FDA Advisory Committee Meeting on Cardiovascular Outcome Trials

October 24-25, 2018; Silver Spring, MD; Full Report – Draft

Executive Highlights

  • The FDA Advisory Committee meeting on the agency’s 2008 guidance on the assessment of cardiovascular risk for new type 2 diabetes drugs wrapped up today, following two days of presentations and discussion on CVOT design and utility. The committee cast an extremely close vote of 10-9 (yes-no) on the sole voting question: Should an unacceptable increase in cardiovascular risk be excluded for all new drugs to improve glycemic control in patients with type 2 diabetes, regardless of the presence or absence of a signal for cardiovascular risk in the development program? In other words: should a CVOT be required for all new type 2 therapies? We heard criticism from multiple stakeholders that the question was designed to force one to vote “yes” – not to mention that the wording was confusing by any definition. Exactly what it means to “exclude” and what is “unacceptable” are left up for interpretation, which multiple committee members addressed in their votes. We note that seven of eight cardiologists voted “yes” and four of six endocrinologists voted “no” – an intriguing pattern. Of note, we’re not entirely clear what FDA’s timeline or intended action items are based on this vote; agency employees in the room were ambiguous on this front (Dr. Chong said “We’ll be getting back to the field …”) and there was not a sense of urgency.

  • Despite the discrepancy in votes, which likely stemmed at least in part from the question’s confusing wording, committee member agreed on multiple points. As Dr. George Grunberger put it, “If you turned off the screen, you wouldn’t know how anyone voted ­– everybody said exactly the same thing. My neighbor voted ‘yes’ and said exactly what I’m going to say now.” While there were nuances between opinions, the committee reached an encouraging near-consensus that CVOT requirements, design, and conduct can be meaningfully simplified to create a far less burdensome and expensive guidance – find far more on this below.

  • We were encouraged by the relative consensus throughout the day although several aspects of the meeting made it less productive than it otherwise may have been. We would have liked, for example, to see an incredibly valuable communication to FDA from the ADA read out loud at OPH. As well, we would have liked to have seen other professional groups weigh in who would normally be at meetings like this, such as AHA, ACC, AACE, Endocrine Society, AADE, and others. Certainly, the short time between meeting’s announcement and its start may not have given relevant organizations an opportunity to fully prepare and actively participate.  

    • The last guidance, in 2008, was made final before input from professional advocacy groups, manufacturers, payers, and patient groups had looked at it – given the very valuable and varied input today, we hope that the next guidance will have an open comment period.    

  • Discussion today focused on the role of phase 2 and 3 trials in demonstrating CV safety, the role of pre- and post-market studies and statistical goalposts, and the streamlining of CVOT design. The possibility of designing development programs to accumulate enough CV events to demonstrate safety, likely through enrolling some higher-risk patients and conducting longer follow-up, was a point of significant discussion that many came to favor. Complementing this point, there was a lot of push-back against the current “two-step” process of pre- and post-market safety, with separate 95% CI upper limits of 1.8 and 1.3, respectively. Could a midpoint of these be used to adequately established safety pre-approval? Finally, panelists were also keen on the idea of CVOT simplification, particularly through less intensive adjudication procedures (e.g., a single, centralized, standardized committee that could adjudicate multiple studies, though this idea was not fleshed out further than this – would the committee be at FDA? Built by companies?), as well as greater use of technology and less extensive adverse event collection. See below for far more on these points, as well as a rundown on how each committee member voted and why.

  • Also included below are researcher presentations on various aspects of CVOT design and results. Of particular note, former ADA Chief Scientific Officer Dr. Robert Ratner advocated for removal of the CVOT requirement, emphasizing that no CVOT to date has found CV harm, the trials aren’t generalizable, and the ethics of randomizing high-risk patients to a placebo group have come into question. Taking a different stance, Drs. Marc Sabatine and Jennifer Green – two experienced clinical trialists, with the TIMI and DCRI groups, respectively – suggested ways to make CVOTs less burdensome. Both advocated for a continuation of the CVOT requirement, suggesting that technology, less intensive data monitoring, and different approaches to design and outcomes could all reduce time and cost while maximizing ability to demonstrate safety and even identify benefit. Both Dr. Sabatine and Green, as we understand it, are at institutions that are highly immersed in such CVOTs – while there wasn’t disclosure of this, per se, we think that it would be valuable to consider this with their remarks.

The FDA Advisory Committee meeting on the agency’s 2008 guidance on the assessment of cardiovascular risk for new type 2 diabetes drugs wrapped up today, following two days of presentations and discussion on CVOT design and utility. The committee cast an extremely close vote of 10-9 (yes-no) on the sole voting question: Should an unacceptable increase in cardiovascular risk be excluded for all new drugs to improve glycemic control in patients with type 2 diabetes, regardless of the presence or absence of a signal for cardiovascular risk in the development program? We heard criticism from multiple stakeholders that the question was designed to force one to vote “yes” – not to mention that the wording was confusing by any definition. Of note, we’re not entirely clear what FDA’s timeline or intended action items are based on this vote; agency employees in the room were ambiguous on this front (Dr. Chong said “We’ll be getting back to the field …”) and there was not a sense of urgency.

