Canagliflozin in combination therapy for treating type 2 diabetes

The Committee considered the evidence on the clinical effectiveness of canagliflozin compared with other antidiabetic treatments and noted that the manufacturer's submission contained the results from 3 head-to-head trials (DIA3006, DIA3009 and DIA3015) comparing canagliflozin with an active comparator: a DPP‑4 inhibitor (sitagliptin) or a sulfonylurea (glimepiride) in dual therapy and a DPP‑4 inhibitor (sitagliptin) in triple therapy. The Committee recalled that the Evidence Review Group (ERG) had considered these trials to be of generally good methodological quality. The Committee was aware, however, that most data came from the manufacturer's network meta-analyses. The Committee heard from the ERG that the manufacturer's searches had been well conducted and that it was generally satisfied with the manufacturer's approach to the network meta-analyses, although it was initially unclear why the triple therapy meta-analyses had been conducted with 26‑week data instead of 52‑week data. The manufacturer clarified that this was because there were more comparators with data available at 26 weeks than at 52 weeks. The Committee heard that the ERG agreed that using 52‑week data for the triple therapy meta-analysis could make it difficult to construct a suitable network. The Committee also heard from the ERG that results for the 26‑week and 52‑week dual therapy meta-analyses were similar, and considered that this lessened the uncertainty around using the 26‑week data for triple therapy. The Committee concluded that the results of the manufacturer's network meta-analyses provided an appropriate basis for making decisions about the clinical effectiveness of canagliflozin compared with other antidiabetic therapies.

The Committee discussed the generalisability of the clinical trial populations to the NHS patient population in England that was potentially eligible for treatment with canagliflozin. The Committee was satisfied that the populations in the dual therapy and triple therapy trials were generalisable, but had concerns about the population in the DIA3008 insulin sub-study. This was because the sub-study was part of the CANVAS trial, which only included patients with a history of, or estimated, high risk of cardiovascular disease. The Committee heard from the clinical specialists that, in general, by the time patients have tried different oral treatments and progressed to insulin (around 8 to 10 years from diagnosis), they would be likely to have an increased risk of cardiovascular disease compared with the general population. The Committee was persuaded that patients on insulin for type 2 diabetes would be at increased risk of cardiovascular disease as a result of the condition, and concluded that the results of all the canagliflozin trials (including the DIA3008 insulin sub-study) were generalisable to the population likely to receive canagliflozin in clinical practice in England.

The Committee considered the clinical effectiveness of canagliflozin as dual therapy in combination with metformin. The Committee noted that the evidence came from 2 clinical trials and a network meta-analysis. It noted that head-to-head trials had compared canagliflozin at doses of 100 mg and 300 mg with glimepiride (a sulfonylurea) and with sitagliptin (a DPP‑4 inhibitor). The Committee considered that, on the basis of current clinical guidelines and clinical practice, SGLT‑2 inhibitors would only be considered for use as dual therapy in combination with metformin when sulfonylureas were unsuitable, and that therefore sulfonylureas were not a relevant comparator. However, the comparison with sitagliptin was relevant to the decision problem. For sitagliptin, the Committee considered that the data suggested that canagliflozin has a broadly comparable efficacy to sitagliptin in reducing HbA_1c, and greater efficacy in reducing body weight and systolic blood pressure. The Committee concluded that, on the basis of the results of the clinical trials and network meta-analyses (see sections 3.13 to 3.16), canagliflozin as part of dual therapy in combination with metformin appeared to provide broadly comparable glycaemic control to the other antidiabetic drugs, including the key comparators DPP‑4 inhibitors and dapagliflozin, and may result in greater weight loss and lowering of blood pressure than DPP‑4 inhibitors.

The Committee discussed an alternative dual therapy combination that was specified in the NICE scope: canagliflozin in combination with a sulfonylurea. The Committee noted that the manufacturer had not made a case for canagliflozin in combination with sulfonylurea in its submission because most people with diabetes would initially use metformin. The Committee agreed that most patients would start on metformin monotherapy, but was aware that, if metformin was unsuitable, some patients would receive a sulfonylurea instead. However, the Committee concluded that it was unable to make a recommendation on the dual therapy combination of canagliflozin and a sulfonylurea because the manufacturer had not provided any clinical data.

