Memorandum

Novo Nordisk’s Tresiba receives Complete Response Letter asking for pre- approval cardiovascular outcomes trial; dig in, Novo Nordisk – February 11, 2013

Executive Highlights

  • The FDA has issued a Complete Response Letter for Tresiba (degludec) that delays its approval until Novo Nordisk completes a cardiovascular outcomes trial.

Novo Nordisk held a conference call this morning to discuss the FDA’s Complete Response Letter (CRL) for Tresiba (insulin degludec) and Ryzodeg (insulin degludec/aspart), which the company received on February 8. In the CRL, the FDA asks for additional cardiovascular data from a CV outcomes trial (CVOT), presumably to elucidate whether the safety signal seen in the phase 3 CV meta-analysis represents true clinical harm. The FDA also asks Novo Nordisk to resolve the Warning Letter related to a cGMP inspection of a filling plant in Denmark, though we do not believe this was the main issue of the CRL (Novo Nordisk responded to the Warning Letter in December, but noted in Q&A that it has not heard from the FDA since then). We assume Novo Nordisk remains very confident about the favorable benefit:risk profile of Tresiba, and we expect that the company will address these concerns with the FDA to limit Tresiba’s regulatory delay.

The CRL came as a surprise given that the FDA’s endocrinologic and metabolic disease advisory committee voted 8-4 in favor of approval (however, two of the three cardiologists voted against approval) – while panelists unanimously voted to require a CVOT, most felt that a post-approval trial was acceptable (details below). Since deviating from the committee’s vote discredits (to some extent) both the FDA and the panelists, we presume the FDA harbors significant concern regarding Tresiba’s potential CV risk. We are curious whether the FDA would find this CV signal unacceptable for any insulin, or whether it would allow this risk for candidates that provide a more drastic improvement over existing drugs than Tresiba provides over Lantus and Levemir (further discussion below).

During Q&A, management disclosed that the CRL did not provide guidance on conducting the CVOT; as such, they refrained from discussing the CVOT’s design and timeline, emphasizing that further dialogue with the FDA is necessary to gain clarity on this front. Management noted that while this pre-approval CVOT will likely not take as long as Victoza’s roughly-five-year CVOT LEADER, data is not expected before 2015 (and could potentially come later). Novo Nordisk declined to speculate on the FDA’s CV risk benchmark, as well as whether the trial will enroll both type 1 and type 2 patients. As a reminder, Novo Nordisk had previously proposed a design for a post-approval CVOT (7,500 type 2 patients over five years), though it is uncertain whether aspects of this trial design will apply to the pre-approval CVOT (we imagine the pre-approval trial will enroll more patients to accrue events more quickly).

The FDA’s decision has negative implications for the broader diabetes field on two fronts. First, a higher safety standard or more aggressive “digging” for safety concerns will almost certainly discourage future investment and innovation in the insulin and broader metabolic arena. Second, the CV evaluation process reflects the need for greater transparency and dialogue on behalf of the FDA, a concern that increases drug development risk. We elaborate on both issues in the section below titled “Implications for the Diabetes Arena.” Ultimately, we hope that both sides collaborate and avoid the pre-approval CVOT – there is little rationale that this is good for the field and we believe there are a number of tools that FDA can employ to reduce risk to patients. We know the agency can make the right call.