That said, every committee member agreed on multiple points. As Dr. George Grunberger put it, “If you turned off the screen, you wouldn’t know how anyone voted ­– everybody said exactly the same thing. My neighbor voted ‘yes’ and said exactly what I’m going to say now.” While there were nuances between opinions, the committee reached an encouraging near-consensus that CVOT requirements, design, and conduct can be meaningfully simplified to create a far less burdensome and expensive guidance – find far more on this below. There was also discussion on pieces that will be difficult to decode:

  • Most want a more generalizable population to participate in trials. However, this becomes more nuanced should a post-market requirement be removed. In this case, “sicker” patients (those with established CVD) would need to be more heavily included in safety-oriented phase 2 and phase 3 trials for which there is little precedent.

  • Is it ethically fair to give patients placebo? Given the new ADA/EASD consensus statement which recommends a GLP-1 agonist or SGLT-2 inhibitor for those with established CVD, can CVOTs ethically incorporate a placebo arm?, That said, the “reality” of care today is that many patients do not take the optimal therapy (regardless of prescription) and many of those who are prescribed optimal therapy cannot afford it. Moreover, active comparators will make it challenging to show improvement in A1c – and is A1c the only outcome measure to consider? A new study in Diabetes Care by Drs. Roy Beck and Rich Bergenstal posted online October 23, 2018 suggests otherwise.

  • Better surveillance tools sound very positive to some – others say they will simply never be good enough or available. The same for remote monitoring.

Among the physicians on the committee, eight specialized in cardiology and six in endocrinology – and a member’s specialty was very predictive of how they voted. Among cardiologists, seven voted “yes” and only one voted “no”; among endocrinologists, two voted “yes” and four voted “no.” While the primary theme to emerge at the end of the day was that everyone basically agreed on the more specific points of how CVOT guidance could and should change; however, the side on which these opinions were grounded – a “yes” vs. a “no” vote – did seem to carry a broader-scale meaning to individuals.

Broadly, these groups were approaching the question in a different way, and our theory is that differences in cardiology vs. endocrinology practice may speak to this. As one of our colleagues put it, “Cardiologists wanted more data so they can generalize, and endocrinologists wanted more data so they can individualize” – a difference that may reflect how CVD vs. diabetes are managed differently, though we certainly aren’t cardiology experts. Our perception, carried over from the first day of the meeting, remained that there some important gaps in background knowledge and interprofessional understanding. For example, one cardiologist on the committee challenged the need for additional diabetes therapies, of which there are now many classes and drugs within each class. He argued that endocrinologists always say there’s a need for more glucose-lowering agents but, to a cardiologist, there’s already an array of drugs and it doesn’t seem like we need any more.  To be sure, this doesn’t represent the view of all cardiologists, but we were glad that NIH’s head of diabetes Dr. Judith Fradkin interjected to explain that hyperglycemia is a genuine problem that leads to hospitalizations, and there remains no optimal drug for any patient. There are still many challenges with side effects, and we believe Dr. Fradkin was also alluding to the fact that diabetes is a progressive disease, making it important to have more options in the treatment arsenal (especially those that prompt weight loss and bring less risk of hypoglycemia). 

Outcomes beyond A1c arose multiple times during Q&A and discussion sessions, a great sign of progress. In fact, the committee was so interested in discussing the importance of hypoglycemia and time in range that FDA representatives had to interject in and re-orient the conversation. We certainly acknowledge that FDA is holding this Advisory Committee meeting to gather thought leader opinions on CVOTs; on the other hand, given the limitations of A1c and potential links between time-in-range and complications, it was discouraging to see FDA proactively and reactively suppress this important line of conversation. This has real implications for studies –if CVOTs had to do a CGM sub-study, we might learn a lot beyond A1c data alone! 

Below, you’ll find details on the most important points of debate to emerge from the day’s discussion, summaries of speeches from the Open Public Hearing session, and a rundown on each committee member’s vote and reasoning. Also included below are highlights on thought leader presentations to the committee from the first day of the meeting.

