3 The manufacturer's submission
The Appraisal Committee (appendix A) considered evidence submitted by the manufacturer of rivaroxaban and a review of this submission by the Evidence Review Group (ERG; section 10).
3.1 The key clinical evidence in the manufacturer's submission came from EINSTEIN‑PE, an international, event-driven, open-label, assessor-blind, non-inferiority study. The study included 4832 people in an intention-to-treat population. Treatment duration was 3, 6 or 12 months and this was determined by a study investigator before randomisation based on the risk profile of each person and local preferences. Patients were randomised to either rivaroxaban 15 mg twice daily for 21 days followed by 20 mg once daily for the intended treatment duration, or to enoxaparin (a low molecular weight heparin [LMWH]) 1.0 mg/kg twice daily until anticoagulation was established plus a vitamin K antagonist (either warfarin or acenocoumarol), which was dose adjusted to maintain the international normalised ratio (INR) within a therapeutic range of 2.0 to 3.0 with a target of 2.5. Enoxaparin was administered for at least 5 days and was stopped when the INR was more than 2.0 on 2 consecutive measurements at least 24 hours apart. There was an advised overlap with the vitamin K antagonist of 4 to 5 days. Patients were assessed during their intended treatment duration, followed by a 30-day observation period. The manufacturer noted that there was a difference in the dose of enoxaparin used in the trial and the dose covered by the European and UK licence (that is, 1.5 mg/kg once daily for at least 5 days and until adequate oral anticoagulation is established).
3.2 Out of the whole study population, 5.2% were allocated to receive 3 months of treatment, 57.4% to 6 months of treatment and 37.4% to 12 months of treatment. The median time from onset of symptoms to randomisation was 4 days. The EINSTEIN‑PE study allowed a limited amount of treatment before randomisation. A similar proportion of patients in the rivaroxaban arm (92.5%) and the LMWH/VKA arm (92.1%) received pre-randomisation anticoagulation (p=0.62, post-hoc binomial test). Among those who received pre-randomisation anticoagulation, 62.5% of patients received anticoagulation for 1 day (the maximum duration permitted was 48 hours).
3.3 In the intention-to-treat population, the mean age was 58 and approximately 53% of the patients were male. Around 25% of patients in both treatment arms had a concurrent deep vein thrombosis. Pulmonary embolism was unprovoked in 65% of patients receiving rivaroxaban and 64% of patients receiving LMWH with a vitamin K antagonist (hereafter referred to as LMWH/VKA). Approximately 5% of people in both treatment arms had active cancer, and 19% and 20% of people in the rivaroxaban and LMWH/VKA treatment arms respectively had experienced a previous venous thromboembolism. EINSTEIN‑PE excluded people with a creatinine clearance of less than 30 ml/min and people for whom rivaroxaban was not suitable or who had contraindications to enoxaparin, warfarin or acenocoumarol. A total of 555 (11.5%) patients discontinued treatment; the number of people who discontinued was similar in both treatment groups (p=0.07).
3.4 The primary efficacy end point for EINSTEIN‑PE was symptomatic recurrent venous thromboembolism, which was a composite end point comprising recurrent deep vein thrombosis or pulmonary embolism. This included both fatal and non-fatal pulmonary embolism, and unexplained death for which a pulmonary embolism could not be ruled out. In the intention-to-treat population (rivaroxaban n=2419; LMWH/VKA n=2413), rivaroxaban met the pre-specified non-inferiority criterion, which required the upper bound of the 95% confidence interval of the hazard ratio to be less than 2. Symptomatic recurrent venous thromboembolism events occurred in 50 (2.1%) patients in the rivaroxaban arm compared with 44 (1.8%) patients in the LMWH/VKA arm (hazard ratio [HR] 1.12; 95% confidence interval [CI] 0.75 to 1.68).
3.5 The primary safety outcome for EINSTEIN‑PE was clinically relevant bleeding, that is, major bleeding and clinically relevant non-major bleeding in the safety population, which consisted of all patients who had received at least 1 dose of the study drug (rivaroxaban n=2412; LMWH/VKA n=2405). There was no difference between rivaroxaban and LMWH/VKA in clinically relevant bleeding that was experienced by 249 (10.3%) and 274 (11.4%) patients in each treatment arm respectively (HR 0.90, 95% CI 0.76 to 1.07). The proportion of patients who experienced major bleeding was statistically significantly lower with rivaroxaban (1.1%) than with LMWH/VKA (2.2%) (HR 0.49, 95% CI 0.31 to 0.79, p=0.003). In the intended treatment period, there were a similar number of deaths in the rivaroxaban arm (58 deaths) and the LMWH/VKA arm (50 deaths) (HR 1.13, 95% CI 0.77 to 1.65). Treatment-emergent adverse events (other than bleeding and recurrent venous thromboembolism) were similar between the treatment arms. Approximately 5% of patients in the rivaroxaban arm and 4% in the LMWH/VKA arm discontinued treatment because of an adverse event.
