The HypoAna study was one of the most striking presentations we saw today and represented a big time victory for patients. The two-year crossover trial included 159 type 1 patients with recurrent severe hypoglycemia (>2 episodes in the past year). Patients used either analog insulins (detemir and aspart) or human insulin/NPH, and then crossed over to the other study arm. Significantly, treatment with analog insulin resulted in a 29% reduction (p<0.05) in the rate of severe hypoglycemia. This corresponds to an absolute rate reduction of 0.5 severe hypoglycemia episodes per patient-year, meaning the number needed to treat with insulin analogs to avoid one episode of severe hypoglycemia is just two patients per year! We’re already stamping our envelopes to send this data to payers and governments.
The discussion then turned to SAVOR-TIMI 53, with Dr. Deepak Bhatt (Harvard University, Cambridge, MA) re-presenting the data that was first revealed at ESC earlier this month. He confirmed that further analyses of SAVOR will be presented at the American Heart Association’s upcoming annual conference in November. Dr. Itamar Raz (Hassadah Medical Center, Jerusalem, Israel) next discussed sub-analyses of SAVOR data, including a notable finding that saxagliptin was statistically safe in terms of severe hypoglycemia for most patients, but not in patients taking sulfonylureas and with a baseline A1c below 7%. A more detailed breakdown of the pancreatitis seen during the trial revealed that, although there was a numerical imbalance in pancreatitis incidence between the saxagliptin and placebo groups, the cases in the saxagliptin cohort were generally shorter in duration and did not impact treatment 60% of the time. Dr. Naveed Sattar wrapped up the session with some bold commentary on the two trials. He stated that the increase in hospitalization for heart failure seen in SAVOR, along with trends seen in EXAMINE and a trial on Novartis’ Galvus (vildagliptin), may hint at a potential class effect. He also argued that the cardiovascular neutrality in both EXAMINE and SAVOR was completely unsurprising, given the small A1c changes and short trial durations.
Next, Dr. Juris Meier (St. Josef Hospital, Bochum, Germany) provided a thorough overview of the incretin-pancreatitis debate. He highlighted the difficulties involved in investigating a rare phenomenon such as pancreatitis, and acknowledged that a number of mechanisms could link GLP-1 agonists to pancreatitis.
In a new post-hoc analysis of EXAMINE, the hazard ratio for hospitalization for heart failure was found to be 1.19 (95% CI: 0.90-1.58) with alogliptin. Though not statistically significant, it trends in the same direction as SAVOR’s result.
In the afternoon symposium on the risks and benefits of drugs, Dr. Sanjay Kaul (Cedars-Sinai Medical Center, Los Angeles, CA) gave one of the most valuable talks of the conference. His wide-ranging discussion hit on multiple hot topics: the benefit/risk balance and cardiovascular safety of SGLT-2 inhibitors (with a focus on J&J’s canagliflozin and BMS/AZ’s dapagliflozin), the current limitations in drug safety assessment, lessons learned from rosiglitazone, ways to improve the FDA’s CV guidance for diabetes drugs (and his thoughts on why it has not stifled innovation), and plenty more. We could not believe how many gems there were in this talk – read the full details below.
Dr. John Buse (University of North Carolina, Chapel Hill, NC) presented findings from two studies of metformin delayed release (Met DR; Elcelyx’s NewMet), which suggest metformin’s therapeutic action depends on gut epithelial exposure, not plasma exposure. We eagerly anticipate additional trial results for NewMet, as it could provide a similar therapeutic effect as current formulations, but with decreased GI side effects and the potential for use in patients with renal impairment.
- Executive Highlights
- Detailed Discussion and Commentary
- Symposium: Risks and Benefits of Drugs
- Symposium: DPP-4 Inhibitors and CVD
- Results from EXAMINE
- Saxagliptin and Cardiovascular Outcomes in Patients with Type 2 Diabetes
- A Diabetologist’s Interpretation of a Cardiovascular Study
- Symposium: EASD/ADA Symposium: The DCCT/EDIC Study: 30 Years of Progress and Contributions
- Introduction and Overview
- Microvascular Update
- Cardiovascular Update
- The Future
- Symposium: Bariatric Surgery in Type 2 Diabetes – An Update
- Methods and Results of Bariatric Surgery
- Diabetes Prevention and Remission After Usual Care and Bariatric Surgery
- Pathophsyiology of Bariatric Surgery
- Symposium: Pathophysiological Phenotypes in Type 2 Diabetes
- Oral Presentations: Hypoglycemia – Balancing Glucose Control
- The Effect on Insulin Analogs on The Risk of Severe Hypoglycaemia in Patients With Type 1 Diabetes and Recurrent Severe Hypoglycaemia: The HypoAna trial
- Lower Rates of Overall, Nocturnal and Severe Hypoglycaemia During Maintenance Treatment With IDegAsp vs Biphasic Insulin Aspart 30 in Patients With Type 2 Diabetes Mellitus: A Meta-Analysis
- Oral Presentations: Impact of Treatment and Genetic Susceptibility to Comorbidities and Mortality
- Diabetes, Incretin Therapy, Pancreatitis, and Pancreatic Cancer: Meta-Analyses
- Combination Therapy With Metformin Plus Sulfonylureas Versus Metformin Plus DPP-4 Inhibitors and Risk of All-Cause Mortality
- Association Between First-Line Monotherapy With Sulfonylurea Versus Metformin and Risk of All-Cause Mortality
- Oral Presentations: Novel Therapeutic Agents and Insights
- Oral Presentations: Glucose Down The Drain
- The Sodium Glucose Co-Transporter-2 (SGLT2) Inhibitor Empagliflozin Improves Glycemic Control in Patients With Type 1 Diabetes: A Single-Arm Clinical Trial
- Canagliflozin Demonstrates Durable Glycemic Improvements Over 104 Weeks Compared With Glimepiride in Subjects With Type 2 Diabetes Mellitus On Metformin
- Dapagliflozin-Induced Weight Loss Impacts 24 Week HbA1c And Blood Pressure Levels
- Luseogliflozin, A Selective SGLT2 Inhibitor, Added on to Glimepiride for 52 Weeks Improves Glycaemic Control With No Major Hypoglycaemia in Japanese Type 2 Diabetes Patients
- Efficacy and Safety of Tofogliflozin Administered for 52 Weeks as Monotherapy or Combined with Other Oral Hypoglycemic Agents in Japanese Patients with Type 2 Diabetes
- Favourable Gastrointestinal and Genitourinary Safety Profile of LX4211 Added-on to Metformin In a Phase 2b Study
Detailed Discussion and Commentary
Symposium: Risks and Benefits of Drugs
Update on the Evaluation of SGLT-2 Inhibitors
Sanjay Kaul, MD (Cedars-Sinai Medical Center, Los Angeles, CA)
Dr. Sanjay Kaul gave a truly outstanding, opinion-filled talk on SGLT-2 inhibitors, cardiovascular outcomes trials, FDA policy, clinical trial design, and data interpretation. Regarding SGLT-2s, the clear theme of Dr. Kaul’s presentation was uncertainty – he emphasized a vast array of unanswered questions for the drug class, including cardiovascular and renal safety, bone health, risks of malignancy, and volume depletion. He specifically reviewed the cardiovascular profiles of dapagliflozin and canagliflozin, outlining more concerns with canagliflozin based on the data. As he did at the canagliflozin advisory committee, Dr. Kaul came out against the release of interim CVOT data, though he also proposed a number of valuable ways to ensure trial integrity. Speaking more broadly, Dr. Kaul commented that the 2008 cardiovascular risk assessment guidance has hardly stifled innovation, a comment we found reasonable when he compared it to blood pressure medications – but not in diabetes, our major concern. Dr. Kaul spoke out against the FDA’s use of fixed hazard ratios for cardiovascular safety (1.8 and 1.3) – instead, he advocated for flexible assessments based on risk/benefit profiles (i.e., to accept a higher level of risk for a higher level of benefit). This was great to hear from a patient perspective. We appreciated his discussion of rosiglitazone as well, a drug that, in his view, still has insufficient high-quality evidence to incriminate or exonerate it on the cardiovascular safety front. Dr. Kaul highlighted the scary duality implicit in that drug’s story: it became a blockbuster without an established outcome benefit though was virtually killed on insufficient evidence and entrenched opinions. We thought this was one of the best talks at EASD 2013 – more of Dr. Kaul’s opinions and nuanced discussion is below.
Regarding rosiglitazone, the catalyst for the development of 2008 cardiovascular guidance for drug development, Dr. Kaul noted that a lingering uncertainty remains regarding its safety. In his opinion, there is insufficient high-quality evidence to incriminate or exonerate the drug’s effect on cardiovascular safety. The fact that the drug became a blockbuster without an established outcome benefit does not reflect well on drug development; however, it is equally lamentable that the drug was virtually killed on insufficient evidence and entrenched opinions.
Dr. Kaul commented that he thought the 2008 cardiovascular risk assessment guidance has hardly stifled innovation, noting that there are 15 cardiovascular outcomes trials, with two completed (SAVOR and EXAMINE) and many underway. He said there is “quite a rich landscape” in diabetes drug development right now; for comparison, only one new blood pressure medication has been approved by the FDA in the last 15 years.
Dr. Kaul summarized the current benefits of SGLT-2 inhibitors. He noted that SGLT-2 inhibitors have been shown to have “modest” glycemic control (an ~0.7% reduction from a baseline of ~8%), relatively low frequency of hypoglycemia, “modest” weight loss (~1.8 kg [4.0 lbs]), and a blood-pressure-lowering effect on the order of approximately 4.5 mmHg, which could potentially be clinically beneficial. To this regard, he commented that we don’t know whether non-physiologic reductions in blood pressure yield cardioprotective effects.
He also addressed risks of the class: mycotic genital infection (odds ratio=5.0); urinary tract infections (odds ratio=1.4), but not pyelonephritis; polyuria, nocturia, and dysuria; volume depletion, thirst, and increased hematocrit; and increased LDL (benefits and risks were adapted from Vasilakou et al., Ann Intern Med 2013). Dr. Kaul commented that he did not know what the clinical relevance of increased LDL observed with SGLT-2 inhibitor treatment might be.
Dr. Kaul noted that in terms of cardiovascular safety, canagliflozin had an imbalance in stroke as well as uncertainty around an early increase in MACE. Additionally, an increased LDL was observed with canagliflozin and empagliflozin.
In Dr. Kaul’s opinion, the issue of volume depletion with SGLT-2 inhibitors is likely underestimated, noting that in CANVAS’ study protocol design, the earliest assessments of volume depletion were conducted at six weeks. He emphasized that the potential effects of volume depletion on the RAAS system activation should be explored, since the CETP inhibitor torcetrapib (a lipid drug to increase HDL) that increased aldosterone was also associated with a significant increase in cardiovascular events.
Dr. Kaul commented that we have no idea what the effects of SGLT-2 inhibitors are on bone health, given that at present, most evaluations have only been conducted out to one or two years at most. As for malignancy, it remains uncertain what the implications the observed imbalances in bladder and breast cancer with dapagliflozin treatment are. He did not think tumor induction was the likely mechanism, but tumor promotion might be (though, to date, little is known about tumor-promoting agents). During Q&A, Dr. Kaul commented that he believed the imbalances were likely due to detection/ascertainment bias.
In Dr. Kaul’s review of the cardiovascular profiles of SGLT-2 inhibitors, he expressed more comfort with dapagliflozin’s profile than with canagliflozin’s profile. He noted that in the meta-analysis of cardiovascular risk with dapagliflozin treatment: 1) the observed population in trials for dapagliflozin were representative of the intended treatment population; 2) excess cardiovascular risk (HR of 1.8) was ruled out; and 3) numerical imbalances in cardiovascular endpoints consistently favored dapagliflozin versus control. As for canagliflozin, Dr. Kaul pointed out a number of concerns he had with the conducted meta-analysis, including that: 1) there was an early increased MACE risk (whether it is a true risk or a “random high” remains uncertain); 2) the sponsors’ claim of a “lack of association with volume depletion-related AEs” might not have painted the whole picture, since early (less than six week) volume and blood pressure changes were not captured due to protocol; 3) there was limited patient exposure (~1.1 years), that is, the number of events was driven by a high number of patients as opposed to duration of exposure; 4) the impact of imbalance in rescue therapy use remains unclear; and 5) the public release of interim results may have compromised trial integrity.
With the public release of interim results for the cardiovascular outcomes trial for canagliflozin (CANVAS) in mind, Dr. Kaul emphasized that partial unblinding threatens trial integrity and reliability, and proposed a solution to the issue. He noted that disclosure of interim analyses slows enrollment; reduces adherence in the experimental arm (for HR >1.2); encourages cross-ins in the control arm (HR <0.8); and minimizes retention. To use interim analysis results and maintain trial integrity, Dr. Kaul suggested that data access be restricted to an unblinded “firewalled group” formed with input from the regulator, DMC, and steering committee; discussions should occur in a “closed” session to preserve data integrity; and regulators should only disclose publicly if a hazard ratio of 1.8 has been ruled out (implying a point estimate of <1.26 without loss of equipoise).
