Diabetes: glycaemic control in type 2 (drug treatments)

Abstract

Introduction

Diabetes mellitus is a progressive disorder of glucose metabolism. It is estimated that about 285 million people between the ages of 20 and 79 years had diabetes worldwide in 2010, or 5% of the adult population. Type 2 diabetes may occur with obesity, hypertension, and dyslipidaemia (the metabolic syndrome), which are powerful predictors of cardiovascular disease. Without adequate blood-glucose-lowering treatment, blood glucose levels may rise progressively over time in people with type 2 diabetes. Microvascular and macrovascular complications may develop.

Methods and outcomes

We conducted a systematic review and aimed to answer the following clinical question: What are the effects of blood-glucose-lowering medications in adults with type 2 diabetes? We searched: Medline, Embase, The Cochrane Library, and other important databases up to February 2010 (Clinical Evidence reviews are updated periodically, please check our website for the most up-to-date version of this review). We included harms alerts from relevant organisations such as the US Food and Drug Administration (FDA) and the UK Medicines and Healthcare products Regulatory Agency (MHRA).

Results

We found 194 systematic reviews, RCTs, or observational studies that met our inclusion criteria. We performed a GRADE evaluation of the quality of evidence for interventions.

Conclusions

In this systematic review we present information relating to the effectiveness and safety of the following interventions: alpha-glucosidase inhibitors (AGIs), combination treatment (single, double, and triple), dipeptidyl peptidase-4 (DPP-4) inhibitors, glucagon-like peptide-1 (GLP-1) analogues, insulins (including conventional [human] and analogue, different regimens, different length of action), meglitinides, metformin, sulphonylureas, and thiazolidinediones.

Key Points

Diabetes mellitus affects about 6.5% of people aged 20 to 79 years worldwide. In 2010, an estimated 285 million people have diabetes, over 85% of whom have type 2 diabetes.

Type 2 diabetes is often associated with obesity, hypertension, and dyslipidaemia, which are all powerful predictors of cardiovascular disease. For that reason, the treatment of type 2 diabetes requires a multifactorial approach, including lifestyle advice, treatment of hypertension, and lowering of lipid levels.

Without adequate blood-glucose-lowering treatment, blood glucose levels may rise progressively over time in people with type 2 diabetes. Microvascular and macrovascular complications may develop.

Metformin reduces HbA1c effectively compared with placebo.

• The UK Prospective Diabetes Study (UKPDS) RCT found that metformin may be moderately protective against mortality and cardiovascular morbidity, but further high-quality studies are needed.

• We found no evidence to suggest that metformin increases the risk of lactic acidosis.

Sulphonylureas reduce HbA1c by 1% compared with placebo, and they may reduce microvascular complications compared with diet alone. They can cause weight gain and hypoglycaemia. One review found that the risk of hypoglycaemia was highest with glibenclamide compared with other second-generation sulphonylureas.

The effectiveness of the combination of metformin and sulphonylurea on mortality and morbidity is unknown.

Meglitinides reduce HbA1c by about 0.4–0.9% compared with placebo, but robust data are sparse.

Alpha-glucosidase inhibitors reduce HbA1c by about 0.8% compared with placebo. We found no reports of dangerous adverse effects.

Thiazolidinediones reduce HbA1c by 1.0% compared with placebo but may increase the risk of congestive heart failure and bone fractures. Rosiglitazone increases the risk of MI.

• DRUG ALERT: Rosiglitazone has been withdrawn from the market in many countries because the benefits of treatment are no longer thought to outweigh the risks.

Dipeptidyl peptidase-4 (DPP-4) inhibitors reduce HbA1c by about 0.6–0.7% compared with placebo. We found no long-term data on effectiveness and safety.

Glucagon-like peptide-1 (GLP-1) analogues reduce HbA1c compared with placebo and result in weight loss. We found no long-term data on effectiveness and safety.

Combined oral drug treatment may reduce HbA1c levels more than monotherapy, but increases the risk of hypoglycaemia.

Insulin improves glycaemic control in people with inadequate control of HbA1c on oral drug treatment, but is associated with weight gain, and an increased risk of hypoglycaemia.

Adding metformin to insulin may reduce HbA1c levels compared with insulin alone, with less weight gain.

Insulin analogues, short-acting, long-acting, and combined in various regimens, seem no more effective than conventional (human) insulin in reducing HbA1c levels. However, in people presenting with recurrent hypoglycaemic episodes, long-acting insulin analogues may be preferred above human insulin.

Long-acting insulin analogues seem equally effective at reducing HbA1c.

There is lack of evidence about the effectiveness of various insulin analogue regimens after once-daily long-acting insulin has failed.

The effectiveness of insulin basal bolus regimens is not well established.

About this condition

Definition

The term diabetes mellitus encompasses a group of disorders characterised by chronic hyperglycaemia with disturbances of carbohydrate, fat, and protein metabolism resulting from defects of insulin secretion, insulin action, or both. Type 2 diabetes is the most common form of diabetes, and defects of both insulin action and insulin secretion are usually present by the time of diagnosis. WHO recognises diabetes as a progressive disorder of glucose metabolism in which individuals may proceed from normoglycaemia (fasting plasma venous glucose <5.5 mmol/L), impaired glucose tolerance (fasting plasma venous glucose <7.0 mmol/L and plasma glucose between 7.8 mmol/L and 11.1 mmol/L 2 hours after a 75 g oral glucose load, the oral blood glucose tolerance test [OGTT]), impaired fasting glycaemia (fasting venous plasma glucose between 5.6 mmol/L and 7.0 mmol/L), and diabetes.[1] As a consequence of the inability of the body to use glucose as an energy source, blood glucose levels rise and symptoms such as thirst, polyuria, blurring of vision, or weight loss may develop. Diagnosis: Since 1965, WHO has published guidelines for the diagnosis and classification of diabetes. In 2006, WHO decided that the diagnostic criteria should be maintained.[1] In the presence of symptoms, diabetes may be diagnosed on the basis of a single random elevated plasma glucose (11.1 mmol/L or more). In the absence of symptoms, the diagnosis should be based on blood glucose results in the diabetes range taken at different time points, either from a random sample, or fasting (plasma blood glucose 7.0 mmol/L or more), or from the OGTT (plasma blood glucose 11.1 mmol/L or more 2 hours after a 75 g glucose load).[1] Population: For the purpose of this review, we have excluded pregnant women and acutely unwell adults (e.g., after surgery or MI), and people with secondary diabetes (e.g., those with hyperglycaemia based on temporal use of corticosteroids).

Incidence/ Prevalence

It is estimated that about 285 million people between the ages of 20 and 79 years had diabetes worldwide in 2010, or 5% of the adult population.[2] This number will increase to about 438 million in 2030, an estimated prevalence of 7.7%, in the previously mentioned age category. By 2025, the region with the greatest number of people with diabetes is expected to be South-East Asia, with about 82 million people with type 2 diabetes. Incidence and prevalence figures for children and adolescents are unreliable, but there is some evidence that type 2 diabetes is becoming more common in adolescents and young adults, especially in resource-poor countries. The overall estimated prevalence of 6.5% for type 2 diabetes conceals considerable variation in prevalence, which ranges from <2% in some African countries to >14% in some populations.[2]

Aetiology/ Risk factors

By definition, the specific reasons for the development of the defects of insulin secretion and action that characterise type 2 diabetes are unknown. The risk of type 2 diabetes increases with age and lack of physical activity, and the disease occurs more frequently in people with obesity, hypertension, and dyslipidaemia (the metabolic syndrome). Type 2 diabetes also occurs more frequently in women with previous gestational diabetes and certain ethnic groups. There is also evidence of a familial, probably genetic, predisposition.[1]

Prognosis

People with type 2 diabetes have blood glucose levels that have been shown to rise progressively from the time of diagnosis. During the UK Prospective Diabetes Study (UKPDS), HbA1c levels rose in newly diagnosed people with type 2 diabetes, irrespective of the type of treatment given.[3]In 2011, primary care physicians in Denmark, the UK, and the Netherlands succeeded in lowering HbA1c levels in screen-detected type 2 diabetes patients for more than 5 years after diagnosis.[4] Blood glucose levels above the normal range have been shown to be associated not only with the presence of symptoms, but also with an increased risk of long-term microvascular and macrovascular complications. Early treatment of hyperglycaemia in the UKPDS over 9 years resulted in a significant decrease in microvascular complications and a continued reduction in microvascular risk and emergent risk reductions for MI and death from any cause during 10 years of post-trial follow-up.[5] However, in people with longstanding type 2 diabetes, the effects of treating hyperglycaemia are less positive[6] or even absent.[7] Data from a large General Practice Research Database show that both low and high mean HbA1c values are associated with increased all-cause mortality and cardiac events. Both intensive insulin treatment and the risk of hypoglycaemia have been linked to an increased death rate.[8]

Aims of intervention

To control blood glucose levels in order to maximise quality of life and prevent diabetic emergencies, such as ketoacidosis and non-ketotic hyperosmolar coma; to reduce the risk of microvascular and macrovascular complications; all these aims achieved while minimising adverse effects of treatment such as hypoglycaemia, weight gain, or diseases such as cardiovascular disease or cancer.