The Committee discussed the clinical effectiveness of canagliflozin as part of triple therapy in combination with metformin and a sulfonylurea. The Committee noted that the evidence came from 2 clinical trials and a network meta-analysis. It acknowledged the availability of head-to-head trial results for canagliflozin compared with a sitagliptin (a DPP‑4 inhibitor), but was aware that DIA3015 only investigated the higher dose of canagliflozin. The Committee noted that DIA3015 showed that canagliflozin 300 mg statistically significantly reduced HbA_1c, systolic blood pressure and body weight compared with sitagliptin. The Committee was aware that in DIA3002, which compared canagliflozin with placebo in triple therapy, benefits were seen in reducing HbA_1c and body weight, but not systolic blood pressure. The Committee concluded that, on the basis of the clinical trials and the manufacturer's meta-analysis (see section 3.17), canagliflozin as part of triple therapy in combination with metformin and a sulfonylurea gave a comparable HbA_1c reduction compared with a DPP‑4 inhibitor. It further concluded that reduction in weight and systolic blood pressure was greater with canagliflozin than a DPP‑4 inhibitor.

The Committee discussed the clinical effectiveness of canagliflozin as part of triple therapy in combination with metformin and a thiazolidinedione. The Committee noted that the evidence came from 1 clinical trial and a network meta‑analysis. The Committee noted that the clinical trial compared canagliflozin with placebo and not an active comparator, and that it was more effective than placebo in lowering HbA_1c, body weight and blood pressure. It heard from the clinical specialists that the use of thiazolidinediones was decreasing in clinical practice in England because of safety concerns about cardiac problems and bladder cancer, as well as its association with weight gain. The Committee noted that although few people would start taking pioglitazone these days, there would be some people who were still taking metformin and pioglitazone as dual therapy. Although the Committee considered that this was likely to be uncommon in clinical practice, it agreed that the population being small should not prevent the Committee considering canagliflozin as a possible treatment for this group of patients. The Committee concluded that, on the basis of the clinical trials and the manufacturer's meta-analysis, the evidence for canagliflozin as part of triple therapy in combination with metformin and a thiazolidinedione was adequate to confirm that it is clinically effective in this combination.

The Committee considered the clinical effectiveness of canagliflozin as an add‑on treatment to insulin, noting that the evidence came from 1 clinical trial (the DIA3008 insulin sub-study) and a network meta-analysis. The Committee noted that the trial was placebo controlled and only 18 weeks long, and that patients were taking a range of background treatments. The Committee agreed with the manufacturer's opinion that the differences between the patient population of the DIA3008 insulin sub-study and other studies, and the heterogeneity in background treatments across the studies, were limitations of the insulin meta-analysis. Nevertheless, it considered the patient population in the DIA3008 insulin sub-study to be generalisable to the patient population in England (see section 4.6) and the range of background treatments to be typical of clinical trials investigating treatments at this point in the treatment pathway. The Committee concluded that, despite the limitations associated with the methodology of the DIA3008 insulin sub-study and the network meta-analysis, their results showed that canagliflozin as add-on therapy to insulin appeared to be slightly more effective in reducing HbA_1c and body weight than DPP‑4 inhibitors and dapagliflozin.

In light of the dosing options specified in canagliflozin's marketing authorisation, the Committee discussed how dose escalation with canagliflozin might be implemented in clinical practice in England, and how it might relate to its clinical effectiveness. It was aware that canagliflozin 100 mg and 300 mg had been studied separately in the clinical trials, and that there was no clinical trial evidence for the clinical effectiveness of canagliflozin after dose escalation. It heard from the clinical specialists that there was no experience in escalating the drug dose, and that it was not clear how many people would need dose escalation, or when this might occur. The Committee noted the comments received from commentators in response to the appraisal consultation document that data from the USA suggested that 50% of people taking canagliflozin 100 mg escalated to the canagliflozin 300 mg dose. However, the Committee heard from the manufacturer that its observational data showed that 75% of people took 100 mg and 25% of people escalated to the 300 mg dose, and that if dose escalation occurred it would be soon after starting treatment (rather than after prolonged exposure). The Committee concluded that there was uncertainty about the proportion of patients taking canagliflozin 100 mg who would escalate to canagliflozin 300 mg, the timing of this, and the precise clinical effectiveness of canagliflozin when escalating the dose in clinical practice.