3.6 The manufacturer reported a time in therapeutic range for the comparator LMWH/VKA of 62.7% across all 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‑PE between time in therapeutic range and treatment effect.
3.7 Health-related quality of life was not measured in EINSTEIN‑PE. The manufacturer described 2 measures of treatment satisfaction that had been measured: the Anti-Clot Treatment Scale (ACTS) and the Treatment Satisfaction Questionnaire (TSQM). Treatment satisfaction was not used to derive any of the utility values used in the economic analysis.
3.8 The manufacturer presented the results for a number of subgroups considered in EINSTEIN‑PE, which included the 3 subgroups specified in the final scope issued by NICE. These were groups based on underlying risk of bleeding, provoked compared with unprovoked venous thromboembolism, and the presence or absence of cancer. The manufacturer used the intended treatment duration as a proxy for bleeding risk. The manufacturer tested for a statistically significant interaction between each subgroup and the primary efficacy and safety outcomes. The outcomes of the statistical interaction tests are academic in confidence, so the results cannot be presented in this document. The manufacturer also presented graphically the relative efficacy across subgroups, including the 3 subgroups specified in the final scope issued by NICE. The manufacturer presented similar results for the major and clinically relevant non-major bleeding outcome (with the exception of the idiopathic and non-idiopathic groups). The 95% confidence intervals surrounding the hazard ratios for all outcomes overlapped across the subgroups, suggesting a consistency of effect.
3.9 The manufacturer highlighted that in the comparator arm of the trial, patients with cancer had received LMWH/VKA, whereas standard care in the UK is LMWH alone. The manufacturer did not consider it appropriate to conduct a network meta-analysis to estimate the relative effectiveness of rivaroxaban compared with LMWH in people with cancer who had experienced a pulmonary embolism. This was because of the heterogeneity of the studies assessing long-term treatment of venous thromboembolism in people with cancer, which the manufacturer had identified and presented in its submission for NICE technology appraisal guidance 261.
3.10 The manufacturer constructed a Markov model to evaluate the consequences of 3-, 6- and 12-month, and lifelong, treatment with rivaroxaban for preventing recurrent venous thromboembolism in people who experience an acute pulmonary embolism. The model used a time horizon of 40 years and a cycle length of 3 months. To reflect the change in the risk patients experience over time, different risks were applied in cycle 1 (months 0–3), 2 (months 3–6), 3 and 4 (months 6–12) and 5 onwards (12 months onwards). The evaluation was undertaken from an NHS and personal social services perspective, and costs and utilities were discounted at 3.5% per year after the first year.
3.11 For people initially treated for 3, 6 or 12 months, there were 13 health states including death. The lifelong treatment model contained an additional state. People entered the model after their index pulmonary embolism to an on-treatment state in which they received 3-, 6-, 12-month or lifelong treatment with rivaroxaban or LMWH/VKA. They then either stayed on treatment, experienced a recurrent venous thromboembolism (pulmonary embolism or deep vein thrombosis), experienced an adverse event (clinically relevant non-major bleed, major intracranial bleed, or major extracranial bleed), moved to an off-treatment health state, entered a long-term complication state (for example, chronic thromboembolic pulmonary hypertension), or died. The additional state in the lifelong treatment model was for people who had experienced a deep vein thrombosis in the timeframe of the model and who had not stopped treatment for other reasons. All patients who experienced a deep vein thrombosis after their index pulmonary embolism were at risk of post-thrombotic syndrome. There was not a separate health state for post-thrombotic syndrome, rather the consequences of post-thrombotic syndrome were applied as disutilities and costs to patients in both the on- and off-treatment post-deep vein thrombosis health states.
