Memorandum

AZ releases topline DECLARE results: Farxiga superior to placebo on HF/CV death, non-inferior on 3-point MACE – September 24, 2018

Executive Highlights

  • Farxiga achieved superiority over placebo on one of DECLARE’s co-primary endpoints – hospitalization for heart failure and cardiovascular death – making it the first SGLT-2 inhibitor to demonstrate heart failure benefit as a primary endpoint result. On 3-point MACE (CV death, heart attack, and stroke), Farxiga showed non-inferiority vs. placebo, though AZ’s press release reports that results trended in favor of Farxiga (we view this as a positive). No hazard ratios or p-values were provided in the press release. It’s important to note that a p-value <0.025 was required to confer superiority on either of the primary endpoints.

  • AZ will reportedly file for an indication based on DECLARE results with the FDA; no specifics on the potential label change were given. Current SGLT-2 CV indications for J&J’s Invokana (EMA only) and Lilly/BI’s Jardiance (EMA and FDA) are exclusively for those with established CV disease. However, the majority of the population enrolled in DECLARE (~60%) do not have established CV disease but have CV risk factors (primary prevention). Since the superiority found for heart failure and CV death apply to the study’s entire population, we wonder whether regulatory agencies may be persuaded to include primary prevention in an indication. Not only would this be a huge win for at-risk patients, it would also allow for a much wider prescribing base for Farxiga compared to SGLT-2 competitors Lilly/BI’s Jardiance and J&J’s Invokana.

  • DECLARE is perhaps the boldest CVOT we’ve seen to date (at least from a statistical perspective) and certainly the largest (n=17,160) and longest. With respect to size, the primary prevention cohort of DECLARE alone (n=10,189) is bigger than the total enrolled populations of EMPA-REG OUTCOME (n=7,020, <1% primary prevention) or CANVAS (n=10,142, 34% primary prevention). DECLARE is also the longest of the three SGLT-2 CVOTs, with median 4.5 years of follow-up, while EMPA-REG OUTCOME and CANVAS had average follow-ups of 3.1 and 3.6 years, respectively. We very much hope that patients will continue to be tracked.

  • All-important safety data were not shared in AZ’s press release, apart from a broad statement that DECLARE confirmed the well-established safety profile of Farxiga. Full results will be presented November 10 at AHA 2018. Amputations and DKA have been associated with SGLT-2 inhibitors; we would suspect less of this may well be seen since the population was healthier

AstraZeneca today announced highly-anticipated topline results from the DECLARE CVOT for SGLT-2 inhibitor Farxiga (dapagliflozin) – we see these positive findings as a win for AZ, for SGLT-2s as a class, and of course, for diabetes patients. Farxiga achieved superiority over placebo on one of the study’s two co-primary endpoints, a composite of hospitalization for heart failure and cardiovascular death, making it the first SGLT-2 inhibitor to demonstrate heart failure benefit as a primary endpoint result.

On the other co-primary endpoint, three-point MACE (CV death, heart attack, and stroke), Farxiga exhibited non-inferiority over placebo, though AZ’s press release notes that results trended in favor of Farxiga. No hazard ratios or p-values were provided in the press release, and it is important to note that a p-value less than 0.025 was required to confer superiority on either of the primary endpoints. With this in mind, the significant reduction of heart failure and CV death is even more impressive, and the three-point MACE data could still heavily favor dapagliflozin without explicitly meeting the threshold for superiority (more on all of this below).

It’s hard to understate the importance of these results, both for patients with diabetes and heart failure as well as for the SGLT-2 inhibitor class. DECLARE was unique in many ways compared to previously completed SGLT-2 CVOTs – EMPA-REG OUTCOME for Lilly/BI’s Jardiance (empagliflozin) and CANVAS for J&J’s Invokana (canagliflozin). On top of being the first CVOT to include heart failure in a co-primary endpoint, DECLARE was also by far the largest of any SGLT-2 CVOT to date – ~17,000 patients were enrolled, of which the majority (~60%) had never experienced an adverse cardiovascular event (see below). While this makes it difficult to compare the SGLT-2 CVOTs side-by-side (and we urge readers to avoid this), the large primary prevention population and multiple outcomes provide us with an unprecedented, in-depth view of Farxiga’s (and broadly the SGLT-2 class’) non-glycemic effects on a much greater proportion of the diabetes spectrum.

Farxiga is now the third SGLT-2 inhibitor (out of four on the market) to achieve superiority on a primary CVOT endpoint, and the first to do so for heart failure as a co-primary endpoint. Moreover, DECLARE is the third of three SGLT-2 inhibitor CVOTs to read out positively (though the first to not achieve superiority on 3-point MACE – though we believe this is due to the study population being healthier overall). Jardiance is the only member of the class currently with an FDA and EMA CV indication based on results from EMPA-REG OUTCOME (14% RRR on 3-point MACE, p=0.038) while Invokana has received an EMA indication based on results from CANVAS (14% RRR on 3-point MACE, p=0.0158), with a decision from the FDA expected in October.

