Canagliflozin, dapagliflozin and empagliflozin as monotherapies for treating type 2 diabetes

The committee discussed the clinical trials identified in the assessment group (AG) report and the company submissions. It was aware that the AG had not identified any additional trials to those originally identified in the company submissions, and that the AG considered the trials to be generally of good quality. The committee concluded that all relevant trials had been identified, and were of an appropriate quality for decision‑making.

The committee discussed the baseline patient characteristics in the clinical trials. It questioned whether the results were generalisable to UK clinical practice, because several of the trials were done in populations described as 'Asian' (for example, China and Japan). It heard from the clinical experts that baseline measurements such as BMI were likely to vary between these trial populations and the UK population (the UK population is likely to have a higher BMI). This was an important consideration when interpreting secondary outcomes such as weight. However, overall the clinical experts stated they had no concerns about generalisability because:

• there were still people in the clinical trials with high BMIs

• the primary outcome, change in HbA1c, was based on a physiological response to the drug, which would not generally be affected by baseline measurements such as BMI and

• patient outcomes seen in UK clinical practice reflected the positive results seen in the clinical trials.
  
  The committee concluded that the trials were relevant to UK practice and appropriate for decision‑making.

The committee discussed the results of the clinical trials, most of which had compared SGLT‑2 inhibitors with placebo for outcomes such as change in HbA1c and weight. The committee was aware that the SGLT‑2 inhibitors had shown statistically significant improvements compared with placebo for the primary outcome of change in HbA1c. The clinical experts stated that this is the main goal of treatment with medication for diabetes. People had also had reductions in weight compared with placebo, which the clinical experts described as a welcome additional benefit of the SGLT‑2 inhibitors. The committee concluded that the SGLT‑2 inhibitors were a clinically effective treatment compared with placebo.

The committee considered if there was any evidence of differences in effectiveness between the SGLT‑2 inhibitors. It heard from the clinical experts that although it could be advantageous to have the option to increase the dose (as was possible with canagliflozin and empagliflozin), there was no direct evidence available to determine if there are clinically meaningful differences among the SGLT‑2 inhibitors. In response to the appraisal consultation document, Janssen (the company for canagliflozin) stated that although it agreed there was no direct evidence available, the indirect evidence from the AG and company network meta‑analyses (see sections 3.11 to 3.23) demonstrated that canagliflozin was the most effective treatment. The committee was aware that some results showed differences between the SGLT‑2 inhibitors. For example, some of the network meta‑analyses suggested dapagliflozin had a lower HbA1c response than canagliflozin and empagliflozin. However, it heard from the AG and AstraZeneca (the company for dapagliflozin) that the results for dapagliflozin were sensitive to the inclusion of a trial that had different baseline patient characteristics to the others, and when this was removed results were similar to the other SGLT‑2 inhibitors. The committee also noted that the differences in effectiveness between the SGLT‑2 inhibitors in the network meta‑analyses were most pronounced in people having the higher doses of canagliflozin and empagliflozin, which were not licensed starting doses. The committee agreed that the results demonstrated in the higher dose groups, which were derived from clinical trials where patients were able to start on the higher doses, may not be seen in clinical practice, where people eligible for and requiring a higher dose would be required to titrate up to it. Furthermore, the committee heard from Janssen that in clinical practice the majority of patients use the lower, 100 mg, starting dose of canagliflozin. The committee concluded that from the evidence available it was not possible to determine if there are any differences in effectiveness between the SGLT‑2 inhibitors.

The committee discussed the adverse events associated with SGLT‑2 inhibitors. It heard from the clinical experts that genital fungal infections were a concern. However, there is debate about whether this is a treatment or disease‑related effect, and when infections did occur, they were typically one‑off and were not serious. Furthermore, for more serious outcomes such as malignancy, the clinical experts stated that there were no data to suggest an increase in risk associated with SGLT‑2 inhibitors (although long‑term data are needed). The committee also heard from the patient about her experience with SGLT‑2 inhibitors. The patient expert had not had any adverse events. She noted that she had been advised to drink plenty of water, which had probably reduced her risk of having an adverse event, although this resulted in an increased need to pass water, which could be an inconvenience for some people. The committee concluded that the SGLT‑2 inhibitors had an acceptable adverse event profile.

