Semaglutide for reducing the risk of major adverse cardiovascular events in people with cardiovascular disease and overweight or obesity

The company presented evidence for semaglutide in this indication from SELECT, a multinational randomised double-blind placebo-controlled phase 3 trial. It enrolled 17,604 people with a BMI of at least 27 kg/m² and established CVD. Established CVD was defined as having had an MI, or ischaemic or haemorrhagic stroke, or having symptomatic PAD. Symptomatic PAD was defined as having 1 of the following:

• intermittent claudication with an ankle-brachial index (the ratio of systolic blood pressure in the ankle to the arm) of less than 0.85 at rest

• a peripheral arterial revascularisation procedure, or

• an amputation because of atherosclerotic disease.
  
  The trial compared subcutaneous once-weekly semaglutide plus standard care (n=8,803) with placebo plus standard care (n=8,801). Standard care in SELECT included healthy lifestyle counselling and medicines to lower cardiovascular risk. Trial investigators were encouraged to optimise medicines according to treatment guidelines or local clinical practice. Treatments used by people in the trial included lipid-lowering medicines (90.1%), antiplatelet drugs (86.2%), beta-blockers (70.2%), angiotensin-converting-enzyme inhibitors (45.0%) and angiotensin‑2 receptor antagonists (29.5%). The primary endpoint was the time from randomisation to first occurrence of a major adverse cardiovascular event (MACE) after randomisation. This was a composite endpoint consisting of cardiovascular death, non-fatal MI or non-fatal stroke. There was a statistically significant reduction in the risk of first MACE with semaglutide compared with placebo (hazard ratio [HR] 0.80; 95% confidence interval [CI] 0.72 to 0.90). The direction of effect was consistent across other secondary endpoints, including individual components of MACE:

• non-fatal MI (HR 0.72, 95% CI 0.61 to 0.85)

• non-fatal stroke (HR 0.93, 95% CI 0.74 to 1.15)

• coronary revascularisation (HR 0.77, 95% CI 0.68 to 0.87)

• hospitalisation for unstable angina (HR 0.87, 95% CI 0.67 to 1.13)

• heart failure hospitalisation or urgent heart failure visit (HR 0.79, 95% CI 0.60 to 1.03)

• cardiovascular death (HR 0.85, 95% CI 0.71 to 1.01)

• all-cause death (HR 0.81; 95% CI: 0.77 to 0.93).
  
  The company also provided supportive evidence from the STEP‑HFpEF and STEP‑HFpEF DM trials. These trials compared semaglutide with placebo in people with heart failure and a preserved ejection fraction, and a BMI of at least 30 kg/m², with and without type 2 diabetes. But because these trials did not report outcomes directly relevant to this evaluation (that is, a reduction in MACE), the focus of the evaluation was the SELECT trial only. The lead team noted that the population in SELECT was aligned with the final scope and decision problem addressed by the company. It concluded that SELECT was appropriate for informing the lead team's recommendation.

SELECT excluded people with diabetes, but they are not excluded from the marketing authorisation for this indication. The company provided additional evidence to support using semaglutide in people with type 2 diabetes. This included evidence from 4 cardiovascular-outcomes trials, which showed that semaglutide reduced the risk of MACE in people with type 2 diabetes compared with placebo. The EAG noted that the cardiovascular-outcomes trials used different doses or routes of administration for semaglutide, and there were no specific inclusion criteria for previous cardiovascular events or BMI. But it was noted that a substantial proportion of people across the trials were relevant to this indication (that is, people who had had a previous cardiovascular event, and people with overweight or obesity). The lead team noted that there were some limitations with the evidence presented by the company. But it also noted that the evidence showed a similar effect size for people with type 2 diabetes as that seen in SELECT. The lead team concluded that the clinical effectiveness of semaglutide seen in SELECT was likely to be generalisable to people with diabetes. So, it also concluded that this population should not be excluded from the recommendation.

SELECT excluded people who had had any of the following cardiovascular events within 60 days before the day of screening:

• an MI

• a stroke

• a transient ischaemic attack

• hospitalisation for unstable angina.
  