CVOT TRIAL DESIGN AND IMPLICATIONS FOR NOVO NORDISK

  • The design of the pre-approval CVOT is largely uncertain, as Novo Nordisk is waiting on dialogue with the FDA on potential trial requirements. In our view, some considerations remain:
    • 1. The patient population. During Q&A, management declined to comment on whether the CVOT would include both type 1 and type 2 patients. As a reminder, Novo Nordisk previously proposed a design for a post-approval CVOT that included only type 2 patients. The FDA’s CV meta-analysis included four trials in type 1 diabetes and found that CV event rates were generally lower among type 1 patients vs. type 2 patients. For reference, the MACE+ analysis for the type 1 studies showed no increase in CV risk relative to comparators (HR: 0.96 [95% CI: 0.30-3.09]) while the strict MACE analysis showed a 30% increased risk, albeit with a larger confidence interval (HR: 1.30 [95% CI: 0.27-6.29]). In general, Novo Nordisk should have less of a challenge in recruiting participants, as Tresiba’s CVOT will likely enroll patients who take insulin and thus are further along in the disease progression relative to the participants needed for CVOTs assessing oral medications. Management noted during Q&A that setting up and starting recruitment for the CVOT will alone take at least nine months.
    • 2. MACE vs. MACE+ as the primary endpoint. A point of contention during the advisory committee meeting was the merit of using MACE (defined as CV death, non-fatal stroke, and non-fatal myocardial infarction) as the CV endpoint versus MACE+ (the three MACE outcomes plus unstable angina). For background, in the CV meta-analysis, the hazard ratios (HR) were consistently higher for the MACE calculations: the original meta- analysis showed a HR of 1.10 for MACE+ (95% CI: 0.68-1.77) compared to a HR of 1.39 for strict MACE (95% CI:0.76-2.57); the updated analysis showed a HR of 1.30 for MACE+ (95% CI: 0.88-1.93) compared to a HR of 1.67 for MACE (95% CI: 1.01-2.75). TheFDA originally agreed to a meta-analysis using MACE+ but later redefined the endpoint post-hoc to exclude unstable angina (the panelists appeared to see both sides and agreed with the use of both MACE+ and MACE). We would expect the FDA to mandate strict MACE as a primary endpoint, as is typically done for CVOTs.
    • 3. The CV event rate. While the inconclusive nature of the CV data could reflect the limitations of the meta-analysis, it also suggests that any potential CV signal would be very small. We’re curious whether the presumption of a small safety signal (and thus fewer events per patient-years) would require Novo Nordisk to conduct a larger CVOT than typically seen in order to accurately identify and characterize the CV risk.
    • 4. The FDA’s role in trial design. Tresiba’s CRL is the first instance after the 2008 CV guidance for diabetes drugs in which the FDA has specifically requested a pre- approval CVOT for a diabetes medication (Orexigen’s Contrave comes to mind as an example for obesity). Though the FDA’s dialogue with Novo Nordisk will of course be confidential, we’ll be interested in hearing how involved the agency will be in setting the trial design and analysis. This is of particular interest since it is our impression that FDA had originally excluded insulin from the CV guidance.
  • We wonder whether the FDA has considered ways to assess CV risk without delaying Tresiba’s approval until the CVOT’s completion. The first option that comes to mind would be to use the CVOT interim analysis as the basis for approval, similar to the FDA’s approach with Orexigen’s Contrave. As a reminder, the FDA is not only accepting the interim analysis for the Contrave’s resubmission, but announced in January that it would accept the summary report of the interim analysis, rather than the complete clinical study report. A second potential option would be to approve Tresiba with either a strict REMS or a limited indication. During the advisory committee meeting, influential cardiologist Dr. Marvin Konstam (Tufts University School of Medicine, Boston, MA) gave a strong vote against approval, but remarked that he would consider an approval for patients with a high risk for hypoglycemia but a low risk for CV events. While the practicality of this approach is uncertain, the existence of alternative drugs such as Lantus would appear to make a REMS or narrowed indication easier, as patients would not necessarily be “desperate” for Tresiba.
    • During canagliflozin’s recent advisory committee meeting, roughly 80% of the CV data came from the interim analysis of an ongoing CVOT (CANVAS), leading the panelists to discuss the merits of using interim CV data for approval. Highly respected panelist Dr. Sanjay Kaul (Cedars-Sinai Heart Institute, Los Angeles, CA) expressed concern that if a drug is approved while the CVOT is still ongoing, unblinding the CV data could result in patients leaving the trial to seek the favored study treatment from their physician, thus compromising the ability of the trial to keep patients enrolled. We presume that keeping the data blinded could mitigate this problem, but believe this would be hard to do given that the data supporting approval are publically disclosed in the FDA briefing documents published online. See our full report on canaligliflozin’s advisory committee meeting at http://www.closeconcerns.com/knowledgebase/r/aa7bbddc and read our interview with Dr. Kaul at http://www.closeconcerns.com/knowledgebase/r/24a3594d.
  • The option that FDA has appears to have effectively foreclosed is for simply approving the NDA with appropriate labeling, risk management approaches, and hard requirement for a definitive post-approval CVOT. We certainly hope this is not the case and would never expect Novo Nordisk to assume that it is. Indeed, FDA itself expressed the possibility that the overall benefit to risk relationship is positive given offsetting advantages on hypoglycemia. In the case of Contrave, FDA found its way to more moderate expectations for the Contrave trial than initially announced and provided the interim analysis approach. We believe strongly it is not too late for Novo Nordisk to push back by asking for re-consideration of its case,and hope to see more than just a pre-approval CVOT take place. For perhaps Novo Nordisk can take this on but how many other companies can do so? Surely not more than a handful? .

Other points of interest:

  • According to IMS moving annual total data from November 2012, North America holds a ~76% value share (~$5.8 billion) of the worldwide basal insulin market (valued at ~$7.6 billion). Delaying Tresiba’s approval not only cuts profit margins but likely places a launch after the 2015 patent expiration of Lantus, giving Tresiba little time to establish market share before the potential entry of biosimilar glargine products.
  • During Q&A, Novo Nordisk remarked that the CRL will not affect Tresiba’s reimbursement in Europe; the company also plans to move forward in developing IDegLira, with a EU filing still expected in mid-2013. As a reminder, IDegLira is a fixed- ratio combination of insulin degludec/liraglutide, which was shown in phase 3a studies to provided greater A1c reductions compared to Tresiba alone (average A1c reduction of 1.9% at 26 weeks from a baseline A1c of 8.7%; data for Tresiba not provided). We see this decision as an important positive, since one of degludec’s key advantages is its potential for combination with oral medications or faster-acting insulin.
  • Management noted that it will retain the US sales force hired to promote Tresiba and will use it to market Victoza and the modern insulin portfolio. To our knowledge, Novo Nordisk hired an additional 500-600 full-time employees in anticipation of Tresiba’s launch, increasing its US sales force to roughly 2,700 people.

IMPLICATIONS FOR THE DIABETES ARENA

The FDA’s CRL provides a glimpse (albeit blurry) into the Agency’s perspective on CV safety, giving us significant food for thought. Below, we discuss the broader implications for insulin development and diabetes drugs.

