for the Treatment of Advanced Renal Cell Carcinoma (arcc ) - Additional Analysis Using RPSFT Thank you very much for the opportunity to submit additional analysis in relation to the above technology appraisal of everolimus for the treatment of arcc. Background When the data from the pivotal, randomised, placebo-controlled trial (RECORD-1) were analysed, it became apparent that the intention-to-treat (ITT) overall survival (OS) analysis was of limited value in demonstrating the survival benefits associated with everolimus treatment. This was because at the February 2008 data cut-off, 76% of the patients randomised to the placebo arm of the RECORD-1 trial had switched to everolimus treatment following progression. In addition, more than half of these patients switched to everolimus treatment within 8 weeks of randomisation. Therefore the ITT OS results are highly confounded and do not represent a meaningful comparison of treatment effect on survival. In order to correct for this confounding a number of statistical approaches were considered including the (inverse probability of censoring weights) IPCW and the rank preserving structural failure time (RPSFT) method. At that time there was no guidance or precedents that would indicate which approach would be preferred by NICE. In the absence of such information, it was decided at a global level that we would use the IPCW approach to inform our economic analysis. Although, there is academic debate regarding the most appropriate statistical approach, our choice was based mainly on the perceived simplicity of the practical application of a HR directly within the current economic model. The base case results as presented in our submission, are therefore based on the IPCW approach using data from the February 2008 data-cut. No economic analysis using results from a RPSFT approach have been available until now. The global Novartis statistics team instigated an RPSFT analysis based on longer term follow up data from RECORD-1, that has become available (November 2008 data-cut). It should be noted that by the time of this analysis, 81% of the patients randomised to the placebo arm of the RECORD-1 trial had switched to everolimus treatment following progression. Following a review of the ERG Report, and in acknowledgement that a comparison of the results from both the IPCW and RPSFT approaches would be informative, we have now conducted an analysis using the results from the RPSFT method in our economic model which was originally submitted on 30 th September 2009. The results of this analysis are presented below. 1
RPSFT Analysis The RPSFT statistical analysis is fully described in Appendix 1, and is based on the November 2008 data-cut. It was also presented in poster format at the 15 th and 34 th ESMO Congress, Berlin 20 th -24 th September 2009. i At this analysis, based on the ITT analysis the median OS for the everolimus arm was 14.78 months and for the placebo arm the median OS was 14.39 months (HR=0.87, 95% CI: 0.65 1.17, p=0.177). In addition, 81% of patients randomised to placebo had switched to everolimus treatment following progression (compared to 76% reported in the submission for the February 2008 cut-off). A post-hoc exploratory analysis of OS was conducted using the RPSFT method. This is an accelerated failure time model which uses a structural assumption of time proportionality. This method provides an estimate of treatment effect based on randomisation thus correcting for the bias introduced due to patients switching from the placebo arm to the everolimus arm of the trial. Using the RPSFT approach, the corrected median OS for the placebo arm is 10 months versus the unadjusted median OS of 14.4 months. i The results from this analysis estimate a relative survival time that is 1.93 fold longer in the everolimus arm than the placebo arm (95% CI: 0.5 8.5). i These data were used to generate an RPSFT corrected Kaplan Meier OS curve for the placebo arm of the trial. The Kaplan Meier curves of OS based on the raw data for the two treatment arms and the RPSFT corrected OS for the placebo arm are presented in the figure below. Figure 1 - Kaplan Meier Curves of OS based on November 2008 Data-cut i 2