  But these groups are not excluded from the marketing authorisation for this indication. The company explained that excluding people with a recent cardiovascular event is common in cardiovascular-outcomes trials such as SELECT. The company added that including people with a recent cardiovascular event may confound and complicate the determination of primary and secondary endpoints in trials. This is because these groups may be at a higher risk of complications that are not preventable or modifiable by the study drug. The EAG noted that the clinical effectiveness of semaglutide in reducing the risk of MACE was seen shortly after starting treatment, before any substantial effect on weight loss was reported. This shows that there is a mechanism of action independent of weight loss. The lead team noted that the risk of a cardiovascular event is highest shortly after a previous event. So, it thought that it was plausible that semaglutide could be more clinically and cost effective if used earlier. It concluded that the recommendation should not be restricted based on the time since a cardiovascular event.

The company developed a state-transition model. People entered the model in the 'established CVD' health state (see section 3.1). The risk of a further cardiovascular event (first modelled event) was estimated using parametric survival (time-to-event) extrapolations fitted to data from SELECT. First modelled events included MI, stroke, hospitalisation for heart failure, hospitalisation for unstable angina and coronary revascularisation. People who survived the first modelled event entered a 'post-event' health state, in which they were at risk of further events. People in the post-event state who had the same event (for example, a second MI) returned to the same post-event health state. People who had a different event (for example, a stroke after an MI) entered a post-event health state for people with more than 1 modelled event. People had an ongoing risk of further events. The model also tracked the development of chronic kidney disease and type 2 diabetes separately. Cardiovascular-related or non-cardiovascular-related death could occur from any health state. The types of events that were modelled were based on the endpoints in SELECT. The EAG noted that the company had included some events in which the difference between treatment arms did not show statistical significance (see section 3.3). The EAG preferred to exclude these events from its base case. The lead team noted that the benefit of semaglutide in reducing the risk of these events was less certain. It also noted that excluding these events resulted in a moderate increase in the incremental cost-effectiveness ratio (ICER). But it was reassured that results across all endpoints were consistent, and most were statistically significant. The lead team concluded that these events should be modelled in this evaluation.

The company used data from SELECT to determine how long people were on treatment with semaglutide. This was then used to calculate the cost for semaglutide in the model. The company used data for median time to stopping treatment, which was defined as the first time being off treatment for over 35 days. This was then used to determine a constant cycle (4‑week) stopping rate. The exact stopping rate is considered confidential by the company, so cannot be reported here. The EAG noted that it was unclear how well the company's constant stopping rate reflected the stopping of semaglutide over the whole period of SELECT and for the remainder of model time horizon. The lead team noted that it would have been more appropriate to fit a range of parametric extrapolations to the data on stopping semaglutide from SELECT to provide a more accurate extrapolation of treatment stopping. It noted that, without this analysis, the company's estimates of treatment stopping, and so the costs associated with semaglutide, were uncertain. But, given the low ICERs (see section 3.10), this was not likely to affect decision making.

The company modelled the increased risk of cardiovascular events for people who had developed diabetes using hazard ratios based on de Jong et al. (2020). The company explained that it only applied these hazard ratios after the SELECT follow-up period to avoid double counting cardiovascular events observed in the trial. The company also assumed that people who stopped semaglutide had no further benefit and had the same risk of cardiovascular events as people having standard care. The EAG noted that the company's extrapolations already reflected cardiovascular outcomes for a population in which a proportion had developed type 2 diabetes or stopped semaglutide by the end of follow up. So, it preferred not to apply any additional adjustments to cardiovascular-event rates from the end of the trial follow-up period.

The lead team noted that the EAG's approach of not adjusting cardiovascular-event rates after the end of the trial period for people who have stopped semaglutide resulted in a substantial decrease in the ICER. The lead team discussed that the stopping rate for semaglutide may not be constant over time, as assumed by the company (see section 3.7). It added that, without long-term data, it could not be assumed that the stopping rate would not increase. This may have resulted in a smaller reduction in the risk of events for people having semaglutide compared with placebo, after the trial period. The lead team added that there was further uncertainty because the risks of developing type 2 diabetes, having a cardiovascular event and stopping semaglutide are not independent. Trial evidence on the relationship between stopping treatment, cardiovascular events and developing type 2 diabetes would have been helpful to understand whether the adjustment was needed. The lead team concluded that, without further evidence, it was not clear whether a separate adjustment should have been applied for people who developed type 2 diabetes or stopped semaglutide after the observed trial period. The lead team considered both the company's and EAG's approaches in decision making, and noted that the ICERs remained low regardless of which approach was used (see section 3.10).