Statistics from Altmetric.com
The efficacy and safety of sodium glucose co-transporter 2 (SGLT2) inhibitors.
SGLT2 inhibitors are clinically effective in type 2 diabetes for improving glycaemic control.
They also lead to reductions in weight.
SGLT2 appear to be safe in the short-term but longer term data are needed.
Strengths and limitations of this study
Rigorous systematic review by independent group.
Clearly defined protocol with defined inclusions and exclusions.
Searches updated July 2012.
Focus on clinically relevant trials.
Only two trials against active comparators.
No trials of use in triple oral therapy.
No long-term data of SGLT2 safety available.
Type 2 diabetes is one of the most important and prevalent chronic diseases today, with an excess of 2.6 million people affected in the UK in 2010.1 The guidelines on the management of type 2 diabetes from the UK's National Institute for Clinical Excellence, recommend that if lifestyle intervention is insufficient, the first line of drug treatment is metformin, followed by a sulphonylurea, or sometimes a glitazone, before starting on insulin. However, sulphonylureas, glitazones and insulin all cause weight gain which may worsen insulin resistance. The sulphonylureas and insulin can also cause hypoglycaemia. Pioglitazone, now the only glitazone left in use, can cause oedema, heart failure and fractures.
It is estimated that 65% of people with diabetes will die as a result of cardiovascular complications,2 ,3 therefore antidiabetic medications need not only to produce a reduction in HbA1c, but ideally also a reduction in cardiovascular disease mortality.
Glucose is normally filtered in the kidney and is reabsorbed in the proximal tubules. Glycosuria occurs when the renal threshold of glucose (blood glucose of approximately 10 mmol/l (160–180 mg/dl) has been reached. At this threshold the proximal tubule cannot reabsorb all of the filtered glucose, resulting in glycosuria. In total, 98% of the urinary glucose is transported across the membrane of the proximal tubule by sodium glucose co-transporter 2 (SGLT2). A naturally occurring mutation in the SLC5A2 gene, resulting in a defective SGLT2 protein, produces significant glycosuria. Individuals who have this mutation have not been seen to have significant problems related to the glycosuria, such as urinary tract infections (UTIs).4
Therefore a therapeutic option in type 2 diabetics is to mimic the effect of the SLC5A2 mutation and prevent the reabsorption of renal-filtered glucose back into the circulation, thereby reducing hyperglycaemia, without the side effects of weight gain or hypoglycaemia.5
A new class of drugs has been developed to do this, and in this systematic review we review the evidence for clinical effectiveness and safety of the new SGLT2 inhibitor drugs (dapagliflozin, formerly known under the synonym: BMS-512148 and canagliflozin (JNJ28431754)). Since there are existing drugs which are inexpensive and with a long safety record, it is unlikely that SGLT2 inhibitors would be used first line, and we therefore review their role as second or third drugs used in combination therapy in type 2 diabetes.
The key questions for this review are:
How does the clinical effectiveness of the SGLT2 inhibitors compare with that of current pharmacological interventions, when prescribed in dual therapy, for example, metformin plus SGLT2 vs metformin plus sulphonylurea, and in triple therapy, for example, metformin, sulphonylurea and SGLT2 inhibitor versus metformin, sulphonylurea and dipeptidyl peptidase 4 inhibitors (DPP4) such as sitagliptin.
We also considered trials of SGLT2 inhibitors against placebo in dual and triple therapies.
The review of the evidence for clinical effectiveness was undertaken systematically, following the general principles recommended in the Cochrane Handbook for Systematic Reviews of Intervention.6
Randomised controlled trials (RCT) and systematic reviews of trials were used for assessing efficacy. As HbA1c is the main outcome being measured, no trial covering less than 8 weeks was accepted into the review, due to that being the minimum period required for a measureable change in HbA1c levels to be detected due to turnover of red blood cells.
Quality-of-life (QoL) data were also sought. A change in QoL may result from, for example, a reduction in hypoglycaemic episodes, and reduced fear of hypoglycaemia.
