American Diabetes Association 77th Scientific Sessions

June 9-13, 2017; San Diego, CA; Full CANVAS Results – Draft

Executive Highlights

We’re back with our full coverage of the results from the CANVAS trial for J&J’s SGLT-2 inhibitor Invokana (canagliflozin), which were presented to a packed auditorium at ADA 2017 on Monday, immediately after the DEVOTE CVOT results. Read on below for our initial themes related to the results, followed by detailed coverage of the results and discussant presentations. For those looking for more, see our initial breaking news report on the full results, and the full paper published in NEJM (which, by the way, has evidently become THE journal for diabetes CVOTs – the DEVOTE publication can be found in the latest issue as well, and NEJM previously published the SAVOR-TIMI, EXAMINE, TECOS, ELIXA, EMPA-REG OUTCOME, LEADER, and SUSTAIN 6 results. 


  • First, the good news – the CANVAS program demonstrated the cardiovascular superiority of Invokana (canagliflozin), as measured by the primary endpoint of three-point MACE (non-fatal MI, non-fatal stroke, and CV death). In the program, canagliflozin therapy reduced risk of MACE events by 14% (95% CI:0.75-0.97, p=0.0158 for superiority). Further, canagliflozin therapy produced an extremely impressive benefit for a secondary endpoint of hospitalization for heart failure, reducing risk by 33% (95% CI:0.52-0.87). As the second SGLT-2 inhibitor to report CV results – following the highly unexpected and impressive EMPA-REG OUTCOME findings demonstrating a CV benefit for Lilly/BI’s Jardiance (empagliflozin) – it’s perhaps only natural that many in the field are already drawing direct comparisons between the two trials, even though we stress they are not directly comparable for a number of reasons outlined below. Indeed, however, the CANVAS investigators themselves directly compared the hazard ratios and confidence intervals of a number of key endpoints across CANVAS and EMPA-REG OUTCOME – see the Dr. David Matthews “Implications for Clinical Practice” section below for more on this. 
    • Several key differences in both the participant population and CV findings exist between the CANVAS and EMPA-REG OUTCOME trials. In terms of participant population, EMPA-REG OUTCOME exclusively enrolled those with a prior history of CV disease – thus, the generalizability of the results to a lower-risk patient population has been a major question in the 21 months since EMPA-REG OUTCOME was presented at EASD 2015. On the other hand, CANVAS sheds some light on this question – about one-third of participants enrolled in the trial had risk factors for cardiovascular disease but no established history, thus creating a primary prevention cohort. There was no heterogeneity of benefit (p-value for interaction was non-significant) between the primary prevention and secondary prevention cohort in the trial (see Dr. Bruce Neal’s presentation of the Effects on Cardiovascular Outcomes below for more detail on this). Thus, there’s some early suggestion of CV benefit for SGLT-2 inhibitors even in a primary prevention cohort – we expect the DECLARE CVOT for AZ’s Farxiga, with its 17,000+ participant population, will offer additional information on this front when it reports in 2019.
    • In terms of CV endpoints, CANVAS did not demonstrate a CV or all-cause mortality benefit for canagliflozin. While the point estimate for the hazard ratio of both findings was in the “right direction,” the confidence intervals crossed the line of unity and just missed statistical significance. For all-cause mortality, the hazard ratio was 0.87 (95% CI:0.74-1.01, p=0.24) – this was particularly disappointing since this was the second endpoint in the pre-specified hierarchical testing structure and thus the other endpoint results in the study can only be considered exploratory. For CV death, the hazard ratio was 0.87 (95% CI: 0.72-1.06). This contrasts with the strong 38% risk reduction for CV death (p<0.0001) and 32% risk reduction for all-cause death (p<0.001) in the EMPA-REG OUTCOME trial. It’s unclear what kind of label indication the CANVAS findings will be able to support, given that Jardiance is only indicated for CV death rather than for its primary MACE finding (in fact, some on the FDA Advisory Committee felt that the p-value for the primary endpoint in EMPA-REG OUTCOME may not be convincing enough – though the p-value for superiority in CANVAS is more compelling). On the plus side, canagliflozin did not demonstrate a signal for increased stroke as empagliflozin did – in fact, the point estimate for stroke and all other components of expanded MACE (including hospitalization for unstable angina) were on the right side of unity, a reassuring finding indeed.
  • All in all, the CV findings of the CANVAS program offer further reassurance of the “validity” of the EMPA-REG OUTCOME benefit and suggest that the CV benefit, at least, is a class effect for SGLT-2 inhibitors. Given the highly unexpected and unprecedented nature of EMPA-REG OUTCOME, the question of class effect has been one of THE most frequently asked questions regarding SGLT-2 inhibitors and CV benefit over the last two years. The other most popular question is “what is the mechanism of benefit?” A non-inferiority CV finding in CANVAS would have potentially thrown into question the validity of the EMPA-REG OUTCOME benefit and may have led some to suggest the benefit demonstrated in the trial was a statistical fluke. Indeed, many key thought leaders – who often make up diabetes guidelines committees – have expressed the desire for more data from other agents in the class before wholeheartedly recommending SGLT-2 inhibitors (or even just empagliflozin) as a preferred therapy for those with a history of cardiovascular disease. With two positive CVOTs for the SGLT-2 inhibitor class, we imagine we might see more enthusiastic recommendations for the class for CV benefit in future guidelines. Thus far, the ADA 2017 Standards of Care recommend only the consideration of empagliflozin (and GLP-1 agonist liraglutide) in those with established CV disease and the AACE guidelines have not yet changed recommendations based on CVOT data. On balance, we think the overall CANVAS findings appear at this stage to be more positive for Jardiance than for Invokana. We expect this further confirmation of CV benefit will boost the SGLT-2 inhibitor class as a whole and we expect Jardiance in particular will benefit, given its compelling CV benefit (and indication), as well as the lack of worrisome and unexpected safety signals in the trial (more on this below). Indeed, Invokana has already been losing market share largely to Jardiance (while Farxiga share has been relatively sustained) for several quarters now, and we expect this trend may continue in future quarters– this is not at all unexpected given that all “first to market” compounds do lose share. We do believe the unexplained amputation outcome, in particular, will be a focus of manufacturers, and we hope that Janssen is able to share more data on this front. Overall, while it’s too early to say definitively how patients and providers will react to these results, we expect the messaging from all three SGLT-2 inhibitor manufacturers to be positive on CV and mixed on amputation and fractures (we believe there is also more to be known about these risks for other manufactures). 