Outcomes

Mortality (all cause, cardiovascular); morbidity (macrovascular, microvascular), glycaemic control (glycated haemoglobin, e.g., HbA1c); quality of life; adverse effects (including body weight, hypoglycaemia, and other adverse effects). Where possible, we have tried to report clinical outcomes that matter to people, such as mortality and morbidity. However, many studies were underpowered for these outcomes and only reported on laboratory-based outcomes such as glycaemic control (e.g., HbA1c). It should be noted that some adverse effects (e.g., hypoglycaemia), are thought to be under-reported; some hypoglycaemic episodes may be asymptomatic, and even severe episodes may not be recorded in trials.

Methods

Clinical Evidence search and appraisal February 2010. The following databases were used to identify studies for this systematic review: Medline 1966 to February 2010, Embase 1980 to February 2010, and The Cochrane Database of Systematic Reviews 2010, Issue 1 (1966 to date of issue). Note: for new options and comparisons added in the February 2010 update, Medline, Embase, and The Cochrane Library were searched from January 2000 to February 2010. An additional search within The Cochrane Library was carried out for the Database of Abstracts of Reviews of Effects (DARE) and Health Technology Assessment (HTA). We also searched for retractions of studies included in the review. Abstracts of the studies retrieved from the initial search were assessed by an information specialist. Selected studies were then sent to the contributor for additional assessment, using predetermined criteria to identify relevant studies. Quality criteria: Study design criteria for inclusion in this review were: published systematic reviews of RCTs and RCTs in English language, at least single blinded, and containing >20 individuals of whom >80% were followed up. The minimum length of follow-up required was 24 weeks. Studies were at least assessor-blinded: we excluded all studies described as "open", "open label", or not blinded with the exception of insulin studies, where open (non-blinded) studies were allowed. For the comparison of human insulin versus insulin analogues, we included reviews or RCTs that compared agents with similar profiles. Human insulin and insulin analogues were searched for as basal and premixed insulin. It is not possible to blind RCTs of insulin analogues as they differ in appearance to conventional (human) soluble insulin. Lifestyle interventions were excluded as well as studies testing the effects of multiple intervention programmes. In reporting, we have tried to only report analyses in reviews in which all the RCTs attained our minimum methodological criteria for inclusion. However, some analyses included RCTs below these minimum criteria. Where any analysis we have reported has included RCTs below these inclusion criteria, we have specified this (e.g., if RCTs were <24 weeks' duration or if open-label RCTs were included). In general, we only report data from a whole class of drugs if data on the individual drugs were missing or limited. We preferred to analyse different drugs from the same class of drugs separately as the effects of drugs within a class may vary, and readers will therefore have evidence on any individual drug that they may wish to use compared with any other specific drug. However, we have sometimes reported class effects, where appropriate. We included systematic reviews of RCTs and RCTs where harms of an included intervention were studied applying the same study design criteria for inclusion as we did for benefits. In addition, we use a regular surveillance protocol to capture harms alerts from organisations such as the FDA and the MHRA, which are added to the reviews as required. General: To aid readability of the numerical data in our reviews, we round many percentages to the nearest whole number. Readers should be aware of this when relating percentages to summary statistics such as relative risks (RRs) and odds ratios (ORs). We have performed a GRADE evaluation of the quality of evidence for interventions included in this review (see table ). The categorisation of the quality of the evidence (into high, moderate, low, or very low) reflects the quality of evidence available for our chosen outcomes in our defined populations of interest. These categorisations are not necessarily a reflection of the overall methodological quality of any individual study, because the Clinical Evidence population and outcome of choice may represent only a small subset of the total outcomes reported, and population included, in any individual trial. For further details of how we perform the GRADE evaluation and the scoring system we use, please see our website (www.clinicalevidence.com). Since the last version of this review: This update of the review has included further new options (glucosidase inhibitors, thiazolidinediones, glucagon-like peptide-1 [GLP-1] analogues, dipeptidyl peptidase-4 [DPP-4 inhibitors], triple therapy, insulin analogues versus each other) including different combinations of agents. The inclusion criteria for the review have been changed at this update with RCTs now requiring a minimum follow-up of at least 24 weeks, as opposed to the previously specified no minimum length of follow-up apart from for RCTs reporting HbA1c results, which required a minimum follow-up of 3 months. In addition, non-blinded (open) studies involving insulin (whether as monotherapy or in combination with oral agents) have now been included in the review. Hence, the reporting in all the options in the previous version of this review have been revised because of the altered inclusion and exclusion criteria.

Table 1.