3.12 The modelled cohort was adults with an acute pulmonary embolism who matched the licensed indication, the EINSTEIN‑PE trial population and the stated decision problem. Data from EINSTEIN‑PE were used to inform the clinical effectiveness of treatments and to derive the transition probabilities used in the model; this was supplemented with data from the manufacturer's systematic reviews. All drug acquisition costs and resources associated with acute treatment hospital stay, monitoring, recurrent thromboembolic events and adverse events (that is, bleeding events) were included in the model.
3.13 A validated preference-based measure of quality of life was not measured in EINSTEIN‑PE; the manufacturer derived the utility values used in the model through systematic review. The manufacturer assigned a baseline utility value of 0.825 to all patients with pulmonary embolism entering the model, which was taken from a survey of the UK general population using a visual analogue scale rating and adjusted with disutility values for deep vein thrombosis, pulmonary embolism, extracranial bleed, intracranial bleed and post-thrombotic syndrome. All people who had an intracranial bleed moved to a post-intracranial bleed health state in the next cycle of the model after their bleed. The utility value assumed for an intracranial bleed was 0.33, which increased to 0.71 in the post-intracranial bleed state. A disutility due to warfarin therapy of 0.012 was applied in the LMWH/VKA arm.
3.14 Base-case results were presented for the 3-, 6-, 12-month and lifelong treatment durations. For the 3-, 6- and 12-month treatment durations, rivaroxaban dominated LMWH/VKA, that is, rivaroxaban was less costly and more effective (0.027, 0.013 and 0.019 incremental quality-adjusted life years [QALYs] and a £395.80, £213.21 and £133.13 reduction in total costs for the 3-, 6- and 12-month treatment groups respectively). In the lifelong treatment analysis, rivaroxaban was associated with an incremental cost-effectiveness ratio (ICER) compared with LMWH/VKA of £13,252 per QALY gained (0.104 incremental QALYs for an extra £1374.73).
3.15 The manufacturer performed 123 deterministic sensitivity analyses for each of the 4 durations of treatment. For the 3-, 6- and 12-month treatments, the net monetary benefit for rivaroxaban compared with LMWH/VKA was positive for all analyses if the maximum acceptable ICER was £20,000 per QALY gained. Cost effectiveness of lifelong treatment with rivaroxaban was most sensitive to changes in the frequency of INR-monitoring visits, where the ICER increased from £13,252 per QALY gained to £27,914 per QALY gained if people have 3, rather than 5, visits in each quarter after the first. The manufacturer conducted 1 scenario analysis, in which the time horizon was reduced from 40 to 5 years. With a 5-year time horizon, rivaroxaban continued to dominate LMWH/VKA for the 3-, 6- and 12-month treatment durations. For the lifelong treatment duration, reducing the time horizon to 5 years decreased the ICER of rivaroxaban compared with LMWH/VKA to £12,282 per QALY gained. Probabilistic sensitivity analysis demonstrated that there was a 99.9%, 95.9%, 93.7% and 59.1% probability that the base-case ICER for the 3-, 6-, and 12-month and lifelong treatments was lower than £20,000 per QALY gained.
3.16 The ERG considered that overall, the manufacturer's submission provided an unbiased estimate of the treatment effect of rivaroxaban. The ERG stated that it was based on a well-conducted systematic review of clinical effectiveness, which identified 1 relevant randomised controlled trial. It considered the trial to be of reasonable quality with a low risk of bias.
3.17 The ERG raised concerns that the patient population in the trial may not be fully representative of the treatment population in the UK. In particular, it stated that patients with severe renal impairment (a creatinine clearance of 15–29 ml/min) were excluded from the trial. The ERG noted that the summary of product characteristics specifies that rivaroxaban can be used with caution with dose reductions if needed in this group of patients. The ERG stated that as these patients are at higher risk of bleeding and were excluded from the trial, it is possible that the trial may have underestimated the rate of bleeding that may be seen in clinical practice with rivaroxaban.
3.18 The ERG noted that the trial only assessed outcomes up to a 12‑month treatment period; therefore, effectiveness and safety of long-term treatment with rivaroxaban is unknown. The ERG stated that the manufacturer's Kaplan-Meier plot of cumulative venous thromboembolism rates suggested a worsening of the relative hazard of recurrent venous thromboembolism while on rivaroxaban, compared with LMWH/VKA, towards the end of the 12-month treatment period. The ERG commented that it is plausible that the hazard of recurrent venous thromboembolism might worsen further if treatments are compared in the longer term, particularly if adherence to rivaroxaban (which does not need the regular monitoring of vitamin K antagonists) declines. The ERG suggested that the long-term adherence to rivaroxaban may be lower than the 80% plus observed in most of the patients in EINSTEIN‑PE.