  • The more negative headline from CANVAS/CANVAS-R is the near doubling of lower-extremity amputation risk with canagliflozin vs. placebo (HR=1.97, 95% CI: 1.41-2.75, P<0.001). Although it is a low absolute risk overall, the risk appeared to occur across all patients, not just those who had previous amputations. That said, all patients in the study had a degree of elevated risk, as two-thirds of participants had a previous history of a CV events and one-third of participants had at least two cardiovascular risk factors. There were 187 lower limb amputations across these two outcomes trials, occurring at a rate of 6.3/1,000 patient-years in the canagliflozin arm vs. 3.4/1,000 patient-years in the placebo arm. Approximately 71% of these amputations occurred at the level of the toes or forefoot, though some were above the ankle and a smaller number were above the knee. This leads to a murky risk/benefit profile for Invokana on the face of it though there are a number of confounders. On the negative side, Dr. David Matthews shared, for every 1,000 patients treated with canagliflozin over five years, we can expect 23 fewer MACE events, 17 fewer hospitalizations for heart failure, 16 fewer renal composite endpoints (renal death, renal replacement therapy, or 40% reduction in eGFR), and 15 additional lower limb amputations – 10 at the level of toe/forefoot, five above the ankle. What makes this even more challenging is that canagliflozin doubled amputation risk across all risk factors, including prior amputation and baseline peripheral vascular disease. Without any insight into what is mediating canagliflozin’s adverse impact on the lower limbs, the risk seems more likely to be associated with the molecule itself. This means that HCPs won’t be able to tailor prescriptions based on patient disposition; that is to say, if the effect on lower limb amputations was driven by a particular subgroup, such as participants with baseline peripheral vascular disease, this safety issue might be circumvented with a single black box warning not to prescribe Invokana to patients with peripheral vascular disease.
    • As a reminder, this trial included a very broad set of patients. In total, over 10,000 patients were enrolled and, notably, the CANVAS program enrolled both those with established cardiovascular disease at baseline (a secondary prevention cohort) and those with two or more CV risk factors but no history of cardiovascular disease (a primary prevention cohort). This contrasts with the EMPA-REG OUTCOME trial, which enrolled only patients with a history of cardiovascular disease at baseline.
    • We have very little information or insight at this stage as to how this will all play out in real-world prescribing habits and formulary negotiations. Dr. Dan Drucker foreshadowed a slow attrition away from Invokana, with new SGLT-2 starts favoring Lilly/BI’s Jardiance (with demonstrated cardioprotection and no associated amputation risk) or AZ’s Farxiga. On the other hand, Dr. Anne Peters suggested that some providers may keep patients on canagliflozin due to formulary requirements, or because of greater glucose-lowering and weight loss vs. empagliflozin (in her anecdotal clinical experience). While some have suggested a worst-case scenario of HCPs steering clear of SGLT-2 inhibitor products in general, once they learn of lower limb amputation risk, we doubt that will happen. While the EMPA-REG study group did not observe a similar increase in amputations in the EMPA-REG OUTCOME trial, that was a shorter trial with a more homogenous population.
    • While the heightened risk in CANVAS/CANVAS-R appeared fairly early on, at year one, the events were adjudicated very differently – with CANVAS/CANVAS-R prospectively and EMPA-REG retrospectively.
      • This reinforces yet again the need for more standardization on the global regulatory front. We expect to see manufacturers of SGLT-2 agents come together and invest in real-world data collection on this and other safety issues – we’d hate for lower-extremity amputations to define the narrative for this class, given the profound glycemic, CV, renal, and weight benefits and given the lack of standardized data from all trials.
    • This safety data from the CANVAS program also reminds us that strong education on foot care for people with diabetes is lacking. In fact, recent conversations have reinforced that this signal could shine a light on the importance of greater education on foot care for people with diabetes – we hope J&J and others will be a leader in expanding understanding/management of this complication. Currently, we don’t believe patients have much education at all about how to minimize this risk and in fact, perhaps key education could provide significantly more benefit. In general, messaging around this potential safety concern will be crucial not only for J&J, but for Lilly/BI around the Jardiance franchise and for AZ around the Farxiga franchise as well.
  • The positive renal data was a key bright spot in the CANVAS results, and contributes to mounting evidence for a beneficial impact of SGLT-2 inhibitors in general on renal complications of diabetes. Canagliflozin showed impressive signs of renal protection, including a 27% risk reduction for progression to albuminuria and a 40% risk reduction for a renal composite outcome (encompassing renal death, renal replacement therapy, or 40% reduction in eGFR). This comes on the heels of positive renal outcomes in the EMPA-REG OUTCOME trial, where empagliflozin showed a 39% risk reduction (HR: 0.61; 95% CI: 0.53-0.70; p<0.001) for the trial’s main renal endpoint of incident or worsening nephropathy (progression to macroalbuminuria or doubling of serum creatinine accompanied by eGFR ≤45 ml/min/1.73 m2 or renal replacement therapy or death due to renal disease). These results are truly profound given the currently enormous unmet need for new therapies to treat chronic kidney disease, and to this end all three of the main SGLT-2 inhibitors on the market are now being explicitly studied for renal outcomes – canagliflozin in the CREDENCE trial, dapagliflozin in the Dapa-CKD trial, and empagliflozin in a just-announced chronic kidney disease trial from Lilly/BI. 