The Committee discussed the adverse events associated with canagliflozin. It heard from clinical specialists that the risk of hypoglycaemia and weight gain were relatively low compared with some other antihyperglycaemic treatments, which was important to people with type 2 diabetes and their healthcare professionals. Focusing on the adverse events that were typically associated with SGLT‑2 inhibitors, the Committee heard from the clinical specialists that although rates of genital mycotic infections in the clinical trials were much higher than with placebo, these had generally been resolved by topical treatment and that recurrence rates were low. The Committee noted a comment received during consultation about a possible increase in low-trauma fracture rate associated with canagliflozin, but was aware that the European public assessment report described the increase as 'questionable'. The Committee also noted the consultation comments on canagliflozin's lipid profile and potential long-term cardiovascular adverse events, but was aware that the CANVAS safety trial was ongoing. The Committee concluded that the short-term adverse events for canagliflozin (100 mg and 300 mg doses) were manageable and that further data for longer-term outcomes are still needed, particularly for cardiovascular adverse events.

The Committee discussed the manufacturer's ECHO‑T2DM model, which had not been used in previous NICE technology appraisals of technologies used to treat type 2 diabetes. The Committee observed that the manufacturer's model followed the NICE reference case and methodology, noting that the utility values had mainly been derived from a large European dataset using EQ‑5D. It noted the ERG's view that the model had been well validated, and heard from the ERG that ECHO‑T2DM had been validated against the well-established CORE model as part of the Mount Hood challenge. The Committee was aware that the ERG had concerns about the stability of the incremental cost-effectiveness ratios (ICERs) generated using the model, especially when probabilistic uncertainty was included (as it was in the manufacturer's base case). The Committee heard from the manufacturer and the ERG that although the values may vary slightly in each model run, the incremental values were highly consistent. It further heard that the instability of the ICER was primarily because the differences in costs and quality-adjusted life year (QALY) benefits between treatments were very small, which meant that very slight changes in either of these could have a large effect on the calculated ratio. The Committee concluded that, despite some uncertainty about the stability of the results, the manufacturer's economic model was suitable for assessing the cost effectiveness of canagliflozin in combination therapy for treating type 2 diabetes.

The Committee discussed the cost-effectiveness analyses presented by the manufacturer, noting that these did not include all the comparators in the NICE scope. The Committee was aware that more comparators had been included in the network meta-analyses but then subsequently excluded from the cost-effectiveness analysis because the manufacturer had chosen a single comparator to represent each class. On balance, the Committee concluded that the manufacturer had included an adequate range of comparators for the cost-effectiveness analysis of canagliflozin in dual therapy, triple therapy and as an add-on to insulin therapy.

The Committee reviewed the manufacturer's general approach to incorporating clinical data into its economic model using a mixture of data from its meta-analyses and clinical trials, where available. The Committee noted that there was a lack of data in certain loops of the network meta-analyses and that these sometimes indicated slightly more efficacious estimates in favour of canagliflozin compared with some of the direct trial evidence. However, the Committee heard from the ERG that the meta-analyses of clinical effectiveness of canagliflozin and its comparators were sufficiently robust and generally consistent with the direct trial evidence, and had been appropriately incorporated into the model. The Committee therefore concluded that the clinical data incorporated into the manufacturer's model were acceptable for informing the cost-effectiveness estimates.

The Committee discussed specific areas of uncertainty relating to data used for the 2 SGLT‑2 inhibitors, canagliflozin and dapagliflozin, which had been identified by commentators during consultation. It noted that the same values for changes in lipids had been used for the 2 drugs in the model, and was aware that the ERG considered the evidence to show that dapagliflozin had a more favourable profile than canagliflozin. However, it heard that the ERG had explored the effect of using different profiles for these 2 SGLT‑2 inhibitors and that this had little impact on the ICER. The Committee also noted that that the manufacturer's sensitivity analyses showed that varying the lipid values did not strongly influence the ICER. The Committee was also aware that there was some uncertainty in the modelling of hypoglycaemia associated with the 2 SGLT‑2 inhibitors, because of differences in the clinical trials with these drugs, and noted the absence of a tornado plot for canagliflozin compared with dapagliflozin in the manufacturer's sensitivity analyses. However, it heard from the ERG that hypoglycaemia was not a key driver of the cost-effectiveness estimates when comparing canagliflozin with the other treatments. The Committee concluded that although there was some uncertainty about how some clinical outcomes had been modelled, this was acceptable because it would not have a material effect on the decision about cost effectiveness.