3.19 The ERG considered the manufacturer's subgroup analyses. The ERG suggested that the outcomes may be worse in the active cancer group than those seen for other patients because of increased bleeding risk. The ERG noted that the manufacturer's presentation of subgroup data suggested consistency across the subgroups in terms of the relative safety and efficacy. It considered that because of the small numbers of people in the trial who had cancer, differences in the incidence of bleeding may not have been apparent because of the small number of events recorded in each treatment arm. The ERG also noted the 95% confidence intervals around the hazard ratio for recurrent venous thromboembolism in patients with active cancer presented by the manufacturer were wide, and suggested this indicated that there is uncertainty around where the true effect lies and that the manufacturer's analysis of efficacy in the cancer subgroup should be interpreted with caution. The ERG had concerns that the manufacturer had used intended treatment duration as a proxy for both underlying risk of bleeding and provoked or unprovoked pulmonary embolism. However, it suggested that there are no robust markers for determining length of treatment in advance, which suggests that pre-specified treatment durations may not be a good proxy for other variables.
3.20 The ERG considered the structure adopted for the economic model to be reasonable, consistent with current clinical understanding of pulmonary embolism and consistent with the previous economic evaluations of treatments for venous thromboembolism, such as the submission for NICE technology appraisal guidance 261. The ERG also considered that the parameters used in the model were generally appropriate and that the population used in the model, drawn from EINSTEIN‑PE, is broadly representative of the pulmonary embolism population in the UK (with the exception that it did not include people for whom rivaroxaban is contraindicated or people with severe renal impairment who may still be eligible for rivaroxaban).
3.21 The ERG noted that all transition probabilities were treatment-specific in the lifelong model but that there appeared to be an error in the model, because after 36 months, the probability of recurrent venous thromboembolism and bleeding events were the same for rivaroxaban and LMWH/VKA. The ERG stated that the probabilities of these events after 36 months were not explicitly stated in the manufacturer's submission. The ERG believed this to be an unintended error and corrected the model so that the treatment effect of rivaroxaban after 36 months was applied to the LMWH/VKA transition probabilities. After this amendment, the ICER for rivaroxaban compared with LMWH/VKA in the lifelong treatment analysis was reduced from £13,252 per QALY gained to £7072 per QALY gained. For all of the subsequent analyses, the ERG incorporated this correction and referred to this as the amended base case. The amended probabilistic base-case ICER for rivaroxaban compared with LMWH/VKA was £7019 per QALY gained.
3.22 The ERG had concerns about some of the utility values used in the manufacturer's model, particularly as some of the sources of utility values identified and used by the manufacturer were small studies or did not use the EQ-5D instrument that is the preferred measure of health related quality of life in adults in NICE's reference case. The ERG considered the utility value of 0.33 for an intracranial bleed, based on a study of 129 people that had used time trade off rather than the EQ-5D to derive the utility value that the manufacturer had applied for 3 months in the intracranial bleed state model, to be low. The ERG identified a prospective, longitudinal study that suggested a utility value of 0.31 immediately after an intracranial bleed (stroke), increasing to 0.55 after 1 month and to 0.61 by 3 months. As the manufacturer had estimated that rivaroxaban was associated with fewer intracranial bleeds than LMWH/VKA, the ERG stated that a mid-value of 0.55 for the intracranial bleed health-state utility value would be a more conservative assumption. The ERG also questioned the manufacturer's choice of utility values for post-thrombotic syndrome. The manufacturer had used a study of 30 healthy volunteers that had used a standard gamble approach to derive a utility value of 0.93 for severe post-thrombotic syndrome. However, the ERG stated that the study of 129 people, which the manufacturer had used to obtain utility values for some of the health states in its model including the utility value for an intracranial bleed, gave a utility value of 0.86 for post-thrombotic syndrome. The ERG did not consider the manufacturer's choice of utility value for post-thrombotic syndrome to be a conservative assumption.