Detailed Discussion and Commentary

Symposium: The Integrated Results of the CANVAS Program

Background to the Design of the Trials

Gregory Fulcher, MD (University of Sydney, Australia)

Dr. Gregory Fulcher opened the symposium with a brief history of the SGLT-2 inhibitor drug class and of the FDA’s post-marketing CVOT requirement. SGLT-2 inhibitors work by increasing urinary glucose excretion, thereby lowering glucose levels in the blood. This mechanism gives reason to believe that SGLT-2 agents may be cardioprotective: For one, hyperglycemia in itself is a risk factor for CV events. SGLT-2 inhibitors also reduce blood pressure via osmotic diuresis, promote weight loss (by excreting more calories through the urine), and reduce albuminuria – each of these independently is a risk factor for CV morbidity and mortality, and there are hypotheses circulating in the diabetes field that SGLT-2 inhibitors (including canagliflozin) might confer CV benefit by attacking a variety of contributing risk factors. Importantly, this is all speculation at this stage, as research is ongoing to determine mechanism of cardioprotection. Switching gears, Dr. Fulcher described the intention behind the FDA’s post-marketing CVOT requirement, which went into effect in 2008 – these trials should demonstrate that a new anti-diabetic drug is not associated with an “unacceptable increase” in CV risk. The upper bound of the 95% confidence interval for hazard ratio on a primary CV endpoint (most often, three-point MACE consisting of non-fatal MI, non-fatal stroke, and CV death) must be <1.8 pre-approval in order to win marketing authorization. Post-approval, a larger, dedicated CVOT must demonstrate a hazard ratio <1.3 to convincingly show non-inferiority vs. placebo. The FDA does not explicitly encourage demonstration of CV superiority, which would be achieved by a hazard ratio and confidence interval bound under 1.0. Lastly, Dr. Fulcher reminded everyone that canagliflozin is branded as Invokana by J&J, and is approved in 100 mg and 300 mg oral doses.

Methods for the Trials and the Integrated Analyses

Kenneth Mahaffey, MD (Stanford University, Palo Alto, CA)

Dr. Kenneth Mahaffey provided an overview of the trial design and data analysis scheme. The CANVAS program studied a total of 10,142 people across the CANVAS (n=4330) and CANVAS-R (n=5,812) trials. After a two-week placebo run-in period, participants in CANVAS were randomized to 100 mg or 300 mg canagliflozin or placebo and participants in CANVAS-R were randomized to canagliflozin 100 mg (with optional up-titration to 300 mg) or placebo. As enrollment criteria, all participants had an A1c ranging from 7% to 10.5%, an eGFR >30 mL/min/1.73 min2, and were aged >30 years with a history of a prior cardiovascular event or >50 years with at least two cardiovascular risk factors. The primary outcome was 3-point MACE (the composite of cardiovascular death, non-fatal MI, and non-fatal stroke). Secondary outcomes included all-cause mortality and cardiovascular death. Exploratory outcomes included non-fatal MI, non-fatal stroke, hospitalization for heart failure, hospitalization for heart failure or cardiovascular death, total hospitalizations, albuminuria progression, albuminuria regression, and a renal composite endpoint encompassing 40% reduction in eGFR, end-stage renal disease, or renal death. Importantly, the CANVAS program employed a hierarchical statistical testing structure – thus, secondary endpoints were assessed in a pre-specified order and, if a secondary endpoint failed, all endpoint results following that one in the hierarchy would be considered only exploratory. On the statistics front, Dr. Mahaffey explained that across the entire CANVAS program data all participants on canagliflozin were pooled together such that the statistical analyses evaluated a “strategy of canagliflozin treatment” versus placebo, without differentiating between the 100 mg and 300 mg doses in CANVAS or the possible up-titration from 100 mg to 300 mg in CANVAS-R. 

