazolidinediones is their unique mechanism of action as selective and potent inhibitors of PPAR‐γ. PPAR‐γ receptors are present in many tissues like adipose, hepatic and skeletal muscle tissue and control insulin‐responsive genes, which have a wide‐ranging influence. Thiazolidinediones appear to improve markers of inflammation and fibrinolysis, exert beneficial effects on vascular reactivity, improve the lipid profile and fat distribution, and decrease pancreatic β‐cell injury.   Rosiglitazone is a member of the thiazolidinedione group which also encompasses troglitazone (withdrawn due to hepatic toxicity) and pioglitazone. It increases the sensitivity of skeletal muscle, liver and adipose tissue to insulin without directly stimulating insulin secretion from pancreatic ß‐cells, thereby reducing plasma glucose levels and endogenous glucose production (Wagstaff 2002). Differences in the side chain on the main thiazolidine‐structure in comparison to pioglitazone are thought to be responsible for the distinct bioavailability, metabolism and antihyperglycaemic potency of rosiglitazone. Although rosiglitazone appears to be associated with some effects that are not mediated by PPAR‐γ (Yang 2001), binding of rosiglitazone to this receptor seems to be the important component of its mechanism of action. Rosiglitazone has several pharmacodynamic properties which could ameliorate the increased risk of cardiovascular disease in type 2 diabetes mellitus. In clinical studies in patients with type 2 diabetes mellitus, rosiglitazone has been associated with reductions in the levels of small dense low density lipoprotein‐cholesterol (LDL‐C) ‐ despite overall increases in total LDL‐C ‐ and increases in the levels of high density lipoprotein‐cholesterol (HDL‐C). Diastolic and systolic blood pressure are thought to be decreased after rosiglitazone treatment. Some other surrogate parameters indicating especially cardiovascular risk were reported to be positively influenced by rosiglitazone therapy.

Why it is important to do this review

Diabetes has long been recognised as a strong, independent risk factor for cardiovascular disease, a problem which accounts for approximately 70% of all mortality in people with diabetes (Laakso 1999). Prospective studies show that compared to their non‐diabetic counterparts, the relative risk of cardiovascular mortality for men with diabetes is two to three and for women with diabetes is three to four (Manson 1991; Stamler 1993). The increased cardiovascular risk associated with diabetes is reflected in the observation that middle‐aged individuals with diabetes have mortality and morbidity risks that are similar to non‐diabetic individuals who have already suffered a cardiovascular event (Haffner 1998).   Both epidemiological and prospective data have demonstrated that treatment of hyperglycaemia in type 2 diabetes mellitus is effective in reducing the risk of microvascular disease (for example diabetic retinopathy) but is less potent in reducing that of myocardial infarction, stroke and peripheral vascular disease. Treatment of other cardiovascular risk factors, although by definition less prevalent than hyperglycaemia, appears to be more effective in preventing macrovascular disease than treatment of hyperglycaemia. The University Group Diabetes Program (UGDP) study was the first published long‐term investigation of people with type 2 diabetes indicating no reduction of cardiovascular endpoints through improved metabolic control but raised cardiovascular mortality after tolbutamide treatment (UGDP 1982). The study of Ohkubo et al. which included relatively lean Japanese patients with type 2 diabetes, was the first to demonstrate prevention of microvascular complications by intensive glucose control in patients with type 2 diabetes (Ohkubo 1995). This study did not address the question of whether good glycaemic control retards the progression of macrovascular disease. The United Kingdom Prospective Diabetes Study (UKPDS) tested mainly whether intensive glucose control with either a sulphonylurea or insulin influences the risk of micro‐ and macrovascular complications compared with conventional treatment (UKPDS‐33 1998). The 10‐year results of the UKPDS evaluated drug treatment in non obese and obese participants with newly diagnosed type 2 diabetes who were referred to hospital clinics. Over 10 years, HbA_1c was 7.0% in the intensive group compared with 7.9% in the conventional group. The 0.9% difference in HbA_1c between the intensive and conventional groups over 10 years was smaller than the 1.9% difference (9.0% and 7.1%) in HbA_1c in the Diabetes Control and Complications Trial (DCCT). The DCCT studied younger patients with type 1 diabetes and assessed the effects of intensive versus conventional insulin therapy on the incidence of microvascular complications of diabetes (retinopathy, nephropathy, neuropathy) over a mean follow‐up of 6.5 years (DCCT 1993). The risk of retinopathy, for example, was statistically significant reduced by intensive insulin therapy with a number needed to treat (NNT) to benefit of six (six type 1 diabetic patients need to be treated by intensive in comparison to conventional insulin therapy over 6.5 years to avoid one additional patient to develop diabetic retinopathy). The UKPDS had a factorial design meaning that another study investigating intensive versus regular blood pressure control (HDS 1993; UKPDS‐38 1998) was imbedded in the main study. Intensive versus conventional glucose control did not result in a statistically significant difference in diabetes related mortality or macrovascular disease endpoints but reduced the relative risk in the 'any diabetes related aggregate endpoint' (Freemantle 2003). Most of this benefit was due to a reduction in microvascular endpoints including the incidence of retinal photocoagulation, which was assessed by ophthalmologists independent of the study. In the UKPDS, the NNT to prevent one patient developing any of the single endpoints over 10 years was 20 (95% confidence interval (CI) 10 to 500) patients (UKPDS‐33 1998). In contrast to these results, publication of the UKPDS‐34, which focused on obese patients with newly diagnosed type 2 diabetes, found several clinically important differences in macrovascular disease endpoints with 10 years of treatment with metformin (UKPDS‐34 1998). In particular, the absolute risk reduction for the aggregate endpoints was more than 10% and for overall mortality was 7%, giving NNTs of 10 and 14, respectively, over 10 years (McCormack 2003).   The UKPDS was criticised on several grounds especially emphasising hidden biases in interpreting the results of this randomised controlled trial (Ewart 2001; McCormack 2003; Nathan 1998). Stratton et al. in their UKPDS‐35 publication are often cited, who tried to determine the relation between exposure to glycaemia over time and the risk of macrovascular or microvascular complications in the UKPDS patients (Stratton 2000). This publication is an epidemiological re‐interpretation of UKPDS data proclaiming that with each 1% reduction in mean HbA_1c, reductions in risk of 21% for deaths related to diabetes and 14% for myocardial infarction could be observed. The RCT itself, though, did not show significant differences in this respect. Moreover, the UKPDS‐38, investigating tight versus less tight blood pressure control with the use of an angiotensin converting enzyme inhibitor captopril or a β‐blocker atenolol as main treatment, showed relative risk reductions (in the group assigned to tight control compared with that assigned to less tight control) of 24% in diabetes related endpoints, 32% in deaths related to diabetes, 44% in strokes and 37% in microvascular endpoints (UKPDS‐38 1998). Due to the factorial design of the UKPDS with two interventions (improvement in metabolic and blood pressure control) aiming at the same outcomes, a fair interpretation of the data needs investigation of the interaction between the two main treatment strategies (McAlister 2003; Montgomery 2003). UKPDS data should be available to the scientific public to evaluate, among other things, the importance of the individual contribution of improved glucose versus blood pressure control in type 2 diabetes mellitus. Unfortunately, until now this has not happened.   Therefore, any new compound in the treatment of type 2 diabetes mellitus, like rosiglitazone, should not only be evaluated with regards to surrogate outcomes (for example reductions in fasting plasma glucose or HbA_1c) but information is urgently needed for the influence of any antidiabetic agent especially on cardiovascular endpoints, which is the greatest problem in the therapy of type 2 diabetes mellitus. Quite a number of health technology assessment reports, (narrative) reviews, systematic reviews and meta‐analyses analysed interventions with rosiglitazone in diabetes (Bloomgarden 2005; Boucher 2002; Boucher 2003; Chiquette 2004; Cox 2004; Czoski‐Murray 2004; Kreider 2002; Lebovitz 2002; Malinowski 2000; Mukhtar 2005; NICE 2000; NICE 2003; NICE 2003b; Wagstaff 2002; Wellington 2005). All of them either suffer from methodological problems like insufficient quality assessment of primary studies, focus on surrogate outcomes or are out‐of‐date. This systematic review tries to collate all available data from RCTs of rosiglitazone treatment and evaluates how many studies investigated patient‐oriented outcomes like mortality, cardiovascular endpoints, adverse events and health‐related quality of life.

A Cochrane review on the effects of pioglitazone treatment has already been published (Richter 2006). For changes to the published protocol see Appendix 12.

As this review contributes to the ongoing critical appraisal of RCTs investigating the risk‐benefit ratio of thiazolidinedione use by the German Institute for Quality and Efficiency in Health Care ('Institut fuer Qualitaet und Wirtschaftlichkeit im Gesundheitswesen ‐ IQWiG), additional data (for example raw data from pharmaceutical companies often provided to IQWiG) of relevance might be included in further updates.

Objectives

To assess the effects of rosiglitazone in the treatment of type 2 diabetes.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials.

Types of participants

Adult persons (18 years or older) with type 2 diabetes mellitus. To be consistent with changes in classification and diagnostic criteria of type 2 diabetes mellitus through the years, the diagnosis should have been established using the standard criteria valid at the time of the beginning of the trial (for example ADA 1997; ADA 1999; WHO 1980; WHO 1985; WHO 1998). Ideally, diagnostic criteria should have been described. If necessary, authors' definition of type 2 diabetes mellitus was used. It was planned to subject diagnostic criteria to a sensitivity analysis.

Types of interventions

Therapy with rosiglitazone for a minimum of 24 weeks. The following comparisons were acceptable for evaluation:

rosiglitazone versus placebo;
rosiglitazone versus another oral antidiabetic medication (meglitinide analogues, metformin, pioglitazone, sulphonylureas);
rosiglitazone in combination with an oral antidiabetic medication or insulin versus a combination of an oral antidiabetic medication or insulin (agents and treatment schemes had to be identical).

Excluded interventions

Combination therapies consisting of different compounds in the treatment arms (for example rosiglitazone plus metformin versus uptitration of metformin or rosiglitazone plus gliclazide versus gliclazide). Another Cochrane review will investigate rosiglitazone‐metformin combination therapies including different treatment regimens of these compounds. Furthermore, dipeptidyl peptidase‐4 (DPP‐4) inhibitors for type 2 diabetes mellitus are excluded, since these are the topic of another Cochrane review (Richter 2007), as well as glucagon‐like peptide analogues for type 2 diabetes mellitus (Cochrane review, Snaith 2007).

Types of outcome measures

Primary outcomes

mortality (all‐cause mortality; diabetes related mortality (death from myocardial infarction, stroke, peripheral vascular disease, renal disease, hyper‐ or hypoglycaemia or sudden death));
morbidity (all‐cause morbidity as well as diabetes and cardiovascular related morbidity, for example angina pectoris, myocardial infarction, stroke, peripheral vascular disease, neuropathy, retinopathy, nephropathy, erectile dysfunction, amputation);
adverse events (for example hypoglycaemia, congestive heart failure, oedema).

Secondary outcomes

health‐related quality of life (using a validated instrument);
costs;
metabolic control as measured by glycosylated haemoglobin A1c (HbA1c).

Covariates, effect modifiers and confounders

compliance;
co‐morbidities (for example myocardial infarction, stroke);
co‐medication (for example antihypertensive drugs, aspirin);
age.

Timing of outcome measurement

Outcomes were assessed in the medium (24 weeks to less than 12 months of treatment) and long term (12 months or more of treatment).

Search methods for identification of studies

Electronic searches

We used the following sources for the identification of trials:

The Cochrane Library (issue 1, 2007);
MEDLINE ‐ OVID interface (until April 2007);
EMBASE ‐ OVID interface (until April 2007).

We also searched databases of ongoing trials: Current Controlled Trials (www.controlled‐trials.com ‐ with links to other databases of ongoing trials).   The described search strategy (see for a detailed search strategy Appendix 1) was used for MEDLINE. For use with EMBASE and The Cochrane Library this strategy was slightly adapted.

Additional key words of relevance were not identified during any of the electronic or other searches. If this had been the case, electronic search strategies would have been modified to incorporate these terms. Studies published in any language were included.

Searching other resources

We tried to identify additional studies by searching the reference lists of included trials and (systematic) reviews, meta‐analyses and health technology assessment reports identified.

Data collection and analysis

Selection of studies

To determine the studies to be assessed further, two authors (BR in combination with all the other authors) independently scanned the abstract or titles, or both sections of every record retrieved. All potentially relevant articles were investigated as full text. Interrater agreement for study selection was measured using the kappa statistic (Cohen 1960). Where differences in opinion existed, they were resolved by a third party (other authors). If resolving disagreement was not possible, the article would have been added to those 'awaiting assessment' and authors would have been contacted for clarification. An adapted QUOROM (quality of reporting of meta‐analyses) flow‐chart of study selection is attached (Moher 1999).

Dealing with duplicate publications

In the case of duplicate publications and companion papers of a primary study, we tried to maximise yield of information by simultaneous evaluation of all available data. In cases of doubt, the original publication (usually but not always the oldest version) obtained priority.

Data extraction and management

For studies that fulfilled inclusion criteria, two authors (BR in combination with all the other authors) independently abstracted relevant population and intervention characteristics using standard data extraction templates (for details see Characteristics of included studies and Appendix 2 to Appendix 16) with any disagreements to be resolved by discussion, or if required by a third reviewer. The data extraction form was pilot tested prior to use and modified. Any relevant missing information on the trial would have been sought from the original author(s) of the article, if required.

Assessment of risk of bias in included studies

Two authors (BR in combination with all the other authors) assessed each trial independently. Possible disagreement was resolved by consensus, or with consultation of a third reviewer in case of disagreement. We planned to explore the influence of individual quality criteria in a sensitivity analysis (see under 'sensitivity analyses'). Interrater agreement for key quality indicators (concealment of allocation, blinding, attrition rates) was planned to be calculated using the kappa statistic (Cohen 1960). In cases of disagreement, the rest of the group was consulted and a judgement was made based on consensus.

