3.1 The key clinical evidence in the manufacturer's submission came from 2 trials (EINSTEIN-DVT and EINSTEIN-Ext). EINSTEIN-DVT was an open-label non-inferiority study that compared rivaroxaban (15 mg twice daily for 3 weeks, then 20 mg once daily for 3, 6 or 12 months) with enoxaparin followed by a vitamin K antagonist (either warfarin or acenocoumarol) for treating patients with acute symptomatic deep vein thrombosis without any symptoms of pulmonary embolism, and for preventing recurrent deep vein thrombosis and pulmonary embolism. Enoxaparin was given until a vitamin K antagonist had brought the international normalised ratio (INR) into the target range, and was then stopped. Based on individual patient risk factors, patients were either assigned to 3, 6 or 12 months of treatment as determined by the treating physician. EINSTEIN-Ext was a randomised placebo-controlled superiority trial that compared rivaroxaban (20 mg once daily; n=602) with placebo once daily (n=594) in patients with confirmed symptomatic deep vein thrombosis or pulmonary embolism that had been treated for 6 or 12 months with a vitamin K antagonist (warfarin or acenocoumarol) or rivaroxaban up to the moment of randomisation. Patients were recruited if the risks and benefits of further anticoagulation were finely balanced, that is, there was 'clinical equipoise' for the decision to continue anticoagulation.
3.2 The manufacturer's submission noted that about 60% of patients recruited into EINSTEIN-Ext were assigned to 6 months of treatment, about 40% were assigned to 12 months of treatment and 28% had previously used rivaroxaban. The manufacturer also noted that some people were excluded from the EINSTEIN-DVT and EINSTEIN-Ext trials, such as those with a creatinine clearance of less than 30 ml/min, clinically significant liver disease, high blood pressure (systolic more than 180 mmHg or diastolic more than 110 mmHg), active bleeding or at high risk of bleeding.
3.3 The primary efficacy endpoint was a composite of deep vein thrombosis or pulmonary embolism (symptomatic, recurrent venous thromboembolism). Pulmonary embolism included both fatal and non-fatal pulmonary embolism. The primary safety endpoint was a composite of major bleeding and other clinically relevant non-major bleeding ('clinically relevant bleeding') for EINSTEIN-DVT and major bleeding for EINSTEIN-Ext. A range of secondary composite endpoints were also included.
3.4 In EINSTEIN-DVT, the primary efficacy endpoint of symptomatic recurrent venous thromboembolism occurred in 2.1% (n=36) of patients in the rivaroxaban group compared with 3.0% (n=51) in the enoxaparin and vitamin K antagonist group (hazard ratio [HR] 0.68; 95% confidence interval [CI] 0.44 to 1.04, p<0.001 for non-inferiority and p=0.076 for superiority). The overall HR for rivaroxaban was 0.97 (95% CI 0.76 to 1.22, p=0.77) for the primary safety endpoint of clinically relevant bleeding and 0.67 (95% CI 0.44 to 1.02, p=0.06) for death from all causes. Recurrent deep vein thrombosis occurred less frequently in patients treated with rivaroxaban than with enoxaparin and a vitamin K antagonist (14 compared with 28). Pulmonary embolisms (fatal and non-fatal) did not differ between treatment groups.
3.5 The manufacturer reported a time in therapeutic range for the comparator enoxaparin and a vitamin K antagonist of 57.7% across all centres and 59.7% in western European centres. The manufacturer highlighted that guidelines from the National Patient Safety Agency and the Scottish Executive Health Department recommend a time in therapeutic range of at least 60%. It also noted there was no statistical interaction observed in EINSTEIN-DVT between time in therapeutic range and treatment effect.
3.6 In EINSTEIN-Ext, patients taking rivaroxaban experienced fewer recurrences of venous thromboembolism (1.3%, n=8) than patients taking placebo (7.1%, n=42) (HR 0.18, 95% CI 0.09 to 0.39, p<0.0001). The numbers of clinically relevant non-major bleeding events were significantly higher in the rivaroxaban arm than in the placebo arm (32 patients [5.3%] compared with 7 patients [1.2%], p<0.001). There were more major bleeding events in patients taking rivaroxaban (4 patients compared with no patients), although this did not reach statistical significance.