Effects on Cardiovascular Outcomes

Bruce Neal, MB, ChB, PhD (The George Institute for Global Health, Sydney, Australia)

Dr. Bruce Neal presented the cardiovascular results from the integrated CANVAS and CANVAS-R program. The room waited with bated breath as Dr. Neal discussed the patient populations, baseline characteristics, and metabolic outcomes in the program (detailed below). Finally, Dr. Neal presented the primary three-point MACE (non-fatal MI, non-fatal stroke, CV death) endpoint: canagliflozin therapy was associated with a 14% risk reduction compared to placebo (HR=0.86, 95% CI: 0.75-0.97, p<0.0001 for non-inferiority, p=0.0158 for superiority) – and the room prompting burst into applause and cheers, interrupting Dr. Neal mid-sentence. With a rueful smile, Dr. Neal acknowledged that he had waited eight years for this moment – and the CV results at least were certainly worth the wait! By individual study, the hazard ratio point estimate for the primary outcome were fairly consistent across CANVAS (HR=0.88, 95% CI:0.75-1.03) and CANVAS-R (HR=0.82, 95% CI: 0.66-1.01), though the results were not significant for either trial individually (likely due to a lack of power). The trial just missed superiority for all-cause mortality (HR=0.87, 95% CI:0.74-1.01, p=0.24) – since this was the second endpoint in the pre-specified hierarchical testing structure, the other endpoint results can only be considered exploratory.

  • Hospitalization for heart failure: Very notably, canagliflozin was associated with a substantial and impressive 33% reduction in risk for heart failure (HR=0.67, 95% CI: 0.52-0.87). Canagliflozin also reduced risk for a composite endpoint of CV death/heart failure hospitalization by 22% (HR=0.78, 95% CI:0.67 to 0.91). Notably, the Kaplan-Meier curve for hospitalization for heart failure especially appeared to diverge early and drastically, mirroring the results we saw in the EMPA-REG OUTCOME trial for Lilly/BI’s Jardiance (empagliflozin).

  • Individual components of MACE primary outcome: The hazard ratio for individual components of MACE trended in the right direction favoring canagliflozin, but none of them reached statistical significance even though three-point MACE did. For non-fatal MI, the hazard ratio was 0.85 (95% CI: 0.69-1.05). For non-fatal stroke, the hazard ratio was 0.90 (95% CI: 0.71-1.15). And for CV death, the hazard ratio was 0.87 (95% CI: 0.72-1.06).

  • Canagliflozin produced pronounced effects on a number of metabolic and cardiovascular risk factors. The mean A1c difference between the canagliflozin and the placebo groups was 0.58% – canagliflozin produced a large drop in A1c within the first few months post-randomization that attenuated over the six year follow-up period. Canagliflozin also produced a mean placebo-adjusted systolic blood pressure reduction of 3.93 mmHg, which occurred early in the trial (within the first year) and was sustained throughout the six-year follow-up. Finally, participants on canagliflozin experienced a mean placebo-adjusted weight loss of 1.6 kg (3.5 lbs) that also occurred in the first year and was sustained throughout the follow-up period.
  • Together, CANVAS and CANVAS-R enrolled 10,142 patients (4330 in CANVAS and 5813 in CANVAS-R). Across the program, 96% of participants completed the study and end-of-trial vital status was known for over 99% of participants. Mean follow-up across the program was 188 weeks – CANVAS initiated first and mean follow-up in that trial was 296 weeks, while mean follow-up in CANVAS-R was 108 weeks. At the end of the trial, 71% of canagliflozin and 70% of placebo patients remained on their assigned intervention.
  • Baseline characteristics were well-matched between the combined canagliflozin arms and placebo. Mean age across the study was 63 years old, with a mean diabetes duration of 14 years. Very notably, about two-thirds of participants had a history of cardiovascular disease at baseline (and one-third had CV risk factors but no history of CV disease). 90% of participants in the canagliflozin group had hypertension at baseline (vs. 90% in the placebo group) and 14% of participants randomized to canagliflozin had heart failure at baseline (vs. 15% of the placebo group).
    • In terms of demographic baseline characteristics, about one-third of participants were female, nearly 80% of participants were white, and about 60% of participants were from North America or Europe. This is fairly typical for a global CVOT, though greater racial/ethnic and geographic representation in trials overall is needed, and we’d like to see this emphasized to a greater degree, particularly given the extent to which under-represented minorities have an outsized risk of type 2 diabetes and type 2 diabetes complications.
    • Participants in the CANVAS program were on a variety of background diabetes and cardiovascular medications. Nearly 80% of participants were taking metformin at baseline, about half were on insulin, 42%-44% were taking SUs, and 12%-13% were taking DPP-4 inhibitors. Notably, only 4% of participants in each arm were taking a GLP-1 agonist at baseline – we would love to see an analysis of cardiovascular and other outcomes in patients on dual canagliflozin and GLP-1 agonist therapy, but we assume that the small number of participants in this subgroup would make such an analysis difficult. In terms of cardiovascular medications, nearly 75%-80% of participants were on RAAS inhibitors, statins, and antithrombotic agents, respectively. A little more than half were on beta blockers and a little less than half were taking a diuretic.
  • Dr. Neal also presented several subgroup analyses that probed the potential for heterogeneity of effect among different patient populations. The p-values for interaction were all non-significant for age (<65 years old or ≥65 years old), sex, race (White, Black/African American, Asian, or other), and region (North America, Central/South America, Europe, and Rest of the world). Similarly, there was no significant interaction for various risk factors, including BMI, blood pressure control, duration of diabetes, A1c, or eGFR. The primary outcome benefit was also consistent regardless of CV disease, peripheral vascular disease, heart failure, or amputation history.
    • In fact, the only potential heterogeneity of effect was observed in two background therapy subgroups. Those taking beta blockers at baseline were significantly more likely to benefit from canagliflozin therapy than those who were not (HR=0.75 for those on beta blockers, 95% CI: 0.64-0.88 vs. HR=1.04 for those not on beta blockers, 95% CI:0.85-1.28, p=0.01). Furthermore, the heterogeneity of benefit was even more pronounced among the subgroups of those taking or not taking a diuretic at baseline: the HR for those on a diuretic was 0.66 (95% CI:0.56-0.79) and the HR for those not on a diuretic was 1.11 (95% CI: 0.93-1.34) – the p-value for interaction was a highly-significant p<0.001. There was no significant difference in benefit among subgroups of patients taking insulin, statins, antithrombotic agents, or RAAS inhibitors at baseline.  