Check out our full analysis of the results, the importance of the trial’s unique design, and our questions below:

Cardiovascular Outcomes

  • We believe that AZ will file for an indication based on DECLARE results with the FDA; no specifics on the potential label change were given. In particular, we’ll be interested to hear which patient population Farxiga may be indicated for. All other CV indications for diabetes therapies are exclusively for those with established CV disease, whereas the majority of the population enrolled in DECLARE were only at-risk (see below). Increasingly, thought leaders have been highlighting lack of generalizability from diabetes CVOTs, since the typical enriched patient population featured in these trials doesn’t accurately reflect the real-world diabetes patient population.

    • At ACC 2018, for instance, Dr. Lawrence Leiter called for diabetes CVOTs that start with “less sick” patients and expose them to an advanced agent for a longer period of time, while a poster at the same meeting found that only 21% of people with type 2 in the Diabetes Collaborative Registry would have been eligible for EMPA-REG OUTCOME (and only 30% for CANVAS and 46% for DECLARE, which is still low).

    • Since the superiority found for heart failure and CV death apply to the study’s entire population, we wonder whether regulatory agencies may be persuaded to include patients with CV risk, but not CVD, in an indication. Not only would this be a significant win for at-risk patients, it would also enable a much wider prescribing base for Farxiga compared to competitors Jardiance and Invokana.

    • That said, much work remains to be done in shifting the paradigm toward preventative medicine. As far as we can tell, few prescribers and even fewer patients understand the notion of taking a medicine to prevent adverse cardiovascular events, and widespread education on the subject would be required to maximize Farxiga’s penetration into the primary prevention market. Of course, AZ is well-equipped to help disseminate this education, with a robust portfolio of CV & diabetes products.

    • We wouldn’t be surprised if we saw DECLARE data stratified according to primary and secondary prevention populations in the full results presentation – November 10 at AHA 2018. While the results from the co-primary endpoints apply to the whole population, these breakdowns would provide much desired clarification as to where Farxiga’s benefit (and by proxy the whole SGLT-2 inhibitor class) lies. As it stands, SGLT-2s seem to be cardioprotective on heart failure in primary prevention patients but may only be cardioprotective on MACE in secondary prevention patients. We’ve heard this notion from several experts ever since a CANVAS post-hoc found heart failure benefit but not MACE benefit associated with Invokana in primary prevention patients (see Dr. Kenneth Mahaffey’s presentation from AHA 2017). That said, some thought leaders like Toronto’s Dr. Subodh Verma have pointed out that subclinical heart failure is quite prevalent in a diabetes patient population, meaning people who are classified as "primary prevention" for MACE (haven't yet had a heart attack or stroke) could still be "secondary prevention" due to their subclinical heart failure.

  • Perhaps the best-case scenario from DECLARE would have been a significant benefit on both 3-point MACE and heart failure/CV death. To be sure, there is a clear advantage to getting a CV indication for MACE on the product label, and this is why sponsors enrich CVOT populations. It’s hard for us to imagine FDA endorsing a MACE benefit on Farxiga’s US label no matter how “borderline” DECLARE results turn out to be. That said, we maintain that DECLARE results are a major win for AZ, especially given that thought leaders predicted a non-inferiority result for the MACE endpoint. And, having heart failure protection reflected on the product label could boost the Farxiga business as well, while also bringing more awareness to the diabetes/heart failure overlap.

  • As we see it, non-inferiority on the MACE endpoint does more to support cardioprotection as a class-effect of SGLT-2 inhibitors than it does to refute it. The evidence certainly favors CV benefit as a class effect, and most thought leaders have espoused this view, although there are notable exceptions (e.g. Dr. Sanjay Kaul at CMHC West). By evidence, we mean EMPA-REG OUTCOME for Jardiance, CANVAS for Invokana, AZ’s CVD-REAL + CVD-REAL 2, and J&J’s EASEL; the last three of these placed heart failure at the forefront of their real-world design. However, seeing as CANVAS and EMPA-REG demonstrated superiority on their 3-point MACE primary endpoint, Farxiga’s non-inferiority on this endpoint might be the first thing competitors point out. In our view, as long as the results align (also considering forthcoming meta-analyses), we certainly find that evidence points in the direction of CV benefit as a class effect. In this sense, there are parallels between the DECLARE story and the EXSCEL story for AZ’s GLP-1 agonist Bydureon (exenatide), and we’d point out that meta-analyses of GLP-1 CVOTs have supported a cardioprotective class effect.

