Rivaroxaban for the prevention of venous thromboembolism after total hip or total knee replacement in adults

The manufacturer's submission compared rivaroxaban with enoxaparin, a low molecular weight heparin (LMWH), using direct evidence from randomised controlled trials (RCTs), and with dabigatran, using RCT evidence in a mixed-treatment comparison. Outcomes analysed included: incidence of deep vein thrombosis (DVT); incidence of pulmonary embolism (PE); mortality; adverse effects of treatment including bleeding events; post-DVT complications including post-thrombotic syndrome; length of hospital stay; and health-related quality of life. The manufacturer's submission did not include analysis of outcomes at the site of the orthopaedic intervention, such as joint infection. There was no comparison with fondaparinux.

The manufacturer conducted a systematic review that identified six RCTs comparing rivaroxaban with other therapies for the prevention of VTE. Four RCTs met the inclusion criteria (RCTs involving patients aged 18 years or older having hip or knee replacement, comparing rivaroxaban with other therapies including placebo): RECORD 1 and RECORD 2 recruited people having total hip replacement surgery and RECORD 3 and RECORD 4 recruited people having total knee replacement surgery. The other two trials were excluded because they were phase 2, dose-ranging studies. The primary endpoint for all four included trials was a composite comprising any DVT, non-fatal PE and death from all causes.

The RECORD 1 (n=4,541) and RECORD 2 (n=2,509) trials were multicentre, prospective, double-blind, parallel-group design RCTs comparing rivaroxaban with enoxaparin for the prevention of VTE after total hip replacement surgery. In RECORD 1, rivaroxaban was administered at a dosage of 10 mg once daily for 35 days starting on the day of surgery. Enoxaparin was administered at a dosage of 40 mg starting 1 day before surgery and for 35 days thereafter. For this study, the manufacturer reported a statistically significant difference in the incidence of the composite primary endpoint between rivaroxaban and enoxaparin based on a 'modified' intention to treat (MITT) analysis. The primary endpoint occurred in 1.1% of the rivaroxaban group compared with 3.7% of the enoxaparin group; relative risk reduction (RRR) was 70% (95% confidence interval [CI] 49 to 82, p<0.001). The manufacturer further reported the results of the secondary endpoint: major VTE (a composite of proximal DVT, non-fatal PE and VTE-related death) for RECORD 1. In this study, in the MITT population, major VTE occurred in 4 (0.2%) patients receiving rivaroxaban compared with 33 (2.0%) patients receiving enoxaparin; the RRR was 88% (95% CI 66 to 96, p<0.001).

RECORD 2 was a comparison of 35 days of prophylaxis with rivaroxaban 10 mg daily compared with a shorter course (15 days) of enoxaparin prophylaxis at 40 mg daily. The manufacturer reported a statistically significant difference in the incidence of the composite primary endpoint between rivaroxaban and enoxaparin in the MITT analysis; 2.0% in the rivaroxaban group compared with 9.3% in the enoxaparin group (RRR 79%, 95% CI 65 to 87). The secondary endpoint, major VTE, occurred in 6 (0.6%) patients receiving rivaroxaban compared with 49 (5.1%) patients receiving enoxaparin (p<0.0001).

The RECORD 3 (n=2,531) and RECORD 4 (n=3,148) trials were multicentre, prospective, double-blind, parallel-group design RCTs comparing rivaroxaban with enoxaparin for the prevention of VTE after total knee replacement surgery. In both RECORD 3 and RECORD 4 rivaroxaban was administered at a dosage of 10 mg once daily for 10 to 14 days starting on the day of surgery. In RECORD 3 the comparator was enoxaparin at a dosage of 40 mg once daily, starting the day before surgery and for 10 to 14 days thereafter. The MITT analysis showed a statistically significant difference in the incidence of the composite primary endpoint: 9.6% in the rivaroxaban group compared with 18.9% in the enoxaparin group (RRR 49%, 95% CI 35 to 61). Major VTE occurred in 9 (1.0%) patients receiving rivaroxaban compared with 24 (2.6%) patients receiving enoxaparin (RRR 62%, 95% CI 18 to 82; p=0.02).

In RECORD 4 the comparator was enoxaparin at a higher dosage of 30 mg twice daily starting 1 day before surgery and continuing for 10 to 14 days thereafter. In this study, the composite primary outcome occurred in 6.9% and 10.1% of the rivaroxaban and enoxaparin groups, respectively (p<0.012). RECORD 4 found a lower incidence of major VTE in patients treated with rivaroxaban compared with enoxaparin.

The manufacturer used the per protocol population to test for non-inferiority of rivaroxaban compared with enoxaparin. The MITT analysis was used as supportive analysis in the test for superiority of rivaroxaban over enoxaparin. The manufacturer presented a range of meta-analyses that pooled the RECORD 1 and RECORD 2 studies for total hip replacement and the RECORD 3 and RECORD 4 studies for total knee replacement. Because of a lack of head-to-head trials between rivaroxaban and dabigatran, the manufacturer used an indirect comparison methodology that compared rivaroxaban and dabigatran, with enoxaparin as the common comparator. The analysis compared the incremental effect of rivaroxaban over enoxaparin to the incremental effect of dabigatran over enoxaparin. The comparison of these incremental effects allowed the indirect estimation of the incremental effect of rivaroxaban over dabigatran. The manufacturer stated that the indirect comparison methods used in this analysis were widely published and ensured that randomisation from the original trials was preserved. The results of these analyses were submitted to NICE in confidence and are not presented in this document.