  • The FDA’s decision suggests a rising standard for CV safety. While the FDA is certainly aware that its decisions influence drug development, the CRL made us take pause on whether the FDA places enough consideration on the long-term implications for diabetes drug development. While it is too soon to say whether pre-approval CVOTs will become standard for insulin therapy, an inclination toward such a requirement could discourage insulin development, especially for smaller companies. Novo Nordisk’s situation makes a strong, likely lasting, impression: despite conducting the largest development program for an insulin therapy and despite its expertise and history in diabetes, the company appeared to be caught off-guard about the FDA’s CV evaluation, a risk that smaller companies may not be willing to take. We’ll be curious to see whether the FDA’s decision regarding a new drug class such as SGLT-2 inhibitors (J&J’s canagliflozin is currently under review) also reflects a higher bar for CV safety. In ignoring its advisory panel’s advice, future committee votes are likely to be more conservative irrespective of the metabolic indication under consideration.
  • We’re interested in the FDA’s focus on CV risk, given that the EU and Japanese regulatory authorities did not voice strong concerns when approving Tresiba. Novo Nordisk has consistently remarked that the FDA reviewed the same materials as the regulatory agencies in the EU and Japan (where Tresiba has been approved). While the FDA is presumablymore cautious about CV safety due to Avandia, this concern is not unique to the US: Avandia has been discontinued in the EU while its use is heavily regulated in the US.
  • We’re curious whether the perception of Tresiba as only an “incrementally-better” drug influenced the FDA’s risk tolerance – i.e., would the FDA have allowed this CV risk profile for a “next-generation” insulin? The advisory committee meeting on CV guidelines for obesity drugs included discussion on using a “sliding scale” for risk acceptance, where the drug’s efficacy determined the stringency for safety (drugs with greater efficacy could gain approval with a worse safety profile than drugs with lower efficacy). While advantageous in its flexibility (and leniency in getting better drugs to patients sooner), this approach would require greater dialogue with the FDA so that sponsors do not have to “guess” at the safety threshold that corresponds with their drug’s efficacy.
    • While Tresiba provides only incremental progress over Lantus, its approval in the EU and Japan indicates that regulators see value in its benefits. We share this perspective, with three main thoughts: 1) the suboptimal adherence and outcomes associated with insulin therapy (50% of basal users are not at the 7% A1c goal) indicate the need for even incremental improvement – indeed, there is a lot to be said for moving patients in the right direction; 2) Tresiba’s advantage may likely be magnified when it is combined with better fast-acting insulins or with oral medications (i.e., IDegLira; discussed above); and 3) waiting for a “next-generation” drug sets a higher regulatory bar that could promote innovation, but would also make drug development more risky.
  • During Q&A, management remarked that the FDA’s decision suggests that the Agency is becoming less willing to exempt insulins from the CV guidelines that regulate oral antidiabetic medications (and which requires drugs to rule out a pre-specified degree of CV risk before approval). To our knowledge, the FDA has historically held insulin to a different (in many views, lower) standard than oral medications because insulin is a physiologic replacement. The key question is where the standard for insulins now lies – we note that the upper limit of the 95% confidence interval was close to 1.8 for the meta-analysis of MACE+ outcomes (upper limits of 1.77 and 1.93 for the original and updated analysis, respectively), but was much higher for the strict MACE analysis (upper limits of 2.57 and 2.75 for the original and updated analysis, respectively).
    • For background, Novo Nordisk and the FDA previously agreed that Tresiba’s phase 3 program would not be powered rule to out a pre-specified margin of CV risk; however, at the end-of-phase-2 meeting, the FDA asked Novo Nordisk to prospectively analyze its phase 3 trials using the process outlined in the 2008 CV guidance for diabetes drug. The decision to focus on CV safety during the advisory committee came as a surprise, and to our knowledge, Novo Nordisk was not aware of the FDA’s approach in interpreting the CV safety signal – the signal itself is interesting in that the data is inconclusive (the confidence intervals for the meta-analysis were quite large) and the signal was not accompanied by corresponding changes in CV risk factors such as blood pressure and lipids.
  • Safety analysis aside, the apparent unpredictability of the FDA’s CV assessment bodes poorly for future insulin developers. Higher pre-approval safety standards may increase the resources needed for drug development, but could also benefit the sponsor (and of course patients), since greater safety data supports commercialization and since such resources would likely have been devoted to safety trials following an approval. In contrast, poortransparency on the part of the FDA increases the risk of drug development, with little advantage to any party. We’re curious whether greater transparency and dialogue during Tresiba’s drug development could have led to more conclusive CV data, which would have benefited the FDA and Novo Nordisk, and of course HCPs and patients.

BACKGROUND PERSPECTIVE

  • The CRL came as a surprise given that the endocrinologic and metabolic disease advisory committee voted 8-4 in favor of Tresiba’s approval during the November 8 meeting. While the panelists acknowledged the CV risk and thus voted 12-0 to require a CVOT, most panelists felt that a post-approval trial was appropriate given that degludec’s benefits (i.e., lower hypoglycemia and a more stable action profile over existing options) appeared to outweigh the incremental reassurance a pre-approval CVOT would provide. Of note, all four panelists who voted against approval (including two of the three cardiologists at the meeting) cited potential cardiovascular risk as the basis for their decisions. For perspective, the FDA last disagreed with the committee’s positive vote in 2010 (for Orexigen’s Contrave) and Novo Nordisk management remarked during the call that roughly 90% of drugs that receive a positive advisory committee vote subsequently gain FDA approval.
    • As a reminder, the advisory committee meeting centered on the true risk of the apparent CV signal (discussed below), with the panelists generally accepting the sponsor’s viewpoint that Tresiba’s CV risk was inconclusive – in its presentation, Novo Nordisk emphasized that the phase 3 program was not designed to assess CV safety and that the number of patients and CV events in the meta-analysis was low (only 12.7% of the initial 6,389 patients enrolled in the degludec and degludec/aspart arms and 7.7% of the initial 3,461 patients enrolled in the active comparator arms remained under examination at the end of full two years in extension studies). For full details on the CV discussion between Novo Nordisk, the FDA, and the panelists, please see Appendix A, which includes our CV-related coverage from the advisory committee meeting.
    • During the call, management pointed out the absence of a plausible biological mechanism that could explain why Tresiba would have CV effects that differ from other insulins.
  • During the advisory committee meeting, Novo Nordisk proposed a design for a post- approval CVOT that would randomize 7,500 type 2 patients to either degludec or another insulin, both added to standard care. The proposed primary endpoint was time until first MACE event (CV death, non-fatal MI, or non-fatal stroke). The post-approval CVOT was expected to last five years given the anticipated event rate of 2% per year (which at the time seemed on the fast side, as the event rate observed in Tresiba’s phase 3 program was roughly 1%).