GRADE evaluation of interventions for diabetes: glycaemic control in type 2

Important outcomes	Mortality, morbidity (macrovascular, microvascular), glycaemic control, quality of life, adverse effects (body weight, hypoglycaemia), other adverse effects
Number of studies (participants)	Outcome	Comparison	Type of evidence	Quality	Consistency	Directness	Effect size	GRADE	Comment
What are the effects of blood-glucose-lowering medications in adults with type 2 diabetes?
1 (350)[9]	Mortality	Metformin v placebo	4	0	0	–2	0	Low	Directness points deducted for small number of events (1 event only) and short-term follow-up for this outcome (24 weeks)
2 (377)[9]	Morbidity	Metformin v placebo	4	0	0	–2	0	Low	Directness points deducted for small number of events (2 events only) and short-term follow-up for this outcome (24–26 weeks)
12 (1587)[9]	Glycaemic control	Metformin v placebo	4	0	–1	0	0	Moderate	Consistency point deducted for statistical heterogeneity
10 (1162)[9]	Body weight	Metformin v placebo	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for composite outcome (BMI/weight change)
3 (805)[9]	Hypoglycaemia	Metformin v placebo	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for small number of events (13 in total)
1 (419)[9]	Mortality	Metformin vsulphonylurea	4	0	0	–2	0	Low	Directness points deducted for small number of events (1 event only) and short-term follow-up for this outcome (29 weeks)
1 (419)[9]	Morbidity	Metformin v sulphonylurea	4	0	0	–2	0	Low	Directness points deducted for small number of events (1 event only) and short-term follow-up for this outcome (29 weeks)
1 (4360)[13]	Morbidity	Metformin v glibenclamide or rosiglitazone	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for high loss to follow-up
At least 18 (at least 2376)[9] [10]	Glycaemic control	Metformin v sulphonylurea	4	–1	–1	0	0	Low	Quality point deducted for weak methods. Consistency point deducted for statistical heterogeneity
At least 4 (unclear)[9] [10]	Body weight	Metformin v sulphonylurea	4	–1	0	0	0	Moderate	Quality point deducted for incomplete reporting of results
At least 8 (at least 1624)[9] [10]	Hypoglycaemia	Metformin v sulphonylurea	4	–2	0	0	0	Low	Quality points deducted for weak methods and incomplete reporting of results
At least 7 (at least 6454)[13] [9] [10]	Glycaemic control	Metformin v thiazolidinediones	4	0	–1	–1	0	Low	Consistency point deducted for statistical heterogeneity. Directness point deducted for high loss to follow-up in 1 large RCT
At least 6 (unclear)[9] [10]	Body weight	Metformin v thiazolidinediones	4	–1	0	0	0	Moderate	Quality point deducted for incomplete reporting of results
2 (223)[14] [9]	Glycaemic control	Metformin v alpha-glucosidase inhibitors	4	0	–1	0	0	Moderate	Consistency point deducted for statistical heterogeneity
2 (223)[14] [9]	Body weight	Metformin v alpha-glucosidase inhibitors	4	0	–1	–1	0	Low	Consistency point deducted for significant heterogeneity. Directness point deducted for composite outcome (BMI/weight change)
2 (454)[9] [15] [16]	Glycaemic control	Metformin v meglitinides (nateglinide or repaglinide)	4	–1	0	0	0	Moderate	Quality point deducted for weak methods in 1 RCT
2 (454)[9] [15] [16]	Body weight	Metformin v meglitinides (nateglinide or repaglinide	4	–2	0	0	0	Low	Quality points deducted for weak methods in 1 RCT and incomplete reporting of results
2 (454)[9] [15] [16]	Hypoglycaemia	Metformin v meglitinides (nateglinide or repaglinide)	4	–2	0	0	0	Low	Quality points deducted for weak methods in 1 RCT and incomplete reporting of results
1 (751)[9]	Glycaemic control	Metformin v insulin	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for alteration of regimen during trial
12 (unclear, at least 607)[10]	Glycaemic control	Metformin plus second-generation sulphonylureas v sulphonylureas alone	4	–1	–1	–1	0	Very low	Quality point deducted for incomplete reporting of results. Consistency point deducted for statistical heterogeneity. Directness point deducted for short follow-up in some RCTs
10 (unclear)[10]	Body weight	Metformin plus second-generation sulphonylureas v sulphonylureas alone	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for short follow-up in some RCTs
9 (unclear, at least 607)[10]	Hypoglycaemia	Metformin plus second generation sulphonylureas v sulphonylureas alone	4	–2	0	0	0	Low	Quality points deducted for incomplete reporting of results and for combining 3 arms of trial into 1 comparison group
11 (unclear)[10]	Glycaemic control	Sulphonylurea v placebo	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for short-term RCTs
1 (40)[10]	Body weight	Sulphonylurea v placebo	4	–1	0	–1	0	Low	Quality point deducted for sparse data. Directness point deducted for no statistical analysis between groups
14 (at least 1106)[33] [34] [35] [10]	Glycaemic control	Sulphonylurea v thiazolidinediones	4	–2	0	0	0	Low	Quality points deducted for incomplete reporting of results and weak methods
8 (at least 1106)[10] [33] [34] [35]	Body weight	Sulphonylurea v thiazolidinediones	4	–2	0	0	0	Low	Quality points deducted for incomplete reporting of results and weak methods
8 (at least 1106)[10] [33] [34] [35]	Hypoglycaemia	Sulphonylurea v thiazolidinediones	4	–2	0	0	0	Low	Quality points deducted for incomplete reporting of results and weak methods
4 (unclear)[10]	Glycaemic control	Sulphonylurea plus metformin v placebo plus metformin	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for limited number of comparators (glibenclamide)
3 (unclear)[10]	Body weight	Sulphonylurea plus metformin v placebo plus metformin	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for limited number of comparators (glibenclamide)
3 (unclear)[10]	Hypoglycaemia	Sulphonylurea plus metformin v placebo plus metformin	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for limited number of comparators (glibenclamide)
1 (672)[33]	Glycaemic control	Glimepiride plus rosiglitazone v rosiglitazone alone	4	–1	0	0	0	Moderate	Quality point deducted for weak methods
1 (672)[33]	Body weight	Glimepiride plus rosiglitazone v rosiglitazone alone	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for no statistical analysis between groups
1 (672)[33]	Hypoglycaemia	Glimepiride plus rosiglitazone v rosiglitazone alone	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for no statistical analysis between groups
3 (1009)[10] [36]	Glycaemic control	Sulphonylurea plus metformin v thiazolidinediones plus metformin	4	–2	0	0	0	Low	Quality points deducted for incomplete reporting of results and results for 2 arms combined in analysis in 1 RCT
1 (314)[36]	Body weight	Sulphonylurea plus metformin v thiazolidinediones plus metformin	4	–2	0	0	0	Low	Quality points deducted for incomplete reporting of results and results for 2 arms combined in analysis
2 (908)[10] [36]	Hypoglycaemia	Sulphonylurea plus metformin v thiazolidinediones plus metformin	4	–2	0	0	0	Low	Quality points deducted for incomplete reporting of results and results for 2 arms combined in analysis
1 (159)[37]	Glycaemic control	Glimepiride plus metformin plus thiazolidinedione v placebo plus metformin plus thiazolidinedione	4	–1	0	0	0	Moderate	Quality point deducted for sparse data
1 (159)[37]	Quality of life	Glimepiride plus metformin plus thiazolidinedione v placebo plus metformin plus thiazolidinedione	4	–2	0	0	0	Low	Quality points deducted for sparse data and incomplete reporting of results
1 (159)[37]	Body weight	Glimepiride plus metformin plus thiazolidinedione v placebo plus metformin plus thiazolidinedione	4	–1	0	–1	0	Low	Quality point deducted for sparse data. Directness point deducted for BMI results (no direct weight analysis)
1 (168)[37]	Hypoglycaemia	Glimepiride plus metformin plus thiazolidinedione v placebo plus metformin plus thiazolidinedione	4	–1	0	0	0	Moderate	Quality point deducted for sparse data
2 (453)[15] [10]	Glycaemic control	Repaglinide v placebo	4	–2	0	–2	0	Very low	Quality points deducted for incomplete reporting of results and weak methods. Directness points deducted for baseline differences in 1 RCT and unclear dropouts
1 (56)[10] [15]	Body weight	Repaglinide v placebo	4	–2	0	–1	0	Very low	Quality points deducted for weak methods and sparse data. Directness point deducted for poor follow-up
3 (unclear)[10] [15]	Hypoglycaemia	Repaglinide v placebo	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for short follow-up in 2 RCTs
8 (unclear, at least 225)[38] [39] [10]	Glycaemic control	Repaglinide v sulphonylurea	4	–2	0	0	0	Low	Quality points deducted for incomplete reporting of results and weak methods
7 (unclear)[38] [39] [10]	Body weight	Repaglinide v sulphonylurea	4	–2	0	0	0	Low	Quality points deducted for incomplete reporting of results and weak methods
7 (unclear)[38] [39] [10]	Hypoglycaemia	Repaglinide v sulphonylurea	4	–2	0	0	0	Low	Quality points deducted for incomplete reporting of results and weak methods
4 (unclear, at least 1026)[10] [15]	Glycaemic control	Nateglinide v placebo	4	–2	0	0	0	Low	Quality points deducted for incomplete reporting of results and weak methods
2 (1026)[10] [15]	Body weight	Nateglinide v placebo	4	–2	0	0	0	Low	Quality points deducted for incomplete reporting of results and weak methods
2 (1026)[10] [15]	Hypoglycaemia	Nateglinide v placebo	4	–2	0	0	0	Low	Quality points deducted for incomplete reporting of results and weak methods
2 (1166)[15]	Glycaemic control	Nateglinide plus metformin v placebo plus metformin	4	–2	0	–1	0	Very low	Quality points deducted for incomplete reporting of results and weak methods. Directness point deducted for no statistical analysis between groups
2 (unclear, at least 312)[15]	Body weight	Nateglinide plus metformin v placebo plus metformin	4	–2	0	0	0	Low	Quality points deducted for incomplete reporting of results and weak methods