The main safety endpoint in the RECORD trials was the incidence of treatment-emergent major bleeding. The manufacturer reported the rates of major bleeding for patients treated with rivaroxaban and enoxaparin, respectively, as follows. RECORD 1: 0.3% versus 0.1%, p=0.178; RECORD 2: 0.1% versus 0.1%, p=0.98; RECORD 3: 0.6% versus 0.5%, p=0.77; and RECORD 4: 0.7% versus 0.3%, p=0.11. The incidence of non-major bleeding (comprising clinically relevant non-major bleeding, haemorrhagic wound complications and other non-major bleeding) was similarly low, for rivaroxaban and enoxaparin, respectively, as follows. RECORD 1: 5.8% versus 5.8%; RECORD 2: 6.5% versus 5.5%; RECORD 3: 4.3% versus 4.4%; and RECORD 4: 10.2% versus 9.2%.

The ERG reviewed the literature search strategy and concluded that it effectively identified literature relevant to the decision problem and used relevant search techniques for systematic review and appraisal. The ERG was satisfied that the RECORD trials were of adequate methodological quality. It commented that reporting and interpretation of the safety data were good, and concluded that the manufacturer's submission appeared to contain an unbiased estimate of the effectiveness of rivaroxaban in relation to the main comparator, enoxaparin.

The manufacturer submitted an economic model assessing the cost effectiveness of rivaroxaban compared with enoxaparin in VTE prevention after hip and knee replacement. The model comprised three modules: prophylaxis, post-prophylaxis, and long-term complications. The first two modules represented the acute phase in the form of a decision tree, and the third component represented the chronic phase and was developed as a Markov process. The prophylaxis part of the model was informed by clinical trial events (first 35 days for total hip replacement and 14 days for total knee replacement post-surgery). The post-prophylaxis component reflected the risk of symptomatic VTE events within the first 3 months, and the long-term complications component extrapolated any long-term complications resulting from symptomatic VTE events.

Key assumptions in the economic evaluation included the assumption that asymptomatic VTE events would not incur any costs and did not have an impact on quality of life. It was assumed that all recurrent VTEs were DVTs. If the clinical trial or indirect comparison did not show a significant difference between the two arms, the probabilities were assumed to be equal in the model. It was also assumed that all PEs in the post-prophylaxis module were non-fatal.

The manufacturer presented base-case analyses based on RECORD 1 and RECORD 2 separately to reflect the different comparator regimens. The base case for the total knee replacement indication did not include an analysis based on RECORD 4 because the higher dosage of 30 mg twice daily for enoxaparin reflected practice in the US rather than the UK. However, the manufacturer presented pooled analysis by total hip replacement, total knee replacement and for all trials together.

The results showed that rivaroxaban dominated enoxaparin in both total hip replacement (RECORD 1 and 2) and total knee replacement (RECORD 3). Deterministic sensitivity analysis for RECORD 1 showed that rivaroxaban generally dominated enoxaparin. However with a shorter treatment duration for enoxaparin (but maintaining 35 days' treatment benefits), the incremental cost-effectiveness ratio (ICER) for rivaroxaban compared with enoxaparin was £14,616 per QALY gained. When the benefits were also adjusted to reflect a reduced duration of prophylaxis, the ICER was £914 per QALY gained. In RECORD 2 the main variable that affected the ICER was excluding long-term complications caused by VTE events. This resulted in an ICER of £58,337 per QALY gained. The manufacturer attributed this change to lower prophylaxis drug costs for enoxaparin compared with rivaroxaban. The treatment duration for rivaroxaban was longer (35 days) compared with enoxaparin (15 days) in this trial, and therefore the prophylaxis drug costs for enoxaparin were lower. The indirect comparison with dabigatran showed that rivaroxaban dominated dabigatran in both total hip replacement and total knee replacement.

The ERG considered that the approach to the economic modelling was reasonable. However, it noted that some potential events had been excluded from the model. The possibility of further VTE events other than DVT in the longer-term model was not considered and neither was the possibility of intracranial haemorrhage (a health state associated with marked disutility).

The ERG noted that the conclusions on the cost effectiveness of rivaroxaban were dependent on the assumptions made about parameters that were not statistically significant, and on the appropriateness of pooling data. If all parameters where the p value was greater than 0.05 were set equivalent for rivaroxaban and the comparator, then assuming that all trials were pooled, rivaroxaban dominated the comparators. When the observed data were used and total hip replacement and total knee replacement were pooled separately, rivaroxaban did not always dominate enoxaparin and dabigatran in the total knee replacement indication. The ERG considered it more appropriate to model with observed data than by setting parameters to be equivalent when there was no statistically significant difference. The ERG also considered that pooling the data from all trials was reasonable in the circumstances. However, accepting these points, the differences in costs and QALYs gained across all analyses were extremely small.

Full details of all the evidence are in the manufacturer's submission and the ERG report.