There are a number of remaining unanswered questions about SGLT-2 inhibitors. Regarding durability, it remains to be seen whether the efficacy of drugs in the class will wane with normalized glucose levels. As for safety and tolerability, SGLT-2 inhibitors have unproven long-term safety (especially in the elderly [>75 years], those with renal impairment [GFR <45-60 ml/min/1.73 m2], and those on loop diuretics). He emphasized that the use of SGLT-2 inhibitors should be further investigated in patients with renal impairment, in particular with regards to nephropathy and increased volume depletion. Finally, he also commented that SGLT-2 inhibitors have not been substantially evaluated in minority populations.
Reflecting on cardiovascular safety assessments performed to date and their limitations, Dr. Kaul proposed a number of ways to enhance the integrity of non-inferiority safety trials. Dr. Kaul emphasized the need for such trials to: reduce the ineligibility rate; optimize adherence in the experimental arm (to decrease bias to the null); avoid cross-ins in the control arm (to decrease bias to the null); maximize retention and avoid protocol deviation and drop out (to decrease bias to the null); enroll, capture data, and adjudicate events in a timely manner; minimize exposure of the control arm to treatments that increase cardiovascular risk (to decrease bias to the null); restrict the primary endpoint to stringent MACE – cardiovascular death, non-fatal myocardial infarction, or non-fatal stroke (to decrease ascertainment error and misclassification); adjust for multiplicity if interim data is used for the final analysis (to reduce type 1 error); not publicly disclose results of interim analyses (to increase trial integrity). Dr. Kaul also noted that placebo-controlled trials should be preferred over active-controlled trials (as one would not necessarily know the safety profile of the active comparator). In addition, extension phases of trials add noise (given that they are often unblinded); as such Dr. Kaul advocated for extensions to be blinded.
Dr. Kaul questioned whether we should continue to allow disease-oriented surrogate endpoints to be the standard for regulatory approval, or demand patient-oriented health outcomes benefit for both microvascular and macrovascular complications. He commented that he is not persuaded by the statement that microvascular benefits are unequivocally demonstrated with oral agents for type 2 diabetes, deeming laserphotocoagulation and reduced albuminuria as examples of unvalidated surrogate endpoints. Instead, he would like to see studies with concrete patient outcome benefits – for example, in renal failure, amputations, and blindness.
Dr. Kaul argued against the FDA’s use of fixed hazard ratios (1.8 and 1.3) with regards to cardiovascular safety. Instead of fixed margins, Dr. Kaul proposed that allowable risk should be determined based on the amount of benefit, as opposed to trial feasibility. For example, he would be willing to accept a greater potential increase in harm for a therapy that lowered A1c by 1.5-2.0%, whereas for a therapy that lowered A1c by 0.5%, he might not even be willing to accept a hazard ratio of 1.3. Furthermore, he advocated for benefit/risk assessment to incorporate the patient’s view of acceptable or tolerable risk, and for it to evolve from the mindset to “ensure drug safety” to ensure “favorable risk/benefit.” Finally, he emphasized the need to develop more tools for communication of benefit/risk not only to patients, but also to the scientific community, academia, and regulators.
Dr. Kaul explained that an asymmetry exists in how drug efficacy and safety are evaluated – RCTs vs. non-clinical data/PK-PD studies/retrospective review. Efficacy is assessed in randomized controlled trials, in which endpoints are anticipated and prespecified. These trials are designed with adequate power to assess the endpoint, and the endpoints have clear adjudication, and are precisely measured/quantified. In contrast, safety is typically assessed using a combination of non-clinical data, pharmacokinetic/pharmacodynamic studies, meta-analyses, observation databases, and monitoring programs (e.g., FDA AERS and Sentinel). This stems from a major limitation in randomized controlled trials: there is limited exposure to treatment and narrow study populations, so safety issues cannot be fully captured. In contrast to efficacy assessments, safety issues are often unanticipated, and as such are not prespecified for measurement, or adjudicated proactively (O’Neill, Drug Information Journal 2008).
Questions and Answers
Q: Surely when we prescribe to patients, they are exposed to risks other than cardiovascular risk. What do you think of things like bladder cancer or breast cancer?
A: As I mentioned fleetingly, there was an imbalance in one of the drugs. I believe it’s probably related to ascertainment bias. In order for sponsors to clearly adjudicate this uncertainty, it would require a 30,000-patient trial, which in my opinion, and probably the majority of yours, is not possible to do pre-approval. That should be evaluated in a post-approval, phase 4 type of study.
GLP-1 Based Therapies and Pancreatic Disease: What’s the Clinical Evidence?
Juris Meier, MD (St. Josef Hospital, Bochum, Germany)
In a balanced manner, Dr. Juris Meier detailed the issue of incretin therapy and pancreatic disease, tracing it from its early beginnings to the most recent and cutting-edge data. Dr. Meier analyzed the main trials that gave birth to the controversy. He explained that Elashoff et al.’s FDA-AERS analysis was weakened by a potential over-reporting bias, a lack of validation, and a dearth of information on potential confounders. Dr. Robert Butler’s 2013 morphological study, Dr. Meier noted, had small and poorly matched groups. Singh et al.’s retrospective population-based case-control study faces the same limitations as any retrospective study, according to Dr. Meier (i.e., the potential for residual confounders, etc.). A meta-analysis of incretin therapy randomized control trials conducted by Drs. Meier and Michael Nauck (Diabeteszentrum Bad Lauterberg, Harz, Germany) showed no significant safety signal. Due to the low incidence of pancreatitis, Drs. Meier and Nauck estimated that only a study with 40,000 patients in each group would be powerful enough to elucidate a statistically significant 20% risk signal — a likely unrealistic prospective study size. Dr. Meier concluded with a series of practical recommendations. He pressed not to use GLP-1 agonists in patients with chronic pancreatitis, a history of acute pancreatitis, or a family history of pancreatic cancer. However, he sees their use being acceptable in patients with cholecystolithiasis or hypertriglyceridemia. Additionally, he recommended against routine surveillance of amylase or lipase levels unless symptoms of pancreatitis appear, because abnormal results may not be indicative of an adverse event.
Dr. Meier opened his presentation by walking attendees through the incretin and pancreatitis controversy while offering his own commentary. He noted that Elashoff et al.’s FDA-AERS database study in 2011 has been heavily criticized for its likely over-reporting bias, lack of validation, and incomplete information about confounding factors. More reliable case-control studies have been performed since, according to Dr. Meier; however, they are still limited by their retrospective nature. Dr. Meier brought up Dr. Peter Butler’s (in)famous pancreatic morphological study, which has been criticized for its small, poorly matched groups and likely accidental inclusion of type 1 diabetes patients. SAVOR-TIMI 53 and EXAMINE, the respective cardiovascular outcomes trials for BMS/AZ’s Onglyza (saxagliptin) and Takeda’s Nesina (alogliptin), both had statistically insignificant numerical imbalances in the number of pancreatitis cases. While these numbers may illustrate trends, Dr. Meier stated, they cannot prove or disprove the association.
To help clarify the issue, Dr. Meier and Dr. Nauck conducted a meta-analysis of data from trials of all incretin therapies currently available in Europe. Some individual trials trended towards an increased thought insignificant risk, and the pooled estimate was neutral.
Dr. Meier next discussed mechanistic factors behind GLP-1 agonism, which in theory could impact the pancreas. For example, he pointed to the increase in lipase secretion seen during GLP-1 agonist therapy (Steinberg et al., Gastroenterology 2012) as a somewhat troubling phenomenon that merits further study. Dr. Meier worried that much of the data on lipase activity is not publicly available. Potential mechanistic explanations of the link between GLP-1 agonism and pancreatitis include the direct stimulation of pancreatic GLP-1 receptors in acinar tissue, the indirect effects of GLP-1 agonists on exocrine pancreatic function, and the activation of afferent neural pathways. These or other mechanisms could inhibit pancreatic outflow; inhibit bile motility; or lead to gallstones, which in turn could lead to pancreatitis. Dr. Meier discussed a smattering of early evidence from mechanistic studies that lend credence to these ideas. A study showed that exenatide reduced patients’ gallbladder ejection fraction, and radiolabeling data indicating that GLP-1 receptors are located near exocrine cells of the pancreas.
Dr. Meier ended with a series of practical recommendations. He suggested against the use of GLP-1 agonists in patients with chronic pancreatitis, a history of acute pancreatitis, or a family history of pancreatic cancer. He advocated the immediate discontinuation of treatment if pancreatitis is suspected. However, patients with cholecystolithiasis and hypertriglyceridemia, he said, can continue using the drug class. Dr. Meier dissuaded providers from conducting routine surveillance of amylase and lipase levels unless clinical symptoms manifest themselves (a recommendation which received attention during Q&A). Finally, he said, any and all recommendations should be reviewed after the publication of the ongoing outcomes trials.
Questions and Answers
Q: I had a patient on Byetta [exenatide] with no symptoms, and I did an amylase test and found a nearly tenfold elevation. After discontinuing the drug, the results returned to normal. What would you say about those situations?
A: I think it is a very valid point. We don’t have a clear answer on how to proceed in such cases. I would concur that if a patient has a tenfold elevation, I would also withdraw treatment. I would probably not withdraw treatment if the elevation were below threefold.
Q: Why did you say that there is no use in screening for enzymes? I had two patients with an increase in those enzymes, and then I stopped the GLP-1 analog and they went back to normal. Chronic pancreatitis can have no symptoms.
A: I think the problem is that if you perform routine monitoring of amylase or lipase, you will see elevated levels in 20% to 30% of cases. If you have these abnormalities in the absence of clinical symptoms, what do you do? This may lead to ultrasounds, CTs, and even MRIs, which could be unnecessary in the absence of clinical symptoms. I would prefer performing lab studies only if we see symptoms.
Q: Just one point of clarification – I agree that we shouldn’t be measuring lipase and amylase in the clinical setting, but should we be measuring them in trials?
A: Definitely yes.
Symposium: DPP-4 Inhibitors and CVD
Results from EXAMINE
William White, MD (University of Connecticut, Farmington, CT) and Simon Heller, MD (University of Sheffield, Sheffield, United Kingdom)
Dr. William White and Dr. Simon Heller presented results from EXAMINE, the cardiovascular outcomes trial for alogliptin (Takeda’s Nesina). The main findings from the study were initially presented at the European Society of Cardiology’s annual conference earlier this month – see our ESC 2013 Day #1 Report at http://www.closeconcerns.com/knowledgebase/r/2a8925c5 for our full coverage. In light of SAVOR’s finding that saxagliptin increased hospitalization for heart failure, EXAMINE investigators performed a post-hoc analysis of heart failure hospitalization in EXAMINE, which did not pre-specify hospitalization for heart failure as an endpoint (the pre-specified heart failure endpoint in was a composite of CV death plus hospitalization for heart failure, which EXAMINE investigators have noted is the standard in heart failure trials). EXAMINE’s pre-specified CV mortality and hospitalization for heart failure analysis found that the HR was 0.98 (95% CI: 0.82-1.21) with alogliptin treatment. The new post-hoc analysis showed that the HR for hospitalization for heart failure alone was 1.19 (95% CI: 0.90-1.58), while CV mortality alone was 0.84 (0.64-1.10). Although the hospitalization for heart failure result was not statistically significant, it trends in the same direction of SAVOR’s result.
Questions and Answers
Q: I have some problems with the numbers on heart failure. In Table 1 in the publication on September 2nd, you state that 28% of all patients in EXAMINE had preexisting congestive heart failure, amounting to some 750 patients with heart failure in each arm. Now you report a history of heart failure only in 400-some patients in each arm. Did you change the definition for this new analysis?
Dr. White: The number in the paper was at the time of randomization, so that could have included post-ACS heart failure. What we are reporting now is before the primary ACS event. That’s why the numbers are higher in the paper. We thought that from the standpoint of evaluating morbidity, it made more sense.
Q: Did you do a post hoc exploratory analysis to look at patients using sulfonylurea?
Dr. White: We did a subgroup analysis looking at those who were taking or not taking sulfonylurea, and there was no heterogeneity.
Q: Did you look at microalbuminuria between groups?
Dr. Heller: We did not measure microalbuminuria.
Comment: The only concern I have is with pancreatitis. I don’t think it’s appropriate to do statistics when you have 12 and five cases. The only thing we can say is that there is an imbalance, consistent with the concerns we have had. It is an imbalance that we need to take into consideration.
Dr. Heller: Alternative etiologies seemed to explain some of those cases. I agree that the numbers are small, but I think we could still conclude that these events were relatively unusual, which is reassuring, and that’s probably as far as we can go.