QUESTIONS AND ANSWERS FROM THE CALL

Q: Regarding the potential CV outcomes trial, I think the LEADER trial was a trial that was going to take about five years and had over or around 8,000 patients, with an interim analysis potentially around three years. Could you give us some sort of framework to help us understand the potential design? I know you haven’t talked to the FDA yet, but what is the potential design for an outcomes trial? Can you also comment on whether the FDA would like you to rule out the confidence interval of 1.3 in that trial before approval or 1.8? Obviously that would have implications on the size and duration of the trial.

A: The LEADER trial as you know is a large-scale post-marketing required trial of the classic CV trials that are done for diabetes products once they’re on the market. Hence, it has to rule out the upper level of the confidence interval of 1.3 with a 95% likelihood. That is in a post-marketing situation. I will not, at this point, discuss the upper level specifically for this trial because we’ll first need to discuss and agree on the appropriate design of the trial with the agency. That said, the execution of such a trial takes significant amounts of time, not necessarily the five years you’re mentioning that relates to a confidence interval of 1.3. But nonetheless, setting up and starting recruitment for such a trial will alone take at least nine months. Therefore, we do not expect to deliver the requested data during this year. The duration of the trial, as I mentioned, of course, depends on the design and the confidence level that must be achieved, pending discussions with the agency. But currently, it seems unlikely that we’ll be able to generate the data also in 2014.

Q: What are your thoughts on what other regulators may do with the information that the FDA is requiring from this trial? What would they perhaps communicate to their doctors and physicians?

A: I think the agency, according to the Complete Response Letter, will need the data to become less concerned or not concerned about what they perceive to be a cardiovascular signal. This is unlike the regulatory authorities in other countries such as Europe and Japan that have seen the same data but have not seen this cardiovascular signal.

Q: Regarding the sales force, you mentioned that you would just focus your efforts around existing products. Just to get a confirmation – are all the reps you have hired for Tresiba in the US going to remain intact? And hence, will you have significant pressure on your sales force to market other products?

A: Yes, we intend to continue to put pressure on our modern insulins in the United States. We have a very strong momentum behind Levemir, our long-acting modern insulin, which you saw in the fourth quarter of last year. So this will be continuing with full force.

Q: On IDegLira in Europe, you do mention that you don’t see any effects outside of North America. So should we still expect you to be able to file in Europe around midyear for IDegLira in Europe?

A: Yes, there’s no plan for changing that timeline. As you know, Tresiba is approved in Europe and hence we will move along with the filing of IDegLira as indicated earlier.

Q: On the regulatory requirements for new insulin products, do you see this development as part of a trend and heralding additional new safety requirements for all future new insulin products in the US market?

A: I do not see that it’s a trend outside of the US. I think that outside of the US, there is the notion that insulin is insulin and the kind of data that have been provided by us on all aspects, including adjudicated cardiovascular instances – one from a diabetes-oriented phase 3a trial program – suffices, probably also in the future.

In regards the US FDA, it is indeed an observation – as I noted during my conference call a few minutes ago – that the Agency clearly is not satisfied with the degree or amount of data that comes from such a program as we have delivered. So I could foresee that the US regulators are maybe shying a little bit away in the future – but this is pure speculation – from exempting the injectable insulins from the CV outcome guidelines that normally do require that you rule out the 1.8 upper limit of confidence interval pre-registration. This has not been demanded for the insulins as per the guidelines. That could change, but you should probably pose the question to the US Agency.

Q: To follow up from a question earlier, regarding the post-marketing trial that you had committed to at the advisory committee meeting – what was the duration of that trial and the size that you were committing to?

A: Yes the post-marketing study that we suggested at the advisory committee meeting on November 8 was related to the post-marketing situation where you are to rule out the 1.3 upper confidence level of risk. And that included, henceforth, approximately 9,000 patients studied for up to five years. We suggested this as a post-marketing requirement. There is of course a difference depending which level of uncertainty you rule out pre-authorization. But those were the numbers we presented at that point. [Editor’s note: our report from the advisory committee meeting indicates a proposed study population of 7,500 participants]

Q: Rightly or wrongly, the market’s probably going to have some sentiment read across regarding the Victoza and obesity, the requirement for pre-approval study, etc. Can you update us on your thoughts and timeline on that?

A: We have been, on multiple occasions, in dialogue with the FDA about how this GLP-1 agonist should be analyzed regarding cardiovascular outcomes. And it is a question of basically needing to analyze all the pre-existing data from the LEAD program as well as the SCALE program, plus of course phase 2b studies and other activities that have been ongoing for Victoza and liraglutide both before and after its current marketing authorization. And I think there is additionally, of course, the opportunity to potentially interim analyze the existing LEADER trial which is destined to complete around 2016. But there is a plan with the agency and also a plan that does not require specific pre-approval CV outcome activities, as I know we’ve discussed before, and this is what we expect to adhere to.

Q: Under a best-case scenario, what’s the quickest time you might be in a position to generate the data the FDA required regarding degludec?

A: We will not comment specifically on the timelines. We have to discuss that further with the agency. But we’re not speaking the timelines of a full-blown LEADER trial as the one we are conducting right now, which will complete only after a total of about five years. So as we have learned today, there are still uncertainties about how the regulatory agencies are going to react. So inasmuch as we currently don’t believe there will be a pre-approval CV requirement for obesity, as we have learned, things are uncertain. So please do bear that in mind.

Q: Can you just give us a sense for the CV study that is being required? Should we think of it as a study similar to what alogliptin had, which was a full kind of scale phase 3, or should we think of it as the one that Contrave did where interim analysis from the study was okay to meet the FDA requirements? So just based on your understanding of what the FDA is asking for today, can you give us a sense for that?