Adults, inclusive of any ethnic origin, over 18 years of age, who have been diagnosed with type 2 diabetes, defined using the WHO diagnostic criteria.7
Within those participant groups, we aimed to look at the effects in the following subgroups:
Prior medications: metformin, sulphonylureas, insulin, DPP4 inhibitors (the gliptins)
Patients with a duration of diabetes
Less than 2 years from diagnosis
3–9 years’ duration
Diagnosis for 10 years or longer
The hypothesis regarding duration is that since the mode of action is unrelated to insulin secretory function, the effect should not vary by duration of disease. Type 2 diabetes is often a progressive disease with diminishing β cell capacity.
Any use of SGLT2 inhibitors (dapagliflozin and canagliflozin) in dual or triple therapy, in addition to other interventions including, but not restricted to: metformin, sulphonylureas, insulin and gliptins, compared to placebo or another active antidiabetic medication in combination with the same antidiabetic co-medication as in the SGLT2 inhibitor group. We have focused on doses likely to be used in clinical practice, namely 10 mg/day for dapagliflozin.
The outcomes sought werePrimary outcome
Glycaemic control as reflected in HbA1c.
Change in weight (kg) or body mass index (BMI).
Change in QoL.
Adverse effects, including hypoglycaemia and UTI.
Search methods for identification of studies
We searched the following sources:
The Cochrane Library, all sections
NHS health technology assessment (HTA)
Science Citation Index Expanded (SCI expanded)
On-going Trials Registers
American Diabetes Association—Conference Abstracts
European Association for the Study of Diabetes—Conference Abstracts
Federal Drug Agency
European Medicines Agency (EMEA)
Scrutiny of bibliographies of retrieved papers
We searched for articles published since 2006, as this was the first recording of dapagliflozin on Ovid Technologies (OVID). An example of the SGLT2 dapagliflozin specific MEDLINE search strategy performed via the OVID interface is listed below:
1 or 2 or 3 or 4 or 5
(sodium glucose adj6 inhibitor*).mp.
(sodium-glucose adj6 inhibitor*).mp.
Sodium-Glucose Transporter 2/
sodium glucose-cotransporter 2.mp.
Reference lists of previous systematic reviews were checked for any trials not captured by the searches. The main search was carried out in October 2011. A search update in PubMed was carried out in July 2012.
Data collection and analysis
Two reviewers selected studies independently using the defined inclusion and exclusions criteria above. Any resulting discrepancies were resolved by discussion, with minimal third-party mediation required.
A standardised data extraction form was used. Data extraction was by one reviewer, checked by a second. Discrepancies were resolved by discussion, with involvement of a third reviewer when necessary.
The quality of the individual studies was assessed by one reviewer using the Cochrane Risk of Bias tool6 and checked by a second reviewer. Quality was rated as ‘high’ if at least the first three criteria were fulfilled (adequate sequence generation, allocation concealment and blinding) and not more than one of the others was rated ‘unclear’. Quality was rated as ‘low’ if these first three or any other four criteria were rated as unclear or inadequate. All the others were rated as ‘medium’ quality. Any disagreements were resolved by discussion.
Data synthesis and analysis
The data analysis has been reported according to the guide set down within the Cochrane Handbook for Systematic Reviews of Interventions.6 Meta-analysis was carried out for comparing HbA1c and weight results for 10 mg dapagliflozin versus placebo in the intermediate term (12–26 weeks) and longer term (48–52 weeks) using a random effects model (inverse variance method) using the Cochrane Review Manager 5 software. Results were expressed as weighted mean differences (WMD) with 95% CI. Heterogeneity was assessed using the I2 statistic. Where necessary, SDs were calculated from CIs or SEs as described in the Cochrane Handbook. In cases where means and measures of variation were only given in graphs but not in numerical form, values were estimated from graphs.
No meta-analysis using active comparators was performed due to clinical heterogeneity. Only two trials had active comparators, glipizide and sitagliptin, which have different modes of action and different effects on weight and hypoglycaemia risk.
The results of the literature search are shown in figure 1. After exclusions, made according to the study protocol, eight RCTs published in full, including 29 study arms, remained for analysis.
The characteristics and results of the included studies are shown in table 1.