Q: Were investigators allowed to combine multiple medications, including insulin, on top of alogliptin? If so, why is it that such a low percentage of patients achieve an A1c of less than 7%?
Dr. Heller: The investigators were free to use whatever antidiabetic medications they so chose, and medications increased more in the placebo group, but you’re right, they didn’t match the A1c levels in both groups. I can’t speak for many investigators around the world, but clearly it is an issue, of inertia I presume.
Q: Do you think the percentage of patients who achieved an A1c of less than 7% may have affected the outcome?
Dr. Heller: Of course it’s worth remembering that the patients in the trial were at very high risk. Other intensive control trials suggest very aggressive and low A1c may have adverse consequences, so I’m sure that would be in the investigators’ minds when trying to improve A1c in this group. We don’t have data to conclude whether this affected the outcome.
Saxagliptin and Cardiovascular Outcomes in Patients with Type 2 Diabetes
Deepak Bhatt, MD (Harvard University, Cambridge, MA)
Dr. Deepak Bhatt took the stage to present the main results of the SAVOR-TIMI 53 trial, which investigated the cardiovascular effects of BMS/AZ’s DPP-4 inhibitor Onglyza (saxagliptin). This data was first presented at the European Society of Cardiology’s annual conference earlier this month along with the results of EXAMINE — see our ESC 2013 Day #1 Report at http://www.closeconcerns.com/knowledgebase/r/2a8925c5 for our full coverage of those presentations. The investigators designed a trial powerful enough to potentially demonstrate cardiovascular superiority, given that earlier studies had hinted at a slight cardioprotective effect. However, after two years of treatment, saxagliptin showed no significant effect on the composite primary composite outcome of cardiovascular death, myocardial infarction, and ischemic stroke. The DPP-4 inhibitor led to a somewhat modest A1c improvement of 0.5% (0.3% placebo-adjusted) from a baseline of 8%, although Dr. Bhatt noted that the modest difference was largely due to the escalation of treatment in the placebo cohort. He mentioned that there was a statistically significant increase in hospitalization for heart failure (a much-discussed signal during the weeks since ESC), but noted that the increase was relatively small in an absolute sense (0.7%), that it was not associated with an increase in mortality, and that most cases of hospitalizations were in patients with high levels of BNP (a protein associated with heart failure risk). No significant differences were seen between saxagliptin and placebo in subgroup analyses by age, baseline A1c, pre-existing CVD, or concomitant therapy, among other criteria. He ended by noting that further analyses of SAVOR data are ongoing, and will be presented at the American Heart Association’s annual conference on November 16-20 in Dallas.
A Diabetologist’s Interpretation of a Cardiovascular Study
Itamar Raz, MD (Hadassah Medical Center, Jerusalem, Israel)
Next, Dr. Itamar Raz discussed interesting secondary analyses of SAVOR-TIMI 53 data. In patients with A1c >7.0% at baseline, saxagliptin was significantly more effective at lowering A1c in patients also on metformin, insulin (alone or in combination), or sulfonylureas (the improvement was non-significant for saxagliptin monotherapy). Dr. Raz next provided an expanded analysis of patient renal function. Significantly more patients in the treatment arm progressed to a better category of renal function (defined by albumin/creatinine ratio) than in the control arm (11% vs. 9%), and significantly fewer patients from the treatment arm saw their renal function category worsen (13% vs. 16% with placebo), indicating that saxagliptin may prevent the deterioration of (or even improve) renal function. The first round of SAVOR data had shown that saxagliptin modestly but significantly increased the incidence of both minor and major hypoglycemia. The investigators conducted a thorough examination of hypoglycemia occurrence, and found that there was no risk increase for most patients, but a significant increase in patients on sulfonylureas (which are known to cause hypoglycemia) and/or those with a baseline A1c below 7%. Turning to pancreatitis, study adjudicators found 26 definite or possible acute pancreatitis events in the saxagliptin group compared to 25 in the placebo group (17 vs. nine cases of “definite pancreatitis”). However, the mean event duration was shorter in the saxagliptin group, and patients continued with saxagliptin treatment in 60% of cases. A scatterplot of the data shows that there was no temporal clustering of cases, as would be expected if the drug was directly causing pancreatitis.
Dr. Raz began by presenting more detailed data on saxagliptin’s glycemic efficacy. Significantly fewer patients on the drug received new or increased doses of other diabetes drugs during the trial than in the placebo group, regardless of baseline A1c. Next, we saw data on the mean change in A1c segmented by concomitant therapy — patients on saxagliptin combination therapy with metformin, insulin, or sulfonylurea saw significantly greater drops in A1c than saxagliptin placebo patients on the same concomitant antidiabetic drugs. The difference in A1c change between the saxagliptin and placebo cohorts was not significant but trended towards an effect. Notably, this combination therapy analysis was limited to patients with baseline A1cs above 7% — we imagine that the investigators used this cutoff because patients with lower baseline A1c likely had smaller change, making it difficult to demonstrate statistical significance.
Saxagliptin had a modest but statistically significant beneficial effect on renal function. Prior to the trial, the investigators pre-determined a series of categories based on albumin/creatinine ratio (<30, 30-300, and >300 mg/G). Patients underwent urinary albumin excretion tests yearly and at the end of the trial. Only 13% of patients in the saxagliptin arm shifted to a worse category, compared to 16% of the placebo group (p<0.001). Around 11% of saxagliptin patients saw an improvement in their renal function category, compared to 9% of placebo patients (p<0.001).
The SAVOR study group conducted in in-depth analysis of hypoglycemia, perhaps in response to the finding of a significant increase in both mild and severe hypoglycemia in the saxagliptin arm. The study used a fairly broad definition of minor hypoglycemia: in addition to glucose measurements below 54 mg/dl, the definition included episodes with symptoms that were resolved after carbohydrate consumption with no blood glucose data. Severe hypoglycemia was defined as any event that required the assistance of another person (regardless of blood glucose). This analysis demonstrated that saxagliptin did not have a significant impact on overall hypoglycemia unless the patient was on sulfonylureas, or if they had a baseline A1c below 7% and were on combination therapy involving insulin. The only significant increase in major hypoglycemia was in saxagliptin patients who were treated with sulfonylureas and had a baseline A1c below 7%. This analysis should somewhat assuage fears about the early hypoglycemia data, as sulfonylureas’ impact on hypoglycemia is well known, and most patients with low baseline A1c had less of margin of error during treatment.
Dr. Raz concluded by presenting a more in-depth analysis of the pancreatitis seen during SAVOR. As with some other incretin safety trials, the overall incidence of pancreatitis was low (making statistically significant differences unlikely) but there was a slight numerical imbalance. Adjudicators found 24 cases of any pancreatitis event and 17 cases of definite pancreatitis (compared to 21 and nine cases in the placebo group, respectively). However, Dr. Raz noted that the mean duration of pancreatitis cases was far less in the saxagliptin group compared to the placebo group (14 and 44 days, respectively), and that 60% of pancreatitis patients in the treatment arm continued uninterrupted on saxagliptin, suggesting that patients and providers did not see the drug as a causal factor. Dr. Raz also displayed a scatterplot of pancreatitis cases graphed against time (cases were fairly evenly distributed along the entire study time course) — he argued that if saxagliptin was causing pancreatitis, cases would be clustered towards the beginning of the trial. While this study was neither large nor long enough to definitely disprove the possibility of a pancreatitis effect, we found Dr. Raz’s arguments in this section convincing.
Questions and Answers
Q: Regarding the hypoglycemia you saw in the group on sulfonylurea with low A1c, could that have been a protocol problem?
A: As you saw, we recruited patients with A1cs below 6.5% — they comprised 8% of our patient total. We discussed the issue, and at the end of day we decided not to change the protocol. But this is something that will be discussed.
Q: Did most of the hypoglycemia cases happen early in the trial?
A: That is a good question – I don’t have the answer yet. Most probably happened during the earlier stages of the trial.
Q: I have a question about the increase in hospitalization for heart failure. The background treatment was different than in EXAMINE: fewer patients were on beta blockers and ACE inhibitors. Did you analyze the data by baseline cardiovascular medications? Also, did you see if there was an increase in BNP with sitagliptin compared to placebo?
A: We do have BNP data, more specifically pro-BNP levels, since BNP can be affected by DPP-4 inhibitors. We hope to have that data analyzed by AHA. It is important to keep in mind that the overall hazard ratio for the secondary endpoints was 1.02, so no sign of harm. The p-value for hospitalization was significant, but it is hard to interpret the significance of one item like that when the overall hazard ratio is neutral. An analysis looking for differences in sub-characteristics, including cardiovascular medications, is being done. We showed the preliminary BNP data here because we were looking for the group at highest risk, but in terms of clinical descriptors, the biggest risk factor for the development of heart failure was previous heart failure, and that was true in both arms. EXAMINE did not find a significant excess in heart failure – I commend the investigators for doing a lot of analyses. It takes a lot of effort. There are two explanations I can think of: one is that finding in our trial in spurious; the other is that the same effect exists in EXAMINE but the trial was not powerful enough to pick it up, and that if it was as powerful as SAVOR, the finding would have been significant there as well. But overall the data is reassuring to some extent, as the hospitalization for heart failure was isolated and did not lead to an increase in other cardiovascular outcomes.
Q: Did you analyze whether the occurrence of hypoglycemia was in any way related to the occurrence of the primary cardiovascular endpoint?
A: We’re in the process of evaluating the relationship between hypoglycemia and the primary and secondary endpoints. It’s tricky to do those analyses though, because as previous analyses have shown, hypoglycemia is associated with a number of things, from cardiovascular death, to non-cardiovascular death, to cancer. What is necessary is to do time-dependent analyses, to examine the temporal relationships.
Q: Does the expertise of providers at the different sites have an effect in hypoglycemia – to be up-front, was there more hypoglycemia with cardiologists than with endocrinologists?
A: We don’t have the answer for that, as we didn’t do these analyses yet. One thing we can learn from SAVOR is that for patients who have A1c less than 7%, you shouldn’t add saxagliptin to a sulfonylurea. Most patients who had established cardiovascular disease came to us through cardiologists, and most of those people had an A1c between 7% and 8%. I believe we will find that most hypoglycemia was related to multiple risk factors.
Naveed Sattar, MD, PhD (University of Glasgow, Glasgow, UK)
Dr. Naveed Sattar ended the session with an energetic, no-holds-barred interpretation of SAVOR and EXAMINE results. He noted that the FDA’s new guidelines “force” study designs that may be less than ideal (i.e., they enroll higher risk patient populations to promote expediency), but that this is a reality of the times. He brought up the apparent paradox between the results of earlier DPP-4 inhibitor trials, which (when meta-analyzed) showed promise of a cardioprotective effect, and the results of SAVOR and EXAMINE, which were neutral. Dr. Sattar stated emphatically that we should not be surprised by this result, as A1c reductions as modest as those seen in SAVOR and EXAMINE (0.3 – 0.35%) couldn’t have an appreciable effect on cardiovascular risk in a mere two years. To be successful in such a time period, he argued that DPP-4 inhibitors would need to have pleiotropic effects on blood pressure or lipids. Overall, he argued, CV management is more effective when pursued through agents such as statins and blood pressure medications than through glucose lowering. In his view, statin use in SAVOR was suboptimal at 78%. He found the pancreatitis data from both trials reassuring, although he acknowledged that they are not enough to disprove a possible effect. The hypoglycemia results of SAVOR, in his mind, indicate that A1c targets should be relaxed in patients with cardiovascular disease. He found the reduction of progression to albuminuria seen with saxagliptin modest and generally unsurprising. He emphatically stated that the medical community must take the hospitalization signal in SAVOR seriously — given the similar but non-significant trend seen in EXAMINE, as well as data showing that the DPP-4 inhibitor Galvus (vildagliptin) had a slight effect on ventricular size, Dr. Sattar argued that we cannot rule out a class-wide effect.
Symposium: EASD/ADA Symposium: The DCCT/EDIC Study: 30 Years of Progress and Contributions
Introduction and Overview
Bernard Zinman, MD (University of Toronto, Toronto, Canada)
Dr. Bernard Zinman provided a thorough overview of the design and results of the Diabetes Control and Complications Trial (DCCT) and the Epidemiology of Diabetes Interventions and Complications (EDIC), highlighting the beneficial effects of intensive therapy on microvascular and neurologic complications. DCCT was designed to evaluate how intensive glycemic control can prevent or reduce the progress of long-term complications, measured mainly through the incidence of retinopathy. Dr. Zinman reviewed the results of the study, highlighting the 2% drop in A1c in the intensive arm. Turning to EDIC, Dr. Zinman highlighted the “metabolic memory” phenomenon that allowed patients in the intensive group to experience the continued benefit of risk reduction past the end of DCCT. Dr. Zinman specifically noted the importance of this data collection for evaluating cardiovascular outcomes, since patients with type 1 diabetes provide a “pure glucose model” for examining the role glycemia plays in CVD (implying that patients with type 2 diabetes are NOT a good model to test this hypothesis). Dr. Zinman expressed his continued gratitude to the participants of the study and emphasized the incredible 95% retention of the surviving participants. For more information on DCCT and EDIC, please see page 226 of our ADA 2013 full report at http://www.closeconcerns.com/knowledgebase/r/94f937d8.