Measures of treatment effect

Dichotomous data

Dichotomous outcomes (for example stroke yes/no) were planned to be expressed as odds ratios (OR) or relative risks (RR) with 95% confidence intervals (CI).

Continuous data

Continuous outcomes (for example metabolic control as measured by glycosylated haemoglobin A1c (HbA1c) were planned to be expressed, if possible, as mean differences with 95% CI.

Time‐to‐event data

Time‐to‐event outcomes (for example time until death) were planned to be expressed as hazard ratios (HR) with 95% CI.

Unit of analysis issues

Different units of analysis (for example OR and RR) were planned to be subjected to a sensitivity analysis.

Dealing with missing data

Relevant missing data were planned to be obtained from authors. Evaluation of important numerical data such as screened, eligible and randomised patients as well as intention‐to‐treat and per‐protocol population was carefully performed. Drop‐outs, misses to follow‐up and withdrawn study participants were investigated. Issues of last‐observation‐carried‐forward (LOCF) were critically appraised and compared to specification of primary outcome parameters and power calculation.

Assessment of heterogeneity

In the event of substantial clinical or methodological or statistical heterogeneity, study results were not planned to be combined in a meta‐analysis. Heterogeneity was identified by visual inspection of the forest plots, by using a standard χ²‐test and a significance level of α = 0.1, in view of the low power of such tests. Quantification of heterogeneity was also examined with I², ranging from 0% to 100% including its 95% confidence interval (Higgins 2002). I² demonstrates the percentage of total variation across studies due to heterogeneity and was used to judge the consistency of evidence. I² values of 50% and more indicate a substantial level of heterogeneity (Higgins 2003). When heterogeneity was found, we attempted to determine potential reasons for it by examining individual study characteristics and those of subgroups of the main body of evidence.

Assessment of reporting biases

Funnel plots were planned to be used in exploratory data analyses to assess for the potential existence of small study bias. There are a number of explanations for the asymmetry of a funnel plot, including true heterogeneity of effect with respect to study size, poor methodological design of small studies and publication bias (Sterne 2001). Thus, this exploratory data tool may be misleading (Lau 2006; Tang 2000; Thornton 2000) and we did not place undue emphasis on this tool.

Data synthesis

Data were planned to be summarised statistically if they were available, sufficiently similar and of sufficient quality. Statistical analysis was planned to be performed according to the statistical guidelines referenced in the newest version of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2005). Pooled results were planned to be analysed using primarily a fixed‐effect model. Meta‐regression was planned to be performed using Stata/SE (version 8, Stata Corporation, Texas USA) to determine whether various study‐level characteristics (for example follow‐up interval, duration of the intervention, total attrition, year of publication) affected the between‐group changes in primary outcomes. We planned to examine interaction terms for all models.

Subgroup analysis and investigation of heterogeneity

Subgroup analyses were planned to be performed only if one of the primary outcome parameters demonstrated statistically significant differences between treatment groups. The following subgroup analyses were planned:

gender (female versus male);
age (depending on data but especially older versus younger patients);
patients with or without co‐morbidities (for example heart attack, stroke, peripheral vascular disease);
patients with or without co‐medication (for example antihypertensive drugs, statins, aspirin).

Subgroup analyses were planned to be mainly used to explore clinical or methodological or statistical heterogeneity.

Sensitivity analysis

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

repeating the analysis excluding unpublished studies;
repeating the analysis taking account of study quality, as specified above;
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: diagnostic criteria, language of publication, source of funding (industry versus other), country.

The robustness of the results was also planned to be tested by repeating the analysis using different measures of effects size (risk difference, odds ratio etc.) and different statistical models (fixed and random‐effects models).

Results

Description of studies

Results of the search

The initial search identified 6058 records with eight additional publications from reference lists; from these, 40 full papers were singled out for further examination. The other studies were excluded on the basis of their abstracts or titles because they were not relevant to the question under study (see Figure 1 for details of the amended QUOROM (quality of reporting of meta‐analyses) statement). After screening the full text of the selected papers, 32 publications describing 18 studies finally met the inclusion criteria.

QUOROM (quality of reporting of meta‐analyses) flow‐chart of study selection

Most studies of at least 24 weeks rosiglitazone treatment were published in the years 2005 to 2007 (10 trials), with the first study was published in 2001.

Assessment of publication bias inter‐rater agreement

Inter‐rater agreement for study selection, that is qualifying a study as 'included' or 'potentially relevant' was 95%.

Included studies

For details see Characteristics of included studies.

Interventions

Comparisons

Ten of the 18 included publications investigated rosiglitazone monotherapy versus another monotherapy (12 monotherapy arms), eight publications evaluated the combination of rosiglitazone with another glucose‐lowering intervention versus a comparable combination.

Monotherapy

Five study arms compared rosiglitazone to placebo.
Three study arms investigated rosiglitazone versus metformin, two versus glyburide and one each versus repaglinide or pioglitazone.

Combination therapy

Eight publications investigated rosiglitazone combination therapy versus a similar combination with another compound.
Two studies evaluated glimepiride and metformin combination, and one glibenclamide plus metformin, pioglitazone plus metformin or pioglitazone plus glimepiride, respectively.
Three publications reported on triple combination comparisons (sulphonylurea or glimepiride plus metformin plus insulin).

Number of study centres

Number of study centres ranged between one and 488, the multicentre design was dominant with a median of 31 study centres. Seven trials involved a substantial number of more than 40 study centres (Garber 2006; Goldberg 2005; Hanefeld 2007; Kahn 2006; Lebovitz 2001; Phillips 2001; Rosenstock 2006b ).

Country and location

Ten studies were performed in the USA and Canada , six in Europe, one in Latin America, and one in China, Korea an Taiwan, respectively (summarising to more than 18 studies due to multinational designs).

Setting

Eight publications presented some details about the study setting, like recruitment of participants.

Treatment before study

If stated, most studies specified that pharmacotherapy like sulphonylureas, metformin or both were used by participants before entering the study. In two studies participants were treated by diet, exercise or both, only (Hällsten 2002; Kahn 2006).

Methods

Duration of the intervention

Median treatment duration lasted 26 weeks, the longest trial had a median duration of four years (Kahn 2006).

Duration of follow‐up

Treatment duration and follow‐up were identical in all studies, no post‐intervention follow‐up was reported.

Run‐in period

Ten studies described run‐in periods. These usually lasted four weeks where previous antidiabetic medication was stopped, titration of new medication started or a placebo intervention initiated.

Language of publication

All included studies were published in English.

Participants

Who participated

Study participants were mainly white individuals with type 2 diabetes mellitus, in two studies the entire cohort was pharmaco‐naive (that is, people treated with diet only ‐ Hällsten 2002; Kahn 2006).

Inclusion criteria

Investigators specified various inclusion criteria, such as diet non‐responders, sulphonylureas or metformin, or both failures and certain glycosylated haemoglobin A1c (HbA1c) levels.

Exclusion criteria

Investigators specified various exclusion criteria. Nine of 18 included studies stipulated specific exclusion criteria for the severity of congestive heart failure (NYHA (New York Heart Association) classification): Seven studies mentioned NYHA class III or IV and two studies NYHA I or above (including the biggest trial, the ADOPT (A Diabetes Outcomes Progression Trial) study ‐ Kahn 2006).

Diagnostic criteria

Twelve studies provided some details of diagnostic criteria for type 2 diabetes mellitus.

Co‐morbidities

Only four studies presented data on co‐morbidities (Goldberg 2005; Jung 2005; Stocker 2007; Sutton 2002 ).

Co‐medications

Six of the 18 included studies reported co‐medications, either glucose‐lowering drugs or medication for other disorders, or both (Derosa 2004; Derosa 2006b; Jung 2005; Kahn 2006; Ko 2006; Stocker 2007).

Outcomes

Primary outcomes

Most studies investigated HbA1c and lipid parameters (such as total cholesterol, high‐density and low‐density lipoprotein cholesterol, triglycerides) as primary endpoints.

Secondary and additional outcomes

Most studies evaluated lipid parameters, fasting and non‐fasting plasma glucose, adverse events, insulin, HbA1c, C‐peptide and indicators for insulin resistance as secondary outcomes.

Missing data

For this version of the review no author was contacted for additional data. As this review contributes to the ongoing critical appraisal of RCTs investigating the risk‐benefit ratio of thiazolidinedione use by the German Institute for Quality and Efficiency in Health Care ('Institut fuer Qualitaet und Wirtschaftlichkeit im Gesundheitswesen ‐ IQWiG), additional data (for example raw data from pharmaceutical companies often provided to IQWiG) of relevance might be included in further updates.

Excluded studies

Twenty‐two publications had to be excluded after careful evaluation of the full publication. Main reasons for exclusion were trial duration of less than 24 weeks or non comparable treatment regimens (for details see 'Characteristics of excluded studies').

Risk of bias in included studies

For details on methodological quality of included studies see Appendix 13 to Appendix 16.

Overview

All included trials were of a parallel study design. No crossover studies or factorial trials fulfilling the inclusion criteria were detected. Two of the 18 included studies primarily specified a non‐inferiority or equivalence design (Hanefeld 2007; Sutton 2002) with both trials specifying a 95% confidence interval (CI) of equivalence. The other studies investigated superiority or inferiority of rosiglitazone versus comparator compounds.   Interrater agreement for the key quality indicators randomisation, concealment of allocation and blinding was 95%.

Allocation

All included studies were randomised controlled clinical trials of parallel design and randomised individuals. The method of randomisation was somewhat specified in five studies (Derosa 2004; Derosa 2006b; Goldberg 2005; Kahn 2006; Stocker 2007), four studies specified a randomisation ratio other than 1:1, that is randomisation numbers were a‐priori not equal between rosiglitazone and comparator drugs (Hanefeld 2007; Kahn 2006; Phillips 2001; Raskin 2004).   Four studies particularized concealment of allocation (Derosa 2004; Derosa 2006a; Kahn 2006; Stocker 2007).

Blinding

Eleven studies had a double‐blind, five studies an open‐label design and two publications (Jung 2005; Ovalle 2004) did not lay down information on blinding. No publication reported checking of blinding conditions.

Incomplete outcome data

Screened and randomised patients

Nine studies or 50% of publications reported numbers of screened patients (Garber 2006; Goldberg 2005; Hanefeld 2007; Kahn 2006; Lebovitz 2001; Phillips 2001; Rosenstock 2006b; Stocker 2007; Yang 2002), ranging from 120 to 6676 screened patients with a median of 643 participants.   Altogether approximately 3888 participants were randomised to rosiglitazone treatment and 4544 to control therapy, summing up to 8432 individuals taking part in the included studies. A single study contributed 52% of randomised individuals (Kahn 2006).

Discontinuing participants and attrition rates

Six studies described discontinuing participants and provided some details about the reasons for terminating the trial (Goldberg 2005; Hanefeld 2007; Ko 2006; Rosenstock 2006b; Stocker 2007; Sutton 2002).   Discontinuation rates in the rosiglitazone arms varied between five and 40% (between four and 44% in control groups), with five studies reporting high drop‐out rates above 20% (Hanefeld 2007; Kahn 2006; Lebovitz 2001; Raskin 2004; Sutton 2002).   Discontinuation rates between intervention and control groups were dissimilar in six studies (Garber 2006; Hanefeld 2007; Jung 2005; Rosenstock 2006b; Stocker 2007; Sutton 2002). Five studies did not report details on attrition rates.

Intention‐to‐treat and per‐protocol analyses, missing data

Thirteen of the 18 included studies reported an intention‐to‐treat analysis, three trials a per‐protocol evaluation and two both (Goldberg 2005; Sutton 2002). Intention‐to‐treat was clearly defined in 11 studies.   Six studies used the last‐observation‐carried‐forward (LOCF) imputation method for missing data (Hanefeld 2007; Lebovitz 2001; Phillips 2001; Rosenstock 2006b; Sutton 2002). For example, a study of 12 months duration could extrapolate missing glycosylated haemoglobin A1c (HbA1c) values for randomised patients and declare these as endpoints, if the first post‐randomisation HbA1c value (for example after three months) was available. Two studies used other methods for imputation. A clear definition of the LOCF population was provided by one study, only (Lebovitz 2001).

Other potential sources of bias

Definition of primary endpoint and secondary endpoints

Ten studies clearly defined primary endpoints in association with power calculations, mostly one outcome, with one study presenting more than one parameter (Derosa 2006b).   The number of secondary endpoints, if stated as such, varied between two and 16. The total number of detailed endpoints in the included studies ranged from seven to 17 with a mean of seven endpoints. Only four studies adjusted for multiple outcomes, repeated measurements, or both (Derosa 2004; Derosa 2006b; Ko 2006; Phillips 2001).

Power calculation

Seven studies showed details of power calculation, the calculated number of participants per group ranged from 40 to approximately 1394.

Compliance measures

Five of the 18 included studies tried to investigate patients' compliance with the recommended treatments (Derosa 2004; Derosa 2006a; Derosa 2006b; Hällsten 2002; Stocker 2007).

Funding

Ten studies reported commercial funding, six publications did not indicate possible funding sources (Derosa 2004; Derosa 2006b; Ko 2006; Lebovitz 2001; Phillips 2001; Sutton 2002).

Publication status

Sixteen studies were published in peer review journals, none was circulated as a journal supplement.

Effects of interventions

Baseline characteristics

For details of baseline characteristics see Appendix 2, Appendix 3, Appendix 4 and Appendix 5.