Points of Discussion

  • No panelists furthered Dr. Robert Ratner’s view that a CVOT should only be required if there is a safety signal in phase 2/3 (see below). However, many did propose that phase 2 and phase 3 programs could be designed differently going forward to give an adequate assessment of cardiovascular safety. On the one hand, there was general agreement that phase 2 and 3 studies historically have not given enough data to exclude the potential for cardiovascular risk with a new therapy. Typically, only a small number of MACE events (<100 in total) occur in these programs, which is usually enough statistical power to exclude an upper 95% CI limit of 1.8, but not to exclude 1.3 (i.e., the 2008 guidance requirements). On the other hand, there also hasn’t been a hint of a CV risk signal in the phase 2 or 3 program for any diabetes therapy on the market today – a point that many made in questioning the necessity and validity of the CVOT requirement. It is worth noting that the phase 3 SUSTAIN program for semaglutide adequately demonstrated CV safety (below 1.3) because it included a smaller CVOT, SUSTAIN 6 (more on this below). Some panelists took a more aggressively conservative stance on the issue; Dr. Thomas Wang went as far to say that no useful CV info can be taken from phase 2 and 3 studies. Rather than not showing evidence of CV harm, the studies were fundamentally underpowered to do so, in his opinion. The committee’s views on this point were summed up when Dr. Burman said, “A standard phase 2/3 trial for safety and efficacy is not generally adequate to detect a relevant CV signal,” a sentiment that was strongly echoed by Dr. Budnitz during the voting question. All of this informed a point that the committee well converged upon: perhaps phase 2 and 3 programs should contain an adequate number of CV events in order to demonstrate safety premarket. This would require some combination of: (i) larger study populations; (ii) longer duration of follow up; and/or (iii) enrollment of more high-risk patients; Dr. Yanovski pointed out that eligibility criteria would need to be expanded. Dr. Grunberger developed this concept further: “The idea would be to broaden the phase 3 trials, the population in those, and the risks we’re discussing. They should include heart failure, kidney, and peripheral disease. More robust, larger phase 2 and 3 trials can get better data, and if there is a signal then do a dedicated [CVOT]. If there’s no signal, there’s no reason to require a dedicated CVOT.” There are, however, important questions to work out. For example, Dr. Low Wang pointed out that primary vs. secondary CVD is different from a pathophysiological perspective, and the inclusion of high-risk patients with established CVD doesn’t necessarily answer a question for the whole population. Of course, we acknowledge that obtaining a definitive signal in low-risk individuals might require massive trials, especially if one is shooting for benefit, and there is little reason to believe that, if safety is demonstrated in a high-risk population, a drug would not be safe in a lower-risk population. Of course, these issue of generalizability also apply to all trials – whether they are done pre- or post-market, this question of high vs. low-risk is still an issue.

  • In light of the potential value of phase 2/3 trials (see above), several panelists posited that the current, two-step approach to CV outcomes could be shortened to a one-step, pre-market process. (Discussion Questions 2C and 2D). According to the 2008 guidance, premarket phase 2/3 studies must rule out a hazard ratio of 1.8 from the upper bound of the 95% confidence interval for the estimated risk ratio prior to approval. Moreover, a reasonable point estimate is required to assure safety at this stage; as worded in the guidance: “it would not be reassuring to find a point estimate of 1.5 (a nominally significant increase) even if the 95% upper bound was less than 1.8.” Once accomplished, a post-market CVOT must definitively show that the upper bound of the confidence interval is less than 1.3. However, we were reminded that, should a hazard ratio of 1.3 be ruled out in premarket studies, no post-market CVOT is currently required; one example of this is Novo Nordisk’s semaglutide, for which a “smaller” CVOT (SUSTAIN-6) was conducted in phase 3 and actually demonstrated superiority on the 3-point MACE with an upper boundary of 0.95 – well below 1.3. Also, to our understanding, enough data was gathered on dapagliflozin after the SGLT-2 inhibitor received an initial CRL in order to exclude 1.3. With this context, and in light of the potential value of phase 2/3 trials (see above), several panelists posited that the current, two-step approach to CV outcomes could be shortened to a one-step, pre-market process. If this were to occur, the premarket 1.8 upper boundary limit would need to be lowered to ensure CV safety. Notably, this would require a greater number of events to be accumulated during phase 2 and phase 3 trials and, thus, larger and/or higher-risk populations to be enrolled in these studies. While many on the committee agreed on lowering the boundary limit for a premarket CV trial, no consensus was reached as to what the new bar should be – several advisers made arbitrary suggestions of 1.5 for the upper boundary and 1.2 as a reasonable point estimate limit. In fact, it was noted several times that 1.8 and 1.3 were also arbitrarily chosen ten years ago; to this end, multiple advisers advocated for more rigorous reasoning should the guidance be change. As it stands, the upper bound is determined by statistical testing but set on what is considered “unacceptable risk,” which was arbitrarily set ten years ago according to the committee. (Certainly an 80% increased risk, if actually observed, is quite significant and would likely doom any drug.) Toward the end of the discussion, Dr. Grunberger put forth an interesting suggestion: Why not ask patients and prescribers what an acceptable level of risk might be? While this thought was not discussed due to its late introduction, we were thrilled to see this patient-centric viewpoint expressed during a highly technical debate and hope the FDA solicits opinions from the larger diabetes community in introducing any changes to the guidance. Of course, this idea needs further refinement; for example, if they said 0% risk is acceptable given that diabetes increases CV risk already then an HR of 1.01 would have to be ruled out. Altogether, we’re not sure if this is a good idea, especially because interpretation of relative vs. absolute risk is really critical (i.e., a 1.25x risk sounds scary, but not as much so if it is a 2.5% vs. 3.1% event rate).

