Statistical Analysis of Prothena’s VITAL Phase III Study

In December 2014, Prothena (NasdaqGS: PRTA) announced the launch of the VITAL Phase III trial of NEOD001 as a treatment for newly diagnosed AL amyloidosis with cardiac involvement. Based on positive results from a Phase I/II trial and the announced details of Prothena’s VITAL trial, we conducted a statistical analysis to determine the most likely hazard ratio for the trial, number of patients needed to find a statistically significant result, and the trial’s chances of success. Based on our scenarios, the trial has between 72% to 81% chance of reporting positive results within 18 months after the final patient is enrolled.

Click Here to Download the Full Report



Figure 1. Key Conclusions from Statistical Analysis of VITAL Phase III Trial

figure-1

Overview of Statistical Analysis. The Phase III VITAL trial is based on positive results from a Phase I/II trial where 50% of patients had a cardiac response and an additional 21% had stable disease. Based on the available Phase I/II results and the announced details of Prothena’s VITAL trial, including α=0.05 with 90% power (β=90%), we conducted a statistical analysis to better understand the likelihood of success for Phase III. The details and key assumptions used for this analysis are discussed below. We modeled two different scenarios – a conservative case (Scenario A) that assumes no additive benefit between NEOD001 and standard of care treatment, and Scenario B, which assumes an added benefit for the two treatments. The primary conclusions of this analysis are outlined in the table in Figure 1.

Figure 2. Predicted Kaplan-Meier Survival Curves for VITAL Phase III Trial

figure-2

Predicted Survival Curves for the VITAL Phase III Trial – SoC and Two Treatment Scenarios. In order to give an overview of the potential results for the Phase III trial of NEOD001, we show the SoC survival as described by Kumar and compare it to predicted survival curves for two treatment scenarios. The colored bands surrounding the survival curves in Figure 2 indicate the projected size of the error bars given the Phase III statistics. The scenarios are as follows, and are consistent through the rest of this discussion:

  • Scenario A (NEOD001): This is the conservative case. For modeling this possibility we assume that NEOD001 provides all patient benefit in the trial and there is no additional effect from the SoC. Because NEOD001 has a novel mechanism of action that is orthogonal to currently available treatments, we view this as a very conservative estimation.
  • Scenario B (SoC + NEOD001): This case assumes that there is an additive benefit between the SoC and NEOD001. For cardiac non-responders, relative rates of cardiac stabilization and disease progression remain unchanged from Scenario A.

Explanation of Hazard Ratios. One of the keys to determining the possibility of Phase III success for NEOD001 is the predicted hazard ratio for the treatment. The hazard ratio is a way to measure the expected benefit between two different treatment arms and is an expression of their relative efficacy. For example, in the present discussion, a hazard ratio of 0.5 would indicate a 50% reduction in risk of experiencing a cardiac-related hospitalization or death. The graph in Figure 3 gives a weighted probability of hazard ratios for the scenarios outlined above, with the peak of each line indicating the most likely hazard ratio for the scenario. It is worth noting that not all patients in the Phase I/II trial received the optimal dose of NEOD001, which could potentially lead to better efficacy than predicted by this model.

Figure 3. Predicted True Hazard Ratio

figure-3

The predicted hazard ratio is based on Prothena’s reported Phase I/II clinical trial results. As the Company reports additional patient outcomes from this trial, it will be possible to refine this analysis and more precisely assess the predicted hazard ratio for Phase III. The table in Figure 4 is based on the plot in Figure 3 and shows the hazard ratios at different confidence intervals for Scenario A and Scenario B. In other words, this table tells you the percentage chance that the hazard ratio is below a given value.

Figure 4. Upper Bounds for the Hazard Ratio at Given Confidence Limits

figure-4

Analysis Reveals Number of Events and Patient Requirements. One of the useful aspects of this analysis is that it can reveal the number of patients needed to adequately power the trial and, subsequently, the number of events needed to reach statistical significance. Again using the same scenarios as described above, along with the weighted average of all possible hazard ratios, we created the table in Figure 5. The analysis reveals that in a 236 patient trial and using our conservative assumptions, 54 events would be required to reach statistical significance. Under the more optimistic and potentially more realistic Scenario B, 28 events will be required for the trial to reach statistical significance. This graph allows you to individually analyze and determine the number of events needed based on your assumption of a given hazard ratio.