Heart Failure and SGLT-2 inhibitors

  • The hype around SGLT-2s for heart failure began with significant risk reduction for heart failure hospitalization in EMPA-REG OUTCOME (35%), and it’s only grown since. Notably, the EMPA-REG OUTCOME result led to a positive opinion from CHMP to include additional heart failure data from the study on all empagliflozin-containing therapies (Jardiance, Synjardy, and Glyxambi), in addition to the 35% reduction in hospitalization for heart failure with empagliflozin vs. placebo already on these European product labels (the FDA label does not mention heart failure data at all). CANVAS found a similar risk reduction for heart failure hospitalization with J&J’s Invokana vs. placebo (33%). Importantly, however, this data all pertained to secondary endpoints, meaning that a specific indication for heart failure could not be included in the label. Needless to say, we’re eager to hear how Farxiga measures up to these (impressive) results from competitors, though it seems at this point that protection against heart failure is a class effect of SGLT-2 inhibitors – Dr. Verma has described SGLT-2 inhibitors as a potential “sweet spot” for heart failure management (in and outside the context of type 2 diabetes), advocating for further investigation.

    • Heart failure has for too long been ignored as a serious diabetes complication. Diabetes and prediabetes have both been inextricably associated with heart failure – people with type 2 diabetes are 2-2.5 times more likely to develop congestive heart failure than people without diabetes – and nearly six million Americans live with the condition, leading to nearly one million hospitalizations each year in the US. However, less focus has been placed in treating heart failure to date since the condition is chronic, and not as immediately traumatizing or visible as heart attack, stroke, or CV death. Heart failure was left out of FDA’s 2008 CVOT guidance. All this said, a first-in-class indication for heart failure could be a gamechanger for patients and prescribers.

    • KOLs have suggested that SGLT-2 inhibitors’ mechanism of cardioprotection is likely centered around reduced frequency and severity of heart failure. At WTCD 2017, Dr. Francesco Giorgino proposed this as one explanation why empagliflozin’s three-point MACE benefit in EMPA-REG was driven by CV death rather than one of the atherosclerotic component endpoints (stroke, MI).

  • The DECLARE results bode very well for Dapa-HF, DELIVER, and the EMPEROR studies of SGLT-2s in chronic heart failure patients (with or without type 2 diabetes). AZ initiated the Dapa-HF study in February 2017 to study Farxiga in heart failure with reduced ejection fraction (HFrEF), and very recently added the DELIVER CVOT in heart failure with preservered ejection fraction (HRpEF). Lilly/BI are also conducting dedicated outcomes trials of Jardiance in both forms of heart failure (reduced and preserved ejection fraction) in the EMPEROR HF program. As Dr. Mikhail Kosiborod articulated at CODHy, HFpEF is closely associated with adiposity and insulin resistance, which makes it especially relevant in diabetes/obesity care. Moreover, HFpEF is less responsive to traditional myocardium-targeting heart failure drugs, and SGLT-2s could fill this need. Rates of HFpEF have climbed alongside the diabetes and obesity epidemics, and it’s high time we ramped up our therapeutic toolkit to address this cardiovascular complication. We do continue wonder why AZ’s initial outcomes study in heart failure was focused on HFrEF rather than HFpEF. The company is also sponsoring PRESERVED-HF in patients with heart failure with preserved ejection fraction, but notably, this is not an outcomes trial and only measures biomarkers rather than adjudicated HHF events.

Safety and Non-Cardiovascular Outcomes

  • All-important safety data were not shared in AZ’s topline press release, apart from a statement that DECLARE confirmed the well-established safety profile of Farxiga.  Amputations concerns with SGLT-2 inhibitors have remained at the forefront of discussion since CANVAS read out at ADA 2017. A provocative cohort study from ESC identified a class-wide lower limb amputation signal with SGLT-2s compared to GLP-1s (HR=2.23, 95% CI=1.37-3.91). Notably, only 1% of the study population was on Invokana – the only SGLT-2 inhibitor with an FDA black box warning for lower limb amputation (though EMA has included information about the amputations observed with canagliflozin in all SGLT-2 -labels, in a more conservative approach). We imagine due to the broader study population and not as many sick patients, better safety data may well come from DECLARE, though we do not believe the safety data should be compared, apples to apples.

  • We’re also eager to see renal data from the study, as kidney protection could be another key class-wide benefit of SGLT-2 inhibitors. We can’t wait to see the results, especially in light of the recent news on J&J’s CREDENCE (Invokana in DKD), which was stopped a year early for meeting its primary endpoint ~one year ahead of schedule (more below). DKD is notoriously challenging to treat, and SGLT-2s could fill an enormous unmet need in slowing down the decline of renal function. Moreover, the results bode well for Lilly/BI’s EMPA-KIDNEY (Jardiance in CKD) and for AZ’s Dapa-CKD (Farxiga in CKD), both of which enroll patients with and without type 2 diabetes.