3.7 The manufacturer reported a mixed treatment comparison for the subgroup of patients with cancer. This compared the relative effectiveness of rivaroxaban with dual low molecular weight heparin (LMWH) and a vitamin K antagonist, long-term LMWH compared with LMWH and a vitamin K antagonist, and rivaroxaban compared with long-term LMWH. The manufacturer provided 3 analyses. The primary analysis used data from a systematic review of long-term anticoagulation in patients with cancer reported by Akl et al. (2011) and from the whole EINSTEIN-DVT trial population. Secondary analysis 1 used data from a trial by Lee et al. (2003) evaluating the LMWH dalteparin for the prevention of recurrent venous thromboembolism in patients with cancer and the data from the whole EINSTEIN-DVT trial population. Similarly, secondary analysis 2 used data from Akl et al. (2011) and effectiveness data from the cancer subgroup of EINSTEIN-DVT.
3.8 Results from the primary analysis indicated that for patients with active cancer, the venous thromboembolism recurrence hazard ratio for rivaroxaban compared with long-term LMWH was 1.44 (95% credible intervals 0.07 to 31.4). Secondary analysis 2 showed that rivaroxaban was less effective than LMWH at preventing venous thromboembolism recurrence (HR 1.32, 95% credible intervals 0.06 to 32.3) but induced fewer major bleeding events (odds ratio 0.24, 95% credible intervals 0.00 to 9.44). However, the manufacturer noted that the mixed treatment comparison had credible intervals with wide margins for the efficacy and safety of rivaroxaban compared with long-term LMWH.
3.9 The manufacturer reported adverse events from EINSTEIN-DVT and EINSTEIN-Ext that were experienced in at least 4% of any treatment group. The most common adverse events across both EINSTEIN trials were headache, pain in extremity, nasopharyngitis and nosebleed. The reported incidences of post-thrombotic syndrome and chronic thromboembolic pulmonary hypertension were low in both arms of EINSTEIN-DVT and EINSTEIN-Ext.
3.10 The manufacturer's submission used a Markov-based model for the economic evaluation of rivaroxaban within its licensed indication for treating deep vein thrombosis and preventing recurrent thromboembolic events. Two analyses were presented: a primary analysis comparing rivaroxaban with LMWH and a vitamin K antagonist over 3, 6 and 12 months, and a cost-minimisation analysis for patients with active cancer, which used dalteparin (a LMWH) as the comparator. The manufacturer also presented a further exploratory cost-effectiveness analysis for patients with active cancer at the request of the ERG.
3.11 The Markov model comprised 11 health and treatment states and patients entered the model after a diagnosis of deep vein thrombosis. The model relied on the control arm of EINSTEIN-DVT to derive the probabilities of recurrent venous thromboembolism, bleeding and discontinuation rates. The probabilities in the rivaroxaban arm were calculated by applying the appropriate hazard ratio or risk ratio to the probability in the control arm. The 3-month discontinuation rate was assumed to be 1.9%. Probabilities for long-term complications and risk of mortality were taken from both EINSTEIN-DVT and literature reviews. Drug and resource costs were derived from relevant UK sources ('British national formulary' [BNF], NHS Reference Costs 2009–10 and Personal Social Services Research Unit [PSSRU] 2010) and generally reflected UK clinical practice. The model did not include monitoring for patients treated with rivaroxaban or LMWH. It assumed 9 visits in the first 3 months, followed by 5 visits thereafter (every 3 months) for patients treated with a vitamin K antagonist. It also assumed that 66% of visits for INR monitoring would take place in primary care and 34% in secondary care. For primary care, the manufacturer assumed INR monitoring would be delivered equally by a GP and a nurse (50/50 split). The estimated annual cost of INR monitoring, including transport costs, was £656 in the first year and £540 thereafter.
3.12 A validated preference-based measure of quality of life was not used in the EINSTEIN-DVT trial, so the economic model submitted by the manufacturer used utility values sourced from literature reviews. The manufacturer assigned a baseline utility value of 0.825 to all patients with deep vein thrombosis entering the model, which was taken from a survey of the UK general population using a visual analogue scale rating (Kind et al. 1998) and adjusted with disutility values for deep vein thrombosis, pulmonary embolism, extracranial bleed, intracranial bleed and post-thrombotic syndrome.
3.13 The base-case results included all the drug acquisition costs, resources associated with monitoring, and costs associated with adverse events (that is, bleeding events) and were presented by intended treatment durations (3, 6 and 12 months). Treatment with rivaroxaban dominated treatment with LMWH and a vitamin K antagonist across all treatment durations, that is, rivaroxaban was less costly and more effective compared with LMWH and a vitamin K antagonist (0.02 incremental QALYs for all treatment durations and cost savings of £163 at 3 months, £124 at 6 months and £33 at 12 months).