2 (643)[15]	Hypoglycaemia	Nateglinide plus metformin v placebo plus metformin	4	–2	0	–1	0	Very low	Quality points deducted for incomplete reporting of results and weak methods. Directness point deducted for no statistical analysis between groups
1 (395)[50]	Glycaemic control	Nateglinide plus rosiglitazone v placebo plus rosiglitazone	4	–1	0	0	0	Moderate	Quality point deducted for weak methods (unclear randomisation/allocation concealment)
1 (395)[50]	Body weight	Nateglinide plus rosiglitazone v placebo plus rosiglitazone	4	–1	0	0	0	Moderate	Quality point deducted for weak methods (unclear randomisation/allocation concealment)
3 (852)[40] [41] [42]	Glycaemic control	Nateglinide plus metformin v sulphonylurea plus metformin	4	0	0	–2	0	Low	Directness points deducted for highly selected population in 1 RCT (extension study) and high loss to follow-up in 1 RCT
2 (639)[41] [42]	Body weight	Nateglinide plus metformin v sulphonylurea plus metformin	4	0	–1	–1	0	Low	Consistency point deducted for conflicting results. Directness point deducted for high loss to follow-up in 1 RCT
3 (852)[40] [41] [42]	Hypoglycaemia	Nateglinide plus metformin v sulphonylurea plus metformin	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for no statistical analysis between groups in 1 RCT
1 (81)[51]	Glycaemic control	Nateglinide plus insulin plus metformin v placebo plus insulin plus metformin	4	–1	0	0	0	Moderate	Quality point deducted for sparse data
1 (81)[51]	Body weight	Nateglinide plus insulin plus metformin v placebo plus insulin plus metformin	4	–2	0	0	0	Low	Quality points deducted for sparse data and incomplete reporting of results
1 (81)[51]	Hypoglycaemia	Nateglinide plus insulin plus metformin v placebo plus insulin plus metformin	4	–1	0	0	0	Moderate	Quality point deducted for sparse data
2 (385)[14]	Mortality	Acarbose v placebo	4	–1	0	–1	0	Low	Quality point deducted for weak methods in 1 RCT. Directness point deducted for small number of events (9 in total)
1 (unclear)[57]	Morbidity	Acarbose v placebo	4	–2	0	–1	0	Very low	Quality points deducted for weak methods and incomplete reporting of results. Directness point deducted for unclear outcome
22 (2831)[14]	Glycaemic control	Acarbose v placebo	4	–1	–1	0	0	Low	Quality point deducted for weak methods. Consistency point deducted for significant heterogeneity
14 (1451)[14]	Body weight	Acarbose v placebo	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for inclusion of RCTs with <24 weeks follow-up
8 (596)[14]	Glycaemic control	Acarbose v sulphonylurea	4	–2	–1	0	0	Very low	Quality points deducted for high risk of bias and lack of blinding. Consistency point deducted for significant heterogeneity
5 (397)[14]	Body weight	Acarbose v sulphonylurea	4	–2	0	0	0	Low	Quality points deducted for high risk of bias and lack of blinding
1 (179)[14]	Glycaemic control	Acarbose v nateglinide	4	–1	0	0	0	Moderate	Quality point deducted for sparse data
1 (169)[14]	Body weight	Acarbose v nateglinide	4	–1	0	0	0	Moderate	Quality point deducted for sparse data
1 (661)[53]	Glycaemic control	Acarbose v vildagliptin	4	–1	0	0	0	Moderate	Quality point deducted for unclear randomisation and allocation concealment
1 (661)[53]	Body weight	Acarbose v vildagliptin	4	–1	0	0	0	Moderate	Quality point deducted for unclear randomisation and allocation concealment
4 (444)[54] [55] [56] [57]	Glycaemic control	Acarbose plus metformin v placebo plus metformin	4	–3	0	–1	0	Very low	Quality points deducted for weak methods in 1 RCT, incomplete reporting of results, and no ITT analysis in some RCTs. Directness point deducted for poor compliance in 1 RCT
4 (838)[57] [58] [59] [60]	Glycaemic control	Acarbose plus sulphonylurea v placebo plus sulphonylurea	4	–1	0	–2	0	Very low	Quality point deducted for incomplete reporting of results. Directness points deducted for poor compliance in 1 RCT, people on drugs other than sulphonylureas (25%) in 1 RCT, and poor trial completion in 1 RCT
2 (unclear)[59] [60]	Body weight	Acarbose plus sulphonylurea v placebo plus sulphonylurea	4	–1	0	–2	0	Very low	Quality point deducted for incomplete reporting of results. Directness points deducted for people on drugs other than sulphonylurea (25%) in 1 RCT, poor trial completion in 1 RCT, and no statistical analysis reported in 1 RCT
4 (1088)[14]	Glycaemic control	Miglitol v placebo	4	–1	–1	0	0	Low	Quality point deducted for weak methods. Consistency point deducted for statistical heterogeneity
1 (162)[14]	Body weight	Miglitol v placebo	4	–2	0	0	0	Low	Quality points deducted for sparse data and weak methods
1 (90)[14]	Glycaemic control	Miglitol v sulphonylurea	4	–2	0	0	0	Low	Quality points deducted for sparse data and weak methods (moderate risk of bias; no ITT analysis; only completer baseline data)
1 (90)[14]	Body weight	Miglitol v sulphonylurea	4	–2	0	0	0	Low	Quality points deducted for sparse data and weak methods (moderate risk of bias; no ITT analysis; only completer baseline data)
1 (153)[61]	Glycaemic control	Miglitol plus metformin v placebo plus metformin	4	–1	0	–1	0	Low	Quality point deducted for sparse data. Directness point deducted for restricted population (previous poor control on metformin only)
1 (153)[61]	Body weight	Miglitol plus metformin v placebo plus metformin	4	–1	0	–2	0	Very low	Quality point deducted for sparse data. Directness points deducted for restricted population (previous poor control on metformin only) and no statistical analysis between groups
1 (133)[62]	Glycaemic control	Miglitol plus glibenclamide plus metformin v placebo plus glibenclamide plus metformin	4	–1	0	–1	0	Low	Quality point deducted for sparse data. Directness point deducted for restricted population (previous poor control on metformin plus glibenclamide)
At least 56 (at least 35531)[66] [65] [68] [69] [70] [71] [72]	Mortality	Thiazolidinediones v placebo or other oral hypoglycaemic agents	4	–1	0	–2	0	Very low	Quality point deducted for weak methods of some RCTs. Directness points deducted for inclusion of short term RCTs, mixed control group (placebo, active agents), people with pre-diabetes, and composite outcomes (mortality/morbidity)
At least 41 (at least 31,507)[66] [65] [68] [69] [70] [71] [72] [73] [17] [13] [74] [75] [76] [77] [65] [94] [78]	Morbidity	Thiazolidinediones v placebo or other oral hypoglycaemic agents	4	–1	0	–2	0	Very low	Quality point deducted for weak methods of some RCTs. Directness points deducted for inclusion of short-term RCTs, mixed control group (placebo, active agents), people with pre-diabetes, and composite outcomes (mortality/morbidity)
At least 41 (at least 10,423)[80] [65]	Glycaemic control	Thiazolidinediones v placebo	4	–1	–1	0	0	Low	Quality point deducted for incomplete reporting of results. Consistency point deducted for statistical heterogeneity
At least 14 (at least 1019)[88] [80]	Body weight	Thiazolidinediones v placebo	4	–1	–1	0	0	Low	Quality point deducted for incomplete reporting of results. Consistency point deducted for statistical heterogeneity
1 (600)[81]	Glycaemic control	Pioglitazone plus metformin v pioglitazone alone or metformin alone	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for poor trial completion rate
1 (600)[81]	Hypoglycaemia	Pioglitazone plus metformin v pioglitazone alone or metformin alone	4	0	0	–2	0	Low	Directness points deducted for poor trial completion rate and no statistical analysis between groups
1 (509)[82]	Glycaemic control	Rosiglitazone plus metformin v metformin alone	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for unclear clinical importance
Unclear, at least 4 (unclear)[82] [10]	Hypoglycaemia	Rosiglitazone plus metformin v metformin alone	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for no between-group analysis in 1 RCT
1 (227)[83]	Glycaemic control	Rosiglitazone plus glipizide v placebo plus glipizide	4	0	0	–2	0	Low	Directness points deducted for highly selected population (people >60 years, on submaximal sulphonylurea treatment) and poor trial completion rate (65%)
1 (227)[83]	Body weight	Rosiglitazone plus glipizide v placebo plus glipizide	4	0	0	–2	0	Low	Directness points deducted for highly selected population (people >60 years, on submaximal sulphonylurea treatment) and poor trial completion rate (65%)
1 (227)[83]	Hypoglycaemia	Rosiglitazone plus glipizide v placebo plus glipizide	4	0	0	–2	0	Low	Directness points deducted for highly selected population (people >60 years, on submaximal sulphonylurea treatment) and poor trial completion rate (65%)
1 (356)[84]	Glycaemic control	Rosiglitazone plus glibenclamide plus metformin v placebo plus glibenclamide plus metformin	4	0	0	–1	0	Moderate	Directness point deducted for poor trial completion rate (72%)
1 (356)[84]	Body weight	Rosiglitazone plus glibenclamide plus metformin v placebo plus glibenclamide plus metformin	4	0	0	–2	0	Low	Directness points deducted for low trial completion rate (72%) and no statistical analysis between groups
1 (356)[84]	Hypoglycaemia	Rosiglitazone plus glibenclamide plus metformin v placebo plus glibenclamide plus metformin	4	0	0	–2	0	Low	Directness points deducted for low trial completion rate (72%) and no statistical analysis between groups
3 (282)[85] [86] [87]	Glycaemic control	Thiazolidinediones plus sulphonylurea plus metformin v insulin plus sulphonylurea plus metformin	4	–1	0	0	0	Moderate	Quality point deducted for weak methods
3 (282)[85] [86] [87]	Body weight	Thiazolidinediones plus sulphonylurea plus metformin v insulin plus sulphonylurea plus metformin	4	–1	0	0	0	Moderate	Quality point deducted for weak methods
2 (256)[85] [87]	Hypoglycaemia	Thiazolidinediones plus sulphonylurea plus metformin v insulin plus sulphonylurea plus metformin	4	–1	0	0	0	Moderate	Quality point deducted for weak methods
1 (233)[100]	Glycaemic control	Exenatide v placebo	4	0	0	0	0	High