Dr. Zinman also reminded the audience that the difference in A1c levels between the two groups could explain over 96% of risk reduction for all complications in the intensive control group. These complications included: sustained 3-step progression, severe non-proliferative diabetic retinopathy (SNPDR), laser treatment, CSME, albumin excretion rate (AER) >39 mg per 24 hours, and AER >300 mg per 24 hours.
DCCT had a massive impact on diabetes public health for diabetes. Dr. Zinman noted that after DCCT, intensive therapy was advocated as the standard of care for people with type 1 diabetes. Additionally, government and public health agencies advertised DCCT as a way to lower barriers to intensive control, and in many areas, insurance coverage of intensive therapy supplies became mandatory.
William Tamborlane, MD (Yale University, New Haven, CT)
Dr. William Tamborlane gave the microvascular update from DCCT/EDIC, noting that the results have been “incredibly remarkable.” Indeed, the data was clear, consistent, and compelling for the three areas he covered: retinopathy, nephropathy, and neuropathy – it never got old to see slide after slide after slide demonstrating the benefits of intensive therapy on microvascular complications. Dr. Tamborlane’s summary said it all in a few bullet points: 1) intensive therapy reduces all microvascular disease substantially and consistently (it’s rare one can make such an absolute claim in medicine, though in this case, it’s justified!); 2) the benefits of intensive therapy are closely associated with reductions in A1c; 3) the early reduction in microvascular complications during the DCCT has translated to salutary effects on more advanced disease; and 4) metabolic memory applies to all microvascular complications. We note that his presentation condensed the three separate talks we saw in a similarly titled symposium at ADA 2013 – see pages 226-232 at http://www.closeconcerns.com/knowledgebase/r/94f937d8.
Trevor Orchard, MD (University of Pittsburgh Graduate School of Public Health, Pittsburgh, PA)
Dr. Trevor Orchard reviewed the cardiovascular benefits of the DCCT/EDIC trial and highlighted that the benefits of intensive glycemic control have only increased as EDIC continues. He emphasized the risk reduction for cardiovascular events that patients from the intensive control group experienced, such as the reduction in atherosclerosis and clinically important outcomes of cardiovascular death, myocardial infarction, or stroke (57% reduction after 11 years!). Dr. Orchard also noted that they are not done with the full mediation model, and they plan do finish after there are 100 cardiovascular deaths, likely within the next few years. Additionally, there is currently a manuscript in progress on the analysis of mortality from DCCT/EDIC, now that there have been 50 deaths – however, all the data is embargoed until publication. For more information on DCCT and the cardiovascular disease update, please see page 227 of our ADA 2013 full report at http://www.closeconcerns.com/knowledgebase/r/94f937d8.
Gayle Lorenzi, RN, CDE (UCSD, San Diego, CA)
Ms. Gayle Lorenzi closed the session with a look at what’s next in DCCT/EDIC. She highlighted areas of continued exploration/redefinition along with future ancillary studies of gastric emptying, C-peptide, and hearing impairment.
Areas of continued exploration and redefinition include: metabolic memory; triopathy; evidence-based guidelines for the frequency of retinopathy and nephropathy screening; glycemic variability; non-glycemic risk factors and outcomes; cognitive function; and health economics.
Ms. Lorenzi shared details on three future ancillary DCCT/EDIC studies: gastric emptying, residual C-peptide secretion, and hearing in long-standing type 1 diabetes.
Gastric emptying: This pilot study will occur at seven EDIC centers and include 80 participants. A C-spirulina gastric emptying breath test will be administered. The study is tasked with answering two main questions: “What is the prevalence of disturbances in gastric emptying?” and “How do they impact glycemic control?” The results will determine whether to expand the investigation and evaluate the full cohort
Residual C-peptide: In a pilot study presented at ADA 2013, researchers found demonstrable C-peptide in 17% of 58 DCCT/EDIC patients. This upcoming study will examine the full DCCT/EDIC cohort and perform mixed-meal tolerance tests and measured C-peptide using three ultrasensitive assays. Outcomes include A1c over time/insulin dose, hypoglycemia, mediators/risk factors, and long-term complications. The trial will also permit a comparison of the three ultrasensitive assays.
Hearing impairment: A study will examine the incidence and impact of hearing impairment in the DCCT/EDIC cohort. Researchers will examine its relationship to glycemic control.
Symposium: Bariatric Surgery in Type 2 Diabetes – An Update
Methods and Results of Bariatric Surgery
Dimitri Pournaras, MD (Imperial College, London, United Kingdom)
In an opening to the symposium dedicated to bariatric surgery and diabetes management, Dr. Pournaras presented an overview of outcomes from surgical weight loss interventions in patients with diabetes and made several suggestions for future practice. Importantly, he noted that in 2009, diabetes remission was redefined as maintaining a normal A1c <6% and a fasting glucose level <100 mg/dl for at least one year following surgery. Given these modifications to the criteria, diabetes remission rates have significantly dropped from calculations based on previous standards (e.g., for gastric bypass, rates are now 40% vs. ~60% before). Interestingly, Dr. Pournaras interpreted these findings as positive, claiming they support his view that we should be shifting measures of success away from remission rates, which are antiquated, to glycemic control, which is used as a gauge in all other diabetes treatments. Regarding type 2 diabetes recurrence, Dr. Pournaras noted that short-term remission rates are 43%, while long-term are 24%. He argued that we should look to cancer’s combinatorial treatment approach to amend our approach to bariatric surgery and diabetes care – we certainly agree with him that combination therapy, especially early in the course of type 2 diabetes, will result in better outcomes. In Dr. Pournaras’ opinion, our standard evaluation should “not be best medical therapy vs. surgery, but best medical therapy vs. best medical therapy plus surgery.” One study that assessed intensive medical therapy alone vs. combined with surgery found that using both doubled the percentage by which A1c levels were lowered, from 1.5% to 3%. In addition, Dr. Pournaras advocated on behalf of redirecting end goals for surgery to be focused on the reduction of diabetes related co-morbidities and macrovascular complications, as is currently being done in the evaluation of new diabetes drugs. In general, we left with the impression that Dr. Pournaras believes strong efforts should be made to bridge the separation between bariatric surgery and diabetes treatments in the conventional paradigm in order to improve patient outcomes.
The most common weight loss operations – adjustable gastric banding, roux en Y gastric bypass, and sleeve gastrectomy – have shown considerable success and safety, with very low in-hospital mortality rates (0.07% in 2013 as indicated by the NBSR) and significant improvements in A1c levels.
Questions and Answers
Q: On what basis do you decide what type of bariatric surgery a patient should receive?
A: If a patient doesn’t know the answer, we ask that they go back to a patient support group. These are the most educated patients you will ever meet. I listen to patients, they tell me the truth.
Q: Can you predict which patients will go into remission before doing the surgery?
A: I wish I could! Like I said, we need to know who benefits most. Having had diabetes for a long time and having been on insulin lowers your chance of remission, but that doesn't mean that those individuals should be denied surgery – in fact, they may be the ones who benefit the most. It is an area of ongoing study.
Q: Is bypass surgery better than sleeve gastrectomy? Also, with an EndoBarrier liner, how long do you keep it in?
A: The EndoBarrier liner is licensed for 12 months, but you can do repeated therapy. The picture of what happens after is less clear, some put the weight back on and some don’t. In response to your first question, there is no significant difference, though they do work in different ways. My feeling is that with bypass you get more weight loss, but we need more studies for long-term data on gastrectomy. It hasn't been around for very long so there are less data.
Q: Clearly a sleeve can be a different thing to a different surgeon.
A: The technique for sleeve gastrectomy is variable. We have to standardize what we do so that our data can be interpreted in a way that’s useful.
Q: You showed quite impressive data about clinical outcomes. Can you tell us about patient satisfaction and quality of life?
A: Quality of life data are out there. Every study I know shows very good improvements in quality of life. In our experience, these patients do very well. I’ve seen patients who have had pretty severe surgical complications and still do not regret having it. These are some of the most satisfied patients I have seen in surgical practice.
Q: What are advantages and disadvantages, both immediate and late, of bypass surgery and sleeve?
A: The biggest limitation is that there needs to be very close follow up.
Q: Have you explored the intragastric balloon as form of weight loss surgery?
A: We know that as soon as you remove the balloon you gain the weight back. It’s mainly a good option for patients who are too heavy for surgery or too high-risk to undergo another form for surgery.
Q: If you get second thoughts after the operation, is it possible to reverse it?
A: We tell patients that gastric bypass is irreversible, but technically you can reverse it. It is a very demanding and dangerous procedure, however. The sleeve is reversible but you gain the weight back straight away.
Q: Could you do an oral glucose tolerance test after bariatric surgery?
A: It’s safe and we have definitely done it, but we’ve moved away from this method and are using the new criteria for remission.
Q: You have not given any data on any long-term nutritional deficiencies.
A: Data are available. For both banding and gastric bypass, there aren’t any major concerns.
Diabetes Prevention and Remission After Usual Care and Bariatric Surgery
Lars Sjöström, MD, PhD (University of Gothenburg, Gothenburg, Sweden)
Dr. Lars Sjöström, former primary investigator of the Swedish Obesity Subjects (SOS) study, strongly critiqued the current BMI-based eligibility criteria for bariatric surgery. He argued that it is invalid to base the criteria around BMI since baseline BMI does not predict any treatment effects so far examined, including surgery’s impact on diabetes prevention or remission. Instead he believes that eligibility decisions should place more importance on metabolic variables. Indeed, he presented data suggesting that a person’s baseline insulin and glucose levels are better predictors of treatment effect than BMI. Reviewing the SOS’s results, Dr. Sjöström underscored that bariatric surgery is able to robustly prevent type 2 diabetes over the long-term in people who are prediabetic and nondiabetic obese. Additionally, in the short-term, bariatric surgery causes diabetes remission in the majority of people with type 2 diabetes undergoing the procedure. However, after 20 years, ~75% of initially remitted patients have relapsed. It is unclear though, if they remit to a less severe form of type 2 diabetes than they had pre-operation. What research does show, however, is that people are less likely to develop diabetes complications even if they do relapse. Notably, it appears that people who have had diabetes for shorter periods of time are more likely to experience remission. Thus, Dr. Sjöström called for people to treat deteriorating glucose metabolism early, and suggested that people with prediabetes should potentially be recommended for surgery.
Questions and Answers
Q: Thank you for sharing such an impressive result from this landmark study. Given your expertise, if you were to design your SOS today, which procedures would you recommend?
A: Weighing successful treatment and dangers, if I were to design the SOS today, I would randomize between usual care and intensive care and gastric bypass. Of course biliopancreatic diversion has more radical effects on the metabolism.
Q: What did you seen in NAFLD?
A: We have only liver enzymes, so nothing sophisticated there with which to look at the liver.
Q: Are there data in patients with BMI under 30 kg/m2? There are some ethnic groups who show substantial incidence of diabetes in the 25 kg/m2 range.
A: There are some published data between 27 kg/m2 and 35 kg/m2. There are several studies ongoing. I think the end result will show that gastric bypass will influence the diabetic state markedly.
Q: You showed us a quite high relapse rate in the entire operated group. Can you break this out by surgery?
A: We have data, but we had only on the order of 500 patients, so we are not powered to look at the results in the surgical subgroups – especially in gastric bypass, since that was not the main surgery performed. In our study, remission is not significantly different with bypass than the other surgeries, though this could be due to the powering.
Q: Do you think that even if remission is transient it is still having a long-term impact on complications? Or perhaps when you have the relapse the severity of the disease is reduced?
A: We must remember that our definition of diabetes is an arbitrary cutoff and does not describe the whole story. A lady in Madrid has a paper in press where she shows that even in patients who have relapsed after initial remission, other CV risk factors are still improved. So this might be the reason why we see an effect on diabetes complications despite the fact that patients relapsed into diabetes.
Q: Are there any guidelines available for the control of diabetes after bariatric surgery?
A: Most surgeons don't use any diabetes treatment at all during the first years after surgery unless they are part of a study. As diabetes relapses are seen after five to ten years, I think the clinical practice is to treat these patients. The first-line drug, as I understand it, is metformin. I don't think that there are generally accepted rules and there are no randomized studies proving how we should do this best. This certainly begs for more study.
Comment: If BMI is not an indication for the surgery and to operate on people with diabetes does not have very good results, then the main indication to operate is in people with impaired glucose tolerance.