Six studies demonstrated clinically relevant differences between intervention and control groups, for example gender ratio (Garber 2006; Kahn 2006; Ko 2006; Raskin 2004; Rosenstock 2006b; Stocker 2007). More men then women participated in the studies, in the rosiglitazone arms women's involvement ranged between 25% and 57%.   The mean age of patients randomised to rosiglitazone treatment encompassed 47 to 61 years. Studies in established type 2 diabetes patients and providing disease information (N = 13 ) showed a diabetes duration of four to 9 years. The main ethnic group participating in the trials consisted of white people, a few studies included other ethnic populations as well.   Pharmaco‐naive patients usually constituted a minor part of the study participants, but two studies exclusively investigated this group (Hällsten 2002; Kahn 2006), including the largest trial (the ADOPT (A Diabetes Outcomes Progression Trial) study ‐ Kahn 2006).   Most study participants with type 2 diabetes mellitus were also overweight or obese, the mean body mass indices (BMI) in patients randomised to rosiglitazone therapy ranged between 23.3 and 33.6 kg/m² (mean BMI of 29 kg/m²).   Baseline metabolic control as measured by mean glycosylated haemoglobin A1c (HbA1c) varied in the rosiglitazone arms between 6.8% and 9.5%, with a mean of 8.8%.

Primary outcomes

For details of primary outcomes see Appendix 10.

Mortality

No study included mortality as a primary or secondary endpoint.   The ADOPT trial investigated rosiglitazone, metformin and glyburide (glibenclamide) as initial treatment for recently diagnosed type 2 diabetes mellitus by means of a double‐blind RCT involving more than 4000 patients (Kahn 2006). Eligible participants were between 30 and 75 years, with fasting plasma glucose levels between 126 to 180 mg/dl (7.0 to 10.0 mmol/L) and were treated by life style management only. The primary outcome was the time from randomisation to treatment failure. Treatment failure was defined as confirmed hyperglycaemia, that is fasting plasma glucose levels greater than 180 mg/dl on consecutive testing or according to the decision of an independent adjudication committee. Median duration of treatment was 4.0 years for rosiglitazone and metformin and 3.3 years for glyburide. At five years, when around 20% of the original cohort was being followed, the reported cumulative incidence of treatment failure was 15% in the rosiglitazone group and 21%/34% in the metformin/glyburide group, respectively. The mean HbA1c level at four years compared to max. 2g/day metformin and max. 15 mg/day glyburide, was 0.1% and 0.4% less after max. 8 mg/day rosiglitazone therapy. Attrition rates were high in the ADOPT study: 37%, 38% and 44% did not finish the study in the rosiglitazone, metformin and glyburide groups. Mortality data were reported in Table 2 of the publication ('Adverse events, laboratory assessment, concomitant use of cardiovascular drugs, hospitalization, and death'): All‐cause mortality was 34/1456 (2.3%) in the rosiglitazone group, 31/1454 (2.1%) in the metformin group and 31/1441 (2.2%) in the glyburide group.

Morbidity

No study included morbidity like diabetic complications as a primary or secondary endpoint. Eight studies made some statement about the number of participants who died during the course of the trial (Derosa 2004; Derosa 2006b; Goldberg 2005; Hällsten 2002; Hanefeld 2007; Kahn 2006; Stocker 2007; Yang 2002).   The ADOPT trial (Kahn 2006) reported some data in Table 2 of the publication ('Adverse events, laboratory assessment, concomitant use of cardiovascular drugs, hospitalization, and death'):

Hospitalisation for any cause was comparable between the rosiglitazone, metformin and glyburide groups (11.6%, 11.8% and 10.4% of patients, respectively).

Cardiovascular disease [no (%)] of serious / total events was increased in the rosiglitazone compared to the glyburide group:

rosiglitazone 49 (3.4) / 62 (4.3)
metformin 46 (3.2) / 58 (4.0)
glyburide 26 (1.8) / 41 (2.8)

Investigator reported total events [no (%)] of congestive heart failure happened more often in the rosiglitazone compared to the glyburide group:

rosiglitazone 22/1456 (1.5)
metformin 19/1454 (1.3)
glyburide 9/1441 ( 0.6)

Peripheral vascular disease [no (%)] of serious / total events data were as follows:

rosiglitazone 7 (0.5) / 36 (2.5)
metformin 6 (0.4) / 27 (1.9)
glyburide 4 (0.3) / 31 (2.2)

Adverse events

For details of adverse events see Appendix 6, Appendix 7, Appendix 8 and Appendix 9.

The percentage of overall adverse events was comparable between the intervention and control groups, serious adverse events appeared to happen somewhat more often after rosiglitazone treatment (median of 6% versus 4% in the control groups). Median discontinuation rate following rosiglitazone administration was also higher than after control therapy (median of 7% versus 4%).   Three studies evaluated and reported a more pronounced (apparently dose‐related) decrease of haemoglobin after rosiglitazone intake in comparison to other active compounds or placebo. Haemoglobin reductions ranged between 0.5 and 1.0 g/dl.   Eleven studies evaluated body weight and observed an increase up to 5.0 kg after rosiglitazone treatment, four studies described changes in body mass index up to a rise of 1.5 kg/m².   Seven of the 18 included studies showed data on hypoglycaemic episodes: Compared to active monotherapy control rosiglitazone treatment resulted in somewhat lower rates of hypoglycaemia, especially when compared to sulphonylureas. Severe hypoglycaemic events were rarely reported.

Data on the specific adverse event "oedema" were available in nine of 18 studies. Overall, 4739 participants provided information on the occurrence of oedema. The total number of events was 287 in the rosiglitazone and 134 in the control groups. Pooling of the nine studies by means of fixed‐effect meta‐analysis revealed an odds ratio of 2.27 (95% confidence interval (CI) 1.83 to 2.81, P < 0.00001). The test for heterogeneity indicated an I²‐value of 53.4%. The use of a random‐effects model resulted in an odds ratio of 4.62 (95% CI 2.28 to 9.38, P < 0.00001). The robustness of this result was tested by repeating the analysis using the risk ratio as a different measure of effect size, demonstrating a relative risk of 2.10 (95% CI 1.72 to 2.55) for the fixed‐effect model. Since oedema event rates in most studies were below 10%, application of the odds ratio appeared to be the more valid parameter.   We repeated the analysis excluding the large ADOPT study which had a weight of 89.4% in the fixed‐effect model. The odds ratio in the fixed‐effect model now was 6.04 (95% CI 3.31 to 11.02, P < 0.00001) and 6.79 (95% CI 3.76 to 12.25, P < 0.00001). Heterogeneity decreased to an I² of 0%. The point estimate for the ADOPT study only was 1.76 (95% CI 1.39 to 2.22).

Furthermore, the ADOPT study provided additional data on fracture rates:

Men [n] fractures(%)

rosiglitazone 32 (3.95)
metformin 29 (3.36)
glyburide 28 (3.35)

Women [n] fractures(%)

Total

rosiglitazone 60 (9.30)
metformin 30 (5.08)
glyburide 21 (3.47)

Lower limb

rosiglitazone 36 (5.58)
metformin 18 (3.05)
glyburide 8 (1.32)

Upper limb

rosiglitazone 22 (3.41)
metformin 10 (1.69)
glyburide 9 (1.49)

Spinal

rosiglitazone 1 (0.16)
metformin 1 (0.17)
glyburide 1 (0.17)

Secondary outcomes

For details of secondary outcomes see Appendix 11.

Health‐related quality of life

No study investigated health‐related quality of life.

Costs

Only one study reported some data on costs of rosiglitazone therapy (Rosenstock 2006b). Rosiglitazone 8 mg/day plus 2 g/day metformin plus sulphonylurea agents were compared to the combination therapy 10 units/day insulin glargine plus 2 g/day metformin plus sulphonylurea agents. Overall, the estimated mean total cost of glycaemic control over 24 weeks was $235 lower among participants treated with insulin glargine ($1368) compared with rosiglitazone ($1603).

Metabolic control as measured by glycosylated haemoglobin A1c (HbA1c)

Active glucose‐lowering compounds like metformin, glibenclamide, or glimepiride resulted in similar reductions of HbA1c compared to rosiglitazone treatment.

Heterogeneity

Only adverse events (oedema) as one of our primary outcomes could be subjected to meta‐analysis. Heterogeneity as indicated by I² was 53.4% but could be significantly reduced after elimination of the biggest trial by Kahn et al (Kahn 2006).

Subgroup analyses

Not performed due to lack of data.

Sensitivity analyses

Various sensitivity analyses did not change substantially the risk estimates for development of oedema after rosiglitazone treatment.

Publication bias

Not performed due to insufficient amounts of data.

Discussion

Summary of main results

This systematic review shows that published studies of at least 24 weeks rosiglitazone treatment in people with type 2 diabetes mellitus did not provide evidence that patient‐oriented outcomes like mortality, morbidity, adverse effects and health‐related quality of life are positively influenced by this compound. Metabolic control measured by glycosylated haemoglobin A1c (HbA1c) as a surrogate endpoint did not demonstrate clinically significant differences to other oral antidiabetic drugs. One study investigated economic costs of rosiglitazone versus insulin glargine therapy indicating lower costs of insulin glargine treatment. Occurrence of oedema was approximately doubled.

New safety data

The insulin‐sensitising thiazolidinediones pioglitazone and rosiglitazone act as potent inhibitors of the peroxisome‐proliferator‐activator receptor (PPAR) γ. Several PPARs exist with different expressions in various tissues. Activation of PPAR‐γ by thiazolidinediones may cause an increase in bone marrow adiposity and a decrease in osteoblastogenesis, resulting in reduced bone formation. Several publications of animal and human data are available (Ali 2005; Grey 2007; Lazarenko 2007; Schwartz 2006a; Schwartz 2006b; Yaturu 2007). To our knowledge, the ADOPT (A Diabetes Outcomes Progression Trial) ‐ Kahn 2006) study was the first randomised controlled clinical trial which demonstrated increased rates of fractures in women. According to the pharmaceutical company producing pioglitazone, a re‐analysis of the PROactive (Prospective Pioglitazone Clinical Trial In Macrovascular Events) study (Dormandy 2005) showed that 44/870 (5.1%) fractures were observed in pioglitazone treated female patients compared to 23/905 (2.5%) controls.   It is unclear why it took so long to analyse adverse events in an appropriate way. Adverse reactions on fracture rates only showed up in a "Note added in proof" in the New England Journal of Medicine (Kahn 2006) and the PROactive study publication did not mention this side effect at all (Dormandy 2005). For an adequate analysis of possible adverse events of published and unpublished data adverse events information should be freely available to the public and researches alike which should pose no problems with nowadays information technology.   Just before finishing this review a meta‐analysis on the effect of rosiglitazone on the risk of myocardial infarction and death from cardiovascular causes was published (Nissen 2007). Nissen and Wolski analysed 42 trials of rosiglitazone treatment with a study duration of more than 24 weeks. They found in the rosiglitazone group, as compared with the control group, a significant increase of the odds ratio for myocardial infarction of 1.43 (95% confidence interval (CI) 1.03 to 1.98, P = 0.03. The odds ratio for death from cardiovascular causes was 1.64 (95% CI 0.98 to 2.74, P = 0.06). Consequently, the US Food and Drug Administration (FDA), the European Medicines Agency (EMEA) and GlaxoSmithKline issued statements and warnings with regards to this meta‐analysis. Using the data from Nissen and Wolski we performed another meta‐analysis of the myocardial infarction rates for type 2 diabetes only, analysing all studies, rosiglitazone versus monotherapy and rosiglitazone versus combination therapies (in the original publication several other conditions were included as well to investigate the overall cardiovascular risk of rosiglitazone). So far and limited to the sparse data available, we could not confirm significant differences in odds ratios of rosiglitazone versus controls. On the other hand, all odds ratios (with the exception of the comparator glyburide ‐ three studies only) indicated an increased risk of rosiglitazone treatment, albeit not a statistically significant difference.   Moreover, it is disturbing to hear that the manufacturer of rosiglitazone (Avandia) provided the FDA with a pooled analysis of 42 RCTs in which rosiglitazone was compared to either placebo or other antidiabetic therapies in patients with type 2 diabetes. The meta‐analysis suggested that patients receiving short‐term (most studies were of six months duration) treatment with rosiglitazone may have a 30% greater relative risk of heart attacks and other heart‐related adverse events than patients treated with placebo or another antidiabetic therapy. Questions of timing of this information and how it was circled arise. Ongoing trials using rosiglitazone (RECORD) may provide additional data but for a drug which was approved in 1999, the delay in obtaining information about the benefit‐risk ratio is considerable.

The one major ongoing study (RECORD) which eventually could contribute valuable information about the role of rosiglitazone treatment in type 2 diabetes mellitus (for details see Characteristics of ongoing studies).   In the FDA statement 'FDA issues safety alert on Avandia' it is mentioned that "... other published and unpublished data from long‐term clinical trials of Avandia, including an interim analysis of data from the RECORD trial (a large, ongoing, randomized open label trial) and unpublished re analyses of data from DREAM (a previously conducted placebo‐controlled, randomized trial) provide contradictory evidence about the risks in patients treated with Avandia." We do hope that the conduct, analysis and interpretation of this trial will reflect high quality scientific standards and will not resemble the dishonourable events which accompanied the PROactive study (for more details, see Richter 2006). We agree with the commentators on the Nissen and Wolski publication that current drug approval for antidiabetic medications and possibly all new drugs needs to be changed (Psaty 2007). The benefit‐risk ratio of rosiglitazone therapy in type 2 diabetes mellitus needs urgent clarification.