1 (233)[100]	Body weight	Exenatide v placebo	4	0	0	0	0	High
1 (233)[100]	Hypoglycaemia	Exenatide v placebo	4	–1	0	0	0	Moderate	Quality point deducted for incomplete reporting of results
1 (336)[47] [48] [101]	Glycaemic control	Exenatide plus metformin v placebo plus metformin	4	0	0	–1	0	Moderate	Directness point deducted for combining 2 arms in analysis
1 (336)[47] [48] [101]	Body weight	Exenatide plus metformin v placebo plus metformin	4	0	0	0	0	High
1 (336)[47] [48] [101]	Hypoglycaemia	Exenatide plus metformin v placebo plus metformin	4	0	0	–1	0	Moderate	Directness point deducted for no statistical analysis between groups
1 (377)[47] [48] [102] [49]	Glycaemic control	Exenatide plus sulphonylurea v placebo plus sulphonylurea	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for poor trial completion rate (69%)
1 (377)[47] [48] [49]	Body weight	Exenatide plus sulphonylurea v placebo plus sulphonylurea	4	–2	0	–1	0	Very low	Quality points deducted for weak methods and incomplete reporting of results. Directness point deducted for poor trial completion rate (69%)
1 (377)[47] [48] [49]	Hypoglycaemia	Exenatide plus sulphonylurea v placebo plus sulphonylurea	4	–1	0	–2	0	Very low	Quality point deducted for weak methods. Directness points deducted for poor trial completion rate (69%) and no statistical analysis between groups
1 (734)[47] [48] [102]	Glycaemic control	Exenatide plus sulphonylurea plus metformin v placebo plus sulphonylurea plus metformin	4	–1	0	0	0	Moderate	Quality point deducted for non-blinded allocation of sulphonylurea
1 (734)[47] [48] [102]	Body weight	Exenatide plus sulphonylurea plus metformin v placebo plus sulphonylurea plus metformin	4	–1	0	0	0	Moderate	Quality point deducted for non-blinded allocation of sulphonylurea
1 (734)[47] [48] [102]	Hypoglycaemia	Exenatide plus sulphonylurea plus metformin v placebo plus sulphonylurea plus metformin	4	–1	0	–1	0	Low	Quality point deducted for non-blinded allocation of sulphonylurea. Directness point deducted for no statistical analysis between groups
6 (at least 641)[47] [103] [104] [96] [97] [98]	Glycaemic control	Exenatide plus oral blood-glucose-lowering agents v insulin plus oral blood-glucose-lowering agents	4	–2	0	0	0	Low	Quality points deducted for weak methods and incomplete reporting of results
1 (455)[99] [103]	Quality of life	Exenatide plus oral blood-glucose-lowering agents v insulin plus oral blood-glucose-lowering agent	4	0	0	–2	0	Low	Directness points deducted for no ITT analysis and small number of comparators
6 (at least 641)[47] [103] [104] [96] [97] [98]	Body weight	Exenatide plus oral blood-glucose-lowering agents v insulin plus oral blood-glucose-lowering agent	4	–1	0	0	0	Moderate	Quality point deducted for weak methods
6 (at least 641)[47] [103] [104] [96] [97] [98]	Hypoglycaemia	Exenatide plus oral blood-glucose-lowering agents v insulin plus oral blood-glucose-lowering agent	4	–2	0	0	0	Low	Quality points deducted for weak methods and incomplete reporting of results
1 (746)[105] [46]	Glycaemic control	Liraglutide v glimepiride	4	0	0	–1	0	Moderate	Directness point deducted for poor trial completion rate (65%)
1 (746)[105] [46]	Body weight	Liraglutide v glimepiride	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for poor trial completion rate (65%)
1 (746)[105] [46]	Hypoglycaemia	Liraglutide v glimepiride	4	0	0	–1	0	Moderate	Directness point deducted for poor trial completion rate (65%)
1 (1041)[105] [106]	Glycaemic control	Liraglutide plus glimepiride v placebo or rosiglitazone plus glimepiride	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for 3 different doses of 1 active agent v 1 dose of another active agent with different results depending on dose
1 (1041)[105] [106]	Body weight	Liraglutide plus glimepiride v placebo or rosiglitazone plus glimepiride	4	0	0	–1	0	Moderate	Directness point deducted for no statistical analysis between groups
1 (1041)[105] [106]	Hypoglycaemia	Liraglutide plus glimepiride v placebo or rosiglitazone plus glimepiride	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for 3 different doses of 1 active agent v 1 dose of another active agent with different results depending on dose
1 (1091)[107] [105]	Glycaemic control	Liraglutide plus metformin v placebo or glimepiride plus metformin	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for no statistical analysis reported for some arms of trial
1 (1091)[107] [105]	Body weight	Liraglutide plus metformin v placebo or glimepiride plus metformin	4	0	0	–1	0	Moderate	Directness point deducted for different results for different doses compared with placebo
1 (1091)[107] [105]	Hypoglycaemia	Liraglutide plus metformin v placebo or glimepiride plus metformin	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for combined analysis
1 (576)[108] [105]	Glycaemic control	Liraglutide plus metformin plus glimepiride v placebo or insulin glargine plus metformin plus glimepiride	4	0	0	0	0	High
1 (576)[108] [105]	Body weight	Liraglutide plus metformin plus glimepiride v placebo or insulin glargine plus metformin plus glimepiride	4	0	0	0	0	High
1 (576)[108] [105]	Hypoglycaemia	Liraglutide plus metformin plus glimepiride v placebo or insulin glargine plus metformin plus glimepiride	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for no statistical analysis between groups
1 (533)[109] [105]	Glycaemic control	Liraglutide plus metformin plus rosiglitazone v placebo plus metformin plus rosiglitazone	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for poor trial completion rate (75%)
1 (533)[109] [105]	Body weight	Liraglutide plus metformin plus rosiglitazone v placebo plus metformin plus rosiglitazone	4	0	0	–1	0	Moderate	Directness point deducted for poor trial completion rate (75%)
1 (533)[109] [105]	Hypoglycaemia	Liraglutide plus metformin plus rosiglitazone v placebo plus metformin plus rosiglitazone	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for poor trial completion rate (75%)
5 (unclear)[20]	Glycaemic control	Sitagliptin v placebo	4	–1	–1	0	0	Low	Quality point deducted for incomplete reporting of results. Consistency point deducted for heterogeneity
3 (1111)[20]	Body weight	Sitagliptin v placebo	4	0	0	0	0	High
1 (352)[20] [21]	Glycaemic control	Sitagliptin v metformin	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for high attrition rates
2 (867)[20] [110] [111]	Glycaemic control	Sitagliptin plus metformin v placebo plus metformin	4	–1	0	–1	0	Low	Quality point deducted for weak methods in 1 RCT. Directness point deducted for high attrition rates in 1 RCT
1 (355)[20] [21]	Glycaemic control	Sitagliptin plus metformin v metformin alone	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for high attrition rates
1 (207)[20] [43]	Glycaemic control	Sitagliptin plus glimepiride v placebo plus glimepiride	4	0	0	–1	0	Moderate	Directness point deducted for subgroup analysis
1 (337)[20] [113]	Glycaemic control	Sitagliptin plus pioglitazone v placebo plus pioglitazone	4	–1	0	0	0	Moderate	Quality point deducted for weak methods
1 (1135)[114] [20]	Glycaemic control	Sitagliptin plus metformin v glipizide plus metformin	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for high attrition rates
1 (1135)[114] [20]	Body weight	Sitagliptin plus metformin v glipizide plus metformin	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for high attrition rates
1 (1135)[114] [20]	Hypoglycaemia	Sitagliptin plus metformin v glipizide plus metformin	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for high attrition rates
6 (at least 778)[20]	Glycaemic control	Vildagliptin v placebo	4	0	–1	0	0	Moderate	Consistency point deducted for statistical heterogeneity
3 (484)[20]	Body weight	Vildagliptin v placebo	4	0	0	0	0	High
2 (1093)[20] [95] [115]	Glycaemic control	Vildagliptin v thiazolidinediones	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for disparate attrition rates in 1 RCT
2 (1093)[20] [95] [115]	Body weight	Vildagliptin v thiazolidinediones	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for disparate attrition rates in 1 RCT
2 (1015)[20] [22] [23]	Glycaemic control	Vildagliptin v metformin	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for high attrition rates
2 (1015)[20] [22] [23]	Body weight	Vildagliptin v metformin	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for high attrition rates
1 (1092)[44]	Glycaemic control	Vildagliptin v gliclazide	4	0	0	–1	0	Moderate	Directness point deducted for no ITT analysis
1 (1092)[44]	Body weight	Vildagliptin v gliclazide	4	0	0	–1	0	Moderate	Directness point deducted for no ITT analysis
1 (1092)[44]	Hypoglycaemia	Vildagliptin v gliclazide	4	0	0	–2	0	Low	Directness points deducted for no ITT analysis and no statistical analysis between groups
2 (643)[20] [24] [25]	Glycaemic control	Vildagliptin plus metformin v placebo plus metformin	4	–2	0	0	0	Low	Quality points deducted for weak methods and incomplete reporting of results
1 (408)[45]	Glycaemic control	Vildagliptin plus glimepiride v placebo plus glimepiride	4	0	0	–1	0	Moderate	Directness point deducted for poor follow-up
1 (274)[20] [116]	Glycaemic control	Vildagliptin plus pioglitazone v placebo plus pioglitazone	4	–2	0	0	0	Low	Quality points deducted for weak methods and incomplete reporting of results
1 (305)[20] [115]	Glycaemic control	Vildagliptin plus pioglitazone v pioglitazone alone	4	–2	0	0	0	Low	Quality points deducted for weak methods and incomplete reporting of results
1 (2190)[117]	Glycaemic control	Vildagliptin plus metformin v glimepiride plus metformin	4	–2	0	0	0	Low	Quality points deducted for incomplete reporting of results and weak methods (randomisation, no ITT analysis)
1 (2190)[117]	Body weight	Vildagliptin plus metformin v glimepiride plus metformin	4	–1	0	0	0	Moderate	Quality point deducted for weak methods (randomisation, no ITT analysis)