Probability of Phase III Success. Once again employing Scenario A and Scenario B, combined with the weighted average of all possible hazard ratios, it is possible to predict the chances of success for the Phase III VITAL trial. The graph in Figure 6 indicates our predicted probability of success for the trial from the time that the last patient is enrolled. The chart indicates that after 18 months the probabilities do not increase substantially. However, it is worth noting that if the SoC is more effective than predicted that the trial could be longer than expected without affecting the underlying hazard ratio. According to this analysis, Scenario A produces a 72% chance of success 18 months after the last patient in, and Scenario B yields a predicted 84% chance of success.

Figure 5. Sample Size Requirement and Number of Events to Reach Statistical Significance

figure-5

Upside and Risks to this Analysis. The present analysis is based on publicly available information about Prothena’s clinical trials and our best assumptions of patient characteristics. These assumptions were determined to be the most appropriate for the analysis and were made before any calculations were done. Despite doing our best to choose the best underlying facts, uncertainty remains. Some of the uncertainty in this analysis can be categorized as being too conservative, meaning there could be upside based on unknown factors. There are also risks associated with the trial not captured in this analysis. We categorize some of the upside and risk to this analysis below.

Potential Upside:

  • Not all patients in the Phase I/II trial for NEOD001 received the same dose, whereas all patients in Phase III will receive the optimized dose from Phase I/II.
  • The present analysis is based only on Phase I/II results and does not take into account pharmacokinetic analysis, preclinical results, and other information concerning the mechanism of action of NEOD001.
  • Our scenarios do not take into account a potential synergistic effect between NEOD001 and SoC.
  • The data that formed the basis of our analysis are based on survival data. However, the VITAL trial endpoint is a composite of mortality and cardiac-related hospitalization. Therefore, events may accumulate more quickly than predicted.
  • NEOD001 reduces or postpones incidence of cardiac-related hospitalization, hazard ratios would be lower than predicted.

Figure 6. Predicting VITAL Phase III Trial Success

figure-6

Potential Risks:

  • The small sample size of the Phase I/II trial makes it difficult to accurately predict the true hazard ratio for Phase III, and it is possible that Phase I/II results won’t be replicated in Phase III.
  • The underlying calculation of survival with the current SoC may be different in the VITAL trial than predicted by the Kumar study.
  • It is possible that NT-proBNP values are not perfectly predictive of survival benefit.
  • If the SoC treatment performs better than predicted, it is possible that the time that it would take to accumulate the number of required events would be extended. However, this wouldn’t necessarily affect the underlying hazard ratio.

VITAL Phase III Trial Design. This randomized, double-blind, placebo-controlled Phase III trial will evaluate NEOD001 in 236 treatment-naïve patients newly diagnosed with AL amyloidosis. Patients will be randomized 1:1 to receive 24 mg/kg of NEOD001 or placebo via intravenous infusion every 28 days concurrent with standard of care therapy. The composite primary endpoint is an event-based measure of all-cause mortality or cardiac hospitalizations. Secondary endpoints include NT-proBNP as a biomarker of cardiac function, proteinuria levels as a biomarker of renal function, six-minute walk test, and several quality of life measures.

Explanation of Assumptions used in Statistical Analysis of VITAL. When interpreting the data presented herein, it is important to keep in mind several key assumptions that were used to construct these models. The statistical model analyzing the VITAL trial required a prediction for survival in the study for patients receiving only standard of care (SoC) treatment and a method of correlating Prothena’s Phase I/II results with survival. Recently published guidelines for staging in AL amyloidosis provided survival data from 583 patients, which was used to establish the predicted baseline survival for the VITAL study.1 While there are other studies available that provide a slightly different view of the potential baseline population, the Kumar study provides a large dataset that correlates well with the expected Phase III population.

In the Phase I/II trial of NEOD001, Prothena found a 50% cardiac response rate and a 37.9% mean decrease in NT-proBNP levels in cardiac responders. NT-proBNP is a predictive biomarker of mortality in AL amyloidosis. A 2012 study reported the correlation between NT-proBNP response and overall survival for AL amyloidosis patients. For the purposes of this discussion, we assume that the relationship between NT-proBNP and survival is exactly as reported by Comenzo and colleagues. As of September 30, 2014, there were reported data for 14 cardiac-evaluable patients from the Phase I/II trial. Cardiac evaluable means that they had NT-proBNP greater than 650 pg/mL and had NT-proBNP measured at the end of the trial. We used the following reported Phase I/II trial results as a basis for the present analysis.

  • 7 patients experienced a cardiac response, defined as decrease in NT-proBNP of at least 30% and at least 300 pg/mL.
  • 3 patients had stable disease.
  • 4 patients progressed, meaning at least a 30% increase and 300 pg/mL increase in NT-proBNP.



Source: LifeSci Capital, LLC research report

Research analyst certification and important disclosures can be found in the full research report. To access the full research report, please click on the link at the top of the page.