A: I don't think that is the main conclusion. We have the best results in preventing diabetes in people with impaired glucose tolerance. I am not saying that we should not treat the diabetic patients. Even if they relapse it appears that they still avoid more diabetes complications. We have shown strong 20-year results on diabetes complications. I think we should operate on people with diabetes.
Q: You did not mention anything about safety concerns. Also can you comment on bariatric surgery’s cost-benefit?
A: We had five post-operative mortalities in the study. When we started in 1987 a lot of studies reported up to 5% post-operative mortality. We trained the surgeons heavily before they were allowed to recruit patients and that might be why we were able to keep the mortality low. On cost benefit analysis, we have some early analysis on our five — six year data. At that time, there was no cost benefit, but there was also no decreased benefit. If we repeated this study today I think we would have a cost benefit. We published a JAMA paper last year, showing that the use of medications was quite dramatically reduced as compared to the control group. More investigations must be done. It took us 13 years until we saw a significant effect on mortality so I think look at cost benefit at five, ten years is much too early.
Pathophsyiology of Bariatric Surgery
Ele Ferrannini, MD, PhD (University of Pisa, Pisa, Italy)
Dr. Ele Ferrannini described the pathophysiology of Roux-en-Y gastric bypass (RYGB) and biliopancreatic diversion (BPD). Detailing and comparing the two procedure’s effects, he demonstrated that BPD improves insulin resistance more so than RYGB. An ongoing trial suggests that the reason for this could be BPD’s stimulation of the secretion of bile acids. Thus, bile acids could be an interesting treatment option. Dr. Ferrannini also noted that current research suggests that the best way to determine if a person with diabetes will respond to bariatric surgery is their beta cell function – the better a person’s beta cell function the better the treatment’s effect. Dr. Ferrannini therefore echoed Dr. Lars Sjöström’s (University of Gothenburg, Gothenburg, Sweden) call for modifying the eligibility criteria to discount BMI and focus on metabolic parameters (in the case of Dr. Ferrannini specifically beta cell function).
Symposium: Pathophysiological Phenotypes in Type 2 Diabetes
Treatment of Type 2 Diabetes Based on Pathophysiological Knowledge
Stefano Del Prato, MD (University of Pisa, Pisa, Italy)
Reflecting on treatment individualization with the ADA/EASD position statement, Dr. Stefano Del Prato questioned whether HCPs should individualize treatment for a person’s specific type 2 diabetes pathophysiology in addition to the factors explicitly mentioned in the document. He pressed that it is rational to want a pathophysiological treatment because it will treat a specific defect rather than merely compensating for it. Dr. Del Prato noted that the current arsenal of drugs might make it feasible to treat patients pathophysiologically, assigning at least one drug to each component of Dr. Ralph DeFronzo’s (University of Texas Health Science Center, San Antonio, TX) “Ominous Octet” of pathophysiological defects in type 2 diabetes. Dr. Del Prato then touched upon Dr. DeFronzo’s triple therapy (metformin plus pioglitazone plus exenatide twice daily), as it is based on pathophysiology. Dr. DeFronzo’s finding that initiation of triple therapy at diagnosis results in superior glycemic control compared to the conventional step-wise method previously endorsed by the ADA (metformin followed by subsequent addition of a sulfonylurea and then basal insulin) provides further evidence, in Dr. Del Prato’s mind, that type 2 diabetes can be treated pathophysiologically and that it can improve glycemic control compared to a more compensational approach (for more details on Dr. DeFronzo’s triple therapy results, please see page 8 of our ADA 2013 Report – Treatment Algorithms and Strategies at http://www.closeconcerns.com/knowledgebase/r/76f49d51). However, the triple therapy is tailored to the average pathophysiology of type 2 diabetes, not a person’s unique presentation. Dr. Del Prato lamented that our current understanding of type 2 diabetes biomarkers and genetics does not enable practitioners to fully individualize treatment based on a person’s specific pathophysiology.
Dr. Del Prato made the case that it is logical to treat type 2 diabetes in a pathophysiological manner. Dr. Del Prato explained that a compensatory approach (e.g., sulfonylureas) might improve glycemic control in the short-term; however, it does not necessarily control the diseases’ underlying mechanisms. For example, he explained that sulfonylureas (SFUs), DPP-4 inhibitors, and GLP-1 agonists can be used to stimulate beta cell insulin secretion. However, while SFUs induce beta cell apoptosis in cultured humans islets (Maedler et al., J Clin Endocrinol Metab 2005), GLP-1 treated islet cells appear to be protected (Fanila et al., Endocrinology 2003). Given the importance of beta cell mass in type 2 diabetes, Dr. Del Prato suggested that incretin mimetics (unlike SFUs) might be able to modify the pathophysiology of type 2 diabetes.
Dr. Del Prato, therefore, did not list SFUs as a pathophysiological treatment for impaired insulin secretion or any other component of the ominous octet. During Q&A he noted that SFUs would be a pathophysiological option for people with a defect in the beta cell’s potassium channel. Indeed, people with MODY due to certain mutations impacting this channel respond particularly well to SFUs.
Dr. Del Prato showed that each component of the “Ominous Octet” of pathophysiological defects in type 2 diabetes can be addressed by a currently available agent and that more options are in development. Below are the agents he noted for each pathogenic characteristic of type 2 diabetes.
Table 1: Choices for Pathophysiological Treatment of Type 2 Diabetes
Agents in Development
Impaired insulin secretion
DPP-4 inhibitors and GLP-1 agonists
Glucokinase activators, GPR agonists, and IL-receptor antagonists
Decreased incretin effect
DPP-4 inhibitors, GLP-1 agonists, and bile acid sequestrants.
Increased glucose reabsorption
Dual and pan PPARs activators, and 11HSD inhibitors
Decreased glucose uptake
Dual and pan PPARs
Bromocriptine (Santarus’ Cycloset)
Increased hepatic glucose production
Metformin and TZDs
Glucokinase activators and glucagon receptor antagonists
Increased glucagon secretion
DPP-4 inhibitors and GLP-1 agonists
Glucagon receptor antagonists.
- The pathophysiology of type 2 diabetes is believed to vary among patients. Dr. Del Prato supported this hypothesis, citing evidence that Arabic people have substantially lower insulin sensitivity while Japanese people have a higher insulinogenic index (Abdul-Ghani et al., Diab Metab Syndrome 2007).
- Thus, Dr. Del Prato would like for HCPs to be able to personalize their patients’ treatment according to their underlying pathophysiology. For example, he believes a person with more insulin resistance of peripheral tissues than at the hepatic level, might benefit more from a TZD than metformin, since TZDs have a greater impact on insulin-mediated glucose uptake.
- Unfortunately, such individualization is not currently possible due to the lack of validated biomarkers of the disease’s underlying pathophysiology or predictive alleles. α-hydroxyburate might be a biomarker of insulin resistance and the development of type 2 diabetes (Gall et al., PLoS ONE 2010; Ferrannini et al., Diabetes 2013); however, Dr. Del Prato thinks it needs further validation. Similarly a number of potential beta cell biomarkers exist (e.g., insulin, amylin, IA-2, etc.), yet they have also not been validated.
- Dr. Del Prato emphasized the importance of early intervention, citing ACCORD, ADVANCE, and VADT as evidence that late intervention cannot reverse CV damage, while citing UKPDS as demonstrating that early intervention has both positive microvascular and macrovascular effects.
- While discussing the impact metformin has on type 2 diabetes pathophysiology, Dr. Del Prato noted that metformin might act as a GLP-1 enhancer and sensitizer (Cho and Kieffer, Diabetologia 2011). He therefore hypothesized that combining metformin and incretin mimetics, could potentially be synergistic. Indeed, a two-year study of sitagliptin and metformin found that when the agents are taken as an initial combination therapy, patients achieve better glycemic control than with monotherapy with either agent (Williams-Herman et al., Diab Obese Metab 2010). However, it was not clear that the two drugs’ glycemic benefit was multiplicative rather than additive. Additionally, the glycemic control gradually degraded over the two years. Dr. Del Prato noted that some people think that combining exenatide and metformin could be more durable; however, he thinks longer studies with both DPP-4 inhibitors and GLP-1 receptor agonists are needed before any conclusions can be drawn.
- Dr. Del Prato also briefly explained that glucose appears to stimulate the release of GLP-1 from certain pancreatic alpha cells (alpha-TC1/6 cells). If this occurs within the body it would mean that GLP-1 is not solely a gut hormone secreted by the intestinal L cells. Dr. Del Prato’s group is presenting this research in a poster at EASD (#578).
Questions and Answers
Q: One thing you have not addressed is the possibility of looking at the response to treatment in patients and then making a decision on whether to continue drug use?
A: You are right, but I don't see that as being much different from what we are currently doing. I think that nowadays trial and error might be the only option we have, on top of phenotypic characterization. On the contrary I was elaborating on the possibility to individualize treatment since diagnosis on the basis of the main pathophysiologic mechanism(s).
Q: But Stefano don't you think that we forget to stop a drug the patient is not responding to?
A: Oh for sure, but I don't think that has anything to do with the pathophysiology, rather to our own clinical inertia.
Q: We do some personalize insulin, based on a person’s glucose profile. Do you think we can personalize orals around the glucose profile too?
A: Yes we can do that but that is not pathophysiologically driven, rather it is phenotype driven.
Q: Could C-peptide be of any use in determining type 2 diabetes pathophysiology?
A: C-peptide is a potential biomarker for beta cell function. However, I think that C-peptide is a tricky marker as it is still dependent on prevalent glucose levels. We need biomarkers that are totally independent of the environmental condition.
Q: Is there no place for SFUs?
A: I think that if you can identify a prevalent defect at the level, for instance, of the potassium channel as it occurs for some neonatal forms of diabetes SFU are treatment of choice. I understand that this is not a typical condition in type 2 diabetes, but it makes the point for a treatment based on pathophysiologic knowledge. And this was the issue I was trying to elaborate upon: whether or not it is possible to base treatment on the pathophysiology .
Can Genotyping Aid in the Choice of Antidiabetic Therapy?
Leif Groop, MD, PhD (Lund University, Malmö, Sweden)
Dr. Leif Groop explained that although certain genetic variations do appear to impact a person’s response to a given anti-diabetic agent, the effect is too slight to justify genetic screening. Some evidence exists, according to Dr. Groop, that genetic variations have a small impact on the efficacy of sulfonylureas (TCF7L2), metformin (OCT1 and ATM), and DPP-4 inhibitors (CTR1). Instead of pharmacogenetics helping to identify who will respond to a given agent, Dr. Groop remarked during Q&A that it could be more effective at determining who will experience more severe side effects on a drug (at least based on medical experience with other diseases). For example, the odds ratio of developing statin-induced myopathy is 4.5 (95% CI, 4.7 to 61.1) if a person taking statins has a SCLO1B1 variant (SEARCH Collaborative Group, NEJM 2008). Dr. Groop noted that genetics could also help researchers identify new therapeutic approaches.
Genetic studies discovered that people with certain melatonin (MTNR1B) alleles have an increased risk for developing type 2 diabetes (Lyssenko, Nat Genet 2008). Investigators subsequently learned that these risk alleles led to increased expression of melatonin, suggesting that melatonin impairs insulin secretion. Indeed, MTNR1B knockout mice have heightened insulin secretion. Risk genotype carriers of MTNR1B have increased expression of MTNR1B in human islets. Thus, Dr. Groop recommended further investigation into the role of melatonin in glucose homeostasis. Dr. Groop hypothesized that melatonin helps defend the body from nighttime hypoglycemia, since its levels are highest at night.
Questions and Answers
Q: Given the low risk associated with each allele, perhaps we are better off stratifying patients by biological pathways.
A: Absolutely, if you really did know those biological pathways. What I think we really need is better outcome data. When you treat people you just add on additional drugs so it is hard to dissect the effect of each drug. You are right though. Of course all pathways going into impaired insulin secretion could have an effect. I forgot to mention one recent study where they showed that a variant in the CTR1 gene actually predicted response to DPP-4 inhibitors, but again the effect size was modest.
Q: People these days seem to be more concerned about side effects than efficacy. What is the potential role of pharmacogenetics in identifying people with regards to side effect response?
A: I think the odds of that working are greater. My guess would be that pharmacogenetics could help us more in predicting side effects than in determining who will respond. In the cancer field they have very good data that if you have a variant in one of those genes that a drug will have really severe side effects.
