April 14, 2015; Silver Spring, MD; Full Commentary – Draft

Executive Highlights

The FDA EMDAC’s meeting on the results of the SAVOR CVOT for AZ’s DPP-4 inhibitor Onglyza (saxagliptin) and EXAMINE for Takeda’s Nesina (alogliptin) ended with a set of positive votes: panelists nearly unanimously agreed that both drugs demonstrated cardiovascular safety, and consequently, the labels should be updated to reflect the trials’ findings.

Specifically, the panel voted 13-1-1 that data from SAVOR supported an acceptable cardiovascular risk profile for Onglyza. The fifteen-member panel subsequently voted 14-1 that new safety data from SAVOR should be added to Onglyza’s label, and none (except the consumer representative, whom we do not think received the appropriate education for this role) suggested that the FDA should withdraw it from the market. Although these results are a net positive from a regulatory perspective, Onglyza did not emerge from the process unscathed. Panelists voiced that the 27% statistically significant increase in hospitalization for heart failure seen in SAVOR was concerning (though they weren’t sure it is “real”), and heart failure (HF) represented most panelists’ biggest source of anxiety. AZ, for its part, agreed to adding the heart failure data to the label and suggested that the risk was small overall, mostly limited to identifiable high-risk patients (renal impairment, prior history of HF), and could be managed by clinicians. The company is already beginning a 180-patient, six-month mechanistic study to better understand the heart failure signal (such as it is). Regarding the potential increase in all-cause mortality that emerged from an FDA sensitivity analysis, most panelists characterized it as a nagging doubt rather than a definite issue, especially given the lack of an increase in any one specific category of depth that could shed light on an underlying mechanism.

In the afternoon, Takeda’s Nesina received all positive votes for the same questions on CV safety (16-0) and recommendations for FDA action (13 in favor of adding EXAMINE data to the label, three in favor of no label changes). There was extensive discussion on the heterogeneity of the primary MACE endpoint by geography, and it was also noted that only 16% of patients were enrolled in the US/Canada (well under the FDA’s recommendation for 25%-30%). Panelists expressed some confusion on how to interpret the FDA’s detailed subgroup analyses, which were ultimately deemed nothing more than hypothesis-generating. For their part, Takeda presenters made the case that heart failure should be considered as part of a composite including CV death. Most panelists agreed that the very high-risk patient population with which the trial was completed (15-90 days following an event) was not ideal from a generalizability perspective, but the clear neutrality of the primary endpoint provided sufficient reassurance. Most panelists appeared on the fence or unconcerned regarding whether heart failure is a DPP-4 inhibitor class effect.

TECOS for Merck’s Januvia (sitagliptin) will be presented at this year’s ADA and should provide the most instructive data on heart failure yet; this meeting could have been even more productive if it had happened after we saw TECOS results this summer. We are surprised and disappointed so many resources (over 600 “people days” by our estimates – well over 10,000 hours) went into putting this meeting together. It was a valuable day, but would’ve been more so after such a large data set augments the knowledge on this class’ safety – not much would’ve been lost in a three month delay from our perspective.

Please see below for our themes of the day’s meeting along with more detail on the two sessions’ voting results and panelists’ post-voting opinions.

