Dabigatran etexilate for the treatment and secondary prevention of deep vein thrombosis and/or pulmonary embolism

The company presented data from the RE‑COVER and RE‑COVER II trials on the effectiveness of dabigatran etexilate for treating deep vein thrombosis (DVT) and pulmonary embolism (PE). The RE‑COVER trials were multicentre, randomised, double‑blind, double‑dummy controlled studies which were identical in design. The trials were designed to test the non‑inferiority of 150 mg twice‑daily dabigatran etexilate compared with adjusted‑dose warfarin (with a target international normalised ratio [INR] of 2.0 to 3.0) in 5,153 patients with confirmed acute symptomatic venous thromboembolism (VTE), for whom the investigator considered at least 6 months of anticoagulant treatment to be appropriate. In the dabigatran etexilate arm, patients were given a parenteral anticoagulant for at least 5 days; in the warfarin arm, patients were also given parenteral anticoagulant and warfarin was added to achieve an INR of 2.0 to 3.0 patients were then given dabigatran etexilate or continued warfarin for 6 months with a 30‑day follow‑up after treatment had ended. The primary outcome was recurrent symptomatic VTE and VTE‑related deaths in 6 months. VTE was defined as the combined incidence of DVT (detected by venous compression ultrasonography or venography) and PE (detected by ventilation‑perfusion lung scan, pulmonary angiography or spiral [helical] CT). The average age of patients in the RE‑COVER trials was around 55 years. patients in the warfarin arm had an INR of 2.0 to 3.0 (were in the appropriate therapeutic range) for between 57% and 60% of the time across the trials.

The company presented data from 2 trials that assessed dabigatran etexilate for secondary prevention: RE‑MEDY and RE‑SONATE. RE‑MEDY was a randomised, double‑blind trial designed to test the non‑inferiority of 150 mg twice‑daily dabigatran etexilate compared with warfarin (target INR 2.0 to 3.0) in 2,866 patients with confirmed acute VTE which had been successfully treated with an anticoagulant for 3 to 12 months. Of these patients, 1,141 (40%) had previously participated in the RE‑COVER trials. patients in the RE‑MEDY trial had increased risk of recurrent VTE (as determined by a study investigator); according to the company, these patients represented a 'more severely affected patient group than other trial populations'. It was originally planned that patients in RE‑MEDY should have dabigatran etexilate or warfarin for 18 months. However, the protocol was amended to extend the treatment period because a lower‑than‑projected event rate was observed, and patients instead had 6 to 36 months of treatment with a 30‑day follow‑up after the end of the course of treatment. The primary outcome in RE‑MEDY was recurrent symptomatic and objectively confirmed VTE or death associated with VTE (excluding unexplained death). Clinically suspected DVT was objectively verified using pre‑specified imaging studies. The average age of patients in the trial was around 55 years. patients having warfarin had an INR of 2.0 to 3.0 around 62% of the time.

RE‑SONATE was a randomised controlled trial that compared 150 mg twice‑daily dabigatran etexilate with placebo in 1,353 patients who had completed 6 to 18 months of treatment with a vitamin K antagonist for confirmed acute symptomatic VTE. Of this population, 27 (2%) had previously participated in the RE‑COVER trials. RE‑SONATE only included patients for whom there was uncertainty about the need for continued anticoagulation treatment (termed 'clinical equipoise'), indicating that the risks and benefits of extended treatment were unclear. The original study protocol for RE‑SONATE stated that patients should be treated for 6 months, but because the number of VTE events necessary for statistical analysis was reached before some patients had completed 6 months of treatment, patients included in the analysis had dabigatran etexilate or placebo for between 3 and 6 months. All patients in the trial were also followed up for 12 months after treatment ended. The primary efficacy outcome in RE‑SONATE was recurrent symptomatic and objectively confirmed VTE or death associated with VTE (including unexplained death). The average age of patients in the trial was around 56 years.