Oral Presentations: Hypoglycemia – Balancing Glucose Control
Ulrik Pedersen-Bjergaard, MD (Hillerød Hospital, Hilleroed, Denmark)
Dr. Pedersen-Bjergaard presented notable results from the HypoAna study – the two-year crossover trial included 159 type 1 patients with recurrent severe hypoglycemia (>2 episodes in the past year). Patients were randomized to a basal-bolus regimen using analog insulins (detemir and aspart) or human insulin/NPH. Each patient underwent a three-month run-in period, nine months of treatment in one of the arms, a three-month washout period, and then nine months in the other arm. Significantly, treatment with analog insulin resulted in a 29% rate reduction (p<0.05) in severe hypoglycemia (intention-to-treat analysis), which rose to 34% when adjusted for A1c. This corresponds to an absolute rate reduction of 0.5 severe hypoglycemia episodes per patient-year, meaning the number needed to treat with insulin analogs to avoid one episode of severe hypoglycemia is just two patients per year. Striking and victorious results to say the least, and notably, the reduction in hypoglycemia in the analog arm still translated to a 0.13% A1c advantage over human insulin (p<0.05; baseline A1c: 8%). Dr. Pedersen-Bjergaard showed an interesting graph of severe hypoglycemia by time of day – the advantage of analogs only showed up in the nocturnal period. Overall, we were extremely glad to see the real-world ambition of this study and the inclusion of patients with recurrent severe hypoglycemia, those who arguably need analogs the most. We hope this data can be taken to governments and payers who do not cover analog insulin for type 1 diabetes – we think the study clearly supports cost savings with analog insulin.
The HypoAna trial was a two-year, crossover, seven-center study. The protocol randomized 159 patients to treatment with basal-bolus therapy with insulin aspart/detemir or regular human insulin/NPH. Patients had a three-month run-in period with the insulins, followed by nine months of maintenance. There was then a three-month washout period, followed by crossover to the other therapy for nine months. The glycemic target was the maintenance of baseline A1c (8%), as these patients had a high frequency of severe hypoglycemia. The number of daily doses and adjustment of insulin was at the discretion of the local investigator.
Dr. Pedersen-Bjergaard emphasized the key inclusion criterion: patients had to have had two or more episodes of severe hypoglycemia in the previous year. The study’s primary endpoint compared severe hypoglycemia between analogs and human insulin in the nine-month maintenance periods. Severe hypoglycemia was defined strictly as requiring third party assistance. Secondary endpoints included hypoglycemia as measured via CGM. The CGM data was not shown in this presentation. A blinded adjudication of endpoints according to Whipple’s triad and severity was applied.
Fifty-three percent (53%) of the 159 patients were classified as hypoglycemia unaware and 41% had impaired hypoglycemia awareness (based on a Danish questionnaire). Patients had a mean duration of diabetes of 30 years and a baseline A1c of 8%.
While 159 patients were randomized, there were 49 dropouts (36 while on human insulin and 13 while on an insulin analog). While that reflects a fairly high 31% dropout rate, it disproportionately occurred in the human insulin arm. We wonder if the dropout reflects patients seeing worse clinical outcomes with human insulin. We suspect if the dropout rate had been lower, the data would have been even more compelling.
Overall, 64% of participants had an episode of severe hypoglycemia during the study. The mean rate during the maintenance periods was 1.3 episodes per patient per year. A striking 34% of these episodes resulted in a coma.
The prevalence of severe hypoglycemia was 55% with human insulin vs. 39% with analog insulin – this translates to a 29% rate reduction, which rises to 34% when adjusted for A1c.
The baseline A1c of 8% was maintained throughout the study in both groups – the slide showed no noticeable improvement or decrement in either group over the two-year trial. A1c was a very slight 0.13% better during analog treatment (p<0.05), though this was not a point of emphasis.
To us, this study provides clear proof for why A1c does not tell the full story – anyone looking at this study would say patients were miles better off with analog insulin (and far less costly), though that improvement was not at all reflected in A1c. We hope that payers are governments will be very receptive to this type of severe hypoglycemia data, since it is so directly tied with economic data.
Dr. Pedersen-Bjergaard highlighted a Swedish survey that found patients experience more than one episode of severe hypoglycemia per patient-year (Kristensen et al., Diabetes Res Clinc Pract 2012). Further, the survey discovered a disproportionate distribution of events: only 20% of patients had recurrent severe hypoglycemia, though they accounted for up to 90% of all events.
To date, insulin analogs have shown a less clear impact on severe hypoglycemia for three main reasons: 1) high risk patients are excluded from studies; 2) studies are thereby deflated from statistical power; and 3) hypoglycemia events are mainly safety endpoints. These limitations were the rationale for conducting this study in patients with recurrent severe hypoglycemia.
Questions and Answers
Q: Did you think about the ethics of the study? Many of us would use analogs in people who have these problems already.
A: Of course, we all have used analog insulins in high-risk subjects. We have personal experiences indicating that it may be beneficial. Still, we felt that this study was needed, because there are no scientific data to support these experiences.
Q: Very nice paper. You had 50% of your population with hypoglycemia unawareness. How was that diagnosed – a questionnaire? And what did you do during the run-in period and the crossover period – did patients go back to their previous regimen?
A: Hypoglycemia unawareness was classified according to a Danish developed questionnaire, and it was classified in three groups: normal, impaired, and unawareness. This was not an inclusion criteria; it was only a description of the subject. 95% of patients had some degree of impaired awareness. During run-in and crossover, there was no insulin algorithm supplied. It was all at the discretion of the local investigator. Subjects came into the study with a certain insulin treatment, and the investigator shifted to the trial treatment based on their own experience.
Q: Congratulations on this study. Did you recommend a different carb distribution in the different treatment arms?
A: We did no other interventions other than the treatment.
Q: This was a real-world study, so you left it to the people and investigators. Did you collect data on how people used the insulin – twice-daily NPH and so on?
A: I did not show the distribution of insulin regimens for sake of time. We saw the same difference between the treatment arms regardless of how many basal injections they got in the study.
Q: There is a difference in PK and PD in the human insulin regimen. Did you give different instructions on one insulin regimen vs. another?
A: There were no standardized instructions given to the patients. They just followed usual care.
Allan Vaag, MD (Rigshospitalet, Copenhagen, Denmark)
Dr. Allan Vaag presented a post-hoc meta-analysis of two 26-week phase 3 trials comparing Novo Nordisk’s IDegAsp (a premix of 70% insulin degludec/30% insulin aspart) to biphasic insulin aspart (NovoMix: 70% insulin aspart protamine/30% insulin aspart). IDegAsp had a significant advantage over NovoMix in all measures of hypoglycemia: a 19% lower rate of overall confirmed hypoglycemia (p=0.03); a 57% lower rate of nocturnal confirmed hypoglycemia (p=0.0001); and a 39% lower rate of severe hypoglycemia (not statistically significant, as there were only 29 episodes). Dr. Vaag mentioned that in the FDA’s review of insulin degludec, the agency also requested hypoglycemia outcomes during the maintenance period (“when patients had been used to using these novel insulins”). When narrowing the IDegAsp vs. NovoMix data to the maintenance period, the hypoglycemia data looked even stronger – IDegAsp had a 31% lower rate of overall confirmed hypoglycemia (p<0.05), a striking 84% lower rate of severe hypoglycemia (P<0.05), and 62% lower rate of nocturnal confirmed hypoglycemia (p<0.0001). He emphasized that the these definitions of hypoglycemia were “discussed and challenged by the FDA,” but they were prespecified for insulin degludec, and thus, used in this meta-analysis.
Oral Presentations: Impact of Treatment and Genetic Susceptibility to Comorbidities and Mortality
Diabetes, Incretin Therapy, Pancreatitis, and Pancreatic Cancer: Meta-Analyses
Peter Boyle, PhD (International Prevention Research Institute, Lyon, France)
Dr. Peter Boyle explored the potential associations between diabetes, incretin therapies, pancreatitis, and pancreatic cancer through a series of meta-analyses. His presentation was a condensed version of the talk he gave during Sanofi’s corporate symposium on Sunday; for our coverage of the Sunday event, please see page 20 of our EASD Day #2 Report at http://www.closeconcerns.com/knowledgebase/r/f517afa8. Pooling the results of randomized control trials and observational studies produced a strong correlation between diabetes and pancreatitis (an unsurprising finding). A separate meta-analysis of 70 independent studies (including RCTs and observational studies), however, showed no evidence of increased risk of pancreatitis with incretin therapies (summary relative risk=1.08; 95% CI, 0.87-1.34). Dr. Boyle took issue with the over reliance on databases such as the FDA-AERS, which cannot provide information on potential confounding factors and may be subject to significant reporting bias. Dr. Boyle concluded by noting that, at this time, poor data quality precludes the possibility of drawing definitive conclusions about pancreatitis risk with incretin therapy and that large prospective studies are urgently needed to provide satisfactory answers. Dr. Boyle, however, hypothesized that if any risk for pancreatitis exists with incretin mimetics, it probably is small.
Questions and Answers
Q: I had found the data from Peter Butler’s study convincing. Is it possible that there are some sort of subclinical changes caused by these agents that don’t immediately manifest as clinical symptoms?
A: What we are here to discuss is the risk of pancreatitis and pancreatic cancer. I don’t agree that the work of Peter Butler is very robust. I had a privilege to attend a NIDDK meeting in June on this issue [for our coverage of this meeting, please see http://www.closeconcerns.com/knowledgebase/r/8326d0dd and http://closeconcerns.com/knowledgebase/r/e90a832d], and there was strong criticism of Dr. Butler’s study. There is still lot of doubt about the reliability of that work, and it is based on a very small sample size.
Q: You mentioned that we need large population-based studies. Have you worked out what sample size and duration we would need to resolve the issue of pancreatitis and cancer?
A: We would love population studies of about 15 million people followed for four to five years, but we’re not going to get that. I think that a 3,000-4,000 patient trial lasting four to five years should suffice, assuming 100 events per 100,000 patients. Databases are inadequate; we need more direct, clinical information about the etiology of pancreatitis.
Q: Could you comment on the reporting system through which data makes its way into studies? Sometimes databases use passive reporting, and sometimes this data makes its way into scientific papers. As a scientific community, should we be limiting ourselves to randomized control trials and better observational studies?
A: We’ve got to be careful — these database studies are fraught with difficulties, and you have to pay attention to get them right. I think the FDA database has some problems. The databases in England, Denmark, and the Swedish registry are fairly rigorous. Generally, you have to be careful about where data come from.
Craig Currie, PhD (Cardiff University, Cardiff, UK)
Dr. Craig Currie presented one of two BMS/AZ-supported, retrospective analyses his research team performed on the safety of sulfonylureas (SFUs) as either a first line agent or as an add-on to metformin. The researchers looked at the Clinical Practice Research Datalink, a retrospective UK dataset that includes roughly 10% (~10 million) of all patients treated in primary care in the UK. The analysis looking specifically at SFUs in combination with metformin found that being on a SFU plus metformin (n=33,983) was associated with an adjusted hazard ratio of 1.357 (95% CI, 1.076-1.710; p=0.010) compared with combination therapy with a DPP-4 inhibitor and metformin (n=7,864). For context, 84 people on a DPP-4 inhibitor and metformin died and 1,133 people on an SFU and metformin died. All of the subgroup analyses found a numerical imbalance in the favor of DPP-4 inhibitors with regards to all-cause mortality, and a few were statistically significant (females, older than 63 years, and having a Charlson index of at least two). Dr. Currie’s group also conducted two sensitivity analyses: 1) a matched-cohort study using age, gender, year of index exposure, diabetes duration, BMI, serum creatinine, and A1c, and 2) a matched-cohort using a propensity score predicted by a “wide range of candidate co-variables” The adjusted hazards ratio significantly disfavored the SFU plus metformin arm for both the directly matched analysis (aHR=1.85; 95% CI, 1.40-2.45) and the propensity matched analysis (aHR=1.50; 95% CI, 1.09-2.05).
The researchers looked at the Clinical Practice Research Datalink: a retrospective UK dataset, which comprises ~10% (n=10 million) of all patients treated in primary care in the UK. From this dataset, the researchers extracted people with type 2 diabetes who initiated treatment between 2007 (when DPP-4 inhibitors were introduced in the country) and 2013. They criteria required a six-month wash-in period to ensure people had not previously received the studied treatment. The index date was defined as the first prescription date and the end date was when a person died or 90 days after her last prescription.
The main comparative analysis utilized a Cox proportional hazards model to compare time to death. The covariates included were age, gender, diabetes duration, A1c, SBP, HDL, LDL, triglycerides, serum creatinine, BMI, smoking status, and baseline morbidity as measured by contacts with their general practitioner and a Charlson index (which stratifies comorbidity). We were disappointed that Dr. Currie’s group did not use a more robust measure of renal failure than creatinine.
Dr. Currie also conducted two sensitivity analyses. One was a matched-cohort study using the following criteria: age (±2 years), gender, year of index exposure, diabetes duration (±1 year), BMI (±3 kg/m2), serum creatinine (±10 μmol/l), and A1c (±1%). The second analysis was a matched-cohort using a propensity score predicted by a “wide range of candidate co-variables” (Dr. Currie did not specify these).