Potential biases in the review process

We focused on a minimum duration of 24 weeks rosiglitazone therapy in order to have a chance to detect clinically meaningful differences in patient‐oriented parameters. Theoretically, studies of a shorter duration could demonstrate a significant impact on these outcomes but this is highly unlikely, even with regards to important adverse events.   Moreover, it was difficult to separate primary studies from companion papers because the latter quite often did not identify themselves as an additional publication of a parent study; especially authors Derosa et al did not reference multiple publications to each other (for details see 'References of included studies', primary studies are marked by an asterisk).

Authors' conclusions

Implications for practice.

This systematic review shows that published studies of at least 24 weeks rosiglitazone treatment in people with type 2 diabetes mellitus did not provide evidence that patient‐oriented outcomes like mortality, morbidity, adverse effects and health‐related quality of life are positively influenced by this compound. Metabolic control measured by glycosylated haemoglobin A1c (HbA1c) as a surrogate endpoint did not demonstrate clinically significant differences to other oral antidiabetic drugs. Occurrence of oedema was approximately doubled. New safety data on increased rates of fractures and possibly the risk of myocardial infarction and cardiovascular disease should lead to a very cautious approach to rosiglitazone use. If possible, other antidiabetic medications should be employed.

Implications for research.

Patient‐oriented endpoint studies are urgently needed for the management of type 2 diabetes mellitus. The use of proxy indicators like metabolic control is not sufficient to approve drugs which many patients have to take for the rest of their lives. It appears questionable whether new studies with rosiglitazone will be ethical given the fact that less dangerous therapeutic alternatives exist.

Feedback

Dollow, July 2007

Summary

The following query was made on 18 July 2007:

The Cochrane Collaboration has a reputation for robustness of analysis and integrity of data interpretation. Therefore, it was disappointing to read the conclusions made in the recent Cochrane Review written by Richter et al, titled, "Rosiglitazone for type 2 diabetes mellitus."

The authors drew conclusions regarding the impact of rosiglitazone on mortality and morbidity by reviewing a limited number of short term studies (18) primarily designed to assess glycaemic control. This analysis cannot provide a full picture of all the research conducted with rosiglitazone. The conclusions provide no new evidence about the role of rosiglitazone in clinical practice. In addition, the conclusions regarding cardiovascular safety disagree with the authors' own meta‐analysis on myocardial infarction which could not confirm an increased risk.

The studies assessed in the review contained no stratification for baseline cardiovascular risk, leading to unavoidable imbalances between rosiglitazone and control groups. Most importantly the authors fail to include the interim findings of RECORD(1), a prospective long‐ term study primarily designed to evaluate the profile of rosiglitazone with respect to cardiovascular disease. The RECORD(1) data was available as an online publication some six weeks prior to the publication of this review. Its exclusion is surprising and adds question to the robustness of the authors' conclusions.

Questions about the safety of rosiglitazone should be answered by reviewing all relevant evidence, in particular long‐term prospective trials. The conclusion regarding the cardiovascular data from ADOPT(2,3) are puzzling, given that in ADOPT(2,3) all major adverse cardiovascular events (MACE) were analysed and found to be rare in this population and comparable for all treatments ‐ rosiglitazone, glibenclamide and metformin.. Additionally, no excess in mortality with rosiglitazone was seen overall. The significant benefits of rosiglitazone in maintaining the duration of glycaemic response in ADOPT(2) are unfortunately not given similar prominence.

The interim findings of RECORD(1), the only study specifically designed to look at cardiovascular outcomes with rosiglitazone, does not show evidence of a difference in cardiovascular death between rosiglitazone and control groups. Additionally, no significant differences for myocardial infarction between groups were seen.

The totality of the data ‐ including long‐term studies such as ADOPT(2) and RECORD(3) and a real‐world epidemiological analysis of 33,000 patients(5) ‐ show that rosiglitazone has a comparable ischaemic cardiovascular profile to the most commonly used oral anti‐diabetic medicines, metformin and sulphonylureas.

With respect to the analysis of glycaemic efficacy, it is puzzling that the authors excluded a number of studies which are applicable to decisions made in clinical practice, such as Bailey et al(4) in which uptitration of metformin is compared with metformin and rosiglitazone. Additionally, whilst a significant decrease in the rate of hypoglycaemia associated with rosiglitazone is reported in the results section of the review, this is not referred to in the authors' conclusions. Instead, only oedema is mentioned, which is a well recognised side‐effect of thiazolidenedione therapy.

The studies selected for use in a Cochrane systematic review should be appropriate to the purpose of the review. It is therefore difficult to understand how the limited range of studies selected from the much larger number of studies available, allow the authors to draw robust conclusions with respect to morbidity, mortality and health‐outcomes for rosiglitazone. In addition, the conclusions drawn regarding ischaemic cardiovascular safety should be substantiated by the data analysed and not inferred from statistically insignificant odds ratios.

Finally we question the appropriateness of raising comment about the timing of data release to regulatory authorities and regulatory approval requirements in diabetes as part of a systematic review. GSK has actively shared data on rosiglitazone with regulatory agencies worldwide in a timely manner. The company carries out its clinical trials with the highest level of ethical conduct and is committed to patient safety.

References   1. Home PD et al N Engl J Med 2007; 357: 28‐38   2. Kahn SE et al N Engl J Med 2006; 355: 2427‐43.   3. Krall RL Lancet 2007; 369: 1995‐1996   4. Bailey CJ et al Clin There 2005; 27:1548‐1561   5. McAfee AT et al Pharmacoepidemiology Drug Saf; 2007 16: 711‐725

Abbreviations   ADOPT ‐ A Diabetes Outcome Progression Trial   RECORD ‐ Rosiglitazone Evaluated for Cardiovascular Outcomes

Reply

The comments by Dr Dollow are answered in a point‐by‐point fashion:

The Cochrane Collaboration as well as the Metabolic and Endocrine Disorders Group adhere to high quality standards. It is unclear how Dr Dollow defines "integrity of data interpretation". As a matter of course, the discussion and conclusion sections are firmly based upon the data evaluated in our review.

The types of interventions we included had to have a minimum trial duration of 24 weeks. The point that a limited number of studies had a longer duration, for example more than one year, is due to the fact that neither the manufacturer nor the scientific community seems to be interested in the long‐term benefit‐risk ratio of rosiglitazone therapy, but cannot be attributed to the review itself. Furthermore, the bulk of studies investigated glycaemic control as primary efficacy endpoint and not patient‐oriented parameters like mortality, morbidity and health‐related quality of life which again has to be ascribed to the deficiencies of studies but not the systematic review. Our review so far provides the best overview of the risks and (with regard to relevant outcomes) missing benefits of rosiglitazone therapy and therefore is of great importance for clinical practice. We did not perform our own meta‐analysis on myocardial infarction but tried to replicate the findings by Nissen et al using their data in the discussion section of our review (Nissen 2007). Cardiovascular disease and safety in their clinical meaning include more than myocardial infarction, for example increased risk of congestive heart failure following rosiglitazone therapy. Therefore, we stand by the conclusions as stated in the review.

The studies and publications we discovered and assessed in our review ‐ with the exception of the ADOPT (A Diabetes Outcome Progression Trial) ‐ did not investigate cardiovascular risk. That is one of the reasons why Nissen et al (Nissen 2007) had to search the manufacturer's as well as drug authorities web sites. The publication schedule of the Cochrane Library demands from Cochrane review groups to hand in their "module" (all new and updated protocols and reviews) around two months before the publication of the Cochrane Library. Therefore, the interim findings of the RECORD (Rosiglitazone Evaluated for Cardiac Outcomes and Regulation of Glycaemia in Diabetes) study could not be included in our review (Home 2007). Furthermore, these interim data do not provide assurance of the cardiovascular safety of rosiglitazone treatment in type 2 diabetes mellitus (see below).

We agree that questions about the safety of rosiglitazone should be answered by critical appraisal of especially well‐performed long‐term randomised controlled clinical trials. With regard to the cardiovascular data from the ADOPT trial Dr Dollow mentions a letter to the Lancet editor by Dr Krall, Chief Medical Officer of GlaxoSmithKline. It is of interest to note that this letter to the editor which refers to the Nissen et al publication was published in the Lancet and not the New England Journal of Medicine where the study originally was published. The new endpoint MACE (major adverse cardiovascular events, that is all cardiovascular deaths, myocardial infarction serious adverse events (fatal and non‐fatal), and stroke serious adverse events (fatal and non‐fatal)) was not part of the original publication of the ADOPT trial, resulted from a post‐hoc analysis by the manufacturer and "forgot" to mention congestive heart failure which was part of the outcomes contributing to the overall endpoint cardiovascular disease. Here, significant differences between glyburide (glibenclamide) and rosiglitazone were reported, indicating increased cardiovascular disease risk after rosiglitazone therapy, as mentioned in our review. The ADOPT trial was not powered to investigate mortality. The primary outcome time from randomisation to treatment failure as measured by elevated fasting plasma glucose levels was not part of our pre‐specified outcomes but we agree with the accompanying New England Journal of Medicine editorial stating "the choice of time to failure based on a confirmed fasting glucose level of more than 180 mg per deciliter as the primary outcome, rather than one based upon glycated hemoglobin levels, seems anachronistic " (Nathan 2006).

The unscheduled interim analysis from the RECORD trial should not be interpreted as evidence for cardiovascular safety of rosiglitazone therapy (Home 2007). We once again agree with the statements of the associated editorial in the New England Journal of Medicine (Nathan 2007):   "The primary end point of the RECORD trial consists of an aggregate of time to first hospitalization for a cardiovascular event or death from cardiovascular causes" ... "Unfortunately, this interim analysis, performed after a mean of 3.75 years (about 60% of the planned 6‐year duration of the study) fails to provide exculpatory evidence" ... "RECORD extremely underpowered for the primary outcome" ... "the results of this underpowered interim analysis suggest a possible adverse effect of treatment with rosiglitazone on the primary outcome, rather than the benefit that was hypothesized ... considering the low power of the study and the trend for more adverse outcomes in the rosiglitazone‐treated group, it is highly unlikely that the study will ever establish a cardiovascular benefit for rosiglitazone" ... "In the aggregate, however, these analyses support a concern regarding the safety of rosiglitazone" ... "It is reasonable to ask whether physicians should feel comfortable using a drug that might have an 8% excess risk of severe cardiovascular disease or death from cardiovascular causes" ... "Unless further studies provide convincing assurance that treatment with rosiglitazone does not increase the risk of cardiovascular disease, the largely circumstantial evidence of the meta‐analyses and the nonsignificant trend in the current report from the RECORD trial must be taken seriously" ... "The jury may still be out with regard to the cardiotoxicity of rosiglitazone, but when it comes to patient safety, "first, do no harm" should outweigh any presumption of innocence."

As demonstrated above, the totality of data do not show that rosiglitazone has a comparable ischaemic cardiovascular profile to the most commonly used oral antidiabetic medicines. To claim a comparable ischaemic cardiovascular risk profile especially to metformin in obese type 2 diabetes patients appears careless: Contrary to rosiglitazone treatment metformin positively influences patient‐oriented outcomes since the United Kingdom Prospective Diabetes Study (UKPDS) demonstrated that patients allocated metformin had significant reductions for any‐diabetes related endpoint, diabetes‐related death, stroke and all‐cause mortality (UKPDS‐34).

Dr Dollow claims that "authors excluded a number of studies which are applicable to decisions made in clinical practice, such as Bailey et al(4) in which uptitration of metformin is compared with metformin and rosiglitazone". In the 'criteria for considering studies for this review' we clearly exemplified under 'excluded interventions': "Combination therapies consisting of different compounds in the treatment arms (for example rosiglitazone plus metformin versus uptitration of metformin or rosiglitazone plus gliclazide versus gliclazide). Another Cochrane review will investigate rosiglitazone‐metformin combination therapies including different treatment regimens of these compounds." We want to perform another Cochrane review on different combination partners because it does not appear to be adequate to compare interventions with different combination partners neglecting the complicated interplay of various agents. Furthermore, we did not report on a significant decrease in the rate of hypoglycaemia associated with rosiglitazone but stated "Seven of the 18 included studies showed data on hypoglycaemic episodes: Compared to active monotherapy control rosiglitazone treatment resulted in somewhat lower rates of hypoglycaemia, especially when compared to sulphonylureas. Severe hypoglycaemic events were rarely reported." Apart from that, serious adverse events were noted more often after rosiglitazone treatment as were higher median discontinuation rates compared to control therapy.

Our studies selected for this review were indeed appropriate to our objectives. To speak of a "limited range of studies selected from the much larger number of studies available" does not understand our strategy. We especially focused on patient‐oriented parameters like mortality, morbidity, health‐related quality of life and adequately reported on all available study results according to our in‐ and exclusion criteria. Unfortunately, the availability of sound studies is scarce due to the fact that concerning this matter only the ADOPT and the RECORD trial provide some hypotheses about the benefit‐risk ratio of rosiglitazone therapy which does not appear to be positive (see above).

According to Krall (Krall 2007), GlaxoSmithKline performed similar meta‐analyses in 2005 and 2006 and found similar results as Nissen et al (Nissen 2007). We are not aware that the public was adequately informed about these results, otherwise the meta‐analysis by Nissen et al would not have aroused such a huge public interest. It is well know that glycosylated haemoglobin is a relatively poor surrogate for cardiovascular outcomes and these data urgently suggest that we need to change the regulatory pathway for drugs for the treatment of type 2 diabetes to make clinical outcomes, not surrogates, the primary endpoint (Rosen 2007). It would be prudent for one of the biggest pharmaceutical companies in the world being committed to patient care to engage in relevant clinical studies of patient‐oriented outcomes from the very beginning on.