  • Panelists deeply considered trial design and how CVOTs could be streamlined to maximize efficiency and cost-effectiveness while preserving their statistical robustness and clinical value. Dr. Everett explored the need for an independent CV event adjudication committee along with better ascertainment of events of interest in order to truly achieve validity in reporting these events. In his view, “we have to have validity before generalizability – without validity of endpoints, you don’t have any generalizability.” This was well received by participants, and served as a backbone for the entire conversation. To achieve “validity before generalizability,” Dr. Everett advocated for moving away from investigator-reported adverse event monitoring (Dr. Everett specifically mentioned “subtle shades of unblinding that can occur using this method). Instead, independent and centralized committees for multiple trials should be used so that a standardized approach can prevail, which we understand would be more efficient and cost-effective. Dr. Low Wang echoed these sentiments, proclaiming that “there is no substitute for rigorous adjudication of CV endpoints.” She mentioned that she has been on several committees that have served to adjudicate these events; in her experience, these arrangements only work in the context of well-designed and well-conducted trials. This idea – that independent CV committees are dependent on trial design – was echoed by multiple participants, including Dr. Rosenberg, Dr. Budnitz, and Dr. Wasserman. Somewhat provocatively, Dr. Wasserman threw support behind the idea of using large data bases to look at adjudication in “a non-endpoint-committee way” (multiple panelists grimaced at this suggestion). Dr. Wasserman noted the skepticism to this idea but urged participants to keep an open mind to these approaches – after all, the field is evolving, and “we should look at these things futuristically.” Consumer representative Ms. McCollister-Slipp also reinforced this push for non-traditional analysis methods and the use of technology and large datasets as a substitute for complex adjudication processes.

Thought Leader Presentations

Below, we’ve summarized three thought leader presentations from the first day of the meeting, delivered by Drs. Robert Ratner, Marc Sabatine, and Jennifer Green.

  • Should CVOTs continue to be mandatory? Dr. Robert Ratner says no, for four reasons: (i) no CVOT to date has revealed a harmful CV signal; (ii) CVOT populations are not generalizable (only ~22% of adults with type 2 diabetes have established CVD); (iii) no CVOT has demonstrated superiority in a primary prevention population, with some trending towards placebo; (iv) it may not be ethical to randomize patients to placebo given the new ADA/EASD consensus report, which recommends a GLP-1 or SGLT-2 be considered for all patients with established CV disease. Dr. Ratner went on to address two other major questions posed to this Advisory Committee: Where should the FDA go from here, and how can outcome trials be modified to be more feasible? Topping his recommendations on “where FDA should go from here” was a controversial suggestion to return to the pre-2008 regulatory approach. In his words – “that probably isn’t going to happen, and probably shouldn’t happen, but there are a whole lot of simple tweaks to current or previous guidelines that could improve the process.” For example, loosening exclusion criteria could enable shorter trials with a more representative population – “cherry-picking” clearly skews results. Moreover, if the CVOT requirement were removed, it would be the FDA’s responsibility and obligation to continue to demand outcome trials whenever a safety signal is noted in phase 2 or 3 trials (just like all other diseases and complications). Lastly, in the absence of a CVOT mandate, manufacturers who desire a label indication could still be required to conduct a CVOT. On the flip side, manufacturers with a new CV-neutral therapy, for example a new DPP-4, shouldn’t be required to conduct an expensive CVOT. To make outcome trials more feasible, Dr. Ratner argued for (i) different trial designs, including real-world trials (e.g., CVD-REAL), adaptable trials, or registry studies; (ii) different statistical approaches – he argued that Bayesian statistics could dramatically improve the power of CVOTs (thereby reducing the required number of events) based on this paper coauthored by Dr. Sanjay Kaul; and (iii) incorporating outcomes important to people with diabetes. On the last of these, Dr. Ratner referenced a recent patient survey conducted by The diaTribe Foundation and dQ&A, which found CV risk is not among the top five patient priorities factors which make “a big impact” on daily life, regardless of type of diabetes or current medication (slide 13). That said, Dr. Ratner did not allude to slide 26 of this presentation, which states that “less risk of complications” was the #1 goal for new therapies across all types of diabetes and current medication regimens. Dr. Ratner wrapped up his discussion with a powerful reminder to FDA and the audience: “the goal [of diabetes treatment] is to improve the lives of people with diabetes, not just their longevity.”

  • In discussing CVOT design, Dr. Marc Sabatine suggested concrete changes to make CVOTs more “pragmatic” or “streamlined.” While he said it’s critical to maintain the randomization, blinding, and prospective nature of these studies, he suggested the possibility of an open-label design with fewer but higher-volume trial sites, patient contact through mobile devices, and risk-based monitoring – or even an embedding of the trial within the healthcare delivery system. Additionally, data collection could be simplified: Outcomes could be collected through medical records, though more detailed data would be difficult to acquire and adverse events would not be captured as uniformly or thoroughly. For data monitoring, Dr. Sabatine discussed the potential to limit monitoring to items that have a meaningful impact on the trial’s internal validity, as well as the shift to a centralized and risk-based monitoring system. To be clear, Dr. Sabatine was not explicitly advocating for any of these changes but did support their utility in minimizing costs while optimizing the scientific validity of a trial. There is no substitute for a dedicated CVOT if the goal is to definitively demonstrate CV safety; but in reconsidering the agency’s guidance on the issue, it should be considered whether a lower-burden trial design can ensure patient safety while better encouraging innovation and advances in diabetes care.