The baseline characteristics between the SFU plus metformin group and the DPP-4 inhibitor plus metformin group varied significantly for several covariates. The average age was 62.3 years in the SFU and metformin arm vs. 59.6 years in the DPP-4 inhibitor and metformin arm (p <0.001). The baseline A1c was 9.1% in the SFU and metformin arm and 8.5% in the DPP-4 inhibitor plus metformin arm (p <0.001). The mean diabetes duration was 5.1 years in the SFU plus metformin arm and 5.5 years (p=0.010) in the DPP-4 inhibitor and metformin arm. The mean follow-up for each treatment was not reported.
Being a retrospective analysis, this study had several important weaknesses, the magnitude of whose impact attendees sharply questioned during Q&A. As noted above, large differences existed in the baseline characteristics between the two groups. Though Dr. Currie worked to control for these differences, due to the dataset’s limitations, he was not able to control for several key covariates, including race/ethnicity and income. Additionally, it is possible that confounding by indication could have impacted the results – a patient characteristic might cause a physician to prescribe a SFU inhibitor instead of a DPP-4 inhibitor, and make a person more likely to die independent of taking a SFU (though the matched-cohort using a propensity score might have been able to account for both of these variations). Furthermore, the DPP-4 inhibitor and metformin arm had relatively few events (n=84), raising the question of whether the result could have been skewed by chance. We also note that Dr. Currie’s analysis included many different SFUs. It could be that earlier generation SFUs drove mortality risk, while later generation options (i.e., gliclazide) were less harmful. We would be interested in the results of a subanalysis looking at each SFU’s risk independently.
Questions and Answers
Q: One of the problems with these analyses is that you are looking at relatively new drugs and we have not accumulated enough numbers. In terms of mortality the numbers were relatively small.
A: We don't have any control over these data. We get the data and we periodically analyze the data to independently determine if there are any signals. All we can do is wait.
Q: Can you tell us what drugs did you mostly see?
A: The primary DPP-4 inhibitor in this study was Januvia.
Q: I am interested in the cause of the deaths. Do you have data on that, and on hypoglycemia?
A: We tend not to look at hypoglycemia since it is fairly unreliably reported. Similarly, the death report does not come with the primary cause of death. All cause mortality, however, uses all CV, cancer, etc. deaths.
Q: We know there are significant differences between SFUs in terms of CV risk and mortality. So under this view, which are the SFUs you have in your study?
A: That is true. We do have that information. It is possible for us to break it up further.
Q: The current recommendations in the UK are to introduce patients on metformin and then later SFUs. What are the criteria to introduce a person onto a DPP-4 inhibitor and metformin?
A: It guides for physicians doing so when a patient is contraindicated for a SFU. However, nobody takes care of the guidelines anyways. So… [Laughter]
Q: You started your study when gliptins were introduced, so there is a calendar year bias. Did you adjust for that?
A: We did look at calendar year.
Q: You said that we should be taking this to the regulatory authorities. Would you like to comment on the limitations of your study? Including the varied baseline characteristics.
A: This is just one part of the puzzle. I think that the majority of population scientists think that there is a reason to look at SFUs and when I’m on it insulins as well.
Christopher Morgan, MSc (Pharmatelligence, Cardiff, UK)
Dr. Christopher Morgan presented the second of two BMS/AZ-supported, retrospective analyses his and Dr. Craig Currie’s (Cardiff University, Cardiff, UK) research team performed on the safety of sulfonylureas (SFUs). In both studies, the researchers looked at the Clinical Practice Research Datalink, a retrospective UK dataset, which comprises ~10% (n=10 million) of all patients treated in primary care in the UK. The analysis looking specifically at SFUs as a first-line monotherapy, found that being on a SFU (n=15,687), instead of metformin (n=76,811) was associated with an adjusted hazard ratio of 1.580 (95% CI, 1.483-1.684; p<0.001). All of the subgroup analyses found a statistically significant imbalance in the favor of metformin. Like the first trial, the research group also conducted two sensitivity analyses: 1) a matched-cohort study using age, gender, year of index exposure, diabetes duration, BMI, serum creatinine, and A1c, and 2) a matched-cohort using a propensity score predicted by a “wide range of candidate co-variables” The adjusted hazards ratio significantly disfavored the SFU arm for both the directly matched analysis (aHR=1.38; 95% CI, 1.05-1.70) and the propensity matched analysis (aHR=1.90; 95% CI, 1.73-2.09).
The researchers looked at the Clinical Practice Research Datalink, a retrospective UK dataset, which comprises ~10% (n=10 million) of all patients treated in primary care in the UK. From this dataset, the researchers extracted people with type 2 diabetes who initiated treatment between 2000 and 2013. They required a six months wash-in period to ensure people had not previously received the studied treatment. The index date was defined as the first prescription date and the end date was when a person died or 90 days after her last prescription.
As with the first study, the main comparative analysis utilized a Cox proportional hazards model to compare time to death. The covariates included were age, gender, diabetes duration, A1c, SBP, HDL, LDL, triglycerides, serum creatinine, BMI, smoking status, and their baseline morbidity as measured by contacts with their general practitioner and a Charlson index (which stratifies comorbidity). We were disappointed that the group did not use a more robust measure of renal impairment than creatinine.
Dr. Morgan also conducted two sensitivity analyses. One analysis was a matched-cohort study using the following criteria: age (±2 years), gender, year of index exposure, diabetes duration (±1 year), BMI (±3 kg/m2), serum creatinine (±10 μmol/l), and A1c (±1%). The second analysis was a matched-cohort using a propensity score predicted by a “wide range of candidate co-variables” (Dr. Morgan did not specify these).
The baseline characteristics between the SFU plus metformin group and the DPP-4 inhibitor plus metformin group significantly varied for several covariates. The average age was 61 years in the metformin arm vs. 68 years in the SFU arm (p <0.001). Baseline creatinine was 85 in the metformin arm and 98 in the SFU arm (p<0.001). The Charlson index was 1.9 in the metformin arm and 2.2 in the SFU arm, indicating that people receiving SFU monotherapy were significantly more ill at treatment initiation (p<0.001). The mean A1c was 8.6% in the metformin arm and 9.1% in the SFU arm (p<0.001). The mean follow up for each treatment was 3.1 years.
Being a retrospective analysis, this study had several important weaknesses — their magnitude was sharply questioned during Q&A. As with the first study, while the research group worked to control for differences in baseline characteristics between the two groups, due to the dataset’s limitations, they were unable to control for several key covariates, including race/ethnicity and income. Additionally, it is possible that confounding by indication could have impacted the results – a patient characteristic might cause a physician to prescribe a SFU instead of metformin, and make a person more likely to die independent of taking a SFU. Though the matched-cohort using a propensity score might have been able to account for both of these variations. Furthermore, as one attendee noted, “A physician prescribing a SFU first line in this day in age is not the same as a physician prescribing metformin.” Thus, there could have been uncontrolled physician level covariates. We also note that Dr. Currie’s analysis included many different SFUs. It could be that earlier generation SFUs drove mortality risk, while later generation options (i.e., gliclazide) were less harmful. We would be interested in the results of a subanalysis looking at each SFU’s risk independently.
Questions and Answers
Q: Could it be metformin intolerance that biases people to die?
A: These were all patients whose first ever script was for either a SFU or metformin.
Q: Did you control for ethnicity?
A: Ethnicity could not go into the model because it is poorly coded in the dataset.
Comment: You can imagine that a physician prescribing a SFU first line in this day in age is not the same as a physician prescribing metformin.
A: That is true, but it used to be that people were prescribed SFUs over metformin in general.
Comment: The variations seen in the subgroup analyses make me concerned that there were unmeasured confounding factors.
A: All of the adjusted hazard ratios are within one another’s confidence intervals, except for the younger than 63 years group, which was even more likely to die on an SFU. Potentially there are residual confounders. We adjusted as much as we could. We cannot deal with what is unknown. The fact that all three of our analyses came up with a significant difference indicates that there is something in this finding.
Q: Did you control for renal function?
A: We used creatinine as a proxy.
Q: Did you take into account the differences between SFUs?
A: Everybody bar 300 patients was prescribed a second generation SFU. Gliclazide was the most common. I think 12,000 of the patients were on gliclazide.
Q [Dr. Ele Ferrannini (University of Pisa, Pisa, Italy)]: The results of this and the previous study suggest that SFUs will kill more people than DPP-4 inhibitors and metformin. How do you reconcile this finding with trial evidence? In ORIGIN, SFUs did not kill an excess number of people. If I remember correctly, SFUs in the UKPDS were not associated with excess mortality.
A: In UKPDS they were at an increased risk when on an SFU in combination with metformin.
Dr. Ferrannini: That analysis was on only 320 people.
Q: Have you factored in a time dependent analysis? I am interested in if there is an SFU dose effect. That would be one way to get around there being an indication bias in these analyses.
A: That is something we could do.
Oral Presentations: Novel Therapeutic Agents and Insights
Dissociation Between Metformin Plasma Exposure and Its Glucose-Lowering Effect: A Novel Gut-Mediated Mechanism of Action
John Buse, MD, PhD (University of North Carolina, Chapel Hill, NC)
After proposing how metformin’s therapeutic effects might be mediated primarily at the level of the gut, Dr. John Buse discussed findings of two clinical studies of a new formulation of metformin, metformin delayed release (Met DR; Elcelyx’s NewMet), that substantiate this hypothesis. In a single day dosing, four-way crossover trial in healthy subjects (n=20) comparing the pharmacokinetics of Met DR to other formulations, Met DR had significantly reduced bioavailability and lower plasma exposure compared to metformin IR and metformin XR. In a five-day, three-way crossover study comparing the effects of Met DR versus metformin IR in type 2 diabetes patients (n=24), metformin DR and metformin IR treatment led to similar and significant improvements in fasting plasma glucose, GLP-1, and PYY from baseline, with Met DR having markedly lower metformin exposure. Adverse events were fairly even across treatments, with the suggestion of improved GI tolerability with Met DR. Based on the findings from these two studies, Dr. Buse concluded that metformin’s glucose-lowering action depends on gut epithelial exposure, not on plasma exposure. Encouragingly, he stated that Met DR directed to the lower bowel may provide maximum efficacy with improved tolerability at lower doses (≤1,000 mg total daily dose), and that it could potentially be used in patients with moderate to severe renal impairment, as reduced plasma metformin exposure would be expected to reduce the risk of lactic acidosis.
Dr. Buse cited evidence supporting the hypothesis that metformin’s mechanism of action is largely mediated at the level of the gut. Perhaps the strongest available evidence to date, metformin was shown to be acutely effective at lowering glucose when administered intraduodenally, but not when administered intravenously in STZ rats (Stepensky et al., Drug Metab Dispos 2002). In addition, Dr. Buse stated that in other studies, the level of metformin in blood had little relationship with its efficacy as a glucose-lowering agent. He also noted that metformin has been documented to increase secretion of GLP-1 and PYY (secreted by L-cells, which are predominantly found in the lower bowel), raising the possibility that plasma exposure may not be necessary for metformin’s therapeutic activity.
Dr. Buse described metformin delayed release (Met DR), a new formulation of metformin that bypasses the upper bowel to deliver its full dose to the lower bowel, where absorption is poor. Thus, metformin DR plasma absorption is ≤25% in the lower bowel, and more of the delivered dose remains in the lower bowel (the site of action). Met DR also enables a maximally effective dose at lower total doses than current metformin formulations. This is expected to reduce the incidence of nausea (upper GI-tract related), and reduced plasma exposure may allow for use in patients with renal impairment.
In a single day dosing, four-way crossover trial in healthy subjects (n=20) comparing the pharmacokinetics of metformin DR to other formulations, Met DR had significantly reduced bioavailability and lower exposure when compared to metformin IR and metformin XR. Participants in the trial received 24-hour dosing of 1,000 mg metformin IR BID, 1,000 Met DR BID, 500 mg Met DR BID, and 2,000 mg metformin XR QD in a randomized sequence, with a 3-7 day washout period between treatments. Compared to metformin IR and metformin XR, Met DR had lower peaks in plasma metformin concentration; overall, 1,000 mg Met DR BID and 500 mg Met DR BID resulted in 52% and 68% lower relative exposure, respectively, than 1,000 mg metformin IR BID (p<0.0001), and 48% and 65% lower relative exposure than 2,000 mg metformin XR QD (p<0.0001).
In a five-day, three-way crossover study comparing the effects of Met DR versus metformin IR in type 2 diabetes patients (n=24), fasting plasma glucose was significantly and similarly reduced with metformin DR and metformin IR. In the study, participants received five days of 1,000 mg metformin IR BID, 1,000 mg Met DR BID, and 500 mg Met DR BID treatment in a randomized sequence, with seven-day washouts between treatments. Notably, Met DR BID decreased systemic exposure, with respective 45% and 57% decreases for 1,000 mg Met DR BID and 500 mg Met DR BID versus 1,000 mg metformin IR BID (n=19; p<0.0001). All three treatments led to significant improvements from baseline in fasting glucose, GLP-1, and PYY.