Themes

• Most panelists felt the heart failure data did not strongly support a DPP-4 inhibitor class effect. Perspectives ranged from posing the question (“is there a class effect?) to “we don’t know” to “there may be a class effect” to “it’s not at all conclusive” to “I have a mild instinct that it’s a class effect” (the latter was endocrinologist Dr. Morris Schambelan). Due to the hesitation over whether or not HF is indeed a class effect, it’s probably that data on each DPP-4 inhibitor drug will be treated separately moving forward. This came as a bit of a surprise given how similar all the drugs are. That said, should such a signal arise in TECOS, it would certainly be more likely that FDA would adopt more of a blanket warning approach for the entire class. The extent to which this could’ve been a chance finding is of course hard to say. The number of comments mostly rose on this front in the afternoon, probably due to the fact that EXAMINE data was being interpreted in the context of the morning’s discussion on SAVOR (where the HF signal was a bit more pronounced than in EXAMINE). Panelists expressed hope that further mechanistic studies could reveal a link, and certainly, all eyes will be on this data once TECOS comes out at ADA. Many wondered whether the high-risk study populations were driving the heart failure finding, since most events occurred shortly after the trial began; some panelists speculated that the HF signal might not emerge in a longer study with lower risk patients (very logical).
• Panelists openly acknowledged the difficulty in interpreting hypothesis-generating sub-group findings in the context of assessing drug safety. None of the safety issues that have emerged from SAVOR or EXAMINE were primary endpoints that the trial was prospectively designed to assess. In the case of heart failure, panelists seemed more open to considering the signal in SAVOR as real because heart failure hospitalization was a prospectively defined and blindly adjudicated event. However, some of the other issues (such as all-cause mortality in SAVOR or the geographic heterogeneity in EXAMINE) came from post-hoc slicing and dicing of the data. Although the FDA suggested that it was not convinced that the signals it saw were due to chance, during discussion, Dr. Hiatt noted that chance findings are still possible in large clinical trials. In our view, for CVOT policy to have a chance at being sustainable, the FDA cannot raise major red flags for every issue that emerges from a post-hoc analysis. After all, odds are that one out of 20 statistically significant findings from a clinical trial will be due to chance when using a p-value threshold of 0.05.
• Although both Nesina and Onglyza received positive votes from the EMDAC panel, there is room for downside (especially for Onglyza) in addition to upside. Both AZ and Takeda will (likely) get to say that their respective DPP-4 inhibitors have outcomes data showing cardiovascular safety, which is more than most other branded type 2 diabetes drugs can say (that said, we do not think there is worry in most cases – generic SFUs aside). It is hard to know whether the finding of non-inferiority but not superiority will have a positive or negative impression on payers, which tend to be focused on improved outcomes. In terms of downside, the DPP-4 inhibitor class is so homogenous and dependent on a pristine safety and tolerability profile that even the fairly modest safety issues that emerged from SAVOR could have an impact. In terms of efficacy there is little reason to prescribe one DPP-4 inhibitor over another, and label additions on heart failure (likely) and all-cause mortality (less definite, but possible) could give prescribers a reason to avoid Onglyza. With Nesina having relatively stronger safety data compared to Onglyza, such results can play out to be a decent marketing line for Takeda. In the context of increasing payer pressure, however, formulary positioning will likely have more of an impact on prescribing than the data from SAVOR or EXAMINE.
• Both committees voted to change the drugs’ labeling, but a major question is what this language should look like. Many panelists cited the need to inform HCPs about this new safety data, but Dr. Lamont Weide pointed out the balance: “I am concerned we will scare the patients. They hear everything we say, everyone on TV, and everything in the package inserts. Then they come back not taking their medicine” – see below for other quotes on this challenge. We thought this was an important point for the FDA to consider, particularly as it thinks about incorporating any sub-group analyses. We assume the label could read something like, “This drug has no adverse effect on MACE; care should be taken in patients with renal disease; and there is a statistically significant risk of heart failure in those with previous CVD, renal impairment or who are older (>70), but the absolute risk is small.”
• Panelists questioned findings on heart failure and renal impairment and whether the trials are truly answering clinically important questions. For instance, Dr. William Hiatt emphasized that SAVOR and EXAMINE answered the question of CV safety, but cannot convincingly answer questions like renal safety – such data requires longer exposure, something these trials are not currently designed to do. It was noted that the population’s high risk of CV death may also have swapped the trial’s ability to see heart failure due to early deaths. In response, the FDA acknowledged that EXAMINE was designed just as the Agency was building comfort with the 2008 CVOT Guidance. It was also good to hear that the FDA is interested in hearing more input on trial design – we hope this was a sincere plea for help, since these trials could certainly use more expert opinion on design and clinical utility. 
  