Effects on Renal Outcomes

Dick De Zeeuw, MD (University Medical Center Groningen, Netherlands)

Next up, Dr. Dick De Zeeuw presented the effects of canagliflozin on renal outcomes. To complement its impressive cardioprotective effects, canagliflozin also showed signs of renal protection, including a 27% risk reduction for progression to albuminuria (HR=0.73, 95% CI: 0.67-0.79), a 70% increase in the regression of albuminuria (HR=1.70, 95% CI: 1.51-1.91), and a 40% risk reduction for the composite endpoint of renal death, renal replacement therapy, or 40% reduction in eGFR (HR=0.60, 95% CI: 0.47-0.77). Canagliflozin was further associated with an 18% (95% CI: -16% to -20%) reduction in urinary albumin:creatinine ratio (UACR). For the subset of participants with microalbuminuria and macroalbuminuria, canagliflozin decreased UACR by 34% and 36%, respectively. Together, these results suggest a potential renal-protective effect of canagliflozin treatment in people with type 2 diabetes, and Dr. De Zeeuw underscored these findings as a highlight of the CANVAS and CANVAS-R integrated full results. This data is particularly impressive in light of the fact that the CANVAS program enrolled a low renal risk population in which 70% of participants had normoalbuminuria (an albumin:creatinine ratio <30 mg/g) and a sizeable proportion had a normal eGFR (25% and 24% of participants in the canagliflozin and placebo arms respectively had mean eGFR >90 ml/min/1.73 m2, and 56% and 54% had mean eGFR between 60-90 ml/min/1.73 m2). Our curiosity is piqued for J&J’s CREDENCE trial investigating Invokana specifically for renal outcomes in patients with type 2 diabetes and diabetic kidney disease (expected to complete in June 2019).

Effects on Safety Outcomes

Vlado Perkovic, MD (Georgia Institute for Global Health, Sydney, Australia)

The safety data makes CANVAS a more complicated story. Dr. Vlado Perkovic presented these results, the headline being a near doubling of lower limb amputations with canagliflozin vs. placebo (HR=1.97, 95% CI: 1.41-2.75, p<0.001). In total, there were 187 lower limb amputations across the two outcomes trials (CANVAS and CANVAS-R), occurring at a rate of 6.3/1,000 patient-years in the canagliflozin arm vs. 3.4/1,000 patient-years in the placebo arm. Approximately 71% of these amputations occurred at the level of the toes or metatarsals, though some were above the ankle and a smaller number were above the knee. The Kaplan-Meier curves for lower-extremity amputations separate around ~year one and continue to diverge for six years post-randomization. Dr. Perkovic presented a multivariate analysis of common risk factors for amputation, which disappointingly showed consistently higher risk with canagliflozin vs. placebo, regardless of a patient’s prior history of amputation, peripheral vascular disease, etc. (see below for a breakdown of this data). While prior history of amputation leads to the greatest chance of a subsequent amputation, this risk is still intensified to the same degree by canagliflozin vs. placebo. Without any insight into what is mediating canagliflozin’s adverse impact on the lower limbs, most experts with whom we have spoken (admittedly still a small number) say that the risk appears to be associated with the molecule itself. This suggests that HCPs won’t be able to tailor prescriptions based on patient disposition (i.e. if the effect on lower limb amputations was driven by a particular subgroup. For example, if this risk were limited to participants with a risk like baseline peripheral vascular disease, this safety issue might be circumvented with a single black box warning not to prescribe Invokana to patients with peripheral vascular disease. At present, that does not appear to be the case. The FDA initiated an investigation of lower limb amputations based on interim CANVAS results in May 2016, and then just last month issued a boxed warning on all products in the Invokana family, so of course we knew this was likely to be an upsetting outcome regardless of the CV data. 