In the analyses of primary and secondary outcomes in the RE‑COVER, RE‑MEDY and RE‑SONATE trials, the company used what it termed the 'full analysis set'. This comprised patients who were both randomised and had taken at least 1 dose of the study drug (5107 of the 5153 patients randomised in the RE‑COVER trials, 2856 of 2866 in RE‑MEDY and 1343 of 1353 in RE‑SONATE). The company tested for the non‑inferiority of dabigatran etexilate compared with warfarin using a non‑inferiority margin for the upper boundary of the 95% confidence interval (CI) around the hazard ratio (HR) of 2.75, and a margin for the upper boundary of the 95% CI around the difference in risk of 3.6 percentage points at month 6 in the RE‑COVER trials; in RE‑MEDY the corresponding margins were defined as 2.85 for the HR and 2.8% for the risk difference at month 18. In the pooled analysis of the RE‑COVER trials, 60 patients (2.4%) in the dabigatran arm and 55 (2.2%) in the warfarin arm had a recurrent VTE (HR 1.09, 95% CI 0.76 to 1.57, p<0.001 for non‑inferiority). In RE‑MEDY, 26 patients (1.8%) in the dabigatran arm and 18 (1.3%) in the warfarin arm had a recurrent VTE (HR 1.44, 95% CI 0.78 to 2.64, p=0.01 for non‑inferiority). In RE‑SONATE, 3 patients (0.4%) in the dabigatran arm and 37 (5.6%) in the placebo arm had a recurrent VTE (HR 0.08, 95% CI 0.02 to 0.25, p<0.001 for superiority).

In its safety analysis, the company used data from all randomised patients who received at least 1 dose of the study medication. The analysis was based on which drug they took; this meant that it may not have been the drug to which they were randomised. Numerous bleeding outcomes were measured in the 4 trials. Tables 1 to 3 summarise the bleeding events reported throughout the company's submission.

Across the RE‑COVER trials, 9 patients (0.4%) in the dabigatran arms and 5 (0.2%) in the warfarin arms had an acute coronary syndrome event. Of these, 8 patients (0.3%) in the dabigatran arms and 4 (0.2%) in the warfarin arms had a myocardial infarction. In RE‑SONATE, 1 patient in each of the dabigatran and placebo arms had a myocardial infarction. Two patients in the dabigatran arm had a transient ischaemic attack and 1 in the placebo arm had an ischaemic stroke. In RE‑MEDY, 13 patients (0.9%) in the dabigatran arm and 3 (0.2%) in the warfarin arm had an acute coronary syndrome event. Of these, 9 patients in the dabigatran arm and 1 in the warfarin arm had myocardial infarction, and 3 patients in the dabigatran arm and 1 in the warfarin arm had ischaemia or unstable angina.

The company did not identify any head‑to‑head trials comparing dabigatran etexilate with rivaroxaban for treating DVT and PE. The company carried out a meta‑analysis of data from 2 trials, EINSTEIN‑DVT and EINSTEIN‑PE, to estimate the effectiveness of rivaroxaban in treating DVT and PE. EINSTEIN‑DVT and EINSTEIN‑PE were open‑label trials that randomised patients to either rivaroxaban (15 mg twice daily for the first 3 weeks followed by 20 mg once daily) or low molecular weight heparin (LMWH; enoxaparin 1 mg/kg twice daily for at least 5 days) plus a vitamin K antagonist (warfarin or acenocoumarol) for the treatment of recurrent symptomatic DVT or PE (patients in EINSTEIN DVT had an index event of DVT; those in EINSTEIN PE had an index event of PE). Treatment length was 3, 6 or 12 months depending on risk of recurrent VTE, as judged by an investigator at the start of treatment. In the meta‑analysis of the 2 EINSTEIN trials, rivaroxaban was associated with a similar number of recurrent VTE events to LMWH followed by a vitamin K antagonist (HR 0.88, 95% CI 0.54 to 1.43), fewer major bleeds (HR 0.54, 95% CI 0.36 to 0.79), and a similar number of clinically relevant bleeds (HR 0.92, 95% CI 0.8 to 1.06). The company used these estimates of the efficacy and bleeding rates of rivaroxaban (relative to LMWH followed by a vitamin K antagonist) to perform a meta‑analysis of data from the RE‑COVER trials using an adjusted indirect comparison. The confidence intervals for recurrent VTE and major bleeding crossed 1. At the time of the appraisal the company stated that the numerical results of its indirect comparison are academic in confidence and therefore cannot be presented here. The company also presented information on the methods it used to do a network meta‑analysis, but did not present the results in its submission because the evidence network did not include a mixture of head‑to‑head trials and indirect evidence. It therefore did not add additional information to the adjusted indirect comparison.