  • “I think safety is like the word excellence. You strive for it, but you never know if you’re there.” – Dr. Marc Pfeffer (Harvard Medical School, Boston, MA). We thought this quote put the uncertainty of CVOTs and long-term data quite well; a line must eventually be drawn in the sand on when enough data is enough, but it’s hard to know where the appropriate balance lies.
• We did hear consideration of methods other than RCTs for getting answers on some of the questions raised by SAVOR and EXAMINE. The majority of panelists found it reassuring that AZ is already conducting a mechanistic study to explore the heart failure signal from SAVOR. However mechanistic studies do not represent as helpful an option for issues like all-cause mortality in SAVOR, as no single type of death drove the signal, providing no clues on a possible mechanism. Dr. Yves Rosenberg (NIH, Bethesda, MD) in particular drove discussion in the direction of registries and observational studies rather than RCTs. Multiple panelists agreed with the use of observational data, especially in areas such as renal safety for which new RCTs would be too long or costly, since renal impairments take much longer to develop.
• Aside from the highly technical discussion on the data from SAVOR and EXAMINE, we were glad to hear panelists touch upon some bigger-picture questions regarding CVOTs. There was some acknowledgement that two-year trials in highly event-enriched populations with marginal A1c differences between groups are unlikely to provide any meaningful answers on possible cardiovascular benefit (especially since use of statins and other CV drugs was very high, potentially negating any CV effect). It was also widely recognized that the results from SAVOR and EXAMINE cannot be generalized to the patient population with earlier-stage diabetes and without advanced complications. In her post-voting remarks, Dr. Erica Brittain (NIH, Bethesda, MD) put it well when she said that even though SAVOR clearly met the bar set by the FDA’s 2008 Guidance, the trial’s results were still “disappointing” due to the emergence of potential safety issues and a lack of data on long-term CV benefit. During the session on SAVOR, Dr. Thomas Wang (Vanderbilt University, Nashville, TN) noted that the trial’s excess mortality risk is not completely definitive, as little evidence has proven whether this was a true finding or false positive with such a high-risk population. Dr. Lamont Weide (University of Missouri, Kansas City, MO) also shared concerns that the data may be skewed due to the high-risk population. In addition, the morning’s OPH presentations by Dr. Bob Ratner (ADA, Alexandria, VA) and Ms. Kelly Close (diaTribe, San Francisco, CA) questioned the utility of CVOTs due to their lack of representation of low-risk diabetes patients. We were glad to see such a high-level issue brought front and center to the attention of the FDA and hope that this discussion further pressures a revisit to the Guidance, as Dr. John Jenkins (FDA, Silver Spring, MD) suggested at last summer’s FDA public hearing.
• As opposed to EMDAC meetings for drug approvals, in which both benefit and risk are discussed, this meeting only touched upon risk – this led to a somewhat one-sided discussion that we hope does not set a precedent. As industry representative Dr. Mads Rasmussen (Novo Nordisk, Bagsvaerd, Denmark) noted, the actions being considered (i.e. restricting distribution or withdrawing a product from the market) depend on an assessment of both benefit and risk. Non-scientist panelists are likely to be the most adversely impacted by a one-sided discussion. In fact, when explaining her vote in the SAVOR session to withdraw saxagliptin from the market (she was the only panelist to vote for this), consumer representative Ms. Bonnie Arkus noted that no benefits of the drug were mentioned during the meeting and that her vote would have been different if she had seen evidence of a benefit. In any future meetings on CVOT findings, we would hope that the initial presentations include a better summary of a drug’s clinical benefits – DPP-4 inhibitors do offer a lot of value in terms of convenience, less hypoglycemia, once-daily administration, weight neutrality, and relatively few side effects. Generally speaking, some consumer or patient panelists tend to be unprepared or underprepared for EMDAC meetings and we certainly believe that was the case this time.

SAVOR

Voting Question #1

QUESTION: Based on information presented today and in the background materials, do the results of SAVOR demonstrate that the use of saxagliptin in patients with type 2 diabetes has an acceptable cardiovascular risk profile? Explain your rationale and recommend additional studies if you believe these are needed.