A: I think again here, we just have to say that it's a little bit premature for us to speculate on what the outcome of the debate will be with the agency on the cardiovascular study that is required because we haven't had the conversation yet and the CRL does not give us any specific guidance on how to do this. But we would think – if you ask us today, it's probably closer to the kind of cardiovascular trial that Contrave has been doing.

Q: As it relates to the Warning Letter, can you give us a sense of where you stand with that? Have you scheduled a re-inspection with the FDA for that?

A: With regards to the Warning Letter, we've had no further dialogue with the agency on our submission, which was made on December 28 that we have communicated to all of you during our recent road show.  And so we are of the belief that we have responded adequately to the two observations. But this is of course something that is in the hands of the agency. Should the agency want to reschedule a re-inspection, I guess we now have plenty of time for that.

Q: Regarding the impact on Europe, do you think that the negative outcome of US can have any impact on your pricing and reimbursement discussions in Europe?

A: The answer is no. There will be no impact on the pricing. There'll be no impact on our plans. We still believe that Tresiba holds a significant benefit to patients and society and hence, our pricing in Europe is still predicated on the same assumptions. So there is no impact, neither in Europe nor in Japan.

Q: If you had known this two weeks ago, would you still have given the same long-term targets to the market?

A: In regards to the long-term financial targets that we provided at the end of January, it is – as you know -- providing a financial ambition level for Novo Nordisk on a three to five year horizon. Inherent in that ambition level is the expectation of an underlying double-digit growth level in sales measured in local currencies. We now have a significant delay in the potential US launch of Tresiba. It will of course, everything else being equal, make the achievement of the long-term financial targets a bit more challenging. However, as we haven't yet had the discussion of what additional clinical activities that need to take place in the US with the FDA, we don't think it's meaningful for Novo Nordisk to make any adjustments, if required, to the financial ambition levels.

Q: Obviously, degludec was planned as a protection strategy in the long-acting space against potential biosimilars, i.e., Lilly/Sanofi. I wonder if you could discuss how in the short term you view Levemir’s competitive profile. Obviously it's growing well recently, but in the longer term, what other strategic options do you have on the table as you think about your long-acting segment?

A: We believe that Levemir has been doing quite well in recent quarters, and that will be our expectation, to continue. Of course, the situation gets a little bit more fluid when we get around to the point where Lantus goes off patent and where Lilly might be marketing a biosimilar version of Lantus.

We still think that Levemir is a good alternative and we of course hope to finalize the issue of cardiovascular safety surround Tresiba with the US regulators as quickly as possible obviously. And so until we know what that timeline looks like, it's a little bit difficult for me to speculate at this point in time. So you have to pardon me. I can't give you a good answer for how things look in 2015 at this point in time. But once we've talked to the agency, once we have the plan for how we are going to do this, then we'll give you an update.

Q: You mentioned that the lost revenue here in 2013 from Tresiba in the US will be offset by a cost-cutting endeavor but not impacting operating profit. But I think you have hinted a lot during the years that launching new drugs is a net loss in year one. So will you increase spending outside the US?

A: Yes, the expectations that we provided for sales were of course providing a range, and we were assuming that if we were at the high end of the range, we would launch Tresiba early in the year, and if we were at the low end, it would be later in the year, as we were aware toward the end of January that there was uncertainty as for the timing of the Tresiba launch in the US. Consequently, the profit impact will not be significant, and we will still maintain the expanded US sales force that was hired in 3Q12. And hence, the overall implication, as I said, on operating profits is largely very limited from the delay in 2013 of Tresiba in the US.

Q: In this safety trial, will you enroll both type 1 patients as well as type 2?

A: Again, we have mentioned a couple of times that we will be in immediate discussions with the agency about the specifics of that trial and then we will get back to all of you as soon as possible, both on the design and the exact proposed timing.