For the prevention of recurrent VTE indication ('secondary prevention'), the company presented an adjusted indirect comparison of dabigatran etexilate compared with rivaroxaban using data from RE‑SONATE and EINSTEIN‑EXT, because both trials compared the active treatment with placebo. RE‑MEDY was not included because warfarin was the comparator. EINSTEIN‑EXT was a randomised, placebo‑controlled trial to assess the efficacy and safety of rivaroxaban 20 mg for the prevention of recurrent symptomatic DVT or PE in patients who had had 6 to 12 months of rivaroxaban or vitamin K antagonist treatment for an acute episode of VTE. In the trial, the rate of recurrent VTE was lower with rivaroxaban than with placebo (HR 0.18, 95% CI 0.09 to 0.39), but the rates of major or clinically relevant bleeding were higher with rivaroxaban than with placebo (HR 5.19, 95% CI 2.30 to 11.70). At the time of the appraisal the company stated that the numerical results of its adjusted indirect comparison of dabigatran etexilate and rivaroxaban were academic in confidence and therefore are not presented here.

Around 4% of patients in the RE‑COVER and RE‑MEDY trials had active cancer at baseline. In the RE‑COVER trials, 10 of the 173 patients (5.8%) in the dabigatran arms with active cancer and 12 of the 162 (7.4%) people in the warfarin arm with active cancer had VTE or VTE‑related death (HR 0.76, 95% CI 0.33 to 1.76). The company did not present the rate of recurrent VTE for patients with active cancer in RE‑MEDY. There were also no trial data for dabigatran etexilate compared with LMWH or rivaroxaban in patients with active cancer. The company presented a direct meta‑analysis of 5 trials comparing treatment of DVT and PE with LMWH or a vitamin K antagonist in people with active cancer. LMWH was associated with fewer recurrent VTE events than a vitamin K antagonist (relative risk [RR] 0.49, 95% CI 0.34 to 0.70) and a similar number of major bleeds (RR 1.05, 95% CI 0.53 to 2.10). The company also presented results from the Cancer DACUS extension study, which compared LMWH with a vitamin K antagonist for prevention of recurrent VTE. In this study LMWH was associated with a similar number of recurrent VTE events and major bleeds to a vitamin K antagonist (RR 0.63, 95% CI 0.34 to 1.18 for recurrent VTE; RR 2.58, 95% CI 0.51 to 13.06 for major bleeds).

The ERG commented on the characteristics of patients in the 4 dabigatran trials (RE‑COVER, RE‑COVER II, RE‑MEDY and RE‑SONATE). It noted that the mean baseline ages in the 4 dabigatran trials were between 53 and 56 years, but that clinical advisers would expect most patients with VTE to be over 65 years, and some to be over 80 years (and so having the 110 mg twice‑daily dose). The ERG commented that few, if any, patients in the dabigatran trials were aged over 80 years, and according to the trial protocols their VTE would be treated with the 150 mg twice‑daily dose. The ERG concluded that there were no clinical trial data for people aged over 80 years receiving 110 mg dabigatran etexilate twice daily for this indication.

The ERG considered that the 'clinical equipoise' population in RE‑SONATE was difficult to define in clinical practice, but that it was likely to comprise patients who were not having ongoing treatment for prevention of recurrent VTE. This population would be considered a different population to that currently treated for secondary prevention in the UK.

The ERG commented on the treatment regimen in the dabigatran trials and noted that there were limited data for patients treated continuously with dabigatran etexilate, starting from the acute phase of VTE through to long‑term secondary prevention. It noted that the company had assumed equal efficacy for the different parenteral therapies in the dabigatran and warfarin arms. The ERG noted that according to NICE's final scope, the analysis should consider both patients who need a limited period of anticoagulation (3 to 6 months) and those who need long‑term anticoagulation (usually life‑long). It noted that the mean duration of treatment in the RE‑COVER trials was 164 days, and that no data were available for patients needing only 3 months of anticoagulation. In the RE‑MEDY AND RE‑SONATE trials, assessing long‑term prevention of recurrent VTE, mean treatment duration was around 16 months and 6 months respectively. Because of this, the ERG noted that long‑term data for safety and efficacy were limited by the varying trial lengths.

The ERG commented that the company had used data from the full analysis set rather than the intention‑to‑treat population, but noted that the patient numbers in both populations were similar. The ERG also commented that the results for the primary outcome in the intention‑to‑treat population in the RE‑COVER and RE‑MEDY trials (provided by the company in response to the clarification questions) were similar to those in the full analysis set population. The ERG noted that less than 10% of patients were lost to follow‑up in the 4 dabigatran trials, and the numbers of discontinuations were well balanced between treatment groups.