References:   HOME 2007: Home PD, Pocock SJ, Beck‐Nielsen H, Gomis R, Hanefeld M, et al. New England Journal of Medicine 2007;357:28‐38.   KRALL 2007: Krall RL. Cardiovascular safety of rosiglitazone. Lancet 2007; 369:1995‐6.   NATHAN 2006: Nathan DM. Thizolidinediones for initial treatment of type 2 diabetes? New England Journal of Medicine 2006;355:2477‐80.   NATHAN 2007: Nathan DM. Rosiglitazone and cardiotoxicity ‐ weighing the evidence. New England Journal of Medicine 2007;357:64‐6.   NISSEN 2007: Nissen SE, Wolski K. Effect of rosiglitazone on the risk of myocardial infarction and death from cardiovascular causes. New England Journal of Medicine 2007;356:2457‐71.   ROSEN 2007: Rosen CJ. Rhe rosiglitazone story ‐ lessons from an FDA Advisory Committee Meeting. New England Journal of Medicine 2007;357: Published at www.nejm.org August 8, 2007.   UKPDS‐34: UK Prospective Diabetes Study Group. Effect of intensive blood‐glucose control with metformin on complications in overweight patients with type 2 diabetes. Lancet 1998;352:854‐65.

Contributors

Comments made by Dr Stuart Dollow, Vice President and UK Medical Director GlaxoSmithKline (stuart.c.dollow@GSK.com).

Bernd Richter replied to the comments on behalf of the review authors for the review.

What's new

Date	Event	Description
6 November 2008	Amended	Converted to new review format.

Characteristic	Derosa 2004	Derosa 2006a	Derosa 2006b
	I1: rosiglitazone 4 mg + glimepiride 4 mg   C1: pioglitazone 15 mg + glimepiride 4 mg	I1: rosiglitazone 4 mg + metformin 3000 mg   C1: pioglitazone 15 mg + metformin 3000mg	I1: rosiglitazone 4 mg + metformin 1500 mg   C1: glimepiride 2 mg+ metformin 1500 mg
Sex [%]	I1: female 48; male 52   C1: female 53; male 447	I1: female 48; male 52   C1: female 50; male 50	I1: female 48; male 52   C1: female 51; male 49
Age [years], mean (SD)	I1: 54 (5)   C1: 53 (6)	I1: 56 (4)   C1: 55 (5)	I1: 54 (4)   C1: 52 (5)
Ethnic groups [%]	I1: white 100   C1: white 100	I1: caucasians 100   C1: caucasians 100	?
Duration of disease [years], mean (SD)	I1: 6 (3)   C1: 5 (2)	I1: 5(4)   C1: 6(4)	I1: 5 (3)   C1: 4 (3)
Body mass index [kg/m2], mean (SD)	I1: 24.3 (0.7)   C1: 24.4 (0.8)	I1: 26.4 (1.4)   C1: 26.9 (1.2)	I1: 26.6 (1.3)   C1: 26.8 (1.5)
Pharmaco‐naive patients [%]	I1: none   C1: none	I1: none   C1: none	I1: none   C1: none
HbA1c [%], mean (SD)	I1: 8.0 (0.8)   C1: 8.2 (0.7)	I1: 8.1 (0.9)   C1: 8.2 (0.8)	I1: 8.0 (0.7)   C1: 7.9 (0.6)
Co‐morbidities [%]	?	?	I1: hypertension 42   C1: hypertension 47
Notes	patients who completed the study	./.	./.

Footnotes ? = unclear; I = intervention; C = control; SD = standard deviation; SE = standard error; ITT = intention‐to ‐treat

Characteristic	Garber 2006	Goldberg 2005	Hanefeld 2007	Hällsten 2002	Jung 2005
	I1: rosiglitazone 4 mg + metformin 2000   C1: glibenclamide 5 mg +metformin 1000 mg	I1: rosiglitazone 8 mg + diet   C1: pioglitazone 45 mg + diet	I1: rosiglitazone 4 mg + placebo   I2: rosiglitazone 8 mg + placebo   C1: glibenclamide + placebo	I1: rosiglitazone 8 mg   C1: metformin 2 g   C2: placebo	I1: rosiglitazone 4 mg + glimepiride 4 mg   C1: metformin 1000 mg + glimepiride 4 mg
Sex [%]	I1: female35; male 65   C1: female 44; male 56	I1: female 45; male 55   C1: female 46; male 54	I1: female 32; male 68   I2: female 42; male 58   C1: female 30; male 70	I1: female 29; male 71   C1: female 38; male 62   C2: female 29; male 71	I1: female 57; male 43   C1: female 54; male 46
Age [years], mean (SD)	I1: 56   C1: 56	I1: 56.3 (11.3)   C1: 55.9 (10.5)	I1: 60.4 (8.2)   I2: 60.6 (9.2)   C1: 60.1 (8.3)	I1: 58.6 (7.5)   C1: 57.8 (7.9)   C2: 57.7 (7.1)	I1: 60 (8)   C1: 54 (14)
Ethnic groups [%]	I1: white 79; black 6; hispanic/latino 10; asian/pacific islander 3; other 3   C1: white 80; black 5; hispanic/latino 11; asian/pacific islander 3; other 2	I1: white 60; hispanic 32; asian 3; african 3; other 2   C1: white 65; hispanic 29; asian 3; african 2; other 2	I1: white 99; other 1   I2: white 97; other 3   C1: white 99.5; other 0.5	?	I1: korean 100   C1: korean 100
Duration of disease [years], mean (SD)	I1: 6 (5)   C1: 5 (4)	I1: 4.0 (4.6)   C1: 3.9 (4.4)	I1: 5.9 (6.0)   I2: 6.0 (7.0)   C1: 6.4 (6.9)	newly diagnosed	I1: 9 (5)   C1 : 7 (6)
Body mass index [kg/m2], mean (SD)	I1: 32 (5)   C1: 32 (5)	I1: 32.6 (6.6)   C1: 33.7 (12.9)	I1: 28.7 (3.7)   I2: 28.8 (3.7)   C1: 28.7 (3.9)	I1: 29.3 (3.7)   C1: 29.9 (4.0)   C2: 30.3 (4.5)	I1: 23.3 (2.6)   C1: 24.6 (2.4)
Pharmaco‐naive patients [%]	I1: none   C1: none	Total: 8	I1: 42   I2: 38   C1: 38	?	I1: none   C1: none
HbA1c [%], mean (SD)	I1: 8.4 (1.1)   C1: 8.5 (1.2)	I1: 7.5 (1.2)   C1: 7.6 (1.2)	I1: 8.1 (1.3)   I2: 8.2 (1.4)   C1: 8.2 (1.3)	I1: 6.8 (0.8)   C1: 6.9 (0.7)   C2: 6.3 (0.4)	I1: 9.3 (0.9)   C1: 9.0 (0.8)
Co‐morbidities [%]	?	?	?	?	?
Notes	./.	./.	./.	SDs calculated	text table data mismatch

Footnotes ? = unclear; I = intervention; C = control; SD = standard deviation; SE = standard error; ITT = intention‐to‐treat

Characteristic	Kahn 2006	Ko 2006	Lebovitz 2001	Ovalle 2004	Philipps 2001	Raskin 2004
	Kahn 2006     I1: rosiglitazone max. 8 mg   C1: metformin max. 2 g   C2: glyburide max. 15 mg	I1: rosiglitazone max. 8 mg   + (sulfonylurea +/‐ metformin)   C1: "bedtime insulin"   + (sulfonylurea +/‐ metformin)	I1: rosiglitazone 4 mg   I2: rosiglitazone 8 mg   C1: placebo	I1: rosiglitazone 8 mg   C1: insulin 70/30	I1: rosiglitazone 4 mg   I2: rosiglitazone 2 x 2 mg   I3: rosigllitazone 8 mg   I4: rosiglitazone 2 x 4 mg   C1: placebo	I1: rosiglitazone 8 mg   C1: repaglinide 12 mg   C2: rosiglitazone 4 mg + repaglinide 6 mg
Sex [%]	I1: female 44; male 56   C1: female 40; male 60   C2: female 42; male 58	I1: female 43; male 57   C1: female 57; male 43	I1: female 36; male 64   I2: female 33; male 67   C1: female 34; male 66	?	I1: female 41; male 59   I2: female 41; male 59   I3: female 34; male 66   I4: female 35; male 65   C1: female 31; male 69	I1: female 47; male 53   C1: female 38; male 62   C2: female 49; male 51
Age [years], mean (SD)	I1: 56.3 (10.0)   C1: 57.9 (9.9)   C2: 56.4 (10.2)	I1: 56.6 (10.7)   C1: 59.8 (11.2)	I1: 60 (9.8)   I2: 61 (9.5)   C1: 59 (10.9)	I1: 47 (12)   C1: 56 (14.1)	I1: 57.5 (9.9)   I2: 56.8 (9.4)   I3: 58.9 (9.9)   I4: 56.5 (9.7)   C1: 57.7 (9.2)	I1: 56.6 (10.8)   C1: 58.5 (10.1)   C2: 57.5 (10.8)
Ethnic groups [%]	I1: white 87; black 4; asian 3; hispanic 5; other 1   C1: white 89; black 4; asian 2; hispanic 4; other 1   C2: white 89; black 4; asian 2; hispanic 4; other 0.3	Chinese patients	I1: white 75; black 8; other 16   I2: white 73; black 9; other 17   C1: white 74; black 8; other 18	?	I1: white 76; black 13; other 11   I2: white 78; black 8; other 14   I3: white 80; black 7; other 13   I4: white 71; black 11; other 18   C1: white 79; black 9; other 12	I1: caucasian 68; black 13; hispanic 0; other 19   C1: caucasian 63; black 16; hispanic 2; other 19   C2: caucasian 65; black 17; hispanic 3; other 15
Duration of disease [years], mean (SD)	?	I1: 11.8 (7.7)   C1: 13.6 (7.5)	I1: 4.8 (5.8)   I2: 5.4 (6.0)   C1: 4.6 (4.8)	I1: 7.6 (6.3)   C1: 7.6 (4.8)	I1: 5.4 (6.1)   I2: 5.5 (4.9)   I3: 6.1 (6.7)   I4: 5.9 (6.1)   C1: 6.6 (6.9)	I1: 7.4 (6.6)   C1: 7.2 (5.3)   C2: 7.3 (6.9)
Body mass index [kg/m2], mean (SD)	I1: 32.2 (6.7)   C1: 32.1 (6.1)   C2: 32.2 (6.3)	I1: 25.3 (3.8)   C1: 24.0 (2.7)	I1: 30.2 (4.1)   I2: 29.1 (3.9)   C1: 29.9 (4.1)	I1: 31.5 (6.9)   C1: 30.8 (7.6)	I1: 29.9 (4.1)   I2: 30.0 (4.2)   I3: 30.0 (4.3)   I4: 29.9 (4.3)   C1: 29.1 (4.2)	I1: 31.4 (5.2)   C1: 30.4 (4.7)   C2: 32.3 (5.2)
Pharmaco‐naive patients [%]	I1: 100   C1: 100   C2: 100	?	I1: 26.5   I2: 26.6   C1: 28.5	I1: none   C1: none	I1: 22.1 (40/181)   I2: 24.7 (46/186)   I3: 29.3 (53/181)   I4: 25.1 (47/187)   C1: 22.5 (39/173)	I1: none   C1: none   C2: none
HbA1c [%], mean (SD)	I1: 7.4 (0.9)   C1: 7.4 (0.9)   C2: 7.4 (0.9)	I1: 10.1 (1.0)   C1: 9.6 (0.9)	I1: 9.0 (1.5)   I2: 8.8 (1.6)   C1: 9.0 (1.7)	I1: 8.7   C1: 9.0	I1: 8.9 (1.6)   I2: 8.9 (1.5)   I3:8.9 (1.5)   I4:9.0 (1.5)   C1: 8.9 (1.5)	I1: 9.0   C1: 9.3   C2: 9.1
Co‐morbidities [%]	?	?	?	?	?	?
Notes	antihypertensive therapy:   I1: 51%; C1: 51%; C2: 52%	antihypertensive agents:   I1: 55%; C1: 25%   lipid‐lowering agents:   I1: 9%; C1: 4%	ITT population	SDs calculated	ITT population	./.

Footnotes ? = unclear; I = intervention; C = control; SD = standard deviation; SE = standard error; ITT = intention‐to‐treat

Characteristic	Rosenstock 2006b	Stocker 2007	Sutton 2002	Yang 2002
	I1: rosiglitazone until 8 mg + sulfonyurea + max. 2 g metformin   C1: insulin glargine max. 10 U + sulfonylurea + max. 2 g metformin	I1: rosiglitazone 4 mg   C1: metformin 1.7 g	I1: rosiglitazone 8 mg   C1: glyburide less than 20 mg	I1: rosiglitazone 4 mg   C1: placebo
Sex [%]	I1:female 42; male 58   C1: female 55; male 45	I1: female 29; male 71   C1: female 47; male 53	I1: female 25; male 75   C1: female 29; male 71	I1: female 57; male 43   C1: female 62; male 38
Age [years], mean (SD)	I1: 55.3 (11.4)   C1: 55.9 (10.5)	I1: 64 (11)   C1: 65 (10)	I1: 55.1 (9.0)   C1: 56.1 (8.9)	I1: 58.9 (9.4)   C1: 57.8 (8.9)
Ethnic groups [%]	?	?	I1: white 73; black 5; other 22   C1: white 76; black 3; other 21	?
Duration of disease [years], mean (SD)	I1: 8.1 (5.1)   C1: 8.5 (5.8)	?	I1: 5.3 (6.2)   C1: 6.2 (6.3)	?
Body mass index [kg/m2], mean (SD)	I1: 33.6 (6.3)   C1: 34.6 (7.0)	I1: 29.4 (0.7)   C1: 29.7 (0.7)	>= 27:   I1: 67.3%   C1: 65.7%	I1: 25.8 (2.9)   C1: 25.8 (3.5)
Pharmaco‐naive patients [%]	I1: none   C1: none	I1: 24   C1: 28	I1: 21.2   C1: 18.2	I1: none   C1: none
HbA1c [%], mean (SD)	I1: 8.7 (1.0) (ITT)   C1: 8.8 (1.0) (ITT)	I1: 8.5 (0.3)   C1: 8.5 (0.2)	I1: 9.1 (1.7)   C1: 9.5 (1.6)	I1: 9.5 (1.1)   C1: 9.7 (1.4)
Co‐morbidities [%]		cardiovascular disease   I1: 4.4   C1: 6.4	hypertension:   I1: 7.7   C1: 7.0	?
Notes	ITT population:   for baseline characteristics (112:105)   for HbA1c (112:104)	data on statin, aspirin, beta‐blocker, calcium‐channel blocker, angiotensin receptor blocker , ACE inhibitor and sulfonylurea use	./.	./.