  • J&J now faces the very challenging task of promoting a favorable risk/benefit profile for its SGLT-2 inhibitor. Will patients risk a toe (or possibly lower limb) for the chance to prevent MI, stroke, and CV death with other in-class choices available? Will providers prescribe Invokana when there’s another SGLT-2 inhibitor with the same demonstrated risk reduction for three-point MACE and no signal for amputations? How will this all be reflected on formularies? We’ve heard mixed opinions from thought leaders so far: Some say they’ll keep patients on Invokana, while others advocate for switching to Jardiance. Some suggest (with hope!) that the lower limb amputation risk could be manageable if further investigations identify a confounding variable apart from the canagliflozin molecule itself – we do think it’s critical to separate out the confounders, like, for example, adjudication. Others are wary that this risk may apply to empagliflozin as well and the EMPA-REG OUTCOME trial just wasn’t long enough to pick up on it. We’re hopeful that this latter point isn’t the case, since the heightened amputation risk with canagliflozin vs. placebo appears early on, beginning at year one. We’ll be heartbroken if amputation risk comes to define the narrative surrounding first-in-class SGLT-2 inhibitor Invokana or the drug class more broadly, given demonstrated CV superiority and renal benefits. We can foresee a future, however, where busy HCPs steer clear of any product in this class because of concerns surrounding lower limb amputation. Or, perhaps this risk will hurt the Invokana business but will encourage patient switches to Jardiance or Farxiga (AZ’s dapagliflozin). We’ll be watching these trends closely. Bottom line: This is a major safety concern that can’t be ignored, nor would we expect it to be by providers, patients, payers, or regulatory agencies. Amputation is a particularly vivid diabetes complication for patients and in the public eye, and we expect J&J will have to be exceptionally thoughtful in how it proceeds. Dr. List suggested to us that this signal could shine a light on the importance of greater education on foot care for people with diabetes – we hope J&J will be a leader in expanding understanding/management of this complication. Currently, we don’t believe patients have much education at all about how to minimize this risk and in fact, perhaps key education could provide significantly more benefit.
  • In a separate call with us, Dr. Robert Cuddihy (VP of Medical Affairs for Cardiovascular and Metabolism) underscored Janssen’s continued commitment to the Invokana franchise, and suggested that amputation risk could be managed with improved patient education on foot care and regular screening. As he pointed out, amputations didn’t appear out of nowhere, but were usually preceded by infection or another warning sign. Moreover, the patient population enrolled in CANVAS was particularly prone to this complication (with high CV risk corresponding to an elevated risk of lower-extremity amputations as well). In contrast, a metaanalysis of phase 3 canagliflozin studies found a much lower base rate of lower limb amputations in the background diabetes population vs. the high-risk diabetes population (0.6/1,000 patients vs. 2.4/1,000 patients, respectively). On the other hand, Dr. Cuddihy acknowledged that there’s something quite visceral about an amputation (vs. a heart attack) that may affect an individual patient’s decision on whether or not to start or continue Invokana treatment. Further post-hoc analyses of CANVAS and CANVAS-R will certainly provide more answers. Again, Dr. Cuddihy emphasized that J&J received the final data from these outcomes trials with little time remaining before ADA – continued analysis will come with more time.
  • The hazard ratio for low-trauma bone fractures was 1.23 in favor of placebo, but this did not reach statistical significance (95% CI: 0.99-1.52). Fractures were another safety concern to watch out for in CANVAS and CANVAS-R, given the FDA’s strengthened label warning on this front, and it was reassuring to see no statistically significant signal across the two outcomes trials. Interestingly, there was a statistically significant increase in fracture risk in CANVAS (HR=1.55, 95% CI: 1.21-1.97) but not in CANVAS-R (HR=0.86, 95% CI: 0.62-1.19), and Dr. Perkovic reported a statistically significant p-value of 0.005 for this interaction. Further investigations will have to unpack this and determine how concerned patients/providers should be about canagliflozin and bone health. Said Dr. Neal, “I’m not easily persuaded by things happening by chance, but this is a weird result. We don’t see a fracture signal in EMPA-REG OUTCOME, nor in CANVAS-R, nor in any of the other CANVAS trials, so it’s possibly by chance.”
  • As expected, genital mycotic infections were significantly more likely among females taking canagliflozin (HR=4.37, 95% CI: 2.78-6.88) and among males taking canagliflozin (HR=3.76, 95% CI: 2.91-4.86). DKA was another prespecified point-of-interest, but no statistically significant signal was detected. The DKA event rate was 0.6/1,000 patient-years in the canagliflozin group vs. 0.3/1,000 patient-years in the placebo group (HR=2.33, 95% CI: 0.76-7.17). Of 18 total cases of DKA, five patients were later found to have autoimmune/type 1 diabetes. In general, the concerns surrounding SGLT-2 inhibitors/DKA are more pronounced for a type 1 patient population, so this was a reassuring safety finding for canagliflozin as a type 2 diabetes therapy. That said, we continue to believe that SGLT-2 inhibitors could be advantageous for some patients with type 1, provided there is strong education on ketone monitoring and other DKA risk management.
  • For all other adverse events, there was no significant difference in frequency between the canagliflozin and placebo arms. Serious adverse events occurred at a rate of 104/1,000 patient-years among canagliflozin-treated participants and 120/1,000 patient-years among placebo-treated participants (HR=0.93, 95% CI: 0.87-1.00). There were 1,025 total adverse events that led to study discontinuation – 35/1,000 patient-years vs. 33/1,000 patient-years in the canagliflozin and placebo groups, respectively (HR=1.13, 95% CI: 0.99-1.28). Hospitalization due to any cause occurred at a rate of 119/1,000 patient-years in the canagliflozin group vs. 131/1,000 patient-years in the placebo group (HR=0.94, 95% CI: 0.88-1.00). Data was also collected on UTIs, hypoglycemia, osmotic diuresis, volume depletion, severe hypersensitivity/cutaneous reaction, hepatic injury, venous thromboembolic events, photosensitivity, and acute pancreatitis, with no statistically significant difference in event rates between the canagliflozin and placebo groups.