APPENDIX A: CARDIOVASCULAR SAFETY DISCUSSION FROM CLOSE CONCERNS’ ADVISORY COMMITTEE MEETING REPORT

  • As a reminder, the phase 3 program for degludec was commenced under Novo Nordisk’s agreement with the FDA that degludec and degludec/aspart would not be subject to the FDA’s cardiovascular (CV) guidelines for type 2 diabetes drugs, but that Novo Nordisk would still collect and analyze CV safety data from its phase 3 program. CV safety data submitted at the time of degludec’s NDA submission included 16 phase 3 trials including one completed extension trial. These data demonstrated a 10% increased risk (HR 1.10; 95% CI 0.68-1.77) for cardiovascular death, non-fatal stroke, non-fatal myocardial infarction, and unstable angina (MACE+). The FDA, concerned about a potential CV signal, requested additional data and conducted post-hoc analyses on an updated dataset including the addition of one more phase 3 trials and six more extension studies. The FDA’s post-hoc analysis using the updated dataset found a 1.30 hazard ratio (95% CI: 0.88, 1.93) using MACE+ as the endpoint and a 1.67 hazard ratio (95% CI: 1.01, 2.75) using the FDA’s traditional MACE (MACE+ excluding unstable angina) as the endpoint. Panelists of the advisory committee discussed the reliability of the CV risk assessment based on the original meta-analysis compared to the analysis performed on the updated dataset, the validity of the two endpoints used, and whether the CV safety signal identified represented a clinical concern. For more background please see our preview at: http://www.closeconcerns.com/knowledgebase/r/685e2535.
  • In discussing the datasets used in the CV meta-analyses, speakers representing Novo Nordisk contended that no conclusions about CV risk could reasonably be drawn from the data for a number of reasons. First, the degludec phase 3 program was designed to assess glycemic efficacy, not CV safety, so the trials were not long enough, large enough, or properly powered (through enrollment of a CV-risk-enriched population to allow for sufficient accrual of events) to provide reliable safety data. Secondly, the number of patients remaining in trials at the end of two years was very low: only 12.7%of the initial 6,389 patients enrolled in the degludec and degludec/aspart arms and 7.7% of the initial 3,461 patients enrolled in the active comparator arms remained under examination at the end of full two years in extension studies. Therefore, Novo Nordisk argued that the small number of patients and events, along with the unequal exposure of patients under review, made CV safety analyses highly questionable. Further, enrollment in the extension trials required re-establishing consent after degludec treatment continuation, potentially compromising randomization since patients had a choice of whether to continue in the extension trials. Lastly, enrollment in the extension trials was also preceded by use of NPH (to wash out degludec use and test for immunogenicity) followed by another degludec titration period.
  • The FDA believed the validity of the data used in its meta-analyses were supported by a variety of compelling factors: the strategy for CV risk evaluation was prospectively planned and thoroughly carried out; the extension trials added to the updated dataset were originally agreed-upon to be included in the initial analysis (but were not initially included because at the time of NDA submission they had not yet been completed); the extension trials were controlled and applied the same methodology and rigor as the main trials for the purpose of providing long-term safety data; and retention through the end of extension studies was high(~70%). Furthermore, the updated dataset increased the patient-years of exposure in the analysis from 5444 to 7716 (a 42% increase) and the number of CV events by 60%. Finally, the trend towards CV harm held up when the analysis was performed across different endpoints (MACE vs. MACE+), using different analysis methods (time-to-event, risk difference, and incidence rate difference analyses were performed), and across different subgroup analyses (patients with type 1 diabetes*, patients with type 2 diabetes, insulin glargine-controlled trials only, and trials with only degludec and no aspart).
  • Most EMDAC panelists were skeptical about the authenticity of the cardiovascular signal given the very low sample size and event rate, but also maintained that the potential for harm could not be excluded. Panelists generally agreed that including more data by adding the extension trials to the analysis was appropriate, but erred on the side of caution that data should be interpreted carefully due to the low sample sizes. Thus, not surprisingly, panelists agreed that a definitive conclusion about cardiovascular risk could not be drawn from the available data and would have to wait until a larger, dedicated cardiovascular outcomes trial (CVOT). One Kaplan Meier curve for MACE+ events drew significant attention, as it showed that the degludec and comparator curves were virtually superimposable until ~week 70 when events suddenly stopped occurring in the comparator group for roughly 20 weeks while events continued to accrue at a steady pace in the degludec and degludec/aspart group. Several panelists were skeptical of this curve separation because at week 52, 39% of the initial 9,214 patients remained under investigation, which dwindled to just 12% at week 104. However, most panelists maintained that safety data had to be interpreted much more liberally than efficacy data and that no safety signal should be ignored.
    • Some panelists questioned the lack of a potential mechanism for why degludec would increase CV risk compared to glargine – the associated CV risk factors that were measured [lipids, blood pressure, and ECG] did not track the trend for increased MACE+ risk in the degludec and degludec/aspart group. Such an understanding why degludec might elevate CV risk would of course help in drawing a conclusion about the risk’s clinical relevance. Since Novo Nordisk repeatedly highlighted that the only functional difference between degludec and human insulin was the long pharmacokinetic profile, Dr. William Hiatt implied that that might offer a place to start looking for a mechanism for why there might be a CV signal.
  • Another point of discussion was whether appropriate endpoints were used in the meta-analyses for CV risk; Novo Nordisk contended that the FDA’s post-hoc redefinition of the MACE endpoint to exclude unstable angina skewed the analysis towards harm. Dr. Steven Marso defended Novo Nordisk’s use of the MACE+ endpoint, stating that it had been pre- specified after discussions with the FDA, that each component of the endpoint had been rigorously defined, that the adjudication was performed shrewdly, and that including unstable angina was appropriate given that it is involved in a causal pathway for myocardial infarction; the panel solidly agreed on all fronts. Additionally, the panel agreed that the FDA’s decision to also examine MACE was warranted to potentially unmask any effects of including unstable angina in the composite endpoint. Novo Nordisk proposed a post-approval CVOT: 7,500 patients with type 2 diabetes randomized 1:1 to degludec plus standard of care or another insulin plus standard of care, will be followed for five years. The proposed primary endpoint is time until first MACE (CV death, non- fatal MI, non-fatal stroke) with the trial enrolled to produce an ~2.0% event rate per year (for context, the event rate seen in the degludec phase 3 program was about 1%event rate may be “too low”).
  • Ultimately, the 8-4 vote to approve degludec and degludec/aspart with a post- approval outcomes trial was based on the panelists’ belief that requiring a pre- approval CVOT would delay bringing a meaningfully improved basal insulin to market. Those voting “yes” typically noted that the benefits of a truly once-a-day basal insulin for patient quality of life and adherence outweighed the incremental sense of assurance of ascertaining cardiovascular safety prior to approval. Those voting “no” maintained that, with the current state of uncertainty, the potential to unnecessarily expose patients to this risk outweighed the benefits of the drug beyond available basal insulins. Interestingly, two of the four who voted “no” were cardiologists, one was a biostatistician, and one an endocrinologist. The other five of the total six endocrinologists voted “yes.”
  • We are curious whether this situation will set a precedent for other novel insulins in development or whether it will prompt the FDA to formulate a specific guidance for pre-approval safety requirements for insulins. Even though Novo Nordisk designed its degludec program in compliance with the FDA’s guidance for CV safety, at some points today it seemed like a very real possibility that the panel might vote in favor of a pre-approval CVOT, which could have dramatically changed the course of insulin development for other candidates being developed under the same assumption that they would not have to adhere to specific CV safety guidelines. In the future, we hope that, the FDA can offer clear, consistent, and predictable guidance for companies developing novel insulins, especially on the level of risk that should be considered acceptable. This way, companies can be prepared to address the CV safety issue from the start of clinical development and get new options to patients as efficiently as possible.