  • Dr. Jennifer Green provided a rundown of the wide-ranging effects of the CVOT mandate, discussing “the good, the bad, the ugly, and the future” of FDA guidance on the issue. Her talk was grounded in her opinion that “adequately powered, randomized CV outcomes trials of individual antihyperglycemic agents should continue.” Dr. Green’s general support for the CVOT mandate was not unequivocal, as she did advocate for measures that can be taken to improve the efficiency of these trials. Further qualifying this support, Dr. Green noted that the previously used CVOT model should not be the only acceptable path forward for the future; importantly, future work should focus on addressing barriers to implementation of CVOT data in clinical practice. Novel approaches to trial design, operations, and outcomes should be investigated in order to reduce time and costs and also maximize the ability to identify benefits for patients. Interestingly, Dr. Green also offered support for the idea of diabetes drug approvals going beyond the current sole registration endpoint of A1c. On this front, Dr. Green explained that if a drug meaningfully reduces risk for important complications in people with diabetes (whether that be CV risk, death, end stage kidney disease, or others) – but may work through mechanisms other than A1c lowering – then it should still be approved for people with diabetes. We were definitely glad to hear Dr. Green expand the paradigm for diabetes drugs beyond A1c. Dr. Green was careful to emphasize the lasting importance of A1c considerations and treatment of hyperglycemia in diabetes care – she categorized the devaluation of glycemic control in diabetes management in the CVOT era as part of “the ugly” consequences of recent advances.

  • We note that Dr. Sabatine is Chairman of the TIMI Study Group, and Dr. Green is affiliated with the Duke Clinical Research Institute (DCRI). Both have been involved in the design, conduct, and publication of multiple diabetes CVOTs, which does introduce some potential for bias in their remarks.

Open Public Hearing Summaries

  • AstraZeneca took a strong stance in arguing that conducting CVOTs for diabetes therapies should no longer be required across the board; rather, an outcomes trial should only take place when a member of the same class has given a signal for CV risk or when there was a signal for risk in preclinical or clinical development. Dr. Elizabeth Young opened AZ’s presentation by pointing out that no diabetes CVOT to date has identified an increased risk for MACE, just as no drug submitted for registration in the last decade has shown a signal for adverse CV effects in phase 2/3. As such, there’s little basis for assuming that diabetes drugs increase CV risk. We do think that it’s interesting to consider that there is no formal CVOT requirement for lipid-lowering drugs or anti-hypertensives, two classes of drugs very commonly taken by people with diabetes. She also touched on the unmet need remaining in diabetes as a reason to reduce barriers to innovation, development, and market entry. Given that a CVOT costs ~$200-400 million or more, AZ suggested both alternative data sources (real-world, registries) as well as streamlining of traditional CVOT design. Specifically, this could include centralized adjudication, investigator-assessed endpoints, and/or automated and AI-assisted adjudication. As did others, AZ suggested that MACE is not always the most important endpoint for any given therapy; rather, endpoints should be based on patient needs and a drug’s mechanism. Along similar lines, choosing a study population could be based on mechanism rather than indication; for example, if a drug lowers glucose and reduces fibrosis, a trial could enroll those with either type 2 diabetes or NAFLD/NASH because those patients will later be treated with the drug in clinical practice.

  • Dr. Angelyn Bethel, who recently joined Lilly as a Senior Medical Advisor, delivered a robust and detailed presentation on what the company feels should be required for CV safety demonstration in the future. The company did not advocate for a reversion to pre-2008 procedures but did make a number of very concrete suggestions for revision and then finalization of the 2008 draft guidance. In terms of preapproval safety assessment, Lilly suggested (i) specifying inclusion of high CV risk subgroups for 12-18 months of exposure, (ii) clarifications of best practices for detecting CV safety signals, and (iii) affirmation of the need for safety evaluations without requiring demonstration of benefit. Signal detection, Dr. Bethel clarified, should be flexible on a case-by-case basis for what is “unacceptable” and take multiple inputs into account. In the event that a concerning safety signal is identified, Lilly advocated for adherence to the principles of an individualized safety assessment – that is, one directed by mechanism, preclinical signals, class knowledge, and nature and severity of the signal – but also for primary outcomes beyond MACE. Rather, she said, MACE can be collected as a secondary outcome. While we can understand the appeal of building enough CV events into the phase 3 program for a drug to demonstrate safety, we don’t want to underestimate the logistical and scientific challenges to making this happen, though we certainly think the possibility could be given serious consideration. Of course this would create disproportional barriers for smaller companies, but these barriers would for sure be more manageable than a post-market CVOT. Additionally, Lilly argued for greater consideration of other methodologies, including pragmatic studies that maintain randomization, prospective registries, and use of EHRs for adverse event reporting. Ultimately, Dr. Bethel emphasized that these changes would prevent unnecessary long-term patient exposure for risk assessment; there would be more flexibility for a safety assessment trial or other study, whether for CV purposes or otherwise, to be guided by previous data.