3.23 The ERG was satisfied that the unit costs used in the economic model were relevant and had been obtained using suitable methods. However, it noted that the costs of reversing the effects of rivaroxaban and warfarin in the case of major bleeding or elective surgery had not been included and that these may be significant. The ERG stated that vitamin K, fresh frozen plasma and prothrombin complex concentrate (PCC) are used to reverse bleeding events on warfarin but there is no specific antidote for rivaroxaban. The ERG commented that either activated recombinant factor VIIa or PCC may be considered to manage severe and life-threatening bleeding in patients on rivaroxaban. The ERG's clinical adviser considered that the reversal of bleeding on warfarin is likely to need less PCC than rivaroxaban, and that recombinant factor VIIa may be more effective for reversing rivaroxaban than PCC. The ERG estimated that the cost of treating a patient weighing 70 kg with recombinant factor VIIa is £19,303. The ERG also estimated that the cost of treating a bleed while receiving rivaroxaban with PCC would be £1680, and the maximum cost for treating a bleed on warfarin with PCC concentrate would be £1260.
3.24 The ERG conducted the following exploratory analyses:
reduction in assumed frequency of INR-monitoring visits
reduction in mean LMWH treatment length
reduction in the efficacy of rivaroxaban after 12 months in the lifelong treatment analysis in preventing recurrent venous thromboembolism
higher hazard of major bleed on rivaroxaban in the lifelong treatment analysis
higher utility values for the intracranial bleed state
higher mean age of model population
costs of emergency anticoagulant reversal taken into account in all cases of major bleeding
a multiple assumption scenario (reduction in assumed frequency of INR monitoring visits, with a greater proportion of these in secondary care and a greater proportion of the primary care monitoring visits led by nurses; reduction in mean LMWH treatment length; reduction in the efficacy of rivaroxaban after 12 months in the lifelong treatment analysis; higher hazard of major bleed).
3.25 The ERG noted that the manufacturer had assumed 9 INR-monitoring visits in the first quarter for people receiving LMWH/VKA and 5 in each subsequent quarter, and that this was consistent with what the manufacturer presented for NICE technology appraisal guidance 261. The ERG highlighted that in NICE technology appraisal guidance 261, the Committee had concluded that a less intensive INR-monitoring programme of 6 visits in the first 3 months followed by 3 visits every quarter thereafter was reasonable and relevant (when a deep vein thrombosis was the index thromboembolism). The ERG reduced the frequency of INR-monitoring visits to 6 visits in the first quarter and 3 in each subsequent quarter. The ICER for rivaroxaban compared with LMWH/VKA increased from £7072 in the base case to £17,857 per QALY gained in the lifelong treatment duration analysis, and went from dominating LMWH/VKA in the 12-month treatment duration analysis to having an ICER of £3542 per QALY gained. Rivaroxaban continued to dominate LMWH/VKA in the 6-month treatment analysis. The ERG did not present the effect of a reduced INR-monitoring frequency scenario on the 3-month treatment analysis. For the lifelong treatment duration analysis, the ERG assessed a further scenario of 6 visits in the first quarter and 2 in each subsequent quarter. This assumption increased the ICER for rivaroxaban compared with LMWH/VKA from £7072 to £22,912 per QALY gained.
3.26 Assuming a shorter treatment duration with LMWH from the manufacturer's estimate in the base case (derived from the academic-in-confidence average duration in EINSTEIN‑PE) of either 9, 8 or 6 days was found to have a minimal cost-saving effect in the 6-month treatment duration analysis. Rivaroxaban continued to dominate LMWH/VKA regardless of treatment duration with LMWH. The ERG did not present the effect of assuming the shorter treatment duration with LMWH on the 3-, 12-month or lifelong treatment duration analyses.
3.27 As there was uncertainty surrounding the long-term efficacy and safety of rivaroxaban, the ERG performed scenario analyses that assessed varying efficacy and safety effects of rivaroxaban. The hazard ratio for recurrent venous thromboembolism for rivaroxaban compared with LMWH/VKA was 1.123 for the entirety of the lifelong treatment base case. The ERG assessed 2 scenarios in which the hazard ratio was increased to either 1.5 or 2.0 after 12 months (that is, rivaroxaban was assumed to be increasingly less effective relative to LMWH/VKA). Assuming a hazard ratio of 1.5 for venous thromboembolism after 12 months for the population in the lifelong treatment analysis increased the ICER for rivaroxaban compared with LMWH/VKA from £7072 to £9043 per QALY gained. When a hazard ratio of 2.0 was assumed, the ICER for rivaroxaban compared with LMWH/VKA increased to £14,090 per QALY gained.