The ERG considered the safety profile of dabigatran etexilate to be generally comparable with that of warfarin. The ERG noted that, for the RE‑MEDY trial, the company reported a slightly higher baseline prevalence of cardiovascular risk factors in the dabigatran arm than in the warfarin arm. However, it was unclear whether these small differences in baseline characteristics were linked to the overall incidence of acute coronary syndrome events in RE‑MEDY.

The ERG carried out its own network meta‑analyses for the purpose of comparing dabigatran etexilate with rivaroxaban for acute treatment and for secondary prevention based on the trials presented in the company's submission. For all outcomes, the ERG used data assessed at the end of treatment rather than at the end of the observed period in the trial, and where possible used intention‑to‑treat data in its analysis. The ERG preferred the fixed‑effects model over the random‑effects model, and the odds ratio was used as a summary statistic for all analyses. Over the acute treatment period, rivaroxaban was associated with a similar number of VTE events and major bleeds to dabigatran etexilate (odds ratio [OR] for VTE events 0.837, 95% CI 0.516 to 1.299; OR for major bleeds 0.763 95% CI 0.402 to 1.320). Rivaroxaban was associated with statistically significantly more clinically relevant non‑major bleeds than dabigatran etexilate (OR 1.647, 95% CI 1.234 to 2.114). Over the secondary prevention treatment period, there were no statistically significant differences in the odds of a VTE, major bleeding, or a clinically relevant non‑major bleed between rivaroxaban and dabigatran etexilate (recurrent VTE OR 1.744, 95% CI 0.510 to 4.388; major bleed OR 42, 95% CI 0.329 to 113; clinically relevant non‑major bleed OR 2.133, 95% CI 0.681 to 5.303).

The ERG commented on the subgroup of patients with active cancer in the clinical trials. It noted that the definition of active cancer used in the RE‑COVER and RE‑MEDY trials included some people who may have been in remission for up to 5 years and who would not usually be considered to have active cancer in UK clinical practice. The ERG also noted that in the RE‑COVER and RE‑MEDY trials, people in the control arm with active cancer had a parenteral anticoagulant followed by warfarin. This differs from standard clinical practice in England, whereby these patients would continue treatment with LMWH. The ERG noted that the number of patients with active cancer was small and the event rates low. The ERG carried out a network meta‑analysis for acute treatment in a population with active cancer, but there were no data for clinically relevant non‑major bleeds for any of the comparisons and no data for major bleeding events when comparing rivaroxaban with dabigatran etexilate. The meta‑analysis provided no statistically significant results. The ERG also stated that there were insufficient data to perform an analysis of secondary prevention of VTE in patients with active cancer. As such, the ERG considered that only limited conclusions can be made for the subgroup of patients with active cancer.

The company developed a Markov model with a 1‑month (30‑day) cycle length and a lifetime time horizon (60 years). A 3.5% discount rate was applied for costs and consequences, and the model was from an NHS and personal social services perspective. Patients entered the model after an initial VTE event and had treatment. If a patient had a recurrent VTE, they stopped current treatment and had 6 months' treatment with LMWH followed by warfarin (regardless of their initial treatment). They did not have continued anticoagulation for secondary prevention. Treatment was stopped completely if there was an intracranial haemorrhage, other major bleeding event, or for other reasons (such as end of planned treatment duration, worsening of other pre‑existing conditions or adverse events other than bleeding). In the 'off treatment' health state, patients were assumed to have no risk of bleeding but were at continued risk of recurrent VTE. patients in the model could have up to 2 recurrent VTE events. Those who had a PE (either as the initial, index event or a recurrent PE) could develop chronic thromboembolic pulmonary hypertension; the risk of this complication remained for 2 years after the PE. patients who had a DVT (either as the initial, index event or a recurrent DVT) could develop post‑thrombotic syndrome, the risk for which remained for 5 years after their DVT. The model only included patients with severe post‑thrombotic syndrome; the company stated that the mild form of the syndrome has little detrimental effect on quality of life. patients could die from any cause while in any of the health states. The model used age‑ and gender‑specific UK mortality rates from the Office for National Statistics (2010). The cardiovascular health states (myocardial infarction and unstable angina) and a health state for dyspepsia (indigestion) were only modelled in the sensitivity analyses.

In the model, the average age of the cohort at baseline was 55 years. On entry, 69% of patients had an initial DVT, and 31% had an initial PE (data from pooled RE‑COVER trials). The company modelled 2 base cases: acute treatment of VTE only, and treatment and long‑term prevention of recurrent VTE (secondary prevention). In the acute treatment base case, patients only had acute treatment for initial or recurrent DVT/PE, and did not have any further treatment if they had a recurrent VTE. In the treatment and secondary prevention base case, people had acute treatment and secondary prevention for initial DVT/PE and acute treatment only (with LMWH/warfarin) for recurrent DVT/PE.