  • Novo Nordisk’s Dr. Steven Gough guided attendees through Novo Nordisk’s extensive repertoire of diabetes clinical trials and CVOTs, ending with the company’s ideal scenario for guidance changes. He began with three powerful statements: (i) Novo Nordisk conducts clinical research in 56 countries; (ii) it supplies ~half of the insulin around the world; and (iii) it has developed five medicines in the past four years (Tresiba, Xultophy, Saxenda, Ozempic, and Fiasp, by our count). During a thorough overview of the basic design of and positive results from the LEADER (for liraglutide) and SUSTAIN-6 (for semaglutide) CVOTs, he placed particular emphasis on the trials’ impressive retention rates (97% and 98%, respectively), owing them to Novo Nordisk’s efforts to promote patient engagement through forums to share best practices as well as patient support groups. To our understanding, these rates are the percentage of patients for which data collection was completed, rather than the percentage who stayed on study drug. Dr. Gough detailed Novo Nordisk’s preferences for amendments to the guidance, going through three scenarios: (i) if no CV signal is seen during phase 2 or phase 3, no CVOT should be required post-market; (ii) if a CV signal is seen during phase 2 or phase 3, a CVOT should be required post-market; and (iii) if Novo Nordisk would like to apply for a CV superiority indication, a CVOT should be required. On the last of these, Professor Gough noted that Novo Nordisk would welcome any additional information and guidance on best practices for trial design that can be easily translated to a label.

  • Dr. Jay Edelberg of Sanofi advocated for the adoption of product-specific, risk-based approaches to assessing CV safety for diabetes drugs. He noted that the CVOTs have become a “blanket de facto requirement” following the adoption of FDA’s guidance in 2008 and explained that Sanofi believes the time has come to change this status quo. Dr. Edelberg underscored that scientific evidence accumulated after the mandate has shown no evidence for CV risk for multiple drugs tested across multiple classes. Over 200,000 patients in total have been enrolled in these studies (most of them with high baseline CV risk) and a high degree of certainty has been accumulated that these drugs don’t increase CV risk. An enormous opportunity cost also exists with the CVOT mandate, he said, as it has reasonably discouraged further research in diabetes and depressed patient choice by limiting the number of potential manufacturers with the ability to invest in diabetes. Dr. Edelberg stressed that this stifling of growth within diabetes is antithetical to what the FDA should be doing in looking for policies that encourage innovation. As a result, he concluded that Sanofi requests for FDA to update its practices and adopt product-specific risk-based approaches for CV risk assessment, instead of a sweeping CV mandate that is uniformly applied to all products.

  • Dr. Varuna Srinivasan (Johns Hopkins University, Baltimore, MD) stressed the importance of evaluating safety risks for new drugs and not relying on past evidence that may not be generalizable. Many critics of the CVOT mandate have argued that funds from these trials can be better allocated to other areas; Dr. Srinivasan reminded the audience that the same used to be said about safety trials of hormone replacement therapy. This therapy showed no safety signals in smaller studies; however, the larger Women’s Health Initiative suggested that hormone replacement therapy had risks that outweighed potential benefits. (More recently, the pendulum is somewhere in between – see this NPR article from 2013, which highlights the nuance.) She noted that this research is credited with saving thousands of lives, highlighting the importance of conducting safety research like CVOTs. Dr. Srinivasan also argued that it’s not prudent to rely on old studies and to extrapolate potential class effects to newer drugs. As such, she urged the committee to strengthen the CVOT mandate in order to ensure patient safety.

Commentary from Voting Members


  • Strictly relating to the voting question, a yes vote means the voter believe an unacceptable increase in cardiovascular risk should be excluded for all new type 2 diabetes drugs, regardless of the presence or absence of a signal for cardiovascular risk in the development program (strictly as it relates to the question). However, we note that this is not necessarily a vote for or against post-market CVOTs, as noted in the voter’s explanations:

  • Dr. Kenneth Burman (Georgetown University, Washington, DC) voted yes, stating that the benefits of CVOTs (primarily, ensured CV safety) outweigh other issues such as cost. As he sees it, standard phase 2 and 3 trials are not adequate to detect relevant CV signals with the proper sensitivity and specificity. However, he expressed hope that current CVOTs could be properly modified to improve efficacy, efficiency, and cost through high volume centers, mobile devices, adoption of new endpoints, and risk monitoring schedules.

  • Dr. Brendan Everett (Harvard Medical School, Boston, MA) was compelled to vote yes by the wording of the question – “an unacceptable risk should be excluded.” However, how it is excluded and what is unacceptable he left up for interpretation, opening the door to premarket studies and monitoring of CV endpoints in other dedicated trials (CKD, QoL, etc.) as long as CV outcomes including heart failure are properly measured and adjudicated. His one stipulation – longer term exposure must be required; he found it “outlandish” that a drug that patients take for ten years can get approval with a 52-week study. We’d point out, however, that a ten-year study would mean new drugs never make it to market – what is an acceptable, realistic middle-ground?