1 (2190)[117]	Hypoglycaemia	Vildagliptin plus metformin v glimepiride plus metformin	4	–1	0	–1	0	Low	Quality point deducted for weak methods (randomisation, no ITT analysis). Directness point deducted for no statistical analysis between groups for hypoglycaemic events
1 (510)[20] [118]	Glycaemic control	Vildagliptin plus metformin v pioglitazone plus metformin	4	–1	0	0	0	Moderate	Quality point deducted for weak methods
1 (1134)[26]	Glycaemic control	Vildagliptin plus metformin v vildagliptin alone or metformin alone	4	0	0	0	0	High
1 (727)[27]	Glycaemic control	Saxagliptin plus metformin v placebo plus metformin	4	0	0	–1	0	Moderate	Directness point deducted for poor trial completion (73%)
1 (727)[27]	Hypoglycaemia	Saxagliptin plus metformin v placebo plus metformin	4	0	0	–2	0	Low	Directness points deducted for poor trial completion (73%) and no between-group analysis
1 (555)[119]	Glycaemic control	Saxagliptin plus thiazolidinediones v placebo plus thiazolidinediones	4	–1	0	0	0	Moderate	Quality point deducted for weak methods
1 (555)[119]	Body weight	Saxagliptin plus thiazolidinediones v placebo plus thiazolidinediones	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for no statistical analysis between groups
1 (555)[119]	Hypoglycaemia	Saxagliptin plus thiazolidinediones v placebo plus thiazolidinediones	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for no statistical analysis between groups
1 (1251)[28]	Glycaemic control	Saxagliptin plus metformin v saxagliptin alone or metformin alone	4	0	0	–1	0	Moderate	Directness point deducted for poor trial completion (74%)
1 (1251)[28]	Body weight	Saxagliptin plus metformin v saxagliptin alone or metformin alone	4	0	0	–2	0	Low	Directness points deducted for poor trial completion (74%) and for no statistical analysis between groups
1 (42)[121]	Glycaemic control	Continuation of insulin v metformin or sulphonylurea	4	–2	0	–2	0	Very low	Quality points deducted for weak methods and sparse data. Directness points deducted for restricted population (people with hyperglycaemia hospitalised), different agents used depending on weight, different orders of agents and combinations used in oral group
1 (42)[121]	Body weight	Continuation of insulin v metformin or sulphonylurea	4	–2	0	–2	0	Very low	Quality points deducted for weak methods and sparse data. Directness points deducted for restricted population (people with hyperglycaemia hospitalised), different agents used depending on weight, different orders of agents and combinations used in oral group
1 (42)[121]	Hypoglycaemia	Continuation of insulin v metformin or sulphonylurea	4	–2	0	–2	0	Very low	Quality points deducted for weak methods and sparse data. Directness points deducted for restricted population (people with hyperglycaemia hospitalised), different agents used depending on weight, different orders of agents and combinations used in oral group
4 (1941)[126] [127] [128] [129] [130]	Glycaemic control	Short-acting insulin analogues v conventional (human) insulin	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for unclear clinical importance of between-group difference in 1 RCT
4 (1941)[126] [127] [128] [129] [130]	Hypoglycaemia	Short-acting insulin analogues v conventional (human) insulin	4	–2	0	0	0	Low	Quality points deducted for weak methods and incomplete reporting of results
1 (1024)[142]	Morbidity	Long-acting insulin analogues v conventional (human) long-acting insulin	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for change of regimens at investigators' discretion during trial
At least 12 (at least 2118)[131] [132] [133]	Glycaemic control	Long-acting insulin analogues v conventional (human) long-acting insulin	4	–2	0	–1	0	Very low	Quality points deducted for weak methods and for incomplete reporting of results. Directness point deducted for many studies using NPH differently from usual practice, limiting relevance of results
4 (unclear)[134]	Quality of life	Long-acting insulin analogues v conventional (human) long-acting insulin	4	–1	0	–2	0	Very low	Quality point deducted for incomplete reporting of results. Directness points deducted for limited outcome measurement and unclear clinical relevance in 1 RCT
Unclear (unclear)[132] [134]	Body weight	Long-acting insulin analogues v conventional (human) long-acting insulin	4	–1	–1	–2	0	Very low	Quality point deducted for weak methods. Consistency point deducted for conflicting results. Directness points deducted for heterogeneity and for many studies using NPH differently form usual practice limiting, relevance of results
Unclear (unclear)[134]	Hypoglycaemia	Long-acting insulin analogues v conventional (human) long-acting insulin	4	–1	0	–2	0	Very low	Quality point deducted for weak methods. Directness points deducted for inconsistent results depending on exact analysis performed and for many studies using NPH differently from usual practice, limiting relevance of results
2 (299)[160] [161]	Glycaemic control	Premixed insulin analogues v premixed conventional (human) insulin	4	–1	0	–2	0	Very low	Quality point deducted for weak methods. Directness points deducted for highly selected population in 1 RCT and poor trial completion rate (76%)
1 (121)[161]	Body weight	Premixed insulin analogues v premixed conventional (human) insulin	4	–1	0	–2	0	Very low	Quality point deducted for sparse data. Directness points deducted for highly selected population and poor trial completion rate (76%)
2 (299)[160] [161]	Hypoglycaemia	Premixed insulin analogues v premixed conventional (human) insulin	4	–1	0	–2	0	Very low	Quality point deducted for weak methods. Directness points deducted for highly selected population in 1 RCT and poor trial completion rate (76%)
1 (38)[162] [163]	Glycaemic control	Basic bolus therapy with insulin analogues v twice-daily conventional (human) long-acting insulin	4	–2	0	–1	0	Very low	Quality points deducted for sparse data and weak methods. Directness point deducted for limited generalisability (older population, small number of comparators)
1 (38)[162] [163]	Quality of life	Basic bolus therapy with insulin analogues v twice-daily conventional (human) long-acting insulin	4	–3	0	–1	0	Very low	Quality points deducted for sparse data, incomplete reporting of results, and weak methods. Directness point deducted for limited generalisability (older population, small number of comparators)
1 (38)[162] [163]	Body weight	Basic bolus therapy with insulin analogues v twice-daily conventional (human) long-acting insulin	4	–3	0	–1	0	Very low	Quality points deducted for sparse data, incomplete reporting of results, and weak methods. Directness point deducted for limited generalisability (older population, small number of comparators)
1 (38)[162] [163]	Hypoglycaemia	Basic bolus therapy with insulin analogues v twice-daily conventional (human) long-acting insulin	4	–3	0	–1	0	Very low	Quality points deducted for sparse data, incomplete reporting of results, and weak methods. Directness points deducted for limited generalisability (older population, small number of comparators)
1 (42)[164] [162]	Glycaemic control	Basic bolus therapy with insulin analogues v premixed conventional insulin	4	–2	0	–1	0	Very low	Quality points deducted for sparse data and unclear methods. Directness point deducted for limited generalisability (small number of comparators, selected population)
1 (42)[164] [162]	Quality of life	Basic bolus therapy with insulin analogues v premixed conventional insulin	4	–2	0	–1	0	Very low	Quality points deducted for sparse data and unclear methods. Directness point deducted for limited generalisability (small number of comparators, selected population)
1 (42)[164] [162]	Body weight	Basic bolus therapy with insulin analogues v premixed conventional insulin	4	–2	0	–2	0	Very low	Quality points deducted for sparse data and unclear methods. Directness points deducted for limited generalisability (small number of comparators, selected population) and BMI results only (no direct weight analysis)
3 (1245)[166] [167] [168]	Glycaemic control	Insulin long-acting analogues v each other	4	0	0	–1	0	Moderate	Directness point deducted for different regimens of insulin detemir used between trials (once daily, twice daily) potentially affecting result
3 (1245)[166] [167] [168]	Body weight	Insulin long-acting analogues v each other	4	0	0	–1	0	Moderate	Directness point deducted for different regimens of insulin detemir used between trials (once daily, twice daily) potentially affecting result
3 (1245)[166] [167] [168]	Hypoglycaemia	Insulin long-acting analogues v each other	4	0	0	–1	0	Moderate	Directness point deducted for different regimens of insulin detemir used between trials (once daily, twice daily) potentially affecting result
1 (390)[31]	Morbidity	Insulin plus metformin v insulin	4	0	0	–1	0	Moderate	Directness point deducted for composite endpoint (including mortality and morbidity)
4 (621)[170] [175] [133] [176] [177] [31]	Glycaemic control	Insulin plus metformin v insulin	4	–2	0	0	0	Low	Quality points deducted for incomplete reporting of results and weak methods
3 (408)[170] [175] [133] [176] [177] [31]	Body weight	Insulin plus metformin v insulin	4	–1	0	0	0	Moderate	Quality point deducted for weak methods
4 (621)[170] [175] [133] [176] [177] [31]	Hypoglycaemia	Insulin plus metformin v insulin	4	–1	0	0	0	Moderate	Quality point deducted for weak methods
3 (at least 188)[170] [175] [178] [133] [172]	Glycaemic control	Insulin plus sulphonylurea v insulin	4	–2	0	0	0	Low	Quality point deducted for incomplete reporting of results and weak methods (2 arms combined in analysis, pragmatic addition of insulin)