The company modelled the risk of the combined outcome of recurrent VTE and VTE‑related death, and then split this risk by the proportion that was attributable to DVT, fatal PE and non‑fatal PE. Similarly, it modelled the risk of the composite outcome of major or clinically relevant non‑major bleeds, and split this total incidence by the proportion attributable to extracranial major bleeding events, intracranial major bleeding events, fatal major bleeding events and clinically relevant non‑major bleeding events.

The company used clinical data from the pairwise comparisons in the clinical trials, and did not use the adjusted indirect comparisons or data from a network meta‑analysis in its model. For the acute treatment phase, the company assumed that the baseline risk of each VTE and bleeding composite outcome was the incidence observed in the warfarin arm in the RE‑COVER trials. It multiplied these baseline risks by the hazard ratio from RE‑COVER to derive incidence of each outcome with dabigatran etexilate, or multiplied these baseline risks by the hazard ratio for warfarin compared with rivaroxaban from the pooled EINSTEIN DVT/PE trials to derive the incidence of each outcome with rivaroxaban. To derive the proportion of people having each type of recurrent VTE event or bleeding event, the company applied the proportions observed in the warfarin arm of the RE‑COVER trial, the dabigatran arm in the RE‑COVER trials, and the rivaroxaban arm of the pooled EINSTEIN DVT/PE trials. For the risks of these outcomes during secondary prevention, the company used a similar approach. However, because the rivaroxaban secondary prevention trial EINSTEIN EXT compared rivaroxaban with placebo, the company instead took the baseline risk to be that observed for placebo in RE‑SONATE. It then multiplied this baseline risk by the hazard ratios from RE‑SONATE and EINSTEIN‑EXT, to find the risk of these outcomes while taking dabigatran etexilate and rivaroxaban respectively. The company modelled the risk of these outcomes for dabigatran etexilate compared with warfarin based on the treatment effect and baseline risks observed in RE‑MEDY. As a consequence of this approach, the risks of recurrent VTE or bleeding with dabigatran etexilate during long‑term treatment for secondary prevention differed, depending on whether rivaroxaban or warfarin was the comparator. It also affected other assumptions in the model: for example, dabigatran etexilate treatment compared with warfarin for secondary prevention lasted 18 months (reflecting treatment length in RE‑MEDY), and dabigatran etexilate treatment compared with rivaroxaban for secondary prevention lasted 6 months (reflecting treatment length in RE‑SONATE). The length of acute treatment was 6 months in all comparisons. The company derived the risks of non‑fatal PE, proximal DVT, VTE‑related death and distal DVT after therapy discontinuation from a study by Pradoni et al.

During the acute treatment period, the risk of a recurrent VTE or bleeding was assumed to decrease over time. This was based on the results of the dabigatran trials, which showed that rates of recurrent VTE and bleeds were higher in the months immediately after the index event and gradually declined thereafter. The risk of recurrent VTE and bleeding events was assumed to be constant over the secondary prevention treatment period and in the post‑treatment period. Probabilities of events other than VTE or bleeding in the model were assumed to remain constant over time.

Chronic thromboembolic pulmonary hypertension, or CTEPH, is a complication of PE associated with considerable morbidity and mortality. The probability of CTEPH in the model was taken from a study by Pengo et al. Although post‑thrombotic syndrome was not reported in the trials, its incidence in the model was assumed to be the same for all comparators (based on data from a study by Pradoni et al.).

The company used baseline age‑ and gender‑specific utility values in its model, and adjusted for decreases in quality of life associated with treatment and each health state. The decrement in utility of -0.25 with a recurrent DVT or PE was assumed to be 6 weeks. For any type of bleeding event, a 1‑month decrement in utility was assumed (-0.13 for a major bleed and -0.04 for a minor bleed). An additional decrement in utility of -0.5 was applied for a patient's lifetime if they were disabled as a result of an intracranial haemorrhage. A reduced utility of -0.12 was assumed for 1 month for patients not having treatment and who had CTEPH. patients who had severe post‑thrombotic syndrome while off treatment were assumed to have a reduced utility of -0.07 for their remaining lifetime. The company assumed that taking warfarin would decrease quality of life and so applied a utility decrement of -0.012 while patients had warfarin. Similarly, the company assumed a utility decrement of -0.008 for people having LMWH.