All included trials were double-blind RCTs, and all but one were placebo controlled. Trial durations ranged from 12 to 52 weeks (median 24 weeks). Most trials had longer-term extension periods (not completed/reported in all cases).
Baseline HbA1c levels across the study populations ranged between 7.7% and 8.6% in most trials, but participants in one trial9 had baseline HbA1c levels of 7.2%. Baseline BMI ranged between 31.2 and 36.2 kg/m2, and mean age between 53 and 61 years.
Dapagliflozin was administered orally, with doses ranging from 2.5 to 20 mg, used as once daily preparations. Doses of canagliflozin ranged from 50 to 300 mg administered once daily, with an additional group with 300 mg administered twice daily.
Except for the study by Nauck,11 all studies included a placebo group. Two studies included an active comparator: glipizide (mean dose 16 mg) in the study by Nauck,11 and sitagliptin (100 mg) in the canagliflozin study.16
Most studies included lead in periods (median of 2 weeks) for assessing treatment adherence or stabilising background antidiabetic medication.
All studies reported on HbA1c, fasting plasma glucose (FPG), weight, blood pressure and safety parameters (including urinary or genital tract infections and hypoglycaemia). None of the studies reported QoL parameters.
Quality of included studies
Overall quality ratings are shown in table 1, details of risk of bias assessment are shown in table 2. The reporting quality was rated as ‘high’ in five of the studies,8 ,9 ,11 ,13 ,15 ‘medium’ in two studies14 ,16 and ‘low’ in one study.12
In five of the studies, both reporting of the generation of the randomisation sequence and of allocation concealment were adequate. All studies were at least double blind. Seven studies reported adequate intention-to-treat analysis (using the last observation carried forward method). Completion rates during the main study period were between 78% and 83%. Six of the studies included sample size calculations indicating that sufficient numbers of patients were recruited and included in order to detect a difference in HbA1c between 0.35% and 0.55% (median 0.5%). Seven studies explicitly reported that there were significant no differences in the main baseline characteristics between study groups. All studies were funded by the manufacturers.
Table 1 shows the difference between change from baseline to the main study end between 10 mg/day dapagliflozin and control groups (placebo or active control) for the main outcome measures. Detailed changes from baseline to the main study end or the end of any extension periods reported for all study groups are shown in online supplementary appendix.
Figure 2 shows the results of the meta-analysis of 10 mg/day of dapagliflozin versus placebo for HbA1c for study durations up to 26 and for 48–52 weeks. Figure 3 shows the reductions in HbA1c in the seven study groups of the canagliflozin study16 after 12 weeks of treatment.
Dapagliflozin at a dose of 10 mg/day significantly reduced HbA1c by (WMD) −0.54% (95% CI −0.67% to −0.40%, p<0.00001) after 12–26 weeks of treatment compared to placebo. There was significant heterogeneity, which was eliminated when excluding the only study with a baseline HbA1c<7.5%.9 The WMD in HbA1c for studies with a baseline HbA1c value of >7.5% was −0.59% (95% CI −0.67% to −0.51%). Change from baseline in the 10 mg dapagliflozin groups ranged between −0.39% and −0.96% (main study end), and differences to placebo between −0.29% and −0.69%. HbA1c reductions at 48–52 weeks were similar to those at up to 26 weeks (three studies, WMD −0.54, 95% CI −0.69 to −0.38, p<0.00001).
In the study by Nauck,11 there was no difference in HbA1c reduction between dapagliflozin and glipizide, both reducing HbA1c by −0.52% (95% CI −0.60% to −0.44%).
Canagliflozin reduced HbA1c in a dose-related manner up to 300 mg once daily (HbA1c reductions from baseline ranging from −0.70% to 0.95%) after 12 weeks of treatment, with only a small difference between once daily and twice daily doses at 300 mg (−0.92% SE 0.08 and −0.95% SE 0.08 from baseline, figure 3). The HbA1c reduction from baseline with sitagliptin was −0.74% SE 0.08.