3 (at least 188)[170] [175] [178] [133] [172]	Body weight	Insulin plus sulphonylurea v insulin	4	–2	0	–1	0	Very low	Quality point deducted for incomplete reporting of results and weak methods (2 arms combined in analysis, pragmatic addition of insulin). Directness point deducted for no statistical analysis between groups in 1 RCT
3 (at least 188)[170] [175] [178] [133] [172]	Hypoglycaemia	Insulin plus sulphonylurea v insulin	4	–1	–1	–1	0	Very low	Quality point deducted for weak methods (2 arms combined in analysis, pragmatic addition of insulin). Consistency point deducted for conflicting results. Directness point deducted for no statistical analysis between groups in 1 RCT
2 (850)[79] [179]	Glycaemic control	Insulin plus thiazolidinedione v insulin	4	–1	0	–1	0	Low	Quality point deducted for incomplete reporting of results. Directness point deducted for poor trial completion in 1 RCT
At least 6 (at least 850)[79] [179] [78]	Body weight	Insulin plus thiazolidinedione v insulin	4	0	0	–2	0	Low	Directness points deducted for poor trial completion in 1 RCT and no statistical analysis between groups
At least 6 (at least 1010)[79] [179] [78]	Hypoglycaemia	Insulin plus thiazolidinedione v insulin	4	0	0	–2	0	Low	Directness points deducted for poor trial completion in 1 RCT, inclusion of short-term RCTs, and no statistical analysis between groups in 1 RCT
1 (290)[171]	Glycaemic control	Insulin plus vildagliptin v insulin plus placebo	4	0	0	–1	0	Moderate	Directness point deducted for small number of comparators
1 (290)[171]	Body weight	Insulin plus vildagliptin v insulin plus placebo	4	0	0	–1	0	Moderate	Directness point deducted for small number of comparators
1 (290)[171]	Hypoglycaemia	Insulin plus vildagliptin v insulin plus placebo	4	0	0	0	0	High
1 (41)[175]	Glycaemic control	Insulin plus metformin v insulin plus sulphonylurea	4	–3	0	0	0	Very low	Quality points deducted for sparse data, incomplete baseline data, and incomplete reporting of results
1 (41)[175]	Body weight	Insulin plus metformin v insulin plus sulphonylurea	4	–2	0	0	0	Low	Quality points deducted for sparse data, and incomplete baseline data
1 (41)[175]	Hypoglycaemia	Insulin plus metformin v insulin plus sulphonylurea	4	–2	0	0	0	Low	Quality points deducted for sparse data, incomplete baseline data, and incomplete reporting of results
1 (102)[173]	Glycaemic control	Biphasic analogue insulin plus metformin v biphasic analogue insulin plus repaglinide	4	–1	0	0	0	Moderate	Quality point deducted for sparse data
1 (102)[173]	Body weight	Biphasic analogue insulin plus metformin v biphasic analogue insulin plus repaglinide	4	–1	0	0	0	Moderate	Quality point deducted for sparse data
1 (102)[173]	Hypoglycaemia	Biphasic analogue insulin plus metformin v biphasic analogue insulin plus repaglinide	4	–1	0	0	0	Moderate	Quality point deducted for sparse data
1 (38)[174]	Glycaemic control	Insulin plus sulphonylurea v insulin plus alpha-glucosidase inhibitor	4	–2	0	0	0	Low	Quality points deducted for sparse data and baseline differences
1 (38)[174]	Hypoglycaemia	Insulin plus sulphonylurea v insulin plus alpha-glucosidase inhibitor	4	–2	0	–1	0	Very low	Quality points deducted for sparse data and baseline differences. Directness point deducted for small number of events (3 in total) limiting conclusions that can be drawn
1 (110)[149]	Glycaemic control	Insulin glargine plus metformin v conventional NPH insulin plus metformin	4	–2	0	0	0	Low	Quality points deducted for sparse data and incomplete reporting of results
1 (110)[149]	Body weight	Insulin glargine plus metformin v conventional NPH insulin plus metformin	4	–2	0	0	0	Low	Quality points deducted for sparse data and incomplete reporting of results
1 (110)[149]	Hypoglycaemia	Insulin glargine plus metformin v conventional NPH insulin plus metformin	4	–2	0	0	0	Low	Quality points deducted for sparse data and incomplete reporting of results
2 (2799)[193] [187]	Mortality	Premix analogue insulin v basal analogue insulin	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for small number of events limiting conclusions that can be drawn
2 (2799)[187] [193]	Morbidity	Premix analogue insulin v basal analogue insulin	4	–2	0	0	0	Low	Quality points deducted for incomplete reporting of results and weak methods
At least 13 (at least 5675)[158] [162] [194] [195] [196] [197] [198] [188]	Glycaemic control	Premix analogue insulin v basal analogue insulin	4	–2	–1	0	0	Very low	Quality points deducted for weak methods and incomplete reporting of results. Consistency point deducted for statistical heterogeneity
2 (1035)[193] [188]	Quality of life	Premix analogue insulin v basal analogue insulin	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for assessment of treatment satisfaction only in 1 RCT
At least 6 (at least 1996)[162] [158] [187] [188] [193]	Body weight	Premix analogue insulin v basal analogue insulin	4	–2	–1	0	0	Very low	Quality points deducted for weak methods and incomplete reporting of results. Consistency point deducted for statistical heterogeneity
At least 8 (at least 2479)[162] [158] [187] [188] [193] [197]	Hypoglycaemia	Premix analogue insulin v basal analogue insulin	4	–2	0	0	0	Low	Quality points deducted for weak methods and incomplete reporting of results
1 (708)[193]	Mortality	Prandial (short-acting) analogue insulin v basal analogue insulin	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for significance of results depending on analysis done (mortality or cardiovascular mortality) and no direct analysis between individual arms
1 (708)[193]	Morbidity	Prandial (short-acting) analogue insulin v basal analogue insulin	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for small number of events limiting conclusions that can be drawn
1 (473)[193]	Quality of life	Prandial (short-acting) analogue insulin v basal analogue insulin	4	–1	0	0	0	Moderate	Quality point deducted for weak methods
8 (1529)[194] [195] [199] [193]	Glycaemic control	Prandial (short-acting) analogue insulin v basal analogue insulin	4	–1	–1	0	0	Low	Quality point deducted for weak methods. Consistency point deducted for statistical heterogeneity
6 (1079)[162] [193]	Body weight	Prandial (short-acting) analogue insulin v basal analogue insulin	4	–1	–1	0	0	Low	Quality point deducted for weak methods. Consistency point deducted for statistical heterogeneity
2 (1002);[194] [199] [195] [162] [193]	Hypoglycaemia	Prandial (short-acting) analogue insulin v basal analogue insulin	4	–1	–1	0	0	Low	Quality point deducted for weak methods. Consistency point deducted for statistical heterogeneity
1 (708)[193]	Mortality	Premix analogue insulin v prandial (short-acting) analogue insulin	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for small number of events
1 (708)[193]	Morbidity	Premix analogue insulin v prandial (short-acting) analogue insulin	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for small number of events
3 (740)[162] [194] [195] [200]	Glycaemic control	Premix analogue insulin v prandial (short-acting) analogue insulin	4	–2	0	0	0	Low	Quality points deducted for weak methods and incomplete reporting of results
1 (473)[193]	Quality of life	Premix analogue insulin v prandial (short-acting) analogue insulin	4	–1	0	0	0	Moderate	Quality point deducted for weak methods
3 (740)[162] [194] [195] [200]	Body weight	Premix analogue insulin v prandial (short-acting) analogue insulin	4	–2	0	0	0	Low	Quality points deducted for weak methods and incomplete reporting of results
3 (740)[162] [194] [195] [200]	Hypoglycaemia	Premix analogue insulin v prandial (short-acting) analogue insulin	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for low number of events (no hypoglycaemic events, other events unclear)
2 (1031)[162] [189] [201]	Glycaemic control	Premix analogue insulin v basal bolus analogue insulin	4	–1	0	0	0	Moderate	Quality point deducted for weak methods
2 (1031)[162] [189] [201]	Body weight	Premix analogue insulin v basal bolus analogue insulin	4	–2	0	0	0	Low	Quality points deducted for weak methods and incomplete reporting of results
2 (1031)[162] [189] [201]	Hypoglycaemia	Premix analogue insulin v basal bolus analogue insulin	4	–1	0	0	0	Moderate	Quality point deducted for weak methods
1 (321)[190] [191]	Glycaemic control	Premix analogue twice daily v premix analogue insulin 3 times a day	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for small number of comparators limiting conclusions that can be drawn
1 (321)[190] [191]	Body weight	Premix analogue twice daily v premix analogue insulin 3 times a day	4	–2	0	–1	0	Very low	Quality points deducted for weak methods and incomplete reporting of results. Directness point deducted for small number of comparators limiting conclusions that can be drawn
1 (321)[190] [191]	Hypoglycaemia	Premix analogue twice daily v premix analogue insulin 3 times a day	4	–1	0	–1	0	Low	Quality point deducted for weak methods. Directness point deducted for small number of comparators limiting conclusions that can be drawn
1 (110)[192]	Glycaemic control	Intermediate-acting analogue v long-acting analogue insulin	4	–1	0	–1	0	Low	Quality point deducted for sparse data. Directness point deducted for highly selected population (on metformin and sulphonylurea only, other oral therapy excluded, insulin naive)
1 (110)[192]	Body weight	Intermediate-acting analogue v long-acting analogue insulin	4	–1	0	–1	0	Low	Quality point deducted for sparse data. Directness point deducted for highly selected population (on metformin and sulphonylurea only, other oral therapy excluded, insulin naive)
1 (110)[192]	Hypoglycaemia	Intermediate-acting analogue v long-acting analogue insulin	4	–1	0	–1	0	Low	Quality point deducted for sparse data. Directness point deducted for highly selected population (on metformin and sulphonylurea only, other oral therapy excluded, insulin naive)