The company assumed that patients having warfarin would have an initial anticoagulation clinic visit with a consultant at a cost of £62.56 (NHS reference costs 2012/13), 4 anticoagulation clinic visits during the warfarin titration period, and monthly follow‑up visits during treatment with warfarin. Subsequent visits to the anticoagulation clinic were assumed to cost £27.99 (based on NHS reference costs 2012/13), and were weighted to take into account consultant‑led and non‑consultant‑led appointments. Based on these assumptions, warfarin monitoring over the first year of treatment costs £482.41 (assuming 1 consultant‑led visit and 15 further follow‑up visits to the anticoagulation clinic). patients having dabigatran etexilate or rivaroxaban were assumed to visit an anticoagulation clinic just once.

The company presented base‑case results for the cost effectiveness of dabigatran etexilate compared with LMWH followed by warfarin, for both treating VTE only and for the treatment and prevention of recurrent VTE (secondary prevention). In the deterministic base case for treating VTE, dabigatran etexilate was associated with additional costs of £20 for 0.0239 more quality‑adjusted life years (QALYs) compared with warfarin. This equated to an incremental cost‑effectiveness ratio (ICER) of £862 per QALY gained. Dabigatran etexilate dominated rivaroxaban (that is, was both cheaper and more effective than rivaroxaban; dabigatran etexilate cost £20 less for 0.0003 more QALYs). In the treatment and secondary prevention base case, the company presented 2 pairwise comparisons: 1 for dabigatran etexilate compared with LMWH followed by warfarin, and 1 for dabigatran etexilate compared with rivaroxaban. The deterministic ICER for dabigatran compared with warfarin was £8,319 per QALY gained (incremental costs £458, incremental QALYs 0.0551). Rivaroxaban was dominated by dabigatran etexilate (dabigatran etexilate cost £67 less for 0.002 more QALYs). The company's probabilistic results were similar to the deterministic results.

The company carried out 9 scenario analyses:

• making adjustment for time in therapeutic range while taking warfarin

• allowing patients having rivaroxaban to have initial treatment with LMWH

• removing the utility decrement assumed for warfarin

• shortening the time horizon to 1 year or 6 months

• excluding bleeds while on LMWH before warfarin or dabigatran etexilate

• including the costs and utility decrement of dyspepsia (indigestion)

• including the costs and utility decrement of myocardial infarction and unstable angina

• including data for rivaroxaban from 6‑months' treatment only

• including unexplained deaths in the model.
  
  Of note, removing the utility decrement assumed for warfarin treatment increased the ICER in the treatment‑only base case from £862 per QALY gained to £1,217 per QALY gained. The same scenario increased the treatment and secondary prevention base case ICER from £8,319 per QALY gained to £14,947 per QALY gained. The overall effect of these scenarios was modest, and none increased the ICER for dabigatran etexilate compared with warfarin to over £20,000 per QALY gained.

The ERG had concerns about a number of the assumptions in the model relating to how patients would be treated in clinical practice. In the company's analyses, the length of acute treatment was 6 months. Although the ERG's clinical experts considered this to reasonably reflect clinical practice, the ERG acknowledged that, following publication of NICE's guideline on venous thromboembolic diseases, acute treatment of 3 months is increasingly common. The ERG noted that in the company's model, patients had treatment for prevention of recurrent VTE (secondary prevention) for 6 to 18 months (dependent on the comparator), but some patients in clinical practice in England may receive life‑long anticoagulation treatment. The ERG also noted that in the model, patients with a recurrent VTE did not have secondary prevention; rather, they had only acute treatment. Based on clinical advice, the ERG considered it more likely that a patient with multiple VTE events would receive life‑long secondary prevention unless precluded by the risk of bleeding. The ERG also disagreed that the only anticoagulation treatment option for a recurrent VTE would be LMWH followed by warfarin. It thought that some patients may have rivaroxaban and that patients with cancer would have LMWH. In the model, patients having a major bleeding event stopped treatment permanently. The ERG considered this to mean that patients would not re‑start anticoagulation treatment if they had a recurrent VTE, which may not reflect UK clinical practice. It noted that in NICE technology appraisal guidance on rivaroxaban for the treatment of DVT and prevention of recurrent DVT and PE and rivaroxaban for treating PE and preventing recurrent VTE, patients with a major bleeding event could restart anticoagulation treatment after 1 to 3 months.