Figure 4 shows the meta-analysis of weight change for 10 mg/day of dapagliflozin versus placebo for study durations up to 26 weeks and for 48–52 weeks. Dapaglifozin was associated with a significant reduction in weight. Compared to placebo, weight was reduced by −1.81 kg (WMD, 95% CI −2.04 to −1.57, p<0.00001, no significant heterogeneity) after up to 26 weeks of treatment. Weight reductions ranged from −0.14 to −4.5 kg in the 10 mg dapagliflozin groups and weight change ranged from +1.64 to −1.9 kg in the placebo groups. After 48–52 weeks of treatment, weight was reduced by −2.36 kg (WMD, 95% CI −2.85 to −1.88, p<0.00001, three RCTs) compared to placebo (range +0.69–−4.39 kg for the 10 mg dapagliflozin groups and +2.99 to −2.03 kg for the placebo groups). This reduction was significantly greater than the change at up to 26 weeks (p=0.04).
In the RCT comparing dapagliflozin to glipizide, weight decreased by −3.22 kg (95% CI −3.56 to −2.87) in the dapagliflozin arm after 52 weeks of treatment and increased by +1.44 kg (95% CI +1.09 to +1.78) in the glipizide arm (p<0.0001 between groups).11 In the RCT of canagliflozin, weight was reduced by between −2.3 (SE 0.39) and −3.4 (SE 0.39) kg in the canagliflozin groups with similar reductions of −3.4 kg in the groups receiving 300 mg once and twice daily (vs −1.1 SE 0.29 with placebo and −0.6 SE 0.39 with sitagliptin).16
Wilding et al14 also recorded waist measurement, and reported reductions of 2.5 cm on dapagliflozin 10 mg daily and 1.3 cm on placebo.
Systolic blood pressure
Dapagliflozin produced a reduction in systolic blood pressure at all doses (p values generally not reported) ranging from −1.3 to −7.2 mm Hg in the 10 mg dapagliflozin groups compared to changes of +2 to −0.11 mm Hg in the control groups. Rosenstock et al16 reported a systolic blood pressure reduction in response to canagliflozin ranging from −0.9 SE 1.7 mm Hg with 50 mg once daily to −4.9 SE 1.5 mm Hg with 300 mg once daily (−1.3 SE 1.5 mm Hg with placebo, −0.8 SE 1.4 mm Hg with sitagliptin).
Fasting plasma glucose
A significant reduction in FPG was seen in all dapagliflozin groups compared to placebo, with 10 mg dapagliflozin reducing FPG between −0.86 and −1.47 mmol/l more than control. There was no significant difference between FPG reductions with dapagliflozin versus glipizide in the study by Nauck.11
Canagliflozin reduced FPG by between −0.9 and −1.4 mmol/l (SE 0.20–0.22) with similar effects in the groups receiving 100, 200 or 300 mg once daily or 300 mg twice daily (vs +0.2 SE 0.20 mmol/l with placebo and −0.7 SE 0.20 mmol/l with sitagliptin).16
UTI and genital tract infection
Overall, there was a slight increase in the rate of UTIs when comparing 10 mg dapagliflozin with placebo (risk ratio 1.44, 95% CI 1.05 to 1.98, p=0.02), with a mean rate of 8.8% in the 10 mg dapagliflozin group (range 0–12.1%) and of 6.1% in the control groups (range 0–8.2%).
There was also an increase in genital tract infections when comparing 10 mg dapagliflozin with placebo (risk ratio 3.42, 95% CI 2.19 to 5.33, p<0.00001), with a mean rate of 9.5% in 10 mg dapagliflozin groups (range 0–12.3%) and 2.6% in the control groups (range 0–5.2%).
In most studies, the incidence on UTI or genital tract infections showed no dependence of dapagliflozin dose.
In the canagliflozin study, rates of UTIs ranged from 3.1% to 9.2% in the canagliflozin groups versus 6.1% with placebo and 1.5% with sitagliptin. Corresponding rates for genital tract infections were 3.1–7.8% in the canagliflozin groups, and 1.5% in both the placebo and the sitagliptin groups. There was no evidence of a dose dependence.16
In all cases the reported UTI and genital tract infections were not severe and resolved with simple treatment.