Application of the RPSFT Results in the Economic Model The Kaplan Meier curve for the RPSFT corrected OS for the placebo arm of the study was used to generate the transition probabilities to death ie for following state transitions: progression to death; stable disease without adverse events (AE s) to death and stable disease with AE s to death. As the RPSFT results do not allow differentiation of the conditional probability of death by health state we have assumed the same transition probabilities to death in the placebo arm for each of the states to death. The remaining transition probabilities for the arm are calculated directly from the RECORD-1 trial using the November 2008 data-cut ie for the following state transitions: stable disease with AE s to progression; stable disease without AE s to progression and risk of AE s. All of the transition probabilities for the everolimus arm were calculated directly from the RECORD-1 trial using the November 2008 data-cut. All other base case assumptions in the model remain unchanged. The transition probabilities are presented in Table 1, below. Table 1 RPSFT corrected transition probabilities Per Patient Model : RPSFT-Adjusted Cycle 0 1 2 3 4 5 6 7 8 9 10 1->2 AE Risk 1->3 Progression Risk SD w/o AE 2->3 Prog. Risk from SD w/ae 3->4 Death from PD 1->4 Stable N-Death 2->4 Stable w/ AE N-Death 11 to 18 Per Patient Model Afinitor Cycle 0 1 2 3 4 5 6 7 8 9 10 1->2 AE Risk 1->3 Progression Risk SD w/o AE 2->3 Prog. Risk from SD w/ae 3->4 Death from PD 1->4 Stable N-Death 2->4 Stable w/ AE N-Death 11 to 18 3
The cost-effectiveness results based on the RPSFT approach are presented in Table 2 below. Table 2 Cost-effectiveness results from the RPSFT analysis using the November 2008 cut off Base case with PAS* (IPCW Feb 2008 cutoff) With PAS: RPSFT (Nov 2008 cut-off) Without PAS: RPSFT (Nov 2008 cut-off) 0.607 0.302 0.912 0.454 0.912 0.454 LYG ( 0.841 (10.09 1.265 (15.18 1.265 (15.18 LYG ( 0.426 (5.11 0.639 (7.67 0.639 (7.67 * As presented in the original Novartis submission. Inc LYG ( 0.414 (4.97 0.626 (7.51 0.626 (7.51 Inc cost ( ) cost ( ) Inc cost ( ) ICER for everolimus versus ( /) 0.304 25,222 9,517 15,704 51,613 0.458 36,168 11,824 24,344 53,128 0.458 38,312 11,824 26,488 57,808 The results from the RPSFT analysis give an incremental cost-effectiveness ratio (ICER) of 53,128 with PAS. Conclusion and Discussion As acknowledged in the ERG Report, the appropriate statistical approach to correct for confounding due to crossover is still an area of genuine academic debate. Therefore the opportunity to compare estimates of cost-effectiveness based on two different statistical approaches is of interest as well as providing reassurance that the original IPCW method we adopted does not generate overly favourable results compared to the RPSFT approach. The results generated when applying the RPSFT statistical approach suggest that everolimus, within it s licensed indication, is likely to be cost-effective within acceptable limits when applying the end of life criteria. It should be noted that the longer term, November 2008 data-cut suggests greater survival than the February 2008 data-cut. This means that using the November analysis there are more everolimus patients still alive in the final cycle of the economic model. Unfortunately, there was insufficient time to add further cycles to the model to account for this. However, the overall impact will be to reduce the ICER as there will be greater LYG in the everolimus arm but no further everolimus treatment costs as these are only applicable for the stable disease states. In summary, everolimus fulfils an unmet clinical need as it is a clinically-effective treatment for a small population of arcc patients who do not have any NICE recommended, licensed or effective treatment options. If left untreated these patients are likely to have limited life expectancy and poor prognosis. is the only available oral mtor inhibitor to be licensed for the treatment of arcc and therefore represents an innovative approach to treating arcc. Results from the double-blind, RCT, RECORD-1 trial demonstrated that everolimus reduced the risk of disease progression by 67% (HR=0.33, 95%CI: 0.25-0.43) and improved median progression free survival by 3 months. For these reasons it is anticipated that everolimus will meet the end of life criteria. Furthermore, in order to facilitate early access for 4
patients, Novartis are offering a Patient Access Scheme (PAS) to be considered as part of this appraisal. The Department of Health have confirmed that they are happy for NICE to consider the PAS as part of this appraisal. References i Korhonen P, Haas T, Zuber E, Kay A, Lebwohl D, Motzer R. Overall Survival Among Metastatic Renal Cell Carcinoma Patients Corrected for Crossover Using a Rank Preserving Structural Failure Time (PSFT) Model: Analyses From the Phase 3 Trial. Abstract, P-7155, Joint ECCO 15 th and 34 th ESMO Congress, Berlin 20 th -24 th September 2009. 5