Note from the National Guideline Clearinghouse (NGC): The National Institute for Health and Clinical Excellence (NICE) commissioned an independent academic centre to perform a systematic literature review on the technology considered in this appraisal and prepare an assessment report. The Evidence Review Group (ERG) report for this technology appraisal was prepared by Warwick Evidence, University of Warwick (see the "Availability of Companion Documents" field).
Question 1: Is Once-Weekly Exenatide as Good, or Better Than, Twice Daily Exenatide?
The evidence on this comes from two randomised controlled trials (RCTs) - DURATION 1 and DURATION 5. These were very similar. Full details are given in the industry submission, but in brief:
- The trials were of good quality. Their risk of bias scores were low (see Appendix 1 of the ERG report [see the "Availability of Companion Documents" field]) with the only problem being the inevitable one that blinding of patients is not practical when one arm has twice daily injections and the other has once weekly injections. However the main outcome, glycated haemoglobin (HbA1c), is an objective laboratory measure which would not be affected.
- Both compared twice daily with once weekly exenatide in patients with poor glycaemic control (HbA1c baseline mean 8.3%, range 7.1% to 11.0%) on a mixture of treatments. In DURATION 1 15% were on diet and exercise alone, 43% to 46% were on monotherapy (mostly metformin), and 36% to 39% were on dual therapy. In DURATION 5 16 to 21% were on diet and exercise alone, 43% to 50% were on monotherapy (mostly metformin) and 28% to 40% were on dual therapy. Hence these trials do not reflect the real-life position on exenatide long acting (LA) in triple therapy because a minority were on dual therapy.
- The trials were funded by Amylin and Lilly, which were involved in design, and collection, and analysis of data. Some authors were from the manufacturers.
So the patient groups were mostly not relevant to the decision problem (use of exenatide in triple therapy) but the trials are satisfactory for assessing the effectiveness of exenatide LA versus the twice daily (BD) form.
Exenatide LA versus Liraglutide
DURATION 6 compared exenatide LA with liraglutide 1.8 mg daily in 911 patients. Quality of study as assessed by risk of bias table was good, with the only problem being the impracticality of blinding because of different dosing frequencies.
However, the 1.8 mg dose of liraglutide is not recommended by NICE, on cost-effectiveness grounds. So the competitor for once-weekly exenatide is 1.2 mg liraglutide. Unfortunately, DURATION-6 did not include an arm with 1.2 mg liraglutide. The manufacturer therefore commissioned a network meta-analysis to provide an indirect comparison of liraglutide 1.2 mg with exenatide LA.
The Oxford Outcomes Network Meta-analysis
This was a good quality review, carried out because there was no head to head trial comparing liraglutide 1.2 mg with exenatide LA. One could have made some minor criticisms, such as that the Jadad scoring system for RCTs is now rather out-dated, and the Cochrane risk of bias method might have been better.
One point of difference from the ERG's analysis is the size of difference in HbA1c between the two liraglutide doses. The ERG's meta-analysis estimated that the difference in HbA1c was 0.10%, the Oxford estimate is 0.17%. Most of the difference is explained by the exclusion of LEAD-3 from the ERG's meta-analysis, on the grounds that LEAD-3 was a monotherapy trial – liraglutide alone versus glimepiride alone. (See Figures 7 and 8 of the ERG report [see the "Availability of Companion Documents" field] for comparison of HbA1c change with or without LEAD-3 trial, respectively.)
The ERG accepted the Oxford Outcomes analysis that liraglutide 1.2 mg and exenatide LA are clinically equivalent.
Exenatide LA Versus Insulin
DURATION 3 compared exenatide LA with glargine insulin in 456 patients previously treated with metformin alone (70%) or metformin + sulphonylurea (30%). Baseline HbA1c was 8.3% and body mass index (BMI) 32.
The study was funded by Lilly and Amylin, the companies were involved in design, data collection and analysis, and four of the seven authors were from the companies.
The summary drew on the Cochrane review of the glucagon-like peptide (GLP)-1 analogues. Of which two members of the ERG were authors.
The end-of-study dose of glargine was 31 units, which was greater than in the LEAD-5 study where the dose only reached 24 units. This suggested more effective titration and a more reliable comparison.
Comparison with Sitagliptin and Pioglitazone
The DURATION 2 trial randomized patients to once-weekly exenatide, pioglitazone or sitagliptin. Trial quality was good (see risk of bias table in Appendix 1 of the ERG report [see the "Availability of Companion Documents" field) but the question addressed is not relevant to UK practice. Firstly, the trial recruited patients on metformin monotherapy. As per the NICE guidelines, the ERG would expect the second drug to be sulphonylurea (or oral alternative such as pioglitazone).
Secondly, in routine care, the ERG would expect patients to be tried first on a (relatively) inexpensive oral drug before an expensive injectable.
A more useful RCT would have been in patients not achieving good control on triple oral therapy. For example, if HbA1c was still too high on metformin + sulphonylurea + a gliptin or pioglitazone, would substituting the gliptin or pioglitazone with exenatide achieve good control, given its somewhat greater glucose-lowering effect? The alternative would be insulin.
See Section 4 of the ERG report (see the "Availability of Companion Documents" field) for more information.
The CORE model structure was succinctly summarised within the manufacturer's submission. In brief, the model used patient characteristics such as age, gender, BMI, duration of diabetes, ethnicity, past history of complications (especially cardiovascular disease), and variables such as HbA1c, blood pressure, and blood lipids, to estimate long-term outcomes.
The CORE model permits users to adopt one of two modelling approaches:
- Treatment tree, or
- Treatment line
These trees and lines relate to the switching between therapies. In essence, the treatment tree assumes that after a period of time specified by the user patients fail on one therapy and move onto another. In contrast, the treatment line models an explicit stopping rule based upon an HbA1c cut-off specified by the user, at which point patients fail on one therapy and move onto another.
The manufacturer modelling adopted the treatment tree approach, with the base case in effect assuming that all patients remain on the first treatment for five years, regardless of baseline change in HbA1c. After five years all patients were assumed to switch to glargine.
The adoption of the treatment tree approach was in line with the modelling of Technology Appraisal 203 (see the NGC summary of the NICE technology appraisal Liraglutide for the treatment of type 2 diabetes mellitus). But in response to an ERG clarification question the manufacturer supplied scenario analyses adopting the treatment line approach for the base case analyses.
CORE Model Validation
The CORE model simulations for type 2 diabetes mellitus (T2DM) have been validated for 2nd order validation using epidemiological papers used to construct the CORE model and for 3rd order validation using epidemiological papers not used in the construction of the CORE model. The R2 for these were reportedly 0.975 and 0.875 respectively. The 3rd order validation for T2DM is summarised in Appendix 4 of the ERG report (see the "Availability of Companion Documents" field).
See Section 5 of the ERG report (see the "Availability of Companion Documents" field) for more information.