The ERG stated that it was appropriate to exclude cardiovascular health states from the base case because they are not of direct relevance to the condition of interest for this appraisal. However, it believed that dyspepsia (indigestion) should have been included in the base case due to its being an important side effect associated with dabigatran etexilate treatment.

The ERG noted that the company analysed the data through a series of head‑to‑head and indirect comparisons. The limitation of this approach for the secondary prevention analyses was that the data for dabigatran etexilate varied in each comparison, resulting in a lack of comparability across scenarios. The ERG did not consider that treatment length would differ by intervention, and also noted that discontinuation on dabigatran etexilate for secondary prevention varied if warfarin or rivaroxaban was the comparator.

The ERG noted that the proportions of patients having each type of recurrent VTE event (DVT or PE), type of bleeding event (intracranial and extracranial major bleeding events and clinically relevant non‑major bleeds), and whether a major bleed was fatal differed by treatment arm. This was because the proportions of patients experiencing each outcome were taken directly from observations in each separate clinical trial. The ERG consulted with clinical experts as to whether a difference in these outcomes should be expected with different treatments, and was advised that there was no clinical basis for such a difference.

The ERG noted that in its model the company had used the same decrement in utility associated with DVT and PE as that derived from EQ‑5D data collected in the RE‑COVER trials (-0.25). It noted that this was unexpected, as PE is a more serious condition which might be expected to lead to a larger decrease in quality of life than DVT. It further noted that in NICE technology appraisal guidance on rivaroxaban for the treatment of DVT and prevention of recurrent DVT and PE and rivaroxaban for treating PE and preventing recurrent VTE, the utility values estimated for PE and DVT were 0.63 and 0.84 respectively. However, the ERG agreed that it was appropriate to use EQ‑5D data collected from the RE‑COVER trials in the model, rather than other published estimates of utility.

The ERG noted that the company had assumed a utility decrement for a patient taking warfarin based on data from a study of just 48 patients. It estimated that the utility value of -0.012 may not be robust or generalisable due to the small sample size from which it was derived. Furthermore, an estimate of the utility associated with dabigatran etexilate treatment was not provided by the company. The ERG noted that a head‑to‑head comparison of warfarin and dabigatran etexilate from RE‑LY (a trial of stroke prevention in people with atrial fibrillation) indicated that there was no long‑term difference in the EQ‑5D scores for patients having warfarin or dabigatran etexilate. However, it also noted that the utility decrement used by the company in the current submission was the same as that used in the company submissions for rivaroxaban in NICE technology appraisal guidance for the drug. The ERG concluded that although there may be a difference in utility associated with warfarin, the disutility is likely to be small and may reduce over time as patients adjust to their treatment regimen. It acknowledged, however, that there is a great degree of uncertainty around estimates of disutility while on warfarin.

The ERG noted that for patients with an initial VTE who were treated with dabigatran etexilate, rivaroxaban or LMWH monotherapy, the company had not included an initial appointment where people are prescribed treatment and receive any training on how to take their medication. In contrast, all patients having warfarin were assumed to have an initial appointment at which treatment would be initiated, even those who had been admitted to hospital for their first VTE. The ERG therefore considered that the company had both overestimated the number of appointments for admitted patients who had warfarin (because they will not need an initial anticoagulation visit) and underestimated the number of appointments needed for dabigatran etexilate, rivaroxaban and LMWH monotherapy (because patients diagnosed with VTE outside of hospital will need an initial appointment to discuss anticoagulation treatment). Consequently, the ERG considered that the cost of administering dabigatran etexilate compared with warfarin may be underestimated.

The ERG noted the company's assumption that after an initial titration period in the first month, people taking warfarin would visit an anticoagulation clinic once a month for the remainder of treatment. The ERG heard from clinical experts that although this frequency of monitoring visits may be appropriate for initial acute treatment, visits would be less frequent for people continuing to take warfarin long‑term for secondary prevention (typically once every 3 months). The ERG's clinical experts considered the monitoring schedule for dabigatran etexilate and rivaroxaban to be reasonable, but noted that patients receiving either drug may return to their GP each year. The ERG noted that the company's estimated cost for follow‑up visits to an anticoagulation clinic (£27.99) assumed that some visits would be consultant‑led. The clinical experts advised that in clinical practice, it is typical for nurses to provide the majority of contact at follow‑up visits. The ERG assumed that such visits would instead cost £10.61.