3.28 In the manufacturer's base-case analyses, the hazard ratio for major bleed for rivaroxaban compared with LMWH/VKA was 0.493. The ERG assessed 2 scenarios in the lifelong treatment duration analysis in which the hazard ratio for major bleeds was increased. In the first scenario, the ERG used a hazard ratio for major bleeds of 0.65. This was taken from the EINSTEIN‑DVT trial (one of the key trials supporting the clinical effectiveness of rivaroxaban in the manufacturer's submission for NICE technology appraisal guidance 261) that compared rivaroxaban with LMWH/VKA in preventing recurrent venous thromboembolism in people who had experienced a deep vein thrombosis. In this scenario, the ICER for rivaroxaban compared with LMWH/VKA increased from £7072 to £10,070 per QALY gained for the lifelong treatment duration. In the second scenario, the ERG used a hazard ratio for major bleeds of 0.79, which was the upper limit of the 95% confidence interval surrounding the hazard ratio for a major bleed seen in EINSTEIN‑PE. Applying this hazard ratio, the ICER for rivaroxaban compared with LMWH/VKA increased from £7072 to £14,177 per QALY gained for the lifelong treatment duration.
3.29 Assuming an increased utility value in the intracranial bleed state from 0.33 in the base case to 0.55 (see section 3.22) did not appreciably change the total QALYs and the model outcomes were hardly altered: rivaroxaban continued to dominate LMWH/VKA for 6- and 12-month treatment durations, and in the lifelong treatment analysis the ICER increased from £7072 to £7098 per QALY gained. The ERG did not present the effect of assuming a higher utility value in the intracranial bleed state on the 3-month treatment analysis.
3.30 The ERG noted that the base-case analyses used a population with a mean age of 58 years, which is lower than the mean age of some other pulmonary embolism and deep vein thrombosis patient populations described in the literature. However, assuming a higher mean age of the model population from 58 years in the base case to 65 years did not have a large effect on the cost-effectiveness estimates: rivaroxaban continued to dominate LMWH/VKA for 6- and 12-month treatment durations, and for a lifelong treatment, the ICER increased from £7072 to £7911 per QALY gained. The ERG did not present the effect of assuming a higher mean age of the model population on the 3-month treatment analysis.
3.31 The ERG assessed 3 scenarios in which the costs of emergency anticoagulant reversal were taken into account. The first scenario assumed that all people received PCC in all cases of major bleeding. This scenario had a modest effect on the base-case analyses for the 12-month and lifelong treatment durations: rivaroxaban continued to dominate LMWH/VKA in the 12-month treatment analysis and the ICER decreased from £7072 to £6868 per QALY gained in the lifelong treatment analysis. The second scenario assumed that people who had a bleed while taking LMWH/VKA received PCC, whereas those taking rivaroxaban received recombinant factor VIIa. This scenario resulted in an ICER for rivaroxaban compared with LMWH/VKA of £2328 per QALY gained for the 12-month treatment duration, and increased the ICER from £7072 to £19,642 per QALY gained for the lifelong treatment duration. The third scenario assumed the same as the second scenario in terms of treatments received to reverse major bleeding but also assumed that the risk of a major bleed with rivaroxaban was more similar to the risk experienced on LMWH/VKA (HR 0.65 from EINSTEIN-DVT) and the frequency of INR monitoring for people receiving LMWH/VKA to be 6 in the first quarter and 3 in each subsequent quarter. In this scenario, the ICER increased for the 12-month treatment cohort to £23,364 per QALY gained and to £44,046 per QALY gained for lifelong treatment.
3.32 The ERG's multiple assumption scenario included a reduction in the frequency of INR-monitoring visits with a greater proportion occurring in secondary care (a 50:50 split rather than the 66 primary care to 34 secondary care split as in the manufacturer's base case); a greater proportion of primary care monitoring visits led by nurses (75% rather than 50%); a reduction in LMWH treatment length; a reduction in rivaroxaban efficacy after 12 months in the lifelong treatment duration analysis; and a raised hazard of major bleed. After applying these assumptions, rivaroxaban continued to dominate LMWH/VKA for the 6-month treatment duration. For the 12-month treatment duration, the ICER for rivaroxaban compared with LMWH/VKA was £11,590 per QALY gained, and for the lifelong treatment duration the ICER was £35,909 per QALY gained. The ERG did not present the effects of these multiple assumptions on the 3-month treatment analysis.
3.33 Full details of all the evidence are in the manufacturer's submission and the ERG report.