The committee discussed the structure of the AG and company models, considering if any single model was most appropriate for decision making. It was aware that the AG considered all the models to be of reasonable quality, and it noted that the structure of the models was generally similar, but there were some important differences. For example, the Janssen model assumed a 'linear in segment' (see section 3.30) progression of HbA1c whereas all other models based progression of disease on the UK Prospective Diabetes Study (UKPDS) equations. Furthermore, the AG model assumed that when people intensified treatment, treatments were added, rather than switched (as assumed in all company models at either first or second intensification). The committee heard from the clinical experts that the AG model was the most similar to NHS clinical practice, because clinicians typically retain oral therapies and add another treatment, to reduce the risk of losing control of the disease. Furthermore, the committee noted that NICE's guideline on type 2 diabetes had used the same model used by the AG (OM1, see section 3.27), preferring it to other models because it was based on 1 trial only (the committee was aware this was unlike ECHO‑T2DM which was based on multiple sources) and also because it matched the NICE reference case, was internally and externally validated, and allowed for modelling of additional short‑term outcomes. The committee was aware that in response to the appraisal consultation document, Janssen had described several advantages of the ECHO‑T2DM model over the other submitted models. For example, it stated that ECHO‑T2DM does not double count the treatment effect associated with rescue medication, unlike other models. The committee acknowledged there were different advantages and disadvantages to the different modelling approaches, and it agreed with the AG that all models submitted were appropriate and of a reasonable quality. However it concluded that the AG model was most appropriate for decision‑making, because of its more accurate reflection of treatment intensification.

The committee discussed the quality‑of‑life assumptions in the AG and company models. It noted that there were generally very small quality‑adjusted life year (QALY) differences between the various treatments. It also noted that the AG had presented a number of scenarios varying the assumptions about BMI. 'No BMI' assumed BMI had no impact on quality of life (worst‑case scenario for SGLT‑2 inhibitors), and BMI scenarios 1 to 5 varied the duration of treatment effect on weight loss (where BMI‑1, with weight changes maintained with no rebound to natural history, was the best‑case scenario for SGLT‑2 inhibitors). The committee agreed that weight loss does affect quality of life and agreed with the approach of a disutility of 0.0061 being applied per BMI point greater than 25. The committee noted that the evidence had shown that SGLT‑2 inhibitors do have a significant effect on weight loss, and felt that the AG's BMI‑2 scenario (in which weight gains were maintained and weight losses rebounded to natural history after 1 year) best reflected the treatment effect on weight loss. The committee noted that in response to the appraisal consultation document, Janssen stated it considered the 'BMI‑3' scenario (where weight gains were maintained, and weight losses rebound to natural history at intensification) to be the most appropriate scenario based on expert clinical opinion, although it had noted a lack of evidence to support any assumptions about the duration of weight loss. The committee agreed there is a lack of long‑term evidence to support firm conclusions about the duration of treatment‑related weight change. However it noted that NICE's guideline on type 2 diabetes used the same assumption for the duration of weight gains and losses as that used in scenario BMI‑2, because the clinical evidence on treatment‑related weight‑change was presented at follow‑up of 1 year and 2 years, but was very limited at 2 years. The committee therefore concluded that BMI‑2 was the most plausible scenario, but noted that the small QALY differences between treatments made the ICERs unstable.

The committee discussed the cost‑effectiveness results presented in the AG and company models. It noted that the Janssen model had highly favourable ICERs of £4,400 to £7,900 per QALY gained for canagliflozin 100 mg or 300 mg compared with sulfonylureas, whereas ICERs in the other models were substantially higher (for example, £59,000 per QALY gained for the SGLT‑2 inhibitors compared with the sulfonylureas in the AstraZeneca model, and £71,000 per QALY gained in the AG model using the BMI‑2 scenario). The committee heard from the AG that it considered the most probable driver of the differences between the AG and Janssen cost‑effectiveness results to be the different modelling assumptions used for treatment intensification (see section 3.33). Discontinuing oral therapies when intensifying to insulin meant that patients received the relatively expensive oral therapies for a relatively short period of time in the Janssen model, contributing to lower incremental costs for canagliflozin relative to a sulfonylurea when compared with the AG model. The committee noted that the Janssen model showed total costs for canagliflozin 300 mg were around £500 more expensive than a sulfonylurea, whereas in the AG model total costs were around £5,000 more expensive. The committee concluded there was uncertainty about the reason for the favourable cost‑effectiveness results for canagliflozin compared with a sulfonylurea in the Janssen model, but that the increased costs arising from retaining oral treatments in the AG model (a committee preferred assumption) was likely to be an important contributing factor.