The ERG commented that in the company's acute treatment deterministic base‑case analysis, costs and QALYs varied between all 3 treatment strategies (by £40 and 0.02 per patient respectively). In particular, for dabigatran etexilate compared with rivaroxaban, the estimated total QALYs differed by 0.0003. From this, the ERG reasoned that although dabigatran etexilate dominated rivaroxaban in the company's acute treatment base case, the company's estimates of average total costs and QALYs imply that all 3 treatments strategies would result in similar costs and consequences.

The company did not present a fully incremental analysis for the treatment and secondary prevention base case. The ERG stated that the main disadvantage of using different model parameters for dabigatran etexilate when comparing it with warfarin or rivaroxaban was that the results of each analysis cannot be compared: the costs and QALYs associated with dabigatran etexilate will differ in each case. Although dabigatran etexilate dominated rivaroxaban in the treatment and secondary prevention base case, the ERG noted that there was a QALY difference of less than 0.002 per patient and a cost difference of just £67. These results implied that the treatments are similar.

The ERG identified errors in the company's model. When it presented the company's base‑case analysis with these errors corrected, there was only a modest difference in the resulting ICERs for dabigatran etexilate compared with warfarin. In the acute treatment base case, the corrected errors resulted in an ICER of £831 per QALY gained for dabigatran etexilate compared with warfarin, and dabigatran etexilate was less costly and less effective than rivaroxaban (dabigatran etexilate cost £22 less for 0.0003 fewer QALYs than rivaroxaban). In the treatment and secondary prevention base case, the ICER for dabigatran etexilate compared with warfarin was £9973 per QALY gained and dabigatran etexilate was less costly and less effective than rivaroxaban (dabigatran etexilate cost £23 less for 0.0013 fewer QALYs than rivaroxaban).

To derive exploratory base cases, the ERG combined 16 of the 42 scenarios it had tested (with the exception of 3 for the acute treatment exploratory base case, because they were only relevant to the secondary prevention treatment phase). The scenarios were:

• assuming that 50% of patients with a major bleeding event would return to treatment at a later date

• a 50‑year rather than 60‑year time horizon

• use of the ERG network meta‑analysis for probability of VTE events and probability of bleeding

• assuming a constant proportion of VTE event types across treatments

• assuming a constant proportion of bleeding event types across treatments

• a death rate from intracranial haemorrhage of 39.5%

• an average age of 65 years across the model cohort

• patients had a utility decrement of 0.265 if they had a PE

• the disutility associated with DVT and PE was applied for 30 days

• the initial anticoagulation appointment for warfarin was removed

• monitoring visits assumed to cost £10.61

• cost of intracranial haemorrhage set at £14,777.
  
  In addition, in the treatment and secondary prevention base case the following scenarios were included:

• assumed life‑long treatment for secondary prevention

• monitoring appointments in the secondary prevention period reduced from once a month to once every 3 months

• dabigatran etexilate discontinuation rate assumed to be the same as in RE‑MEDY.

In the ERG's exploratory base case for acute treatment, dabigatran etexilate was associated with an additional cost of £223 and 0.012 more QALYs compared with warfarin (ICER of £18,240 per QALY gained). Dabigatran etexilate dominated rivaroxaban, costing £3 less per patient for an additional 0.0018 QALYs. In the exploratory base case for treatment and secondary prevention, dabigatran etexilate was associated with an additional cost of £3331 per patient and 0.093 more QALYs compared with warfarin (ICER of £35,768 per QALY gained). Rivaroxaban was extendedly dominated by dabigatran etexilate (compared with warfarin, rivaroxaban had additional costs of £2710 and QALYs of 0.020; compared with rivaroxaban, dabigatran etexilate had incremental costs of £620 and QALYs of 0.073).

The ERG commented that for its exploratory acute treatment base‑case analysis, the main factor increasing the ICER for dabigatran etexilate compared with warfarin was the lower costs associated with warfarin monitoring as a result of assuming that follow‑up visits would all be nurse‑led. The cost of a warfarin monitoring appointment was also an important factor in the exploratory base case for treatment and secondary prevention, as was the assumed frequency of warfarin monitoring in the secondary prevention period (once every 3 months rather than once monthly). The ERG commented that a company scenario analysis in which there was no utility decrement associated with warfarin showed that warfarin disutility was a major factor in the company's model. The ERG noted that if, in addition to the assumptions in its exploratory base case, it also assumed no disutility with warfarin in the secondary prevention period, the ICER for dabigatran etexilate compared with warfarin would be £90,000 per QALY gained. This would have such a large effect on the ICER because the ERG's model incorporating its preferred scenarios is increasingly sensitive to changes in QALYs.