3.14 The manufacturer undertook a series of univariate and multivariate deterministic sensitivity analyses to test the robustness of the results by varying most of the parameters used in the economic evaluation. The results were generally sensitive to the cost of monitoring and the hazard ratio for venous thromboembolism. The manufacturer also provided probabilistic sensitivity analyses. These showed that there was a 94.2–98.9% probability of rivaroxaban being cost effective at £20,000 per QALY gained for all treatment durations. The treatment duration of 3 months produced the most cost savings and increased incremental QALYs. The probability of rivaroxaban being the dominant treatment option was 97.1% in patients having 3 months of anticoagulation, 83.9% in those having 6 months and 53.0% in those having 12 months.
3.15 The manufacturer presented a cost minimisation analysis evaluating rivaroxaban in patients with cancer. Patients with cancer were assumed to be treated for 6 months. The cost of rivaroxaban was £4.20 per day for the first 21 days (2 tablets daily), followed by £2.10 per day (1 tablet daily). The cost of dalteparin was £8.47 per day for the first month and £7.06 per day for subsequent months. The total cost saving associated with rivaroxaban compared with LMWH (dalteparin) was £903 for patients with cancer.
3.16 The manufacturer also presented an exploratory cost-effectiveness analysis of the subgroup of patients with cancer. Using treatment effects from the mixed treatment comparison, assuming no INR monitoring cost and using a 6-month treatment duration, rivaroxaban dominated dalteparin (0.0013 incremental QALYs and cost savings of £1085).
3.17 The ERG raised concerns about the applicability of the EINSTEIN trials to UK clinical practice, including that the trials did not fully reflect the UK population with deep vein thrombosis because a number of important patient groups were excluded from both trials. These included patients with high risk of bleeding, creatinine clearance less than 30 ml/min (but not less than 15 ml/min), clinically significant liver disease, high blood pressure (systolic more than 180 mmHg or diastolic more than 110 mmHg) and non-proximal deep vein thrombosis. Specifically, the ERG noted that there are no data to inform decisions about patients with increased risk of bleeding. The ERG also noted that the EINSTEIN trials did not include patients for whom vitamin K antagonists are not appropriate, other than patients with cancer. It noted the population recruited into the EINSTEIN trials excluded a number of important groups relevant to the decision problem.
3.18 The ERG and its clinical advisers considered the comparator (enoxaparin) used by the manufacturer to be appropriate, although the dosage used in the EINSTEIN trials (1 mg/kg twice daily) was not in line with UK clinical practice (1.5 mg/kg once daily). Using the twice-daily dosage may have been unfavourable to rivaroxaban.
3.19 The manufacturer assumed the maximum treatment duration was 12 months for idiopathic deep vein thrombosis or in the presence of permanent risk factors. However, the clinical advisers to the ERG questioned this assumption and stated that it is now common for treatment to extend beyond 12 months, depending on patient characteristics and risk factors. The ERG's clinical advisers estimated that around 20% of people with deep vein thrombosis would have long-term treatment because of an ongoing risk of recurrence of venous thromboembolism.
3.20 The ERG raised concerns about the robustness of the mixed treatment comparison in the cancer subgroup and the way the evidence was synthesised. The ERG noted that the included trials varied in the length of follow-up, and choice and dosage of LMWH also varied across studies. The ERG concluded that the mixed treatment comparison did not provide good estimates of the uncertainty associated with the true treatment effect, but found the point estimate to be reasonable.
3.21 The ERG presented exploratory analyses that corrected certain errors in the model and took into account a less intensive INR monitoring strategy comprising 6 INR monitoring visits in the first 3 months and 3 visits every 3 months thereafter. The results indicated that enoxaparin and a vitamin K antagonist were dominated by rivaroxaban for the 3-month duration group (0.02 incremental QALYs and a cost saving of £51). Compared with enoxaparin and a vitamin K antagonist, the incremental cost-effectiveness ratio (ICER) of rivaroxaban was £3247 per QALY gained for the 6-month treatment duration and £14,902 per QALY gained for the 12-month treatment duration.
3.22 The ERG revised the manufacturer's exploratory analysis in cancer patients to take into account what it considered to be a more plausible – and smaller – distribution of between-study standard deviations (as opposed to the alternative distributions used by the manufacturer). This found rivaroxaban to be less effective than LMWH at preventing venous thromboembolism recurrence. The ERG raised concerns with the limited evidence available in the cancer subgroup, and with the modelling assumptions in the exploratory analysis. The ERG concluded any reliance on the results of the mixed treatment comparison may lead to inaccurate estimates of mean ICERs because they are based on inflated expected values.