• During the vote on Onglyza’s cardiovascular safety, panelists overwhelmingly voted 13-1-1 that SAVOR definitely answered the “rosiglitazone question.” It was good to hear the panel come back to the trial’s success in ruling out excess cardiovascular risk, especially following a discussion that predominantly focused on SAVOR’s more negative secondary findings like heart failure and possibly all-cause mortality. To address those issues, panelists like Dr. Peter Wilson (Emory University, Atlanta, GA), Dr. Rosenberg, and Dr. Nakela Cook (NIH, Bethesda, MD) expressed openness to using observational studies or registries to collect additional data; Drs. Smith and Heckbert stated that they would not recommend additional clinical trials, which appeared reflective of the prevailing opinion. Despite the comfort with saxagliptin’s overall cardiovascular safety profile, Dr. Schambelan suggested that the heart failure signal is probably real, which was in line with what some (but not all) panelists appeared to feel. This vote set the tone for the rest of the votes for the day, which were also overwhelmingly positive.
  
  • In his post-voting remarks, Dr. Hiatt shared that he would be interested in asking more targeted questions about drug safety rather than applying the CV guidance ubiquitously across all diabetes drugs. We thought this remark was spot-on, as it returns the FDA to its core role of assessing benefit-risk and assessing safety on a case-by-case basis rather than using a one-size-fits-all policy. Dr. Hiatt is a member of the core EMDAC, perhaps lending his words more sway with the FDA than if he was invited just for this one panel. Later in his remarks, he also expressed a desire to see a trial confirming that moderate A1c lowering is beneficial in the context of today’s background therapies.

Voting Question #2

QUESTION: Which action do you recommend FDA take regarding the totality of the safety information (Cardiovascular and other) obtained in SAVOR? After the voting is completed, please explain your answer and specify the safety issue(s) of concern, if applicable. Also recommend additional studies if any are needed.

• The panel voted an overwhelming 14-1 to add new safety information to Onglyza’s label, with one panelist (consumer representative Ms. Bonnie Arkus) voting to withdraw the drug from the market. The overall number one reason for adding new safety information was concerns about heart failure, while several panelists also pointed to MACE, all-cause mortality, and renal impairment. Following the morning’s earlier point that higher heart failure risk can be identified in certain patient groups, Drs. Cook  and Wang mentioned that conveying which clinical groups are at risk would be helpful. As we noted earlier, consumer rep Ms. Arkus shared that she heard about no benefits during the meeting, prompting her vote to withdraw the drug. We hope that moving forward, such meetings ensure that non-scientist panelists are receiving adequate preparation and that discussions do not become so one-sided. Please see below for quotable quotes from panelists’ post-voting opinions.
  
  • “There are challenges in how to handle labeling and connecting to actual data. The three issues most significant for me are heart failure risk and while not the highest level of concern, unresolved risks regarding pancreatitis and renal effects. One of the things that makes labeling challenging is that we do know that risk goes up with certain subgroups of patients but risk is not limited to subgroups so this requires some careful thought for labeling.” –Dr. Smith
  • “When thinking about heart failure and what information to add to the label, it’s important to note that hospitalization due to heart failure is a health services outcome, which is different from mortality. How is hospitalization different in China vs. Argentina vs. the US? Does it make sense to include US data or worldwide data in this aspect?” – Dr. Budnitz

EXAMINE

Voting Question #1

QUESTION: Based on information presented today and in the background materials, do the results of EXAMINE demonstrate that the use of alogliptin in patients with type 2 diabetes has an acceptable cardiovascular risk profile? Explain your rationale and recommend additional studies if you believe these are needed.

• The panel unanimously voted (16-0) that Nesina has an acceptable cardiovascular risk profile. Panelists reasoned that EXAMINE fulfilled its primary objective and easily met the threshold set out in the 2008 guidance – this was a fairly expected vote, given the study’s neutral primary endpoint finding. More interesting were panelist comments that questioned the study’s design and general value, as panelists highlighted the limitations of recruiting a high-risk population, the disproportionate geographic distribution of data, and more – see quotable quotes below. More than previous advisory committee meetings, we felt this panel grasped what a major undertaking these trials are. Indeed, many complimented Takeda and AZ for the commitment to conducting the studies.
  