Footnotes ? = unclear; I = intervention; C = control; SD = standard deviation; SE = standard error; ITT = intention‐to‐treat

Characteristic	Mortality	Morbidity	Adverse events	Notes
Derosa 2004     I1: rosiglitazone 4 mg + glimepiride 4 mg   C1: pioglitazone 15 mg + glimepiride 4 mg	not investigated	not investigated	see table 'Adverse events'	./.
Derosa 2006a     I1: rosiglitazone 4 mg + metformin 3 g   C1: pioglitazone 15 mg + metformin 3 g	not investigated	not investigated	see table 'Adverse events'	./.
Derosa 2006b     I1: rosiglitazone 4 mg + metformin 1.5 g   C1: glimepiride 2 mg + metformin 1.5 g	not investigated	not investigated	see table 'Adverse events'	./.
Garber 2006     I1: rosiglitazone 4 mg + metformin 2 g   C1: glibenclamide 5 mg + metformin 1 g	not investigated	not investigated	see table 'Adverse events'	./.
Goldberg 2005     I1: rosiglitazone 8 mg   C1: pioglitazone 45 mg	not investigated	not investigated	see table 'Adverse events'	./.
Hällsten 2002     I1: rosiglitazone 8 mg   I2: metformin 2 g   C1: placebo	not investigated	not investigated	see table 'Adverse events'	./.
Hanefeld 2007     I1: rosiglitazone (2 mg bid) + placebo   I2: rosiglitazone (4 mg bid) + placebo   C1: glibenclamide up to 15 mg + placebo	not investigated	not investigated	see table 'Adverse events'	./.
Jung 2005     I1: rosiglitazone 4 mg + glimepiride 4 mg   C1: metformin 1 g + glimepiride 4 mg	not investigated	not investigated	see table 'Adverse events'	./.
Kahn 2006     I1: rosiglitazone max. 8 mg   C1: metformin max. 2 g   C2: glyburide max. 15 mg	death rates reported but not part of the efficacy outcomes, as defined in the publication of the study design (Diabetes Care 2002):     deaths from any cause [no]:   I1: 34   C1: 31   C2: 31	morbidity rates reported but not part of the efficacy outcomes, as defined in the publication of the study design (Diabetes Care 2002):     cardiovascular disease [no (%)]:   serious / total events   I1: 49 (3.4) / 62 (4.3)   C1: 46 (3.2) / 58 (4.0)   C2: 26 (1.8) / 41 (2.8)     Peripheral vascular disease [no (%)]:   serious / total events   I1: 7 (0.5) / 36 (2.5)   C1: 6 (0.4) / 27 (1.9)   C2: 4 (0.3) / 31 (2.2)	see table 'Adverse events'	./.
Ko 2006     I1: rosiglitazone max. 8 mg   + (sulfonylurea +/‐ metformin)   C1: "bedtime insulin"   + (sulfonylurea +/‐ metformin)	not investigated	not investigated	see table 'Adverse events'	./.
Lebovitz 2001     I1: rosiglitazone 4 mg   I2: rosiglitazone 8 mg   C1: placebo	not investigated	not investigated	see table 'Adverse events'	./.
Ovalle 2004     I1: rosiglitazone 8 mg   I2: insulin (premixed 70/30)	not investigated	not investigated	see table 'Adverse events'	./.
Philipps 2001     I1: rosiglitazone   4 mg o.d.; 2 mg b.i.d.; 8 mg o.d.; 4 mg b.i.d.   C1: placebo	not investigated	not investigated	see table 'Adverse events'	./.
Raskin 2004     I1: rosiglitazone 8 mg   I2: repaglinide 12 mg   C1: repaglinide + rosiglitazone 6 / 4 mg	not investigated	not investigated	see table 'Adverse events'	./.
Rosenstock 2006b     I1: rosiglitazone 8 mg + metformin 2g + sulfonylurea   C1: insulin glargine 10 units/day + metformin 2g + sulfonylurea	not investigated	not investigated	see table 'Adverse events'	./.
Stocker 2007     I1: rosiglitazone 4 mg   C1: metformin 1.7 g	not investigated	not investigated	see table 'Adverse events'	./.
Sutton 2002     I1: rosiglitazone 8 mg   C1: glyburide (mean 10.5 mg)	not investigated	not investigated	see table 'Adverse events'	./.
Yang 2002     I1: rosiglitazone 4 mg   C1: placebo	not investigated	not investigated	see table 'Adverse events'	./.

Footnotes ? = unclear; I = intervention; C = control; o.d. = once daily; b.i.d. = twice daily

Characteristic	Quality of life	Costs	HbA1c [%] (SD)	Notes
Derosa 2004     I1: rosiglitazone 4 mg + glimepiride 4 mg   C1: pioglitazone 15 mg + glimepiride 4 mg	not investigated	not investigated	I1:   end of study data: 6.7 (0.9)   change data:   C1:   end of study data: 6.8 (0.8)   change data:	./.
Derosa 2006a     I1: rosiglitazone 4 mg + metformin 3 g   C1: pioglitazone 15 mg + metformin 3 g	not investigated	not investigated	I1:   end of study data: 6.8 (0.5)   change data: ‐1.3   C1:   end of study data: 6.8 (0.3)   change data: ‐1.4	change data calculated
Derosa 2006b     I1: rosiglitazone 4 mg + metformin 1.5 g   C1: glimepiride 2 mg + metformin 1.5 g	not investigated	not investigated	I1:   end of study data: 6.8 (0.6)   change data:   C1:   end of study data: 7.0 (0.7)   change data:	./.
Garber 2006     I1: rosiglitazone 4 mg + metformin 2 g   C1: glibenclamide 5 mg + metformin 1 g	not investigated	not investigated	I1:   end of study data:   change data: ‐1.1   C1:   end of study data:   change data: ‐1.5	"change data" from abstract
Goldberg 2005     I1: rosiglitazone 8 mg   C1: pioglitazone 45 mg	not investigated	not investigated	I1:   end of study data:   change data: ‐0.6 (1.89)   C1:   end of study data:   change data: ‐0.7 (1.91)	SDs calculated
Hanefeld 2007   I1: rosiglitazone (2 mg bid) + placebo   I2: rosiglitazone (4 mg bid) + placebo   C1: glibenclamide up to 15 mg + placebo	not investigated	not investigated	I1:   end of study data:   change data: ‐0.3   I2:   end of study data:   change data: ‐0.5   I1:   end of study data:   change data: ‐0.7	./.
Hällsten 2002     I1: rosiglitazone 8 mg   I2: metformin 2 g   C1: placebo	not investigated	not investigated	I1:   end of study data: 6.5 (0.75)   change data:   I2:   end of study data: 6.2 (0.72)   change data:   C1:   end of study data: 6.1 (0.37)   change data:	SDs calculated
Jung 2005 I1: rosiglitazone 4 mg + glimepiride 4 mg C1: metformin 1 g + glimepiride 4 mg	not investigated	not investigated	I1:   end of study data: 7.8 (1.1)   change data:   C1:   end of study data: 8.0 (1.1)   change data:	./.
Kahn 2006     I1: rosiglitazone max. 8 mg   C1: metformin max. 2 g   C2: glyburide max. 15 mg	not yet reported but mentioned in the publication of the study design (Diabetes Care 2002)	not yet reported but mentioned in the publication of the study design (Diabetes Care 2002)	I1:   end of study data: 7.1   change data:   C1:   end of study data: 7.3   change data:   C2:   end of study data: 7.4   change data:	estimated from graph (four year data)
Ko 2006     I1: rosiglitazone max. 8 mg   + (sulfonylurea +/‐ metformin)   C1: "bedtime insulin"   + (sulfonylurea +/‐ metformin)	not investigated	not investigated	I1:   end of study data: 9.1 (2.0)   change data: ‐1.1 (1.7)   C1:   end of study data: 8.3 (1.3)   change data: ‐1.3 (1.6)	./.
Lebovitz 2001     I1: rosiglitazone 4 mg   I2: rosiglitazone 8 mg   C1: placebo	not investigated	not investigated	I1:   end of study data:   change data: ‐0.3   I2:   end of study data   change data: ‐0.6   C1:   end of study data:   change data: +0.9	./.
Ovalle 2004     I1: rosiglitazone 8 mg   C1: insulin (premixed 70/30)	not investigated	not investigated	I1:   end of study data: 7.8 (0.5)   change data:   C1:   end of study data: 7.8 (0.3)   change data:	./.
Philipps 2001     I1: rosiglitazone   4 mg o.d.; 2 mg b.i.d.; 8 mg o.d.; 4 mg b.i.d.   C1: placebo	not investigated	not investigated	patients who had received prior oral monotherapy:   I1:   end of study data:   change data: 4 mg o.d. (+0.14); 2 mg b.i.d. (+0.02); 8 mg o.d. (‐0.26); 4 mg b.i.d. (‐0.54)   C1:   end of study data:   change data: +0.98	SDs calculated
Raskin 2004     I1: rosiglitazone 8 mg   I2: repaglinide 12 mg   C1: repaglinide + rosiglitazone 6 mg / 4 mg	not investigated	not investigated	I1:   end of study data: 8.5   change data: ‐0.56 (1.0)   I2:   end of study data: 9.1   change data: ‐0.17 (1.1)   C2:   end of study data: 7.7   change data: ‐1.43 (1.1)	SDs calculated
Rosenstock 2006b     I1: rosiglitazone 8 mg + metformin 2g + sulfonylurea   C1: insulin glargine 10 units/day + metformin 2g + sulfonylurea	not investigated	I1: $ 1,603   C1: $ 1,368	I1:   end of study data:   change data: ‐1.51   C1:   end of study data:   change data: ‐1.66	./.
Stocker 2007     I1: rosiglitazone 4 mg   C1: metformin 1.7 g	not investigated	not investigated	I1:   end of study data:   change data: ‐1.08 (0.14)   C1:   end of study data:   change data: ‐1.19 (0.13)	(SE or SD)?
Sutton 2002     I1: rosiglitazone 8 mg   C1: glyburide (mean 10.5 mg)	not investigated	not investigated	I1:   end of study data: 8.1 (0.3)   change data:   C1:   end of study data: 8.4 (0.2)   change data:	./.
Yang 2002     I1: rosiglitazone 4 mg   C1: placebo	not investigated	not investigated	I1:   end of study data:   change data: ‐0.7 (1.04)   C1:   end of study data:   change data: 0.4 (1.3)	./.

Footnotes ? = unclear; I = intervention; C = control; o.d. = once daily; b.i.d. = twice daily