  • Dr. Fred Kushner (Louisiana State University, Baton Rouge, LA) noted that no surrogate endpoint exists to replace cardiovascular outcomes, meaning that only adequately accumulated events can prove CV safety, especially for new drug classes. To this end, he endorsed CVOTs but acknowledged that a more rigorous, premarket option is not off the table.

  • Dr. James de Lemos (University of Texas Southwestern Medical Center, Dallas, TX) argued that there is no rush to bring new drugs to the diabetes market, as multiple efficacious classes and therapies with CV benefit already exist. As such, we should not compromise the certainty of RCTs (and the value they bring to patients) for a less robust but more efficient method. That said, he acknowledged that as few as 300 endpoints (pre- or post-marketing) might be good enough to exclude an unacceptable risk, and that patient exposure should be maximized in terms of data collection.

  • Patient representative Mr. Richard Lumley was originally on the fence but was persuaded by a late comment from Dr. Wang clarifying that one can vote yes and still want changes to the guideline. He complemented Dr. Wilson and the FDA on a meeting well-run.

  • Dr. Martha Nason (NIH, Bethesda, MD) felt she had to say yes, based on the wording of the question. However, she does not believe that the FDA can continue to mandate post-market CVOTs any longer based on the lack of a CV signal in all completed trials. Instead, she advocated for more inclusive, longer-term, pre-market data to determine CV safety, noting that market dynamics would still incentivize companies to perform CVOTs for label indications.

  • Dr. Connie Newman (New York University, New York, NY) voted yes, with caveats. She stood by double-blinded, placebo-controlled, randomized trials as the gold standard but recommended flexibility on endpoints (e.g., heart failure) and streamlined processes to only gather the data that is necessary. She remained unsure whether new drugs in established classes should require a CVOT, noting that this is where phase 2 and 3 data could be used to help decide.

  • Dr. Yves Rosenberg (National Heart, Lung, and Blood Institute, Bethesda, MD) asserted that both the yes and no votes essentially add up to the same thing, as either side would acknowledge the need for a comprehensive assessment of safety and efficacy of diabetes therapies. In his mind, this assessment should be a more flexible, pragmatic assessment in late phase 2 and early phase 3 studies that encompasses all relevant clinical outcomes, including heart failure.

  • Dr. Thomas Wang (Vanderbilt University, Nashville, TN) argued that the new information learned from CVOTs makes them difficult to replace. As new drugs are added to the diabetes armamentarium, he found it hard to imagine not knowing their effect on CV outcomes in the same amount of detail provided by CVOTs. That said, he did acknowledge the restricted generalizability and other limitations of CVOTs, briefly advocating for consideration of how design and endpoints can be improved.

  • Chairperson Dr. Peter Wilson (Emory University School of Medicine, Atlanta, GA) highlighted the need for guidance in the diabetes field. From the ACCORD trial, to TZDs, to today, he emphasized how guidance, such as the CVOT mandate, has pushed the field forward. However, Dr. Wilson did acknowledge the need for relaxation, simplification, and flexibility within the guidance.


  • Dr. Michael Blaha (Johns Hopkins University, Baltimore, MD) voted no primarily because no CVOTs to date have indicated elevated CV risk. In light of this simple reality, Dr. Blaha finds it hard to continue to justify the mandate. He noted that even without the mandate, manufacturers will still feel competitive pressures to conduct CVOTs in order to demonstrate potential CV benefits. He also supported the proposal pushed by Dr. Everett to focus on investigating these risks in pre-market approval studies.

  • Dr. Daniel Budnitz (CDC, Atlanta, GA) voted no to signal that current guidelines could be improved. He noted that traditional phase 2/3 studies are indeed currently underpowered, and would need to be broadened in order to glean meaningful safety data from these trials. Dr. Budnitz hopes to see an increased emphasis on and attention given to US based patients in clinical trials in order to more readily apply data to practice.

  • Dr. Susan Ellenberg (University of Pennsylvania, Philadelphia, PA) voted no, proclaiming that she is “not convinced that this two-stage approach is the right approach.” She emphasized that by no means is her vote an indication that the field should go back to pre-2008 guidance. However, factors beyond the cardiovascular need to be considered to determine whether a drug is likely to provide more benefit than harm.

  • Dr. Judith Fradkin (CDC, Atlanta, GA) voted no, emphasizing that no CVOTs to date have shown significant CV risk. Dr. Fradkin also noted that CVOTs have not only had an economic impact on pharma, but have also disproportionally shifted attention to traditional MACE events in high-risk populations that don’t accurately represent the diversity of the broader diabetes population. Dr. Fradkin endorsed the solution proposed by Dr. Everett, where phase 3 trials might be enlarged to achieve a more rigorous boundary than the current 1.8 upper boundary pre-market goalpost. Building off of this point, Dr. Fradkin noted that she would support a compound evaluation of traditional MACE events, alongside other events like heart failure and risk factors such as elevated lipids, blood pressure, and weight gain.