APPENDIX B: CARDIOVASCULAR META-ANALYSIS FROM THE FDA BRIEFING DOCUMENTS

  • The briefing documents acknowledge that the FDA and Novo Nordisk previously agreed that the clinical programs for IDeg and IDegAsp would not be powered to rule out a pre-specified margin of CV risk – this information is consistent with previous comments from Novo Nordisk management. The FDA documents do state that at the end-of- phase 2 meeting, the FDA requested that Novo Nordisk collect, analyze, and adjudicate CV data from the clinical trials, using the process outlined in the FDA’s 2008 guidance on assessing CV risk for diabetes drugs (details of the guidance are available in our July 2, 2012 Closer Look at http://www.closeconcerns.com/knowledgebase/r/ac8baf71).
    • Novo Nordisk has previously noted that the safety data provided to the FDA is the same as that given to the regulatory agencies in the EU and Japan. we assume that the favorable opinions given by these respective agencies demonstrates the agencys’ confidence in the CV safety of IDeg and IDegAsp. As a reminder, Novo Nordisk received an approval for both insulins in Japan in September and received a positive opinion from the EMA’s Committee for Medicinal Products for Human Use (CHMP) in October, with a final decision from the European Commission expected by the end of 2012.
  • The FDA briefing documents describe the design of the CV meta-analyses for insulin degludec (IDeg) and insulin degludec/aspart (IDegAsp). At the time of the original NDA filing, Novo Nordisk submitted a CV meta-analysis that included 16 phase 3 trials (11 for IDeg and five for IDegAsp) with data from ~9,000 patients and 5,400 person-years exposure (PYE). According to the FDA, “Results from the initial analysis of the CV safety suggested the potentialfor an increase in CV risk, though a statistically significant finding could not be determined,” prompting the agency to request additional information to perform an updated meta-analysis. The updated analysis included the 16 original trials plus six extension trials (of ongoing phase 3 studies) and one new phase 3 trial – the second analysis evaluated data from a similar number of participants (~9,000; this is expected given that only one new trial was added), but included a 40% increase in PYE (now ~7,700). The primary endpoint was a composite of major adverse cardiovascular events consisting of acute coronary syndrome (including unstable angina pectoris), myocardial infarction, stroke, and CV death – collectively referred to as “MACE+”. Novo Nordisk also conducted an analysis with a stricter endpoint definition consisting only of myocardial infarction, stroke, and CV death – collectively called “MACE”. Information on CV events was censored either seven days or 30 days after treatment discontinuation.
  • At first glance, both the original and updated meta-analyses suggest an increased CV risk with IDeg and IDegAsp though we aren’t sure how to weight (or even interpret) the data given the confidence intervals and given that many fewer patients are included in this analysis compared to the entire set of trials. Based on the MACE+ primary endpoint and data censorship after seven days, the original analysis found a 10% increase in CV with IDeg/IDegAsp compared to active comparators. Notably, however, the 95% confidence interval around the point estimate was extremely wide and indicated that IDeg/IDegAsp’s true CV effects could range from a 32% risk reduction to a 77% risk increase. The updated analysis showed a less-favorable outlook, indicating a 30% increase in CV risk with IDeg/IDegAsp compared to comparators. The 95% confidence internal was again quite large, indicating that the insulins’ true CV effects could range from a 12% risk reduction to a 93% risk increase. A similar statistical analysis based on the MACE endpoint found an increase in CV risk with IDeg/IDegAsp of 39% in the original analysis (95% CI: 0.76- 2.57) and of 67% in the updated analysis (95% CI: 1.01-2.75).

Table 1: Cox proportional hazards analysis results for MACE+ endpoint based on original and updated meta-analyses

 

Original Analysis

Updated Analysis

 

IDeg/IDegAsp

Comparator

IDeg/IDegAsp

Comparator

Censoring after seven days

MACE +

Events

 

53

 

27

 

95

 

37

MACE +

HR (95% CI)

 

1.10 (0.68, 1.77)

 

____

 

1.30 (0.88, 1.93)

 

____

Censoring after 30 days

MACE +

Events

 

56

 

27

 

99

 

39

MACE +

HR (95% CI)

 

1.17 (0.73, 1.87)

 

____

 

1.29 (0.88, 1.89)

 

____

Table 2: Cox proportional hazards analysis results for MACE endpoint based on original and updated meta-analyses

  Original Analysis Updated Analysis
 

IDeg/IDegAsp

Comparator

IDeg/IDegAsp

Comparator

Censoring after seven days

MACE

Events

 

39

 

15

 

70

 

21

MACE

HR (95% CI)

 

1.39 (0.76, 2.57)

 

____

 

1.67 (1.01, 2.75)

 

____

Censoring after 30 days

MACE

Events

 

42

 

15

 

74

 

23

MACE

HR (95% CI)

 

1.50 (0.82, 2.75)

 

____

 

1.61 (1.00, 2.61)

 