Methods	DURATION OF INTERVENTION: 12 months   DURATION OF FOLLOW‐UP:   12 months   RUN‐IN PERIOD: none LANGUAGE OF PUBLICATION: English
Participants	WHO PARTCIPATED:   white patients with type 2 diabetes mellitus and metabolic syndrome   INCLUSION CRITERIA:   white patients of either sex and ages >=18 years; type 2 diabetes according to ADA criteria (duration >=6 months); poor glycaemic control (HbA1c >=7.5% or >=1 adverse effect with diet and oral hypoglycaemic agents (e.g. SU or metformin) given up to the maximum tolerated dose; all patients also diagnosed with metabolic syndrome (National Cholesterol Education Program Adult Treatment Panel III classification; triglyceridaemia (TG >=150 mg/dl) and hypertension (WHO criteria BP >=130/>=85 mmHg); fasting C‐peptide level >1.0 ng/ml   EXCLUSION CRITERIA:   receiving glimepiride, history of ketoacidosis, unstable or rapidly progressive diabetic retinopathy, nephropathy or neuropathy; impaired hepatic function, impaired renal function, severe anaemia; severe cardiovascular disease (e.g. NYHA III or IV congestive heart failure or a history of myocardial infarction or stroke) or cerebrovascular conditions within 6 months before enrolment; women who were pregnant or breastfeeding or of childbearing potential and not taking adequate contraceptive precautions   DIAGNOSTIC CRITERIA:   ADA 2001   CO‐MORBIDITIES:   not stated   CO‐MEDICATIONS:   40.2% receiving antihypertensive drugs; no patient was receiving lipid‐lowering or antiaggregant drugs
Interventions	NUMBER OF STUDY CENTRES:   three   COUNTRY/ LOCATION:   Italy   SETTING:   unclear   INTERVENTION (DOSE/DAY):   rosiglitazone 4 mg once daily (before lunch); +fixed oral dose of glimepiride (4 mg/day divided into 2 doses; before breakfast and before dinner)   CONTROL (DOSE/DAY):   pioglitazone 15 mg once daily (before lunch); + fixed oral dose of glimepiride (4 mg/day divided into 2 doses; before breakfast and before dinner)   TREATMENT BEFORE STUDY:   52.9% poor glycaemic control with metformin; 31% with SUs; 16.1% with glyburide; 14.9% with gliclazide   TITRATION PERIOD:   none
Outcomes	PRIMARY OUTCOMES:   changes in BMI, HbA1c, lipid profile, and lipoprotein variables were the primary efficacy variables SECONDARY OUTCOMES:   fasting and postprandial plasma glucose, insulin levels, insulin resistance (HOMA); blood pressure; adverse events
Notes	AIM OF STUDY:   to assess the differential effect on glucose and lipid variables of the combination of glimepiride plus pioglitazone or rosiglitazone in patients with type 2 diabetes and the metabolic syndrome
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Methods	DURATION OF INTERVENTION:   12 months   DURATION OF FOLLOW‐UP:   12 months   RUN‐IN PERIOD:   none   LANGUAGE OF PUBLICATION:   English
Participants	WHO PARTCIPATED:   Caucasian patients with type 2 diabetes and poor glycaemic control with diet or experiencing adverse effects with diet and metformin, administered up to the maximum tolerated dose   INCLUSION CRITERIA:   patients aged >= 18 years of either sex if they had type 2 diabetes mellitus according to the ADA (duration >= 6 months), and if they had poor glycaemic control   (HbA1c levels > 7.5%) or   experienced adverse effects with diet and metformin, administered up to the maximum tolerated   dose; all patients were diagnosed with metabolic syndrome according to the National Cholesterol   Education Program Adult Treatment Panel III classification and they presented with triglyceridaemia (triglycerides >= 150 mg/dL) and hypertension according to WHO 1999 criteria (systolic/diastolic BP >= 130/ >= 85 mmHg); all patients had a fasting C‐peptide level > 1.0 ng/mL and were overweight (BMI 25.0 ‐ 28.1)   EXCLUSION CRITERIA:   history of ketoacidosis; unstable or rapidly progressive diabetic retinopathy, nephropathy, or neuropathy; impaired hepatic function (defined as plasma aminotransferase and/or gamma‐glutamyltransferase levels higher than the upper limit of normal (ULN) for age and sex], impaired renal function (defined as serum creatinine levels higher than the ULN for age and sex) or severe anaemia; patients with serious cardiovascular disease (e.g. NYHA class I–IV congestive heart failure or a history of myocardial infarction or stroke) or cerebrovascular conditions within 6 months before study enrollment; women who were pregnant, breastfeeding or of childbearing potential and not taking adequate contraceptive precautions   DIAGNOSTIC CRITERIA:   ADA 2001   CO‐MORBIDITIES:   not stated   CO‐MEDICATIONS:   not stated
Interventions	NUMBER OF STUDY CENTRES:   3   COUNTRY/ LOCATION:   Italy   SETTING:   Department of Internal Medicine and Therapeutics, University of Pavia (Pavia, Italy); the ‘G. Descovich’ Atherosclerosis Study Center, ‘D. Campanacci’ Clinical Medicine and Applied Biotechnology Department, University of Bologna (Bologna, Italy); the Diabetes Care Unit at S. Carlo Hospital of Milano (Milano, Italy).   INTERVENTION (DOSE/DAY):   metformin (mean dose 2250 mg/day) + rosiglitazone 4 mg/day (o.d., before lunch)   CONTROL (DOSE/DAY):   metformin (mean dose 2250 mg/day) + pioglitazone 15 mg/day (15 mg o.d., before lunch)   TREATMENT BEFORE STUDY:   diet or diet and metformin, administered up to the maximum tolerated dose   TITRATION PERIOD:   all paitents received metformin beginning with a dose of 1500 mg/day and increasing up to 3000 mg/day, self‐administered for 12 months, this dose depended on the tolerance or glycaemic control of the patients (mean dosage: 2250 ± 750 mg/day)
Outcomes	PRIMARY OUTCOMES:   changes in BMI, HbA1c, lipid profile, lipoprotein variables   SECONDARY OUTCOMES:   (not stated)   "FPG, PPG and HOMA index   were also used to assess efficacy" BMI, HbA1c, fasting and postprandial plasma glucose (FPG, PPG) and insulin levels; HOMA; lipid profile; treatment tolerability
Notes	AIM OF STUDY:   to assess the differential effect on glucose and lipid variables of the combination of metformin plus pioglitazone or metformin plus rosiglitazone in patients with type 2 diabetes mellitus and metabolic syndrome.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Methods	DURATION OF INTERVENTION:   24 weeks   DURATION OF FOLLOW‐UP:   24 weeks   RUN‐IN PERIOD:   during the 1‐week, open‐label lead‐in phase, patients maintained their prescreening dosage of >= 1500 mg/day metformin therapy; LANGUAGE OF PUBLICATION:   English
Participants	WHO PARTCIPATED:   type 2 diabetes patients inadequately controlled on metformin monotherapy.   INCLUSION CRITERIA:   adults (age 20–78 years) with established type 2 diabetes requiring oral therapy; before screening, patients were required to be on a stable dosage of metformin >= 1500 mg/day for >= 8 weeks, HbA1c levels >7.0 and <= 12.0% and BMI >= 23 and <= 45; only patients willing and able to perform self‐blood glucose; women of childbearing potential had to practise acceptable methods of birth control and to have negative pregnancy test results within 72 h of study treatment   EXCLUSION CRITERIA:   marked polyuria and polydipsia with >10% weight loss; the use of any hypoglycaemic agent other than metformin within 8 weeks before screening; anaemia [haemoglobin level: <12.5 g/dl (men) and <11.0 g/dl (women)] and significantly abnormal   renal, cardiac or hepatic dysfunction or disease; pregnant or nursing women and patients with known   sensitivity to any study medications were excluded.   DIAGNOSTIC CRITERIA:   not stated   CO‐MORBIDITIES:   not stated   CO‐MEDICATIONS:   not stated
Interventions	NUMBER OF STUDY CENTRES:   76   COUNTRY/ LOCATION:   USA   SETTING:   not stated   INTERVENTION (DOSE/DAY):   metformin 500 mg plus rosiglitazone 4 mg/day (initial daily dose 1000–2000 mg + 4 mg, depending on previous treatment)   [mean final dose of metformin plus rosiglitazone was 1819 and 7.1 mg]   CONTROL (DOSE/DAY):   metformin‐glibenclamide 500/2.5 mg/day (initial daily dose 1000/5 mg)   [mean final dose of metformin‐glibenclamide tablets was 1509/7.6 mg]   TREATMENT BEFORE STUDY:   patients were required to be on a stable dosage of metformin >= 1500 mg/day for >= 8 weeks   TITRATION PERIOD:   patients were randomly assigned to one of two double‐blind treatments, according to the dose of metformin during the lead‐in phase:   patients receiving 1500 mg/day metformin before screening received metformin‐glibenclamide 1000/5 mg/day (in divided doses) or metformin 1500 mg plus rosiglitazone 4 mg daily (in divided doses); those previously receiving >1500 mg/day were randomly assigned to metformin‐glibenclamide 1000/5 mg (in divided doses) or metformin 2000 mg plus rosiglitazone 4 mg daily (in divided doses) study medications were titrated based on mean daily glucose levels to achieve a therapeutic glycaemic target
Outcomes	PRIMARY OUTCOMES:   change in HbA1c from baseline to week 24 or the last prior blinded visit   SECONDARY OUTCOMES:   changes in body weight and changes in fructosamine, FPG,   2‐h postprandial plasma glucose and fasting insulin levels from baseline to week 24 or the last prior blinded visit; proportion of patients achieving therapeutic glycaemic response (HbA1c levels <7.0% and FPG levels <7 mmol/L) at week 24 or the   last prior blinded visit; safety outcomes included adverse events, particularly hypoglycaemic symptoms;   standard haematology, serum chemistry and urinalysis
Notes	AIM OF STUDY:   to compare the effects of two combination regimens, metformin‐glibenclamide combination tablets versus metformin plus rosiglitazone in patients inadequately controlled on metformin monotherapy.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Methods	DURATION OF INTERVENTION:   52 weeks   DURATION OF FOLLOW‐UP:   52 weeks   RUN‐IN PERIOD:   eligible patients on oral antidiabetic medication stopped treatment 2 weeks before starting a 4‐week, single‐blind placebo run‐in period   LANGUAGE OF PUBLICATION:   English
Participants	WHO PARTCIPATED:   patients with type 2 diabetes   INCLUSION CRITERIA:   FPG = 7.0 ‐ 15.0 mmol/L C‐peptide >= 27 nmol/L; BMI = 22 ‐ 38   EXCLUSION CRITERIA:   patients on insulin therapy or those with diabetic complications requiring treatment, heart failure NYHA III/IV, or serious renal, hepatic (liver function tests > 2.5 times the upper limit of normal). haematologic impairment or women of childbearing potential   DIAGNOSTIC CRITERIA:   see above   CO‐MORBIDITIES:   not stated   CO‐MEDICATIONS:   not stated
Interventions	NUMBER OF STUDY CENTRES:   71   COUNTRY/ LOCATION:   8 European countries   SETTING:   not stated   INTERVENTION (DOSE/DAY):   rosiglitazone as two equal daily doses (i.e. 2 mg bid or 4 mg bid) + placebo   CONTROL (DOSE/DAY):   glibenclamide once daily + placebo   TREATMENT BEFORE STUDY:   patients on monotherapy, combination therapy or diet and exercise only   TITRATION PERIOD:   over the first 12 weeks of treatment, the glibenclamide dose was titrated in 2.5 mg increments (final dose   range = 2.5 ‐ 15 mg) to achieve optimal glycaemic   control a double‐dummy system allowed ‘‘titration’’   of rosiglitazone without a change of dose concomitant medications with potential effects on glucose or lipid metabolism were kept at constant dose   throughout the study
Outcomes	PRIMARY OUTCOMES:   difference between rosiglitazone 8 mg/day and glibenclamide treatment groups with respect to change from baseline in HbA1c after 52 weeks of treatment   SECONDARY OUTCOMES:   (not stated)   lipids, insulin resistance (HOMA), insulin, proinsulin, 32‐33 split proinsulin, safety, adverse effects
Notes	AIM OF STUDY:   to compare the efficacy, tolerability and safety of rosiglitazone with that of   glibenclamide as monotherapy for patientswith type 2 diabetes over a 12‐month treatment period
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Methods	DURATION OF INTERVENTION:   26 weeks   DURATION OF FOLLOW‐UP:   26 weeks   RUN‐IN PERIOD:   4‐weeks with written diet instructions   LANGUAGE OF PUBLICATION:   English
Participants	WHO PARTCIPATED:   patients with newly diagnosed or diet‐treated type 2 diabetes   INCLUSION CRITERIA:   patients with type 2 diabetes, as defined by the WHO criteria and no diabetes complications   EXCLUSION CRITERIA:   fasting plasma glucose value < 6.1 mmol/l or > 11.0 mmol/L after the run‐in period; patients with cardiovascular disease, blood pressure > 160/100 mm Hg, previous or present abnormal hepatic or renal function, antidiabetic medication, anemia, or oral corticosteroid treatment   DIAGNOSTIC CRITERIA:   WHO 1998   CO‐MORBIDITIES:   not stated   CO‐MEDICATIONS:   not stated
Interventions	NUMBER OF STUDY CENTRES:   not stated (1)   COUNTRY/ LOCATION:   Finland   SETTING:   the patients were recruited by advertisement and among clients of the occupational health service in Turku, Finland   INTERVENTION (DOSE/DAY):   rosiglitazone 8 mg/day (4 mg b.i.d.)   CONTROL (DOSE/DAY):   C1: metformin 2 g (1g b.i.d.)   C2: placebo   TREATMENT BEFORE STUDY:   none or diet only   TITRATION PERIOD:   rosiglitazone (2 mg b.i.d. for 2 weeks, thereafter 4 mg b.i.d.), metformin (500 mg b.i.d. for 2 weeks, thereafter 1 g b.i.d.), or placebo
Outcomes	PRIMARY OUTCOMES:   not stated (insulin‐ and exercise‐stimulated skeletal muscle glucose uptake, measured by means of positron emission tomography (PET) during euglycemic‐hyperinsulinemic   clamp and one‐legged exercise)   SECONDARY OUTCOMES:   (not stated)   FPG, insulin, HbA1c, body weight, blood pressure
Notes	AIM OF STUDY:   to compare the effects of treatment with rosiglitazone and metformin on insulin‐ and exercise‐stimulated glucose uptake and perfusion in skeletal muscle tissue in patients with type 2 diabetes.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Methods	DURATION OF INTERVENTION:   6 months   DURATION OF FOLLOW‐UP:   6 months   RUN‐IN PERIOD:   none   LANGUAGE OF PUBLICATION:   English
Participants	WHO PARTCIPATED:   Koreans with type 2 diabetes mellitus who showed poor glycaemic control with glimepiride   INCLUSION CRITERIA:   aged 20‐70 years; secondary treatment failure (HbA1c > 8% on glimepiride 4 mg/day or equivalent dose of other sulfonylureas)   EXCLUSION CRITERIA:   no other severe illnesses including liver failure, renal failure, heart failure   DIAGNOSTIC CRITERIA:   not stated   CO‐MORBIDITIES:   retinopathy ‐   I: 3/14, C: 3/13   proteinuria ‐   I: 2/14, C: 3/13   coronary heart disease ‐   I: 2/14, C: 2/13   CO‐MEDICATIONS   lipid‐lowering agents ‐   I: 5/14, C: 3/13
Interventions	NUMBER OF STUDY CENTRES:   1   COUNTRY/ LOCATION:   Korea   SETTING:   diabetes clinic of the Seoul National University Hospital   INTERVENTION (DOSE/DAY):   rosiglitazone 4 mg/day + glimepiride 4 mg/day   CONTROL (DOSE/DAY):   metformin 1000 mg/day + glimepiride 4 mg/day   TREATMENT BEFORE STUDY:   glimepiride 4 mg/day or equivalent dose of other sulfonylureas   TITRATION PERIOD:   none
Outcomes	PRIMARY OUTCOMES:   not stated   (resistin)   SECONDARY OUTCOMES:   (not stated)   adiponectin, FPG, lipids, HbA1c, plasma insulin, plasma C‐peptide
Notes	AIM OF STUDY:   to see whether improving insulin resistance can modulate circulating resistin levels, the effects of two different insulin sensitizers, rosiglitazone and metformin, on plasma resistin concentrations in Korean participants with type 2 diabetes mellitus were investigated
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Methods	DURATION OF INTERVENTION:   4.0 years (median)   DURATION OF FOLLOW‐UP:   4.0 years (median)   RUN‐IN PERIOD:   4 weeks, placebo + diet/exercise   LANGUAGE OF PUBLICATION:   English
Participants	WHO PARTCIPATED:   people with recently diagnosed type 2 diabetes mellitus, treated with life style management only   INCLUSION CRITERIA:   eligible patients were between the ages of 30 and 75 years, with fasting plasma glucose levels ranging from 126 to 180 mg per deciliter (7.0 to 10.0 mmol per liter) while their only treatment was lifestyle management   EXCLUSION CRITERIA:   clinically significant hepatic disease, renal impairment, a history of lactic acidosis, unstable or severe angina, known congestive heart failure (CHF, New York Heart Association class I, II, III, or IV), or uncontrolled hypertension   DIAGNOSTIC CRITERIA:   "recently diagnosed (i.e., within 3 years)"   CO‐MORBIDITIES:   not stated   CO‐MEDICATIONS:   Antihypertensive therapy [no. (%)]:   I1: 744 (51.1)   C1: 737 (50.7)   C2: 753 (52.3) Lipid‐lowering therapy [no. (%)]   I1: 378 (26.0)   C1: 377 (25.9)   C2: 370 (25.7)
Interventions	NUMBER OF STUDY CENTRES:   488   COUNTRY/ LOCATION:   United States, Canada, and 15 European countries   SETTING:   not stated   INTERVENTION (DOSE/DAY):   rosiglitazone (max 8 mg/day)   CONTROL (DOSE/DAY):   metformin (max 2g/day)   glyburide (max 15 mg/day)   TREATMENT BEFORE STUDY:   diet/exercise   TITRATION PERIOD:   patients received initial daily doses of 4 mg of rosiglitazone, 500 mg of metformin, or 2.5 mg of glyburide for each drug, the dose was increased according to the protocol to the maximum daily effective dose (4 mg of rosiglitazone twice daily, 1 g of metformin twice daily, and 7.5 mg of glyburide twice daily a dose increase was required at each visit if the fasting plasma glucose level was 140 mg per deciliter or more; a dose reduction was permitted if adverse events occurred
Outcomes	PRIMARY OUTCOMES:   time from randomization to treatment failure, which was defined as confirmed hyperglycemia (fasting plasma glucose level, >180 mg/dl) on consecutive testing after at least 6 weeks of treatment at the maximum‐dictated or maximum‐tolerated dose   of the study drug an independent adjudication   committee, whose members were unaware of assignments   to treatment groups, used prespecified criteria (available   at www.nejm.org) to determine whether the primary outcome   was reached in cases in which a confirmatory fasting plasma glucose level had not been obtained, a patient had withdrawn because of an insufficient therapeutic effect, or an additional glucose lowering drug had been administered before the   confirmation of hyperglycemia (according to a protocol   amendment adopted in February 2004) the threshold of more than 180 mg per deciliter for confirmed hyperglycemia was selected to   represent unequivocal failure in the maintenance of adequate glycemic control without incurring undue hyperglycemic symptoms; the threshold of a fasting plasma glucose level of more than   140 mg per deciliter for increasing the dose of a study drug reflected clinical guidelines at the time of study design.   SECONDARY OUTCOMES:   time from randomization to a confirmed fasting plasma glucose level of more than 140 mg per deciliter after at least 6 weeks of treatment at the   maximum‐tolerated dose of a study drug (for patients who entered the study with a fasting plasma glucose level of 140 mg per deciliter or less) other prespecified outcomes were levels of fasting plasma   glucose and glycated hemoglobin, weight, and   measures of insulin sensitivity and beta‐cell function, as determined by homeostasis model assessment (HOMA 2) with the use of the HOMA calculator (www.dtu.ox.ac.uk) Secondary endpoints according to the published study protocol (Diabetes Care 2002): ‐ glycaemic control   ‐ insulin sensitivity   ‐ beta‐cell function   ‐ cardiovascular risk markers   ‐ renal function   ‐ patient reported outcomes (quality of life)   ‐ resource utilization (direct health care costs will be assessed as the number of emergency room visits, number of unscheduled visits to the study physician’s office, number of hospitalizations, and length of stay. Furthermore, indirect economic costs associated   with bed days (days when patients stay in bed for half a day or more) and restricted activity days (days when patients reduce their usual activities, such as housework   or shopping) will be evaluated)   ‐ safety parameters (including hypoglycaemia)
Notes	AIM OF STUDY:   to evaluate the durability of glycemic control in   patients receiving monotherapy with rosiglitazone, metformin or glyburide zhe therapeutic goal was a fasting plasma glucose   level below 140 mg/dl (7.8 mmol/L).
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Methods	DURATION OF INTERVENTION:   one year   DURATION OF FOLLOW‐UP:   one year   RUN‐IN PERIOD:   none   LANGUAGE OF PUBLICATION:   English
Participants	WHO PARTCIPATED:   Chinese patients with type 2 diabetes and conventional oral antidiabetic drugs failure   INCLUSION CRITERIA:   OAD failure was defined as persistent hyperglycaemia with haemoglobin AIc (HbA1c) >= 8.5% for 6 mo or longer despite continuous use of maximal doses of conventional OAD; maximum recommended doses of various OADs were given as follows: glibenclamide 20 mg/d, gliclazide 320 mg/d, glipizide 20 mg/d, and metformin 3 g/d   EXCLUSION CRITERIA:   uncontrolled hypertension   with sitting blood pressure (BP) >200/110 mm Hg and/or a history of myocardial   infarction, cerebrovascular accident, or uncontrolled congestive heart failure during the previous 6 mo, or significant renal impairment (plasma creatinine concentration >= 150 mmol/L)   DIAGNOSTIC CRITERIA:   not stated   CO‐MORBIDITIES:   not stated   CO‐MEDICATIONS:   antihypertensive agents [no (%)]:   I1: 31 (55.3)   C1: 14 (25.0) lipid‐lowering agents [no (%)]:   I1: 5 (8.9)   C1: 2 (3.6)
Interventions	NUMBER OF STUDY CENTRES:   1   COUNTRY/ LOCATION:   Hong Kong, China   SETTING:   Diabetic Clinic and Diabetes Center at AH Nethersole Hospital, in Tai PO, Hong Kong.   INTERVENTION (DOSE/DAY):   rosiglitazone max 8 mg/d   CONTROL (DOSE/DAY):   bedtime isophane insulin   TREATMENT BEFORE STUDY:   OAD ‐ original OAD and other medications remained the same throughout the study patients who fulfilled the inclusion criteria were referred to dietitians and diabetic nursing specialists for reinforcement of their dietary habits, drug compliance, and an understanding of OAD failure; those with HbA1c >=8.5% three months after reinforcement were included   TITRATION PERIOD:   oral rosiglitazone was started at 2 mg/d, insulin was begun at a dose of 6 units administered at night; the insulin dose was titrated 2 to 4 wk later by a diabetic nursing specialist with an increment of 2 to 4 units according to tolerability of the insulin injection and fasting plasma glucose (PG) improvement at 12, 24, 36, and 52 wk, all patients were seen for assessment of tolerability   and compliance with treatment, and for measurement of lipid, glycemic, and other biochemical indices; insulin dosage was adjusted at each visit if this was deemed necessary, with the goal of achieving an HbA1c concentration <7.5%; if the drug was tolerable to patients, rosiglitazone was also increased to the maximum dose of 8 mg   daily, with the goal of reducing HhA1c to <7.5% without the occurrence of significant   hypoglycemia
Outcomes	PRIMARY OUTCOMES:   not stated (differences in HbA1c)   SECONDARY OUTCOMES:   (not stated)   lipids, BMI, FPG, blood pressure
Notes	AIM OF STUDY:   to evaluate the efficacy and tolerability of rosiglitazone in patients with secondary oral anti‐diabetic drug failure and to directly compare rosiglitazone with bedtime insulin
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Study	Reason for exclusion
Bailey 2005	treatment regimens not comparable (rosiglitazone plus metformin versus metformin)
Baksi 2004	treatment regimens not comparable (rosiglitazone plus gliclazide versus gliclazide)
Barnett 2003	treatment regimens not comparable (rosiglitazone plus sulfonylureas versus sulfonylureas plus placebo)
Dailey 2004	treatment regimens not comparable (rosiglitazone plus glyburide/metformin versus placebo plus glyburide/metformin)
Fonseca 2000	treatment regimens not comparable (rosiglitazone plus metformin versus placebo plus metformin)
Fonseca 2003	treatment regimens not comparable (rosiglitazone plus placebo versus rosiglitazone plus nateglinide)
Gomez‐Perez 2002	treatment regimens not comparable (rosiglitazone plus metformin versus placebo plus metformin)
Hubacek 2004	rosiglitazone treatment less than 24 weeks
Kerenyi 2004	treatment regimens not comparable (rosiglitazone plus glibenclamide versus glibenclamide)
McCluskey 2004	treatment regimens not comparable (rosiglitazone plus glimepiride versus glimepiride)
Negro 2005	treatment regimens not comparable (rosiglitazone plus metformin versus placebo plus metformin)
Raskin 2001	treatment regimens not comparable (rosiglitazone plus insulin versus placebo plus insulin)
Reynolds 2002	treatment regimens not comparable (rosiglitazone plus insulin plus life‐style modification versus placebo plus insulin plus life‐style modification)
Rosenstock 2006a	treatment regimens not comparable (rosiglitazone plus glipizide versus placebo plus glipizide)
Tan 2005a	rosiglitazone treatment less than 24 weeks
Tan 2005b	rosiglitazone treatment less than 24 weeks
Vongthavaravat 2002	treatment regimens not comparable (rosiglitazone plus sulfonylureas versus sulfonylureas alone)
Wang 2005	treatment regimens not comparable (rosiglitazone versus "control" without treatment)
Weissman 2005	treatment regimens not comparable (rosiglitazone plus metformin versus metformin)
Wolffenbuttel 2000	treatment regimens not comparable (rosiglitazone plus sulfonylureas versus sulfonylureas plus placebo)
Wong 2005	treatment regimens not comparable (rosiglitazone plus insulin versus insulin)
Zhu 2003	treatment regimens not comparable (rosiglitazone plus sulfonylureas versus sulfonylureas plus placebo)