  • “The study design and inclusion criteria may not have been what was best for giving us the answers we want to have for the broad spectrum of diabetes patients.” – Dr. Herkbert
  • “The study met its objective and showed no increase in CV outcomes. I do have concerns on study design and recruitment and I don’t fault the company for that. Perhaps we need to look at what we want to get, how best to get that info, and how to help the companies.” – Dr. Weide
  • “The trial achieved its primary objective in a manner consistent with the 2008 guidelines. But there are some limitations of the trial. An unanticipated low proportion of US participants. A relatively short duration. But those limitations do not come close to negating the conclusion of an acceptable cardiovascular risk profile.” – Dr. Wang
  • “We as a committee need guidance on how to handle subgroups vis-a-vis the 2008 guidance. Otherwise, there are many opportunities to not meet certain criteria.” – Dr. Wilson

Voting Question #2

QUESTION: Which action do you recommend FDA take regarding the totality of the safety information (Cardiovascular and other) obtained in EXAMINE? After the voting is completed, please explain your answer and specify the safety issue(s) of concern, if applicable. Also recommend additional studies if any are needed.

• The panel voted 13-3 in favor of adding new safety data to Nesina’s label, with three votes supporting no change to the labeling. Drs. Cooke, Wilson, and Wang voted no change to the labeling. Again, heart failure seemed to be the primary reason for adding new safety data although there was obviously some debate over whether the data was adequate to address in the label. Renal impairment, all-cause mortality, MACE, and liver effects were also briefly highlighted as concerns. In addition, post-voting opinions highlighted some of the panelists’ thoughts on heart failure as a potential class effect as Drs. Neaton, Schambelan, Weide, and Budnitz all suggested the hypothesis. As we highlighted earlier, this voting question raised the challenge of labeling with regards to the delicate balance of informing providers vs. “scaring” patients and/or providers. Please see below for more relevant quotable quotes.
  
  • “There is this basic question of what should be in a package insert. Should we include any minor but potentially reasonable effect such as some of the concerns brought up today? There should be something more proven to guide clinicians. It’s better to include the information, especially if couched in the proper preliminary tones. CHF has more potential than the other three – MACE for US, renal effect, and liver effect. The latter three should be included but have a very preliminary, cautionary note.” – Dr. Kenneth Burman (Georgetown University, Washington, DC)
  • “One additional point is the important and difficult communication challenge. The tendency is to come away from this meeting and think there is a whole big set of problems with this drug that have to go in the label. But this is a very different position – there is unconfirmed reason for concern, but it is only appropriate that that’s important to communicate to physicians and prescribers. The challenge is to have them understand what that means as they make their own decisions… The simplest reflex challenge is “there’s a problem.” It’s not a problem; it’s something to be careful about. The sponsor can feel like it’s a negative response but it’s more complex than that.” – Dr. Smith

Panelist Voting History and Results

Panelist History

*Key voting considerations: Our impression of the issue(s) that weigh the heaviest on a panelist’s vote.  **Level of influence: How much a panelist’s opinion appears to sway the votes of fellow members. ***Overall voting history: The count includes the results of votes for/against drug approvals at FDA EMDAC meetings since 2008. This includes the following drugs: Orexigen’s Contrave (bupropion/naltrexone), J&J’s Invokana (canagliflozin), AZ’s Farxiga (dapagliflozin – 2013 and 2011), Novo Nordisk’s Tresiba (insulin degludec), Novo Nordisk’s Victoza (liraglutide) and Saxenda (liraglutide 3.0 mg), Arena’s Belviq (lorcaserin – 2010 and 2012), Vivus’ Qsymia (phentermine/topiramate ER – 2010 and 2012), AZ’s Onglyza (saxagliptin), and MannKind/Sanofi’s Afrezza.

Voting Results and Post-Voting Opinion

SAVOR

• 13-1-1 in favor of acceptable CV risk
• 14-1 in favor of changing labeling to add new safety information (1 for withdrawing from market)

EXAMINE

• 16-0 in favor of acceptable CV risk
• 13-3 in favor of changing labeling to add new safety information (3 for no change in labeling)

Note: Any “---“ indicates panelist was not voting in the session, highlighted agreement with other opinions, or did not provide significant additional value in opinion.

-- by Melissa An, Adam Brown, Manu Venkat, and Kelly Close