  • Dr. George Grunberger voted no, but emphasized the enormous overlap in comments between those voting “no” and those voting “yes.” He noted that his neighbor (Dr. Martha Nason) said the same exact thing he would have said, but that she voted “yes.” Dr. Grunberger signaled that he wants to broaden phase 2/3 trials, along with the definition of CV risk in order to include heart failure and potentially kidney and peripheral disease as well. If a signal is seen in these phase 2/3 trials, then a dedicated trial must be run. Additionally, he argued against the continued use of placebo-controlled trials.

  • Dr. Cecilia Low Wang (University of Colorado, Aurora, CO) voted no, but emphasized that this does not mean she wants to see a return to pre-guidance state of affairs. She called for a broader endpoint than three-point MACE, specifically mentioning the inclusion of heart failure. Dr. Low Wang also advocated for lowering the upper bound of the 95% CI from 1.8 for exclusion of risk in pre-marketing trials (apparently in the case that phase 2/3 data were used to demonstrate safety). In terms of other improvements, she mentioned that patients without established CVD are being missed in the current paradigm and that FDA should look further into using registries and observational data to supplement phase 2 and 3 studies in order to create a larger evidence base when assessing safety.

  • Consumer representative Ms. Anna McCollister-Slipp (VitalCrowd, Washington, DC) voted no on the basis of using other data sources (namely observational studies and registries) to replace CVOTs. Ms. McCollister-Slipp believes that the advent of newer technologies that are adept at curating and analyzing larger data sets will improve the feasibility of CVOT substitutes.

  • Dr. David Robbins (University of Kansas, Kansas City, KS) voted no, emphasizing that industry should be afforded flexibility in light of knowledge collected in the past 10 years of CVOTs. Dr. Robbins underscored that innovation is needed at this key point in time, and that faster approval and lower costs of drugs can be achieved by diverting resources away from massive CVOTs.

  • Dr. Susan Yanovski (NIDDK, Bethesda, MD) voted no and explained that there should be a pathway to approval with a one stage trial that displays reassuring point estimates and no signs of CV risk. Dr. Yanovski also advocated shifting the current paradigm of CV outcomes to include HF and other potential measures not included in traditional three-point MACE.

Summary of Committee Votes

Panelist Name


Overall (Y:N) Vote History for EMDAC Diabetes/Obesity Drug Approval Meetings in Past Nine Years*

Vote on Continuation of CVOT Requirement

Michael Blaha




Daniel Budnitz

Drug Safety

3:2, also voted to keep Avandia REMS


Kenneth Burman




Susan Ellenberg


Has not served on an EMDAC panel for a diabetes/obesity drug over the last ten years


Brendan Everett




Judith Fradkin




George Grunberger


Has not served on an EMDAC panel for a diabetes/obesity drug over the last ten years


Fred Kushner


Has not served on an EMDAC panel for a diabetes/obesity drug over the last ten years


James de Lemos




Cecilia Low Wang




Richard Dan Lumley

Patient Representative



Anna McCollister-Slipp

Consumer Representative

Has not served on an EMDAC panel for a diabetes/obesity drug over the last ten years


Martha Nason

Mathematical Statistician



Connie Newman


Has not served on an EMDAC panel for a diabetes/obesity drug over the last ten years


David Robbins




Yves Rosenberg




Thomas Wang




Peter Wilson (Chairperson)

Cardiovascular Safety



Susan Yanovski




Close Concerns’ Questions

  • How will FDA interpret the 10-9 decision? Given the ambiguous wording of the voting question, will the FDA see the majority “yes” votes as reason to keep the mandate as it stands? There was general consensus that CVOT requirements, design, and conduct should be meaningfully simplified to create a less burdensome and expensive path; how will FDA operationalize and formalize that into guidance?

  • When will FDA respond to the vote? Dr. Chong was rather vague, and no timeline was provided.

  • How would companies adapt strategy if: (i) the status quo as it is was maintained; (ii) some hybrid approach, like expanding phase 2/3 studies; or (iii) the CVOT mandate was completely dropped and we went back to pre-2008 guidance?

  • What have been the full implications of the mandate on industry? How would the world of diabetes drugs look different now, assuming no mandate had been in effect? Since 2008, we’ve seen Amgen, BMS, GSK, Gilead, and Takeda leave diabetes, and other companies including Novartis and Janssen are investing far less. Does the current CVOT mandate dissuade companies’ from entering the market and completely preclude smaller companies from bringing a drug to market on their own? (We believe it does.) How does the breakdown of drug companies in diabetes compare to other fields such as cancer, in terms of number, level of innovation (especially in personalized medicine), and market cap?

  • Has the CVOT mandate at all impacted who is choosing to pursue endocrinology? Prior to the mandate, a 2007 survey found that endocrinology with a diabetes focus and general endocrinology ranked 13th and 14th, respectively, out of 16 medical specialties in terms of interest by medical students. Topping their list of reasons was a perceived inability to change or impact patient behavior. Have CVOTs improved endocrinology’s attractiveness to the best and brightest future doctors?

  • How might industry and regulatory agencies work together to achieve a mutually beneficial decision for patients? We sensed a strong anti-industry bias from some members of the committee.


-- by Ann Carracher, Martin Kurian, Peter Rentzepis, and Kelly Close