Trial name or title	Rosiglitazone Evaluated for Cardiac Outcomes and Regulation of Glycaemia in Diabetes (RECORD)
Methods
Participants	RECORD is a 6‐year, randomised, open‐label study in   type 2 diabetic patients with inadequate blood glucose control (HbA1c 7.1‐9.0%) on metformin or sulphonylurea alone.
Interventions	after a 4‐week run‐in, participants are randomised by current treatment stratum to   add‐on rosiglitazone, metformin or sulphonylurea, with   dose titration to a target HbA1c of <=7.0%; if confirmed HbA1c rises to >= 8.5%, either a third glucose‐lowering drug   is added (rosiglitazone‐treated group) or insulin is started   (non‐rosiglitazone group); the same criterion for failure of triple oral drug therapy in the rosiglitazone‐treated group is used for starting insulin in this group
Outcomes	the primary endpoint is the time to first cardiovascular hospitalisation or death, blindly adjudicated by a central endpoints committee; the   study aim is to evaluate non‐inferiority of the rosiglitazone   group versus the non‐rosiglitazone group with respect to cardiovascular outcomes; safety, tolerability and study conduct are monitored by an independent board
Starting date	recruitment began in April 2001 and was completed in April 2003
Contact information	P. D. Home   School of Clinical Medical Sciences‐Diabetes,   University of Newcastle upon Tyne,   Medical School, Framlington Place,   Newcastle upon Tyne,   NE2 4HH, UK E‐mail: philip.home@newcastle.ac.uk   Tel.: +44‐191‐2227019   Fax: +44‐191‐2220723
Notes	study design and protocol published in Diabetologia 2005;48: 1726–35

PERMALINK

Rosiglitazone for type 2 diabetes mellitus

Bernd Richter

Elizabeth Bandeira‐Echtler

Karla Bergerhoff

Christine Clar

Susanne H Ebrahim

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

Plain language summary

Background

Description of the condition

Type 1 diabetes mellitus

Type 2 diabetes mellitus

Description of the intervention

Adverse effects of the intervention

How the intervention might work

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Excluded interventions

Types of outcome measures

Primary outcomes

Secondary outcomes

Covariates, effect modifiers and confounders

Timing of outcome measurement

Search methods for identification of studies

Electronic searches

Searching other resources

Data collection and analysis

Selection of studies

Dealing with duplicate publications

Data extraction and management

Assessment of risk of bias in included studies

Measures of treatment effect

Dichotomous data

Continuous data

Time‐to‐event data

Unit of analysis issues

Dealing with missing data

Assessment of heterogeneity

Assessment of reporting biases

Data synthesis

Subgroup analysis and investigation of heterogeneity

Sensitivity analysis

Results

Description of studies

Results of the search

1.

Assessment of publication bias inter‐rater agreement

Included studies

Interventions

Comparisons

Monotherapy

Combination therapy

Number of study centres

Country and location

Setting

Treatment before study

Methods

Duration of the intervention

Duration of follow‐up

Run‐in period

Language of publication

Participants

Who participated

Inclusion criteria

Exclusion criteria

Diagnostic criteria

Co‐morbidities