Implications for Clinical Practice

David Matthews, FRCP, DPhil (University of Oxford, UK)

To put the benefit-risk profile into perspective, Dr. David Matthews shared the expected incidence rate for a number of endpoints in the CANVAS program. For every 1000 patients treated with canagliflozin over five years, we can expect 23 fewer patients to experience a MACE event, 16 fewer patients to experience hospitalization for heart failure, and 17 fewer patients to experience the renal composite endpoint (renal death, renal replacement therapy, or 40% reduction in eGFR). In total, 56 fewer patients would be expected to experience on of these cardiovascular or renal events – this is especially notable considering that patients with diabetes face an excess residual cardiovascular risk and that few treatments are available for heart failure or diabetic nephropathy. On the other hand, these substantial wins would be expected to occur at the cost of 15 additional patients experiencing a lower-limb amputation (including 10 toe or forefoot amputations and five above-the-ankle amputations). For comparison, based on EMPA-REG OUTCOME, for every 1,000 patients treated over three years (a shorter time period than the CANVAS figures), clinicians can expect 22 fewer CV deaths, 25 fewer deaths overall, and 14 fewer hospitalizations for heart failure with empagliflozin therapy. The only strong safety signal observed in the trial was genital infections – treatment with empagliflozin in 1,000 patients would be expected to produce 53 additional genital infections over three years.

  • Dr. Matthews highlighted the analysis presented earlier by Dr. Neal a lack of significant interaction between cardiovascular disease status at baseline and the CV benefit for canagliflozin. He emphasized that there’s no clear difference in outcome or size of the hazard ratios or confidence intervals when subgroups of those with and without CV disease are examined. That said, he also underscored that the study is not powered to distinguish between the primary and secondary prevention populations (we may have to wait until the much larger DECLARE trial (n=17,000) for a clearer understanding of this). Overall, Dr. Matthews stated that results hold for a patient population that matches the one in CANVAS, that is one third primary prevention and two thirds secondary prevention.
  • Dr. Matthews also offered a direct comparison of primary and secondary endpoint hazard ratios and confidence intervals between CANVAS and EMPA-REG OUTCOME. While acknowledging the caveat that it’s extremely difficult to draw comparisons across trials due to differences in participant population, trial design, analytic approaches, and drug effects, Dr. Matthews nonetheless presented a helpful slide superimposing the key CANVAS hazard ratios and confidence intervals with that of EMPA-REG OUTCOME (see below). Overall, Dr. Matthews characterized the findings of the two trials as “effectively concordant” – indeed, it appears that the hazard ratios and confidence intervals between the two trials are generally very consistent. In fact, Dr. Matthews pointed out that the hazard ratios and confidence intervals for the primary MACE outcome in the two trials were almost identical. For comparison, the hazard ratio in the CANVAS program was 0.86 (95% CI: 0.75-0.97, p=0.0158 for superiority) and the hazard ratio in the EMPA-REG OUTCOME trial was also 0.86 (95% CI: 0.74-0.99, p<0.001 for non-inferiority, p=0.038 for superiority). Notably, the “nearly insignificant” p-value for the primary endpoint was a sticking point in the FDA Advisory Committee discussion on the Jardiance CV indication, so we’ll be curious to see how the FDA digests the CANVAS data for a label update. Additionally, the CV death component of MACE was not individually significant in CANVAS, whereas the risk of CV death was highly significantly reduced in EMPA-REG OUTCOME and the indication ultimately received for Jardiance was only for CV death. J&J management has already confirmed that the company will submit a Supplemental New Drug Application (sNDA) seeking an expanded indication to the FDA by the end of September – we’ll be following closely to see what kind of language this data will support from a regulatory standpoint (and in guidelines as well).
    • There also appears to be some heterogeneity in the non-fatal stroke results – while neither finding was significant, risk for non-fatal stroke trended toward reduction in CANVAS while there was actually signal for increased stroke in EMPA-REG OUTCOME. The confidence intervals for both were fairly wide and cross the line of unity in both trials, however, so we certainly don’t think any definitive comments can be made on the stroke effect of either drug or the class as a whole.