The Committee discussed how the manufacturer had modelled changes in clinical effectiveness over time. It noted that the annual drift values for biochemical and other risk factors had been based on available data (see section 3.30), but that the same annual drift values for HbA_1c were used for the SGLT‑2 inhibitors, DPP‑4 inhibitors and GLP‑1 analogues. The manufacturer clarified that although there were 4‑year data for the DPP‑4 inhibitors and GLP‑1 analogues, only 104‑week data were available for canagliflozin. The Committee considered this to have implications for the economic modelling, because the manufacturer's deterministic sensitivity analyses showed that results from the model were most sensitive to varying HbA_1c drift. Following comments received during consultation, the Committee heard from the manufacturer that the assumption that canagliflozin had an annual HbA_1c drift value of 0.14% was supported by coefficient durability data from DIA3009, which showed the rate of HbA_1c increase was 0.16% for both canagliflozin 100 mg and 300 mg over 72 weeks. The Committee heard from the ERG that it had received clinical advice that the manufacturer's values were appropriate. Although the Committee was concerned that extrapolating short-term data over a lifetime time horizon contributed uncertainty to the cost-effectiveness estimates, it was persuaded that the clinical data from DIA3009 were not inconsistent with the manufacturer's drift values assumed in the model. The Committee also heard from clinical specialists that sulfonylureas were recognised to have a high drift rate, that the ADOPT study indicated lower drift rates for other treatments, and that that the assumption of broadly equivalent drift rates for other treatments was not unreasonable. In the absence of any long-term clinical data, the Committee concluded that the manufacturer's approach to modelling the long-term effectiveness of canagliflozin was acceptable.

The Committee discussed the manufacturer's scenario for dose escalation. It heard from the clinical specialists that it was impossible to predict with any certainty the number of patients who would be eligible for dose escalation, or when this would occur (see section 4.12). The Committee observed that the summary of product characteristics stated that the higher dose could be used if tighter glycaemic control was needed, but did not provide a definition of this, or specify a time when dose escalation should take place. In the absence of any other information, the Committee accepted that the manufacturer's assumptions for dose escalation in its scenario analysis were not unreasonable. However, it considered that the results were associated with significant uncertainty and should be interpreted with caution, especially because the manufacturer had assumed that clinical effectiveness was the same in patients who had started on the 100 mg dose and escalated to 300 mg as those who had started and continued on the 300 mg dose. Given the absence of evidence to inform any alternative scenarios, the Committee concluded that the cost effectiveness of canagliflozin in combination therapy for treating type 2 diabetes should be informed by considering the cost-effectiveness estimates for the separate 100 mg and 300 mg doses for each treatment comparison.

The Committee discussed the level of uncertainty associated with the cost-effectiveness estimates from the manufacturer's economic model. It was aware of the ERG's opinion that the deterministic sensitivity analyses were incomplete and that the results were not clearly presented. It further noted that the ERG was unclear if the manufacturer's probabilistic sensitivity analyses had been sufficiently explored because the distributional assumptions were not transparent or well described. The Committee noted that the cost and QALY differences between canagliflozin and its comparators were generally stable, but also very small. It concluded that a level of uncertainty associated with the cost-effectiveness estimates generated using the manufacturer's model was inevitable with such small differences in the QALYs gained between canagliflozin and its comparators.