Overall, there was no significant difference in all types of hypoglycaemia between dapagliflozin and placebo groups. Hypoglycaemia, where data permitted, was divided into three categories: severe, moderate and other, corresponding, respectively, to capillary glucose readings of; <3.0 mmol/l (with external assistance required), <3.5 mmol/l, and symptoms suggestive of hypoglycaemia, but without confirming capillary glucose measurement. The incidence of all forms of hypoglycaemia in the dapagliflozin groups ranged from 1.1%16 to 56.6% (ref. 15, any dose of dapagliflozin+insulin±oral anti-diabetes drugs (OAD)).
Wilding et al14, reported more than a doubling of all hypoglycaemic events when dapagliflozin and insulin were compared to placebo and insulin (27% compared to 13%), but with only 16 hypoglycaemic episodes in a total of 71 participants.14 Strojek et al13 reported a small, dose independent, increase in hypoglycaemia from dapagliflozin 2.5, 5 and 10 mg, producing hypoglycaemia rates of 7.1%, 7.5% and 7.9%, respectively, compared to 4.7% for placebo and glimepiride, however again with only a small number hypoglycaemic events, 29 among 592 participants. Nauck et al11 reported that compared to glipizide, dapagliflozin produced a significant reduction in all types of hypoglycaemic events, with an incidence of 3.4% compared to 39.7% (14 vs 162 events).
Rosenstock et al16, comparing placebo to canagliflozin, found a hypoglycaemic event rate of 2% in the placebo group, of 0–6% in the canagliflozin groups (highest rate in the 200 mg once daily group, no dose dependence), and 5% in the sitagliptin group. The severity was not commented on.
Other adverse events
Three studies reported deaths in dapagliflozin groups (Bolinder et al (2011) (one death), Strojek et al (2011) (two deaths), Wilding et al (2012) (two deaths)).9 Causes of death were cardiopulmonary arrest, pulmonary embolism after ischaemic stroke, pneumonia due to oesophageal variceal haemorrhage, cardiogenic shock after aortic valve replacement and coronary bypass surgery, and acute myocardial infarction. None of the events considered to be the result of the study medication. Three deaths were reported by Nauck11 in the glipizide group.
Six studies found similar rates of study discontinuation due to adverse events between the study groups, whereas two studies found slightly higher rates in the dapagliflozin groups (5.6% vs 0% in ref. 9, 9.1% vs 5.9% in ref. 11). Five studies reported small numbers of renal impairment or failure in the different study groups and four of these reported no differences between study groups whereas in the study by Nauck et al11, rates were slightly higher in the dapagliflozin than in the glipizide group (5.9% vs 3.4%). In one study, dapagliflozin was found to have no significant effect on bone formation and resorption or bone mineral density over 50 weeks of treatment.9 ,10
SGLT2 inhibitors, when used in combination therapies, and administered to individuals with type 2 diabetes who had previously reported poorly controlled blood glucose, were shown to be effective in
Improving weight loss in conjunction with advice on lifestyle and diet.
Lowering systolic blood pressure.
Decreasing FPG levels.
Given the mechanism of action of the SGLT2 receptor inhibitors, the incidence and severity of hypoglycaemia would be expected to low.17 Nauck et al11 in one of the largest studies (801 participants), found a significantly higher incidence of hypoglycaemia in the sulphonylurea group, than with dapagliflozin. Hypoglycaemia in patients treated with SGLT2 receptor inhibitors was seen to be greatest when used in combination with insulin.
The present evidence suggests that the optimum dose of dapagliflozin may be 10 mg once daily, since there appears to be little additional benefit from increasing the dose to 20 mg. However, we have, at present, only one study evaluating the 20 mg dose, and then with only 23 patients allocated to that arm.
Implications for future practice
The number of glucose-lowering drugs for type 2 diabetes has been gradually increasing. We now have nine classes, though some contain only a single drug
Meglitinide analogues, nateglinide and repaglinide
The issue that arises is where the SGLT2 inhibitors fit into the therapeutic pathway. Factors to be considered include:
Effect on glycaemic control as reflected in HbA1c reductions.
Effect on weight, compared to other drugs, some of which cause marked weight gain.
Adverse effects, particularly increased genital and urinary infections.