____

  • The high degree of additional CV risk observed with IDeg/IDegAsp would be concerning if based on studies that were designed to assess cardiovascular safety and risk; we’ll be very interested to see how the advisory committee views these results given the large confidence intervals and depending what studies were and were not included, etc. While we presume that the degree of acceptable risk is higher with a life-saving drug such as insulin, the important question is whether the advisory committee will find the incremental benefit of insulin degludec (over insulin glargine and detemir) significant enough to balance the potentially increased CV risk. While it is of course difficult to imagine any benefit in hypoglycemia or flexible dosing that would outweigh a true ~30% increase in CV risk, as indicated by the updated meta-analysis (the additional risk rises to 67% when only strict MACE events are considered), we don’t know, of course, how the meta-analysis will be perceived. As a reminder, for all diabetes drugs but insulins, the FDA requires that the upper limit of the 95% confidence interval fall below 1.8 pre-approval and below 1.3 post-approval. Interestingly, insulin degludec nearly meetings this pre-approval criteria in the MACT+ analysis (upper limits of 1.77 and 1.93 for the original and update analyses, respectively), but falls well above the limit for the perhaps more importance MACE analysis. (upper limits of 2.57 and 2.75 for the original and updated analyses, respectively).
  • FDA changed the meeting time from 8 am to 7:30 am, which suggests that FDA expects a great deal of discussion on a great number of topics. The meta-analysis results will undoubtedly prompt panelists to discuss the meta-analysis itself, what was included, how it was designed, and what it showed, and to characterize the extent to which it should be used to discuss whether a CV outcomes trial should be performed. Following the discussion of the meta- analysis, we assume that discussion will center on whether an outcomes trial should be required and whether it should be performed before or after approval. This decision, as noted, will hinge largely on how panelists view the ability of the meta-analyses to accurately characterize CV risk. The first draft question asks the panelists to evaluate the reliability of the meta-analyses and consider the endpoints chosen, the adjudication process, the patient population, and the design of the clinical program. The documents do clearly state the FDA’s conclusion: “While various scenarios resulted in different values of the HR and RD point estimates, a consistent trend was observed – IDeg/IDegAsp was shown to be associated with an increase in CV risk.”
  • The FDA conducted a sensitivity analysis evaluating the CV risk of IDeg/IDegAsp compared to insulin glargine (IGlar). Of the 17 phase 3 trials included in the updated meta-analysis, 12 used insulin glargine as the active comparator. Results based only on these 12 trials are shown below. Based on the MACE+ primary endpoint and data censorship after seven days, IDeg/IDegAsp was associated with a 54% increase in CV risk compared to insulin glargine, with a 95% confidence internal of 0.99 to 2.40. Based on the MACE endpoint, IDeg/IDegAsp was associated with an 82% increase in CV risk compared to insulin glargine, with a wider (and statistically significant) 95% confidence internal of 1.03 to 3.19. In the briefing document, the FDA noted that compared to insulin glargine, the lower bounds of the 95% confidence intervals are near or above 1.00 for both the MACE and MACE+ endpoints.

Table 3: Cox proportional hazards analysis results for MACE+ and MACE endpoints based on the updated meta-analysis including only IGlar-controlled trials.

 

MACE +

MACE

 

IDeg/IDegAsp

(n=4,397)

IGlar

(n=2,540)

IDeg/IDegAsp

(n=4,397)

IGlar

(n=2,540)

Censoring after seven days

Events (%)

87 (2.0%)

27 (1.1%)

62 (1.4%)

16 (0.6%)

HR (95% CI)

-----

1.54 (0.99, 2.40)

-----

1.82 (1.03, 3.19)

 

  • Additional sensitivity analyses indicated a higher CV risk with IDegAsp than with IDeg; for both treatments, the point estimate was higher for the MACE endpoint vs. the MACE+ endpoint. Eleven of the 17 phase 3 trials included in the updated meta-analysis evaluated IDeg alone and showed that IDeg was associated with a 29% increased CV risk (95% CI: 0.83-2.02) for the MACE+ and a 59% increased CV risk (95% CI: 0.89-2.83) for MACE. Analysis of the remaining six trials investigating IDegAsp revealed a hazard ratio of 1.33 (95% CI: 0.59- 2.99) for MACE+ and a hazard ratio of 1.91 (95% CI: 0.72-5.08) for MACE.
  • Cardiovascular event rates were generally lower among people with type 1 diabetes vs. those with type 2 diabetes. Of the 17 trials included in the updated analysis, four were conducted in patients with type 1 diabetes (with the remaining 13 conducted in patients with type 2 diabetes). Sensitivity analysis showed that in people with type 1 diabetes (and with a seven day censoring window), IDeg/IDegAsp provided a similar – if not slightly better – rate of CV events compared to comparators for MACE+ (HR: 0.96 [95% CI: 0.30-3.09]). However, IDeg/IDegAsp was associated with an increased CV risk compared to comparators in the more strict MACE analysis, though with a very large 95% confidence interval (HR: 1.30 [95% CI: 0.27-6.29]). In people with type 2 diabetes, IDeg/IDegAsp was associated with a 35% increase in CV risk for MACE+ (95% CI: 0.89-2.04) and a 71% risk increase for MACE (95% CI: 1.01-2.90).
  • We first learned about the FDA’s focus on IDeg/IDegAsp’s cardiovascular (CV) risk when the FDA recently published a conflict of interest waiver for pediatric endocrinologist Dr. Charles Stanley (Children’s Hospital of Philadelphia, PA). Specifically, the waiver disclosed the reason for the advisory committee meeting in the “additional facts” section, stating: “this meeting will focus on the cardiovascular safety of two products, insulin degludec/aspart and insulin degludec, as meta-analyses of several clinical trials suggest an excess risk for cardiovascular  events with this insulin over comparators.” The manner in which this news was disclosed was quite surprising to us, as it was written on the second page of a conflict of interest waiver rather than through an open and transparent announcement from the FDA.
    • From a broader perspective, this method of revealing information illustrates how, despite efforts to open communication channels, the FDA’s intentions and actions still remain fairly opaque and unpredictable. From a patient perspective, we hope that the FDA puts greater focus on more open communication with companies and the public, as progress would benefit both the agency (who would receive more specific data from sponsors) and companies (who would get a more concrete idea of the FDA’s expectations, allowing them to more wisely invest in drug development).

-- by Nina Ran and Kelly Close