3.23 After consultation, the manufacturer submitted additional analyses on the cost effectiveness of rivaroxaban in people in whom long-term anticoagulation is intended; that is, people who need anticoagulation for longer than 12 months. The manufacturer also commented further on the characteristics of patients in the 3, 6, and 12 month treatment duration groups in the EINSTEIN-DVT trial.
3.24 The manufacturer's new economic model for long-term use of rivaroxaban used event rates for venous thromboembolism recurrence and bleeding from the 12-month duration group of EINSTEIN-DVT for people treated with a LMWH and a vitamin K antagonist. The long-term risk of venous thromboembolism recurrence (after 1 year) was taken from a meta-analysis review, and results from the whole trial population of EINSTEIN-DVT were used to estimate the treatment effects. The manufacturer presented 2 scenarios. One took into account the manufacturer's assumed INR frequency of 9 visits in the first 3 months followed by 5 visits every 3 months thereafter (first-year costs £656). The other adopted a less intensive INR monitoring programme of 6 visits in the first 3 months followed by 3 visits every 3 months thereafter (first-year cost of £413). The model assumed a discontinuation rate of 3.6% every 3 months based on a review of long-term statin therapy because evidence on adherence to rivaroxaban for longer than 12 months of treatment was not available. The model also included a sensitivity analysis in which the 3-month discontinuation rate was varied from 1.9% to 6.9%.
3.25 The model applied a disutility of 0.012 to warfarin, which was sourced from a study by Marchetti et al. (2001) involving a small group of patients (n=48) attending an anticoagulation clinic. The manufacturer noted that a disutility would not apply to rivaroxaban, citing reasons that included raised levels of treatment satisfaction in comparison with LMWH and a vitamin K antagonist, and that no clinically important adverse events were associated with rivaroxaban that had not already been taken into account in the model. The manufacturer also referred to 'Dabigatran etexilate for the prevention of stroke and systemic embolism in atrial fibrillation' (NICE technology appraisal 249), in which a disutility was applied to dabigatran to account for dyspepsia, an adverse event not associated with rivaroxaban.
3.26 The results from the long-term anticoagulation model showed that the ICER for rivaroxaban compared with LMWH and a vitamin K antagonist was £6037 per QALY gained under the manufacturer's assumed INR monitoring cost of £656 (0.16 incremental QALYs and additional cost of £953). Assuming a lower INR monitoring cost of £413, an ICER of £15,847 per QALY gained (0.16 incremental QALYs and additional cost of £2502) was reported by the manufacturer. The probabilistic sensitivity analysis based on reduced INR monitoring showed that there was a 58% probability that rivaroxaban was cost effective at £20,000 per QALY gained and a 25% probability that rivaroxaban was dominant (more effective and less costly).
3.27 The ERG was generally satisfied with the assumptions made in the manufacturer's long-term economic model but noted that it is uncertain whether the treatment effects assumed in the model would remain fixed over a lifetime. The ERG explored several scenarios based on variations to the manufacturer's long-term anticoagulation model:
Assuming treatment effect from the whole trial population over a lifetime horizon.
Assuming a lower INR cost of £320 for the first year followed by £248 annually thereafter, based on a reduced frequency of visits (6 visits in the first 3 months and then 3 visits every 3 months thereafter with a different GP/nurse consultation ratio than the one used by the manufacturer).
Varying the 3-month discontinuation rate for rivaroxaban from 3.6% as assumed in the manufacturer's long-term model to 1.9% as assumed by the manufacturer in the original submission.
Applying a 0.012 decrement in utility for warfarin and no decrement for rivaroxaban; 0.012 decrement in utility for warfarin and 0.006 decrement in utility for rivaroxaban; and assuming no decrement in utility for both warfarin and rivaroxaban.
Taking into account the above assumptions, the results from the ERG's exploratory analyses yielded ICERs ranging from £19,381 to £38,837 per QALY gained. A deterministic calculation based on a whole-trial treatment effect, the lower cost of INR monitoring, and a utility decrement for warfarin only, indicated an ICER of £19,381 per QALY gained, assuming the same 3-month discontinuation rate of 3.6% for warfarin and rivaroxaban. The equivalent ICER when the discontinuation rate for rivaroxaban was lowered to 1.9%, while keeping the warfarin discontinuation rate at 3.6%, was £25,076 per QALY gained.
3.28 Full details of all the evidence are in the manufacturer's submission and the ERG report, which are available from www.nice.org.uk/guidance/TA261