Type of evidence: 4 = RCT. Consistency: similarity of results across studies.Directness: generalisability of population or outcomes. Effect size: based on relative risk or odds ratio.ITT, intention to treat.

Glossary

High-quality evidence

Further research is very unlikely to change our confidence in the estimate of effect.

Low-quality evidence

Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.

Moderate-quality evidence

Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.

Very low-quality evidence

Any estimate of effect is very uncertain.

Disclaimer

The information contained in this publication is intended for medical professionals. Categories presented in Clinical Evidence indicate a judgement about the strength of the evidence available to our contributors prior to publication and the relevant importance of benefit and harms. We rely on our contributors to confirm the accuracy of the information presented and to adhere to describe accepted practices. Readers should be aware that professionals in the field may have different opinions. Because of this and regular advances in medical research we strongly recommend that readers' independently verify specified treatments and drugs including manufacturers' guidance. Also, the categories do not indicate whether a particular treatment is generally appropriate or whether it is suitable for a particular individual. Ultimately it is the readers' responsibility to make their own professional judgements, so to appropriately advise and treat their patients. To the fullest extent permitted by law, BMJ Publishing Group Limited and its editors are not responsible for any losses, injury or damage caused to any person or property (including under contract, by negligence, products liability or otherwise) whether they be direct or indirect, special, incidental or consequential, resulting from the application of the information in this publication.