The committee discussed whether it could determine the most plausible ICERs for the SGLT‑2 inhibitors compared with the relevant comparators (compared with each other, pioglitazone, sulfonylureas and DPP‑4 inhibitors), using its preferred model (AG model, see section 4.8) and its preferred assumptions about the effect of treatment on BMI (scenario BMI‑2, see section 4.9).

• When compared with pioglitazone, ICERs for all the SGLT‑2 inhibitors were more than £52,400 per QALY gained.

• When compared with sulfonylureas, ICERs for the SGLT‑2 inhibitors were all over £71,000 per QALY gained.

• When compared with DPP‑4 inhibitors, the ICERs for all SGLT‑2 inhibitors were less than £29,300 per QALY gained.

• When the SGLT‑2 inhibitors were compared with each other, the committee agreed that the clinical evidence (see section 4.6) and cost data (all tablets cost the same and had the same frequency of administration) did not support any differences between them.

The committee discussed the most plausible ICERs. For canagliflozin 300 mg, the committee noted that its preferred base case result from the AG had been updated following responses to the appraisal consultation document (to use a corrected value for systolic blood pressure for canagliflozin, see section 3.57). This had reduced the ICER when compared with sulfonylureas or pioglitazone to at least £52,400 per QALY gained, the lowest cost‑effectiveness estimates of the SGLT‑2 inhibitors, but still not within the range usually considered to be a cost‑effective use of NHS resources. The committee also noted that even these ICERs were possibly optimistic, because the clinical‑effectiveness assumptions were based on the more favourable results of the 300 mg dose, which was not the most commonly used dose in clinical practice (see section 4.6). In summary, at ICERs of £20,000 to £30,000 per QALY gained, the SGLT‑2 inhibitors were cost effective compared with DPP‑4 inhibitors, but not cost effective compared with pioglitazone and sulfonylureas. Therefore the committee concluded that canagliflozin, dapagliflozin and empagliflozin as monotherapy were a cost‑effective use of NHS resources, but only when pioglitazone or sulfonylureas were not appropriate treatment options.

The committee discussed the most appropriate wording of the final recommendation. It considered whether to recommend that the least costly SGLT‑2 inhibitor should be used first. However, because all the SGLT‑2 inhibitors cost the same per tablet and have the same dosing frequency, and the clinical data had not robustly demonstrated that any one SGLT‑2 inhibitor was more clinically effective than the others (see section 4.6), it could not determine a meaningful way to differentiate treatments by cost. It also noted concerns raised in response to the appraisal consultation document from Janssen (the company for canagliflozin) that the wording of the draft recommendation implied that a sulfonylurea or pioglitazone should be used before SGLT‑2 inhibitors or DPP‑4 inhibitors, which could affect individualisation of care. The committee discussed whether there was a need to reword the draft recommendation when the final guidance was published. It stated that the existing wording of the recommendation reflected the evidence base, and that the phrasing of the recommendation that SGLT‑2 inhibitors are only recommended where 'a sulfonylurea or pioglitazone was not appropriate' still allowed for clinical choice and individualisation of care; the recommendation does not imply that any treatments have to be tried first if they are not appropriate. The committee agreed that the wording of the recommendation allowed clinicians the freedom to prescribe SGLT‑2 inhibitors when they feel it is appropriate. It also agreed that there was no meaningful way to recommend a hierarchy of SGLT‑2 inhibitors based on cost. It therefore concluded that the final wording of the recommendation should remain the same as the draft recommendation that was consulted on.

The committee considered whether it should take into account the consequences of the Pharmaceutical Price Regulation Scheme (PPRS) 2014, and in particular the PPRS payment mechanism, when appraising canagliflozin, dapagliflozin and empagliflozin as monotherapy. The committee noted NICE's position statement in this regard, and accepted the conclusion 'that the 2014 PPRS payment mechanism should not, as a matter of course, be regarded as a relevant consideration in its assessment of the cost effectiveness of branded medicines'. The committee heard nothing to suggest that there is any basis for taking a different view with regard to the relevance of the PPRS to this appraisal of canagliflozin, dapagliflozin and empagliflozin as monotherapy. It therefore concluded that the PPRS payment mechanism was not relevant for its consideration of the cost effectiveness of canagliflozin, dapagliflozin and empagliflozin as monotherapy.