The Committee considered the most plausible ICERs for canagliflozin in combination with metformin as dual therapy. Based on clinical specialist opinion, the Committee decided that thiazolidinediones and sulfonylureas were not key comparators in this setting (see section 4.3). The Committee also believed that the GLP‑1 analogues were not key comparators because they are recommended only for a very small population (see sections 4.2 to 4.3). The Committee noted that the manufacturer's and ERG's analyses showed that canagliflozin was associated with higher costs and QALYs than DPP‑4 inhibitors and dapagliflozin, but that these incremental differences were small (see sections 3.32 to 3.37 and 3.49 to 3.52).The Committee understood that these low incremental costs and health benefits meant the ICERs could vary dramatically in response to even small changes because of a pronounced effect on the cost-effectiveness ratio. The Committee noted that the ERG had carried out a fully incremental analysis including all dual therapy combinations. However, it noted that this included treatment options that would not be appropriate for all patients in clinical practice. The Committee therefore examined appropriate parts of the incremental analysis, together with relevant pairwise analyses, to evaluate the cost effectiveness of canagliflozin in dual therapy compared with the key comparators (DPP‑4 inhibitors and dapagliflozin). The Committee saw that in some cases there was a very wide range of estimates of the ICER for the same drug comparison. After exploring the reasons for this, the Committee noted that the differences in costs over a 40‑year time horizon were modest (from £1 to £751) and the differences in QALYs were extremely small (from 0.002 to 0.04). The Committee considered that it was important to take these small incremental differences into account when interpreting the ICERs. Overall, the Committee concluded that because of the very small differences in costs and QALYs between canagliflozin and DPP‑4 inhibitors, and between canagliflozin and dapagliflozin, canagliflozin 100 mg and 300 mg in a dual therapy regimen in combination with metformin had been shown to be a cost-effective use of NHS resources. The Committee therefore recommended canagliflozin 100 mg and 300 mg as a treatment option when the alternative treatments would be a DPP‑4 inhibitor or dapagliflozin in line with the recommendations in NICE's guideline on type 2 diabetes (now replaced by NICE's guideline on type 2 diabetes in adults) and NICE's technology appraisal guidance on dapagliflozin in combination therapy for treating type 2 diabetes (that is, if they are at significant risk of hypoglycaemia or its consequences or if a sulfonylurea is not tolerated or contraindicated).

The Committee considered the most plausible ICERs for canagliflozin in combination with metformin and a sulfonylurea as triple therapy. Based on clinical specialist opinion, the Committee decided that the injectable therapies insulin and GLP‑1 analogues were not key comparators because canagliflozin would be most suitable for people who prefer to add a third oral treatment rather than an injectable one, and also because GLP‑1 analogues are only recommended for a specific patient population (see section 4.2). The Committee noted that the manufacturer's results showed that there were small incremental differences in costs and QALYs between canagliflozin and the DPP‑4 inhibitors. Although there was considerable uncertainty around the precise ICERs (especially because only canagliflozin 300 mg was investigated in DIA3015), the Committee concluded that because of the small differences in costs and QALYs between canagliflozin and DPP‑4 inhibitors, canagliflozin 100 mg and 300 mg in combination with metformin and a sulfonylurea in a triple therapy regimen had been shown to be a cost-effective use of NHS resources and should be recommended as a treatment option for people with type 2 diabetes.

The Committee considered the most plausible ICERs for canagliflozin in combination with metformin and a thiazolidinedione as triple therapy. It decided that the injectable therapies insulin and GLP‑1 analogues were not key comparators (see section 4.22). The Committee recalled that canagliflozin had been shown to be more clinically effective than placebo in this setting, and observed that there were small increases in costs and QALYs for the addition of canagliflozin compared with a DPP‑4 inhibitor. The Committee considered that these cost-effectiveness estimates were subject to uncertainty and noted that the model had not incorporated disutilities for the possible long-term adverse effects of thiazolidinediones, such as fractures and bladder cancer. However, it did note that the manufacturer's deterministic ICERs for the 2 doses were within the range which could be considered cost effective. On balance, the Committee concluded that canagliflozin 100 mg and 300 mg in combination with metformin and a thiazolidinedione in a triple therapy regimen had been shown to be a cost-effective use of NHS resources and should be recommended as a treatment option for people with type 2 diabetes.

The Committee considered the most plausible ICERs for canagliflozin as an add-on treatment to insulin. It noted that the NICE scope specified that the comparators were oral agents, and so disregarded the cost-effectiveness estimates for a GLP‑1 analogue. The Committee concluded that canagliflozin 100 mg and 300 mg had been shown to be a cost-effective use of NHS resources compared with DPP‑4 inhibitors and dapagliflozin as an add-on treatment to insulin because of its very small incremental costs and incremental QALYs. The Committee recommended canagliflozin as a treatment option for people with diabetes that is inadequately controlled by insulin with or without other oral antidiabetic drugs.

The Committee discussed whether canagliflozin was an innovative treatment and if there were any additional QALYs that had not been included in the manufacturer's model. The manufacturer said that that there were no health benefits that had not been captured in the model. The Committee concluded that all substantial benefits related to treatment with canagliflozin had been captured in the QALY calculation.