Study Design

  • The decision by AstraZeneca to have co-primary endpoints was bold from a statistical perspective in our view, as a p-value of less than 0.025 was required to confer superiority in either of the outcomes. As we understand it, this decision was made after DECLARE had begun and positive results from EMPA-REG OUTCOME for empagliflozin were released, suggesting heart failure benefit. There is certainly a higher degree of risk with this study design, as it is more difficult to meet a p-value of 0.025 vs. 0.05. For comparison, EMPA-REG’s p-value on the 3-point MACE was 0.038, meaning that superiority would not have been supported should a co-primary endpoint have been established. As such, the superiority on heart failure and CV death found with Farxiga is sure to be extremely significant and compelling, and the 3-point MACE may have heavily favored dapagliflozin without achieving superiority. In general, despite this point, we don’t encourage “apples to apples” comparisons on these trials given the different populations and design.

    • Regarding the non-inferiority found on the 3-point MACE, we are reminded of EXSCEL (AZ’s CVOT for GLP-1 agonist exenatide), which missed out on superiority vs. placebo by the narrowest of margins (p=0.06, upper 95% CI bound of 1.00). This result was widely attributed by KOLs to the study’s low adherence and pragmatic design (no run-in period, concomitant medications and less patient-friendly reconstitution kit allowed), and it is acknowledged that EXSCEL does more to support cardioprotection as a class effect of GLP-1 agonists than refute it. DECLARE is certainly very different from EXSCEL; to our knowledge, it is a much more well-controlled study with a significantly greater follow-up period and stricter guidelines. That said, we’ve heard from thought leaders that superiority over placebo on the 3-point MACE in DECLARE was unlikely because of the large primary prevention population (patients were "less sick" across the board compared to EMPA-REG OUTCOME or CANVAS participants). Dr. Itamar Raz, a DECLARE investigator,  discussed this at WCTD 2017 and predicted this exact result. Because of this, we see a general parallel between EXSCEL and DECLARE in that the design of each made the study less likely to achieve the MACE endpoint. Surely, forthcoming meta-analyses and post-hocs of DECLARE data will shed further light into whether the study’s design contributed to the non-inferior MACE result. We applaud AZ for its commitment to broader study design.

  • DECLARE is a massive study – larger than the previous two SGLT-2 inhibitor CVOTs combined. In total, the program enrollmed 17,160 patients, 6,971 (40.6%) of whom had atherosclerotic CV disease (secondary prevention), and 10,189 (59.4%) of whom had multiple risk factors for CV disease, defined as men older than 55 or women older than 60 with at least one of dyslipidemia, hypertension, or smoking. In fact, the primary prevention cohort of DECLARE alone is bigger than the total enrolled populations of EMPA-REG OUTCOME (n=7,020, <1% primary prevention) or CANVAS (n=10,142, 34% primary prevention). DECLARE is also the longest of the three SGLT-2 CVOTs with median of 4.5 years of follow-up collected, while EMPA-REG OUTCOME and CANVAS averaged follow-ups of 3.1 and 3.6 years, respectively. Needless to say, the duration and population of DECLARE will certainly provide added insight into SGLT-2 inhibitors’ potential in populations and outcomes beyond those addressed in EMPA-REG and CANVAS.

Close Concerns Questions

Q: What were the p-values for each of the primary endpoints? Would dapagliflozin have met statistical significance on three-point MACE were it the only primary endpoint in this study?

Q: What indication will AZ seek with DECLARE data? Could this be the first CV indication for a diabetes product that extends to those patients without established cardiovascular disease? Will any dual indications be sought? For what population will the indication be approved?

Q: What were the hazard ratios of the co-primary endpoints in both the primary vs. secondary prevention populations? Will these analyses at all affect which indication AZ seeks, or what’s approved by FDA and EMA?

Q: When will meta-analyses on the 3-point MACE be undertaken and completed? How do DECLARE results align (or not align) with findings from EMPA-REG OUTCOME and CANVAS?

Q: What was the safety profile of Farxiga? Was there an amputation signal? DKA signal?

Q: What initiatives will AZ undertake to promote Farxiga as a heart failure and primary prevention medicine? Will the company more so target prescribers or patients?

Q: How might DECLARE results play in to the FDA’s CVOT Advisory Committee next month?

Q: What do these results tell us about SGLT-2 inhibitors’ mechanism of cardioprotection? Is the CV benefit (even from EMPA-REG and CANVAS) driven by an advantage on heart failure outcomes? Do these molecules offer cardioprotection independent of glycemic control (in which case, they could be heart failure medicine in their own right, even without a diabetes diagnosis)?

Q: What do the renal data look like, and will DECLARE support any renal-related updates to Farxiga’s product label, either in the US or Europe?

 

--by Peter Rentzepis, Payal Marathe, and Kelly Close