Duration of effectiveness: some other drugs exhibit decreasing efficacy as duration of diabetes increases, especially those that act mainly by stimulating insulin release; the duration of action is unlikely to be affected by remaining levels of endogenous insulin production.
Interactions with other drugs, especially in patients on treatment for comorbidities.
Ease of use, by oral administration rather than injection.
The fear of hypoglycaemia can have a significant impact on the patient's QoL. The studies in this review recruited patients who were poorly controlled on present medications. Future trials might examine the role of the SGLT2 inhibitors in reducing the frequency of hypoglycaemic episodes in patients with good control but at the cost of hypoglycaemia. There is also the potential for their evaluation for use in poorly controlled type 1 diabetes.
Limitations of studies reviewed
There are no long-term data on SGLT2 side effects, both in terms of rare complications yet to be identified, but also on the long-term effects of significant glycosuria on the urinary tract. Two extension studies, published at present only as conference abstracts, reported that weight loss was maintained to 2 years. Del Prato et al,18 in an extension of the Nauck study with 624 of the original 801 participants, reported 2 year weight loss of 3.7 kg on dapagliflozin compared to a gain of 1.36 kg on glipizide. Wilding et al19 in a follow-up of 64% of original participants, reported that by 2 years, weight had increased by 1.8 kg in the placebo group but had decreased by 1.4 kg in the 10 mg dapagliflozin group.
No studies in this review analysed their data by duration of diabetes. It is possible that the SGLT2 receptor inhibitors might be particularly useful in patients with longer duration in whom other agents such as the sulphonylureas may be becoming less effective due to loss β cell capacity.
Data of canagliflozin come from only one paper. Only two studies14 ,15 examined the use of dapagliflozin in triple therapy, with insulin and no trials examined the role of the SGLT2 receptor inhibitors in triple oral therapy.
The costs of the drugs are not yet known, so cost-effectiveness cannot be assessed. The sulphonylureas are now very low cost, so the SGLT2 receptor inhibitors are very unlikely to be cost-effective compared to them. They are likely to be used in patients in whom metformin and sulphonylureas are insufficient or not tolerated, so the main comparators may be the gliptins, which have similar effects on HbA1c, are weight-neutral and which also increase the risk of UTIs, by about 40%.20
Musso et al21 produced a systematic review of SGLT2 inhibitors that included 13 articles. The main reasons for the difference between our own review and that of Musso et al are our focus on a real-world use of SLGT2 inhibitors, and inclusion of recent trials. We excluded studies of less than 8 weeks in duration, while Musso et al analysed studies as short as 2 weeks. In addition, Musso et al included studies with SGLT2 inhibitors as primary intervention, while the present study has only looked at SGLT2 inhibitors as in combination therapy.
Musso et al reached similar conclusions to our own, namely that SLGT2 inhibitors are effective at reducing HbA1c and fasting plasma glucose levels and BMI, while also observing a reduction in serum uric acid and blood pressure. They concluded that there is an increased risk of UTIs with SGLT2 inhibitors, with an OR of 1.34, which is similar to our own findings.
The US Food and Drug Administration (FDA) reviewed dapagliflozin in July 2011.22 They felt unable to approve it without additional safety data, mainly because of concerns about bladder and breast cancer. In the study data, there were nine cases of breast cancer in the dapagliflozin groups and none in the control groups. Some of these cancers occurred not long after dapagliflozin had been started. The absence of breast cancers among the controls was considered unexpected. An analysis by the manufacturers gave a standardised incidence ratio of 1.27 (95% CI 0.58 to 2.41) but this was not sufficient to reassure the FDA committee. There were nine cases of bladder cancer in those taking dapagliflozin and only one in the control groups, though it was noted that in five cases, haematuria had been recorded before dapagliflozin was started. The FDA committee noted that the imbalance might possibly be due to detection bias. The committee voted nine to six against approval.
The SGLT2 inhibitors are effective in lowering raised blood glucose, and as far as can be assessed from short-term results, appear safe. Their cost is not yet known, and so their place relative to other drugs is not yet clear. It is unlikely that dapagliflozin will be used as first-line monotherapy, on cost-effectiveness grounds.
We thank Dr Pamela Royle for help with updating searches.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online appendix
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.