Questions and Answers
Q: Your conclusions were quite bold. How much of the changes in weight with metformin can be explained by changes in GLP-1 and PYY?
A: We certainly can’t see that from these studies. There are ongoing 12-week studies where the company has reported that the primary endpoint was met at four weeks, but there were really no details beyond hat provided. Weight loss with metformin is quite modest, and the mechanisms have not quite been worked out yet.
Q: What about the anti-proliferative and anti-mitogenic effects of metformin in this different formulation?
A: Here with this formulation, we are delivering metformin to the lower bowel. The lower bowel exposure is quite similar to the 2,000 mg dose of metformin, provided differently. Whether plasma exposure is required for some other action other than glucose lowering is unknown. Particularly with regards to anti-proliferative effects, I think those are somewhat under question, and how important they are clinically is unclear.
Q: Can you elaborate on why you see less diarrhea with this new formulation?
A: It would be speculation, but what I would suggest is that we don’t see more diarrhea than we see with previous formulations of metformin. In particular, when we use the 500 mg dose twice a day, the amount of metformin being delivered to the lower GI tract is on par with the current 2,000 mg dose of metformin. I do believe that there is a notion that metformin in the lower bowel may have some relationship to diarrhea as an adverse effect. We think the major difference will be less nausea, because of less delivery of metformin to the upper bowel. I think the pre-formed hypothesis is that rates of diarrhea would be similar, though it might be modestly less.
Oral Presentations: Glucose Down The Drain
Bruce Perkins, MD (Toronto General Research Institute, Ontario, Canada)
This single-arm, open-label, proof of concept, eight-week study explored use of BI/Lilly’s SGLT-2 inhibitor empagliflozin in 42 patients with type 1 diabetes (mean age: 24 years; mean BMI: 25 kg/m2). We first covered results of this study in a poster in our ADA 2013 Full Report; see page 139 at http://www.closeconcerns.com/knowledgebase/r/94f937d8. Results of the treatment period were compared to a two-week placebo run-in period. Mean A1c levels decreased by 0.4% from an 8% baseline, and a stratified analysis found a more pronounced decline in those with the highest baseline A1c (>8.0%). Frequency of symptomatic hypoglycemia was reduced by 33% (0.12 to 0.04 events/day) A range of non-hypoglycemia adverse events were reported by the sample, including two cases of diabetes ketoacidosis that forced the two participants to drop out of the study (though Dr. Perkins attributed these occurrences to the effect of aggressive insulin reductions and not empagliflozin itself – a reminder that more defined insulin protocols may be needed in future studies). Two interesting and currently inexplicable findings are: 1) the 20% drop in total daily insulin needs was explained primarily by reductions in basal insulin (not bolus as might be expected); and 2) the amount of carbohydrate intake in subjects increased dramatically, shooting up from 177 grams per day at baseline to 229 grams per day at week eight (we would guess this has to do with patients’ feeling more freedom to eat less strictly). CGM parameters were also measured but are still undergoing analysis and will be presented at a later date. The data on SGLT-2s continues to look encouraging in type 1 diabetes, and we hope companies aggressively move this class into trials for type 1 diabetes.
Questions and Answers
Q: With regards to the increased intake of carbohydrates, was this suggested by doctors or done spontaneously by patients?
A: There is no question that this was a spontaneous compensation to this therapy. It was a substantial increase. Unfortunately, we didn't expect this and don't have a way to reveal the mechanism. Of course, there substantial glycemic mass is lost through the mechanism of action of this drug.
Q: Could the altered renal dynamic change the excretion of ketones?
A: I do not know, but further studies are going to have to look at blood and urinary ketones, although I do not see a mechanism for their loss.
Gisle Langslet, MD (Oslo University Hospital, Oslo, Norway)
Dr. Gisle Langslet presented the results of a 104-week study of the efficacy and safety of J&J’s SGLT-2 inhibitor Invokana (canagliflozin). According to the abstract, these results represented the longest follow-up of canagliflozin treatment presented to date. People (n=1,450) on stable background metformin were randomized them to one of three therapy groups: canagliflozin 100 mg, canagliflozin 300 mg, or the sulfonylurea glimepiride. Both canagliflozin doses led to a significantly greater reduction in A1c (from an average of 7.8%) than glimepiride (-0.74% with canagliflozin 300 mg, -0.65% with canagliflozin 100 mg, -0.55% with glimepiride, p=0.04). Furthermore, the glycemic improvement seen in both canagliflozin groups was significantly more durable over 104 weeks than that seen with glimepiride Regarding safety, the incidence of adverse events (AEs) was similar across all groups, with more serious AEs seen with glimepiride. Most serious AEs were severe hypoglycemic events, which were more common with glimepiride (41% incidence with glimepiride compared to 7% and 8% in the canagliflozin 100 mg and 300 mg doses, respectively). Genitourinary infections were seen more frequently with both doses of canagliflozin (15% of female canagliflozin patients experienced them), but these effects were generally mild and few led to discontinuation of treatment. Although previous studies have demonstrated the safety and efficacy of canagliflozin, this study was valuable in that it demonstrated the durability of earlier results over nearly two years.
The double-blind, phase 3 trial (n=1,450 people on stable background metformin) randomized people to one of three therapy groups: canagliflozin 100 mg, canagliflozin 300 mg, or the sulfonylurea glimepiride (dose titrated as needed up to 8 mg; mean dose 5.6 mg).
At baseline, patients had an average age of 56 years, average A1c of 7.8%, average BMI of 31, and average disease duration of 6.6 years.
Fasting plasma glucose, body weight, and systolic blood pressure were also significantly decreased in both canagliflozin cohorts compared to the glimepiride group. Both canagliflozin doses led to a 10% increase in HDL cholesterol, and the 300 mg canagliflozin group saw a 14% increase in LDL cholesterol (11% increase for 100 mg dose) — these changes occurred during the first 26 weeks, and were stable from then until the end of the trial.
Questions and Answers
Q: It seems that the difference in fasting plasma glucose was greater than the difference in A1c — is that correct?
A: Yes, at the end of the 104-week period, that is what we saw.
Q: It has been suggested that the changes in cholesterol may be a measure of hemoconcentration. Do you have any data on this from earlier in the therapy?
A: I don’t think we have lipid measurements from early in the study. These changes in cholesterol are seen with most SGLT-2 inhibitors.
Comment: Regarding the LDL rise, which was seen only in the first 26 weeks, the question is whether it is a direct or indirect effect. The fact that this increase is not seen in patients with renal impairment, for whom the drug is less effective, tells me that it is a direct effect.
Q: What is nice about this protocol is that glimepiride could be titrated; this is more clinically realistic. My question is, did you see that the glimepiride dose was increased during the study?
A: During our study, glimepiride could be up and down titrated, and as I said, the mean maximum dose of glimepiride was 5.6 mg.
Dapagliflozin-Induced Weight Loss Impacts 24 Week HbA1c And Blood Pressure Levels
David Sjöström, MD, PhD (Global Brand Physician, AstraZeneca, Molndal, Sweden)
Dr. David Sjöström presented an analysis to determine how much of the A1c and blood pressure improvements dapagliflozin was associated with in a 24-week study were due to the concomitant weight loss. He framed this research question by explaining that it is not clear whether relatively small amounts of weight loss (<5%) lead to significant improvements in CV risk factors like A1c and blood pressure. The baseline characteristics of the study cohort mirror the populations of the six phase 3 studies from which data for this analysis were pooled. The average baseline BMI was ~32 kg/m2 and the mean A1c was 8.3%. In total, 1,066 patients were randomized to receive either 10 mg dapagliflozin or placebo. Treatment with dapagliflozin resulted in significantly greater reductions in A1c, weight, and blood pressure than placebo. Estimating the contribution of dapagliflozin-induced weight loss to the improvements observed in A1c, SBP, and DBP found a significant association between change in body weight and the three dependent variables. Each kilogram of body weight lost was associated with a 0.028 percentage point reduction in A1c, a 0.606 mmHg reduction in SBP, and a 0.253 mmHg drop in DBP. The researchers found that, in a dapagliflozin treated patient, 3 kg of weight loss drove an 8% relative decline in A1c from baseline, a 37% decline in SBP, and a 32% reduction in DBP. Purifying the effect of weight loss independent of dapagliflozin treatment demonstrated that the modest reductions in body weight achievable through SGLT-2 inhibition produce notable improvements in CV risk factors.
Questions and Answers
Q: Why don’t patients disappear if they are continuously excreting calories?
A: The large majority of patients do lose weight on dapa. However, the response is variable. You have a number of different regulatory layers, such as appetite, which would be more individualized, giving us a more varied response. However, overall there does consistently seem to be weight loss of some magnitude. The reason people do not disappear when they keep excreting calories – say you lose approximately 240 calories/day via urine, and if you have a typical intake of 2,400 calories, you are losing 10% of your calories. If you retain the same intake over time, you would lose the part of your body weight that needed the 10% of calories you excreted to be supported. If you put that into thermodynamic equations, it would take one-to-two years to equalize. However, what we have noticed is that patients have a propensity to increase their food intake. There is no threat of disappearing from use of SGLT-2 inhibitors.
Q: Don’t you think that it is difficult to fully disentangle weight loss and changes in blood pressure? How do you differentiate between glucose loss and sodium loss?
A: In our studies, we saw a transient increase in sodium excretion, but over time there was no change in sodium levels. I agree there are some aspects this presentation could not fully answer.
Q: What do you think about SGLT-2 in type 1 diabetes patients?
A: I think that the other speaker in this session will have a nice presentation on this topic.
Nobuya Inagaki, MD, PhD (Kyoto Graduate School of Medicine, Kyoto, Japan)
Luseogliflozin is an SGLT-2 inhibitor being developed in Japan by Taisho Pharmaceuticals. It is reported to be more specific for SGLT-2 (over SGLT-1) compared to dapagliflozin or canagliflozin. This phase 3 study assessed the performance of luseogliflozin 2.5 mg versus placebo on top of stable glimepiride (GLIM) in 221 patients over 24 weeks and then with an open label extension for all patients to 52 weeks (during which dose could be increased to 5 mg). A1c reduction was 0.5% from the baseline (8.1%) and 0.9% with respect to placebo at 24 weeks, and was maintained to a 0.6% reduction out to 52 weeks. Weight reduction for luseogliflozin was 2.2 kg from baseline at 52 weeks, and 1.5 kg at 24 weeks relative to the placebo group. As expected, there was an excess of genitourinary adverse events in the luseogliflozin group (2.7%) but no resultant discontinuations, and there was also double the hypoglycemia in the luseogliflozin group, which Dr. Inagaki implied was because glimepiride doses were not decreased when luseogliflozin was added.
Luseogliflozin is a phase 3 SGLT-2 inhibitor being developed by Taisho Pharmaceutical that is highly selective for SGLT-2 (versus SGLT-1). This study assessed the efficacy and safety of luseogliflozin added to glimepiride (GLIM) over 52 weeks, since sulfonylureas represent the majority treatment in Japan.
The study was a double blind randomized controlled trial of luseogliflozin 2.5 mg versus placebo on top of GLIM for 24 weeks in 221 patients. There was an open label extension to 52 weeks, in which all patients took luseogliflozin, but those with insufficient control could increase the dose to 5 mg. Baseline A1c was 8.1%, participants were 75% male, average BMI was 25 kg/m2 and average diabetes duration was about 7 years.
At week 24, the luseogliflozin arm exhibited an A1c reduction of 0.5% from the baseline (8.1%) and 0.9% with respect to the placebo arm. This was maintained out to 52 weeks with a 0.6% reduction from baseline. Weight reduction for luseogliflozin was 2.2 kg from baseline at 52 weeks, and 1.5 kg at 24 weeks relative to the placebo group. Fasting glucose in the luseogliflozin group was 34 mg/dl lower than the placebo arm at 24 weeks. Other parameters showed slight improvement, including insulin resistance, blood pressure and lipids.
As might have been expected, there was an excess of genitourinary adverse events in the luseogliflozin group (2.7%) but no resultant withdrawals. There was also 8.7% hypoglycemia in the luseogliflozin group, compared to 4.2% in the placebo arm. It seems that glimepiride doses were not decreased when luseogliflozin was added, leading to the excess hypoglycemia.
Questions and Answers
Q: Are there any differences with luseogliflozin compared to dapagliflozin or canagliflozin?
A: This drug has a very high affinity for SGLT-2 and so the dose is very small compared to other drugs. This drug is metabolized in the liver via several pathways, which may lead to differences to the other drugs.
Yukio Tanizawa, MD, PhD (Yamaguchi University, Yamaguchi Prefecture, Japan)
Dr. Yukio Tanizawa presented data on the use of tofogliflozin (Chugai’s SGLT-2 inhibitor) as a monotherapy and in combination with other oral therapies for 52 weeