Memorandum

Dramatic FDA panel clears liraglutide for CV approvability, split down the middle on thyroid cancer issue – April 2, 2009

The liraglutide FDA panel was lively to say the least as the panelists dealt with the complex issues surrounding the liraglutide NDA submission. After a good deal of debate and some tempers flaring on all sides, the panel finally voted that the drug was approvable (8-5) based on cardiovascular risk, but it was split 50/50 on the issue of approvability based on thyroid cancer risk. It’s unclear what the FDA will choose to do with this verdict—there were suggestions of six- or twelve-month additional pre- market studies to rule out additional cancer risk (is that possible?), as well as calls for a stringent post- monitoring program instead. It’s clear that the FDA has a difficult decision on its hands, as there was general agreement that an additional trial would not yield much useful information and a post- marketing program would be difficult due to overzealous detection of and treatment for non- threatening thyroid growths. We’ll be watching the agency with interest for their next move and we are curious how long it will take. Novo Nordisk published a press release soon after the end of the meeting and said that timing for US launch would take place after the FDA completes its review. Novo Nordisk has scheduled an analyst call for 2 am EST to discuss the goings on (8 am Copenhagan time).

More broadly, rumors of a class-based carcinogenic effect for long-acting GLP-1 analogs were floating around the room, although the focus was clearly on liraglutide and no data other than liraglutide’s was presented. Dr. Mary Parks, who heads up review of diabetes drugs at FDA, declined answering questions regarding Byetta and other drugs at the press conference following the meeting (see Q&A at end of this report). The main worry for LAR, we would assume, would be if Byetta safety data was not deemed acceptable to use with LAR. – we had understood that Amylin had an agreement with FDA that this was acceptable. As we understand it, two-year cancer studies for LAR showed an increase in benign tumors in male and female rats, and an increase in malignant tumors with LAR at the highest dose in female rats. We would assume the database would still be acceptable given that Byetta was given both by bolus doses and by constant infusion – while we don’t know how much was by constant infusion, we assume there is not a meaningful difference between constantly infused Byetta and LAR. In an interview with the Wall Street Journal after the meeting today, Dr. Parks said the agency looked back at Byetta’s clinical data and didn't see any risk and hasn't seen any increased risk when looking at post-marketing data, although there is only a relatively small amount of post-marketing data. We see this as positive for Byetta and liraglutide both.

Today’s meeting and yesterday’s meeting for saxagliptin have given us some insight into the way the agency plans to use its new CV guidance. There’s a definite preference for the inclusion of high-risk patients in trials—this issue came up numerous times over the past two days, tied closely to the low event rates observed in both the saxagliptin and liraglutide clinical programs. On the plus side, the positive CV rulings for both drugs indicates to us that the guidance is not an attempt to delay all new diabetes drug approvals as some had feared. There was a respect for the needs of patients with diabetes during the discussion today, and this is a huge positive from our perspective. On the negative side, we imagine requiring companies to include high-risk patients in trials will be challenging for companies who are developing drugs to be used early in disease therapy, such as other DPP-4 inhibitors, SGLT-2 inhibitors, etc.

  • The FDA advisory committee voted 8-to-5 that Novo Nordisk’s liraglutide is approvable based on CV concerns. This included an assessment by the committee that the point estimate of relative risk and associated confidence interval fulfilled December’s guidance (all of the upper bounds fell under a value of 1.8). The debate over the issue was fierce at times, and several members of the panel expressed concerns over the low number of CV events seen in the trial. This specific concern was raised yesterday at the BMS panel as well, although it was even more heated today as the liraglutide data appeared to show even fewer CV events than the BMS data despite the inclusion of a greater proportion of older patients and those with more advanced disease. The low number of events made panelists question the accuracy of the company’s statistical findings. In addition, there were concerns raised about the company’s comparisons of liraglutide to placebo (instead of to total comparator) because these analyses did not always meet the FDA guidance and varied substantially depending on the methodology of the statistical analysis employed. However, the panel was mixed on this issue, and decided to use the comparison versus total comparator as the most important analysis. We should mention that the passage of liraglutide on CV grounds would have been less likely had it not completed the NDA by the time of the release of the agency’s guidance—the panel appeared to be more lenient because of this factor.
  • A post-marketing CV trial is required of liraglutide, as we expected, as it did not even approach the 1.3 threshold to avoid this requirement. The data were so clear on this issue that the question was not even posed to the panel for voting.
  • The panel was split perfectly down the middle on the issue of thyroid cancer risk, with six members voting that liraglutide was approvable, six voting that it was not approvable, and one member abstaining. Put objectively, the FDA will now have to take responsibility for the next step for liraglutide. It is not clear to us how the agency will balance the desire of the panel to see liraglutide approved while also satisfying the equally clear demand that the thyroid issue has to be resolved.
    • The panel dismissed a numerical imbalance in the rates of papillary thyroid cancer. In the data reviewed today, there were six cases of papillary cancer (papillary cancer is the most common type of thyroid cancer) in the liraglutide group compared to one case in the placebo group. As the number of patients in the liraglutide group was much larger than that in the placebo group, the difference in cancer rates between the two groups was small, and the panel did not express much concern over this issue. Our understanding is that this type of growth is quite common and generally does not present a significant clinical risk to the patient. The panel voted 12-to-0 (with one abstention) that liraglutide was approvable despite the papillary cancer observed.
    • The panel was nearly unanimous in its decision that medullary thyroid cancer may pose a risk to patients taking liraglutide, and was split 50/50 on the issue of whether to delay liraglutide approval based on this fact. The committee voted 12-to- 1 in support of the statement that animal findings of medullary cancer might be relevant in humans (despite Novo Nordisk’s assertions to the contrary), and voted 6-to-6 (with one abstention) on the issue of liraglutide’s approvability based on this risk. The risk evaluation was based on animal studies of liraglutide, which (in brief) found that administration of the drug at “low multiples of clinical concentrations” produced papillary cancers in rats and mice of both genders. While we understand that papillary cancer is common in aging rats, it was said to be much less common in mice, and a great deal was made of the fact that liraglutide produced tumors in both species ("It is very rare for a drug that has caused tumors in two species, in both genders at clinically relevant exposures to be approved,” said FDA drug reviewer Karen Mahoney."). Despite the fact that monkey and human studies have beennegative for serious cancers, the committee did not feel that there was enough evidence to convincingly prove that the rodent findings were not applicable to humans. What to do with this data? We assume the committee also did not feel there was enough evidence to say conclusively that the rodent findings were applicable to humans. We feel a trial to test this would be extremely hard to enroll and would have to be of an untenable size and duration. Would patients really enter a trial to test whether they would develop thyroid cancer?
    • Contributing to the panelists’ concern about medullary cancer was the lack of a mechanistic explanation for the cancers that passed the scrutiny of the FDA and the panelists. As we understand it, medullary tumors are the third most common of all thyroid cancers (about 5 to 8 percent)1. The theory provided by Novo Nordisk suggested that chronic activation of GLP-1 receptors in the thyroid might cause increased release of the hormone calcitonin, which might either directly cause tumors or serve as a biomarker for tumor growth. For this reason, a great deal was made of patients’ calcitonin levels. From our analysis and comment from experts on the panel, these levels were quite low in most trial patients, and probably received more attention than they deserved. We were grateful for the efforts of panelist and thyroid cancer expert Dr. Michael Tuttle (Memorial Sloan Kettering Cancer Center, New York, NY) in bringing the committee back to more critical questions.
    • Those panelists voting to deny liraglutide’s approval based on thyroid risk were asked to explain their answers. A number of interesting issues came to the fore, and frankly many panelists seemed unsure about their final vote. In general, the dissenter’s rationale centered around one of two issues: first, many panelists felt that some additional data was needed to clarify the risk of thyroid cancer. It’s not clear yet what form any additional studies might take (see comment above about size, duration, and enrollment issues). Secondly, a number of panelists were not sufficiently convinced of the benefits of liraglutide to be certain that these benefits were worth even a small risk of thyroid cancer. Everyone seemed to agree about the numerical efficacy of the compound, but many panelists weren’t certain that it really provided an incremental improvement over currently existing diabetes therapies. Rebecca Killion (Bowie, MD) and endocrinologist Dr. Kathleen Wyne (Methodist Hospital, Houston, TX) tried to convince them otherwise, but it appears they had mixed results.
    • Those panelists voting YES on liraglutide approval seemed to do so mostly because they didn’t believe that additional short-term data would help clarify patients’ thyroid cancer risk. Most felt that a post-marketing trial and monitoring program would be better able to address the issue. Not exactly a comforting position, but we don’t think that the thyroid problem will pose a large public health risk even if the drug is approved as is, provided that patients are carefully monitored. We aren’t sure, that said, how realistic that is.
  • One of the important questions remaining about the thyroid cancer issue is what form pre- or post-marketing trials might take. Those panelists calling for pre-marketing trials were generally in favor of an additional six- to twelve-month study to monitor calcitonin levels and tumor rates, although Dr. Peter Savage (NIDDK, Bethesda, MD) was in favor of a two-year trial. Still, it’s unclear exactly what additional data even a two-year trial would provide, as the rates of serious thyroid tumors are low and there is no reason to suspect more dramatic calcitonin increases than the small changes seen in the clinical program. A post-marketing trial or monitoring period is equally complicated—while thyroid tumors can be detected through elevated calcitonin levels, a high level of the hormone does not necessarily indicate a serious tumor. Several panelists expressed concern that the widespread monitoring used in a post-marketing program might identify small tumors that pose no risk to the patient, and that these tumors might be extracted by surgeons “just to be safe”. This unnecessary surgery could pose a more serious risk to patient health than if the tumors had never been detected. The panelists were visibly floundering to give adequate answers to this question.
  • More broadly, rumors of a class-based carcinogenic effect for long-acting GLP-1 analogs were floating around the room, although the focus was clearly on liraglutide and no data other than liraglutide’s was presented. Dr. Mary Parks, who heads up review of diabetes drugs at FDA, declined answering questions regarding Byetta and other drugs at the press conference following the meeting (see Q&A at end of this report). The main worry for LAR, we would assume, would be if Byetta safety data was not deemed acceptable to use with LAR. – we had understood that Amylin had an agreement with FDA that this was acceptable. As we understand it, two-year cancer studies for LAR showed an increase in benign tumors in male and female rats, and an increase in malignant tumors with LAR at the highest dose in female rats. We would assume the database would still be acceptable given that Byetta was given both by bolus doses and by constant infusion – while we don’t know how much was by constant infusion, we assume there is not a meaningful difference between constantly infused Byetta and LAR. In an interview with the Wall Street Journal after the meeting today, Dr. Parks said the agency looked back at Byetta’s clinical data and didn't see any risk and hasn't seen any increased risk when looking at post-marketing data, although there is only a relatively small amount of post-marketing data. We see this as positive for Byetta and liraglutide both.
  • Overall, these FDA advisory panel meetings have helped us identify some particular areas that the agency seems to be focusing on in applying its new CV guidance:
    • High-risk patients are important! The issue of low CV event rates has received an unbelievable amount of attention in the last few days, and this underscores the importance of enrolling a significant percentage of high-risk patients in any future drug development program.
    • A subgroup CV analysis of only high-risk patients will be looked on favorably. While subgroup analyses are not officially part of the guidance, panel members repeatedly expressed their interest in this data. In particular, they mentioned subjects over 65 years of age, those with an extended duration of diabetes, and those with prior atherosclerotic disease.
    • CV comparisons versus placebo are better than comparisons versus an active comparator, in general. Drugs currently on the market were not approved with an eye to CV risk, so the use of these drugs as comparators might obscure risk from a new agent.
    • Sponsors should run multiple statistical analyses on their CV data. Additional analyses raise panelists’ confidence in statistical outcomes.
    • In terms of criteria for identifying CV events, the ‘custom MACE’ endpoint may be perceived as more important than the wider ‘SMQ MACE’. This is mostly because participants felt that the custom MACE endpoint was more specific, while the SMQ MACE included extraneous items that obscured the true risk.
    • Interestingly, biostatistics expert Dr. Michael Proschan (NIAID, Bethesda, MD) criticized the statistics of the FDA’s guidance. He said that instead of specifying the 95% confidence interval at 1.8 or lower he would rather specify a required 80% confidence interval at a lower value. He seemed to frame this as a sort of semantic change to make theguidance sound better, and it’s not clear what effect it would have on the actual data needed to meet the FDA requirements, if any.

SUMMARY OF VOTING RESULTS—CV

1) Is there appropriate evidence to conclude that liraglutide rules out unacceptable excess risk relative to comparators, including an upper bound less than 1.8? YES: 8 NO: 5

Dr. Timothy Lesar: Yes.

Dr. Peter Savage: Yes. I have big reservations, however, because the design of these studies was weaker than it should have been. I don’t think we should accept studies without high-risk patients in the future. I think there should be some restrictions on the use of this drug until its safety in higher-risk populations has been determined.

Rebecca Killion: Yes. I think that this particular drug was caught in the interregnum period. I’m excited about this drug, and I think that the CV risk can be safely managed.

Dr. John Teerlink: No. I don’t think that public health is fair, and we don’t have the data to do that adequately. There are low event rates, and I’m uncomfortable with other aspects as well.

Dr. Kathleen Wyne; No. I just don’t think there is enough information to reliably answer the question of CV risk.

Dr. Lynne Levitsky: Yes. I think the company met the guidelines, and there will have to be follow-up.

Dr. Michael Tuttle: Yes. The most important for me is that we shouldn’t keep changing the rules, as long as there’s additional follow-up.

Dr. Eric Felner: Yes. The sponsor came into this after the new guidance, and they did everything they were supposed to do. I think this drug has great potential.

Dr. Kenneth Berman: Yes. I think the total comparator comparison was within the guidance given by the FDA.

Dr. Katherine Flegal: Yes. I based my vote on the total comparator.

Dr. Michael Proschan: No. I don’t think that they ruled out unacceptable cardiovascular risk.

Dr. Jessica Henderson: No. I wasn’t sure that the risk was ruled out.

Dr. Marvin Konstam: No. I need tighter confidence boundaries to show that there’s not unacceptable risk. Because there are so few events in this study, I don’t think you can say that with confidence. I just want to say that I would be willing to accept this level of risk if it truly represents a breakthrough. Is there a way of incorporating the existing data to improve the assessment of risk?

Dr. Michael Proschan (in response): Maybe, it depends on how your prior data is formulated. I think that is worth investigating.

SUMMARY OF VOTING RESULTS—THYROID

1) Do the available data on thyroid C-cell tumors show that this finding is not relevant to humans? Yes: 1 No: 12

Dr. Timothy Lesar: No.

Dr. Peter Savage: No. The animal data was worrisome, and not convinced that adding drug with unknown risk outweighs the tradeoffs. This drug has the potential to impact a lot of people.

Rebecca Killion: No. We just don’t know—I was not clear about these findings. With further study, we will be able to get better clarity on this issue.

Dr. John Teerlink: No. I’m not convinced that what we know about human carcinomas translates to drug-induced carcinomas.

Dr. Kathleen Wyne: Yes. From a scientific perspective, I felt that the data did not suggest that we’d see the same effect in human c-cells as we do in rodent c-cells.

Dr. Lynne Levitsky: No. There’s enough concern about the rodent data that young population might have problems with long exposure.

Dr. Michael Tuttle: No. I don’t think we can comfortably rule out a risk in humans. I wouldn’t be terribly surprised if people developed hyperplasia or cancer years down the road.

Dr. Eric Felner: No. I think this question is ‘unanswerable’. As far as pediatrics, it’s easy to check for thyroid problems, but it doesn’t sound as clear in adults.

Dr. Kenneth Berman: No. The question of possible progression to cancer still exists. The human clinical studies do not address the issue of cancer adequately. I think our goal is to err on the side of caution.

Dr. Katherine Flegal: No. You can’t comfortable rule out some relevance to humans, and there’s an increased risk of unnecessary screening.

Dr. Michael Proschan: No.

Dr. Jessica Henderson: No.

Dr. Marvin Konstam: No. It’s really unknown. Even a clear mechanism wouldn’t reassure me.

 

2) Assuming the remainder of the risk : benefit data are acceptable, do the available data on thyroid C-cell tumors permit approvability of liraglutide? YES: 6 NO: 6 ABSTAIN: 1

Dr. Timothy Lesar: No. I do not believe that the data strongly support risk, but it’s unknown. Also, it’s unclear how we would follow these people and monitor for the effects of this drug.

Dr. Peter Savage: No. I just don’t think we have enough data to be reasonably confident about safety in humans for long-term use. The other issue is that I don’t think it’s the only way to get blood sugar under pretty good control. I haven’t heard enough about this drug’s unique advantages [and if they] outweigh the disadvantages of cancer or screening, which would be out of our control after it went out on the market.

Rebecca Killion: Yes. There’s clearly arguments on both sides, but we’re in that area of the known unknown. I think there are work-arounds here and that the risks are manageable. The risk management probably shifts to the patient and the doctor.

Dr. John Teerlink: No. I was tempted to abstain, but I think it’s impossible to extract the process by which these patients will be screened from the issue of the thyroid cancer occurrence itself. This gives us an opportunity to decide how best to screen.

Dr. Kathleen Wyne: Yes. We have a test that can be easily administered.

Dr. Lynne Levitsky: Yes. A pre-marketing test would be long. This made me worry about exenatide a bit, because we’re using it as if it is a long-term agent. I’m not sure we can be sure about whether there is class risk.

Dr. Michael Tuttle: Yes. I can’t think of more data that would convince me of more safety.

Dr. Eric Felner: Yes. I don’t want to miss out on the benefits of this drug.

Dr. Kenneth Berman: No.

Dr. Katherine Flegal: No.

Dr. Michael Proschan: No.

Dr. Jessica Henderson: Yes.

Dr. Marvin Konstam: Abstain.

 

3) Assuming the remainder of the risk : benefit data are acceptable, do the available data on papillary thyroid cancer permit approvability of liraglutide? YES: 12 NO: 0 ABSTAIN: 1

Dr. Timothy Lesar: Yes. I was not convinced there was a risk.

Dr. Peter Savage: Yes.

Rebecca Killion: Yes.

Dr. John Teerlink: Yes.

Dr. Kathleen Wyne: Yes. I think the summary said it best: it appeared to be an ascertainment bias.

Dr. Lynne Levitsky: Yes.

Dr. Michael Tuttle: Yes. For the same reason, and because detection is easy and routine.

Dr. Eric Felner: Yes.

Dr. Kenneth Berman: Yes.

Dr. Katherine Flegal: Yes.

Dr. Michael Proschan: Abstain.

Dr. Jessica Henderson: Yes.

Dr. Marvin Konstam: Yes.

 

Parks: What would the applicant need to do to address the cancer risk?

Dr. Berman: We don’t need much data, because it’s partly an unanswerable question. I would feel assuaged if there was a 6-12 month extension study documenting calcitonin levels and measure other markers of cancer. I’d feel good if they had an occasional sonogram of the neck as well. I realize no study of several years will answer the question definitively, but I don’t feel comfortable exposing the entire population to excess risk without more data.

Dr. Flegal: I think we need to see if calcitonin levels rise or fall.

Dr. Lesar: Some short-term data would be comforting. I’d like to see what the screening options would be.

Dr. Savage: Some longer data would be good, perhaps two years, to make sure this won’t change in a negative direction. I would also suggest longer-term follow-up on the drug that is in use, because maybe something is going on that we haven’t noticed with exenatide as well.

Dr. Teerlink: I think we need to get a good sense of the screening program. You need to see how many unnecessary thyroidectomies are being done. I’m also not convinced we understand the biology of drug - induced carcinoma.

Dr. Prouschan: I would have to defer to the experts on what would be an acceptable trend in these other markers.

 

Questions and Answers from the Press Session:

Q: The panel had a lot of concerns about the carcinogenicity of Byetta. Do you have any plans to require an analysis of Byetta data?

Parks: We did look back and found no cases of thyroid cancer. There isn’t a signal that we can detect.

Q: Can you talk about the precedent for pre-screening with diabetes drugs, regarding family history?

Parks: There are certainly plenty of drugs where prescreening is indicated. It’s certainly not a novel concept.

Dr. Berman: Calcitonin levels are not recommended in the general population, but if a patient has familial cancer, they are a very good predictor.

Q: Can you explain the two types of thyroid cancer?

Dr. Berman: Papillary cancer is most common, then follicular cancer (unusual), then medullary cancer (about 4% of cases), and anaplastic (most aggressive and rare). Medullary cancer comes in a genetic variety, and the gene is measurable in the blood. If they have the gene, they will certainly get medullary cancer. Here we’re talking about non-familial elevations in calcitonin, and much less is known about those. The issue is whether calcitonin is an indicator of progression to medullary cancer.

Q: What’s your takeaway from the votes for today?

Parks: For the first vote, there were a lot of caveats reflecting the quality of the data. The information given will help us advise companies starting and already underway on how best to approach the agency. The vote on the medullary cancers reflects the complexity of the issue. It was difficult to interpret the data and convey the best path forward. Many of the panelists felt that the company would be unable to better determine risk even with additional studies.

Q: How treatable is medullary cancer?

Dr. Berman: Stage 1 is highly treatable, stage 2 is 80-85% survival, stage 3 has 75% survival, and stage 4 has 50% survival (over 10 years). Familial and non-familial cancers work about the same way mortality- wise.

Q: How would you describe your comfort with pancreatitis in liraglutide patients?

Parks: This is being evaluated as well. We don’t have enough clinical experience with lira to compare it to exenatide. We just don’t have enough exposure for me to comment on that risk of pancreatitis. With Byetta the data is from post-marketing.

Q: Might it be part of the post-marketing surveillance?

Parks: I think it’s safe to say that we would like to see additional data about pancreatitis in post- marketing.

Q: What is the risk of having an unnecessary pancreatectomy?

Dr. Berman: You can do a partial or total thyroidectomy, and here we’d be talking about a total. The risks are general anesthesia, hypocalcemia, and damage to the nerve that controls the voice box. With an experienced surgeon those risks are low.

Q: If a patient takes lira for a year and then stops, the person might stop monitoring. Would they still be at risk for cancer later on?

Parks: I don’t think we have that answer. Let’s just say that we don’t have any data in humans with respect to this issue. I think it’s reasonable to assume that we think this is due to a mechanism of the drug that would not proceed if the drug is withdrawn, but I don’t know if there could be additional mechanisms.

Q: Do you share concerns about exenatide?

Dr. Berman: I can’t really comment.


1 According to EndocrinWeb, “Unlike papillary and follicular thyroid cancers that arise from thyroid hormone producing cells, medullary cancer of the thyroi originates from the parafollicular cells (also called C cells) of the thyroid.” See http://www.endocrineweb.com/camed.html for more information.