Independent Commentary

Clifford Bailey, MD (Aston University, Birmingham, UK)

In his independent commentary, Dr. Clifford Bailey put these integrated CANVAS/CANVAS-R results into context with EMPA-REG OUTCOME for Lilly/BI’s SGLT-2 inhibitor Jardiance (empagliflozin). He posited that CV and renal benefits are highly likely to be class effects for SGLT-2 inhibitors, based on the parallel data in CANVAS and EMPA-REG OUTCOME. The risk reduction for three-point MACE is exactly the same in both trials (14%). Canagliflozin demonstrated a 33% risk reduction for heart failure hospitalization, while empagliflozin demonstrated a 35% risk reduction for this endpoint (p=0.002 vs. placebo). Empagliflozin also significantly reduced risk for diabetic nephropathy by 39%, in comparison to canagliflozin’s 27% risk reduction for progression to albuminuria and 40% risk reduction for the composite endpoint of renal death, renal replacement therapy, or 40% reduction in eGFR. Dr. Bailey mentioned CVD-REAL, the AZ-sponsored real-world analysis that found a significant 39% risk reduction for heart failure hospitalization associated with SGLT-2 inhibitors (including Invokana, Jardiance, and Farxiga) vs. other glucose-lowering drugs (p<0.001). He suggested that all signs so far point to a cardioprotective class effect for SGLT-2 inhibitors, with especially strong signals for a heart failure benefit (though this will have to be determined through RCTs, such as Lilly/BI’s EMPEROR HF program and AZ’s Dapa-HF program). One of the clear bright spots of this new CVOT data is reassurance that EMPA-REG OUTCOME results were not a fluke. For now, these integrated results have created a complicated risk/benefit profile for Invokana, and Dr. Bailey emphasized that the near doubling of risk for lower-extremity amputations is a very real safety concern. Some have said the biggest win from CANVAS may actually manifest for the Jardiance franchise, since empagliflozin’s CV benefits now have more credence without any associated amputation risk – obviously, we would not imagine that anyone would see the troubling safety data as a win for any manufacturer or any patients.

  • Dr. Bailey drew attention to the different patient populations in CANVAS and EMPA-REG OUTCOME. Whereas 99% of EMPA-REG OUTCOME participants has established CV disease at baseline or were at very high CV risk, CANVAS enrolled a much lower proportion of high-risk individuals, only 65%. “The 65% is much more reminiscent of the population you’d see in routine practice,” Dr. Bailey explained, which highlights another important contribution of CANVAS to our understanding of SGLT-2 inhibitors. He argued that the positive CV effects of this therapy class may apply in primary as well as secondary prevention. If this primary cardioprotection is confirmed further by AZ’s CVOT DECLARE, which also enrolls a large subset of lower-risk participants, it could be truly transformative for diabetes care – we could give patients a glucose-lowering, weight-lowering, convenient oral medication that protects stroke, heart attacks, and CV death early on.
  • That the hazard ratio for stroke trended in the right direction in CANVAS is also reassuring, Dr. Bailey suggested. In EMPA-REG OUTCOME, strokes were actually more common in the treatment vs. placebo arm, though the 1.24 hazard ratio did not reach statistical significance (95% CI: 0.92-1.67, p=0.16). In a separate conversation with our team, Dr. Neal suggested that Lilly/BI may have been “unlucky” on this stroke endpoint. Ideally, further investigations of mechanism help elucidate differential effects of canagliflozin and empagliflozin on non-fatal MI, non-fatal stroke, and CV death. For now, we hope the focus remains on the significant risk reduction for three-point MACE and CV events in general. Janssen’s Dr. List announced that a Supplemental New Drug Application (sNDA) for inclusion of the canagliflozin CV data on the Invokana label will be submitted to the FDA by the end of September.

Close Concerns Questions

Q: What are the risk factors for amputation? Are there clear patient populations that should avoid Invokana?

Q: What are the logical next steps to investigate amputation risk using data from databases used in the CVD-REAL study or comparable datasets such as those used by major US insurance companies? Since Invokana has been out the longest, what is the best way to compare compounds?

Q: What is the best way to interpret the different adjudication approaches of amputations used in various CVOTs?

Q: To what extent might the amputation risk be a class effect?

Q: How much will J&J continue to invest in the Invokana franchise? There has previously been discussion of a prediabetes CVOT for the drug, as well as the potential for a phase 3 obesity program for the combination of canagliflozin and phentermine.

Q: What’s going to happen to prescribing habits? Is there a reason an HCP might still prescribe Invokana over Jardiance beyond formulary and how would typical clinicians trade off better glycemic or weight advantages associated (at least for some patients) with the relatively higher amputation risk (despite low overall numbers)? Will the amputation risk is going to deter HCPs from Invokana, specifically, or perhaps from the whole SGLT-2 class? How much more

Q: How will these results impact formularies?

Q: How will patients and providers and professional groups view the trade-off between lowering CV events and increasing chances of a lower limb amputation?

Q: How does the “do no harm” tenet for healthcare providers play into decisions around drugs to use in this class?

Q: What concerns exist over patients enrolled in CANVAS-R who may be at higher than typical risk of amputations?

Q: What components of outcomes trials should be changed to make them more comparable? Would it be a useful investment to have an outcome trial with all SGLTs?


-- by Abigail Dove, Helen Gao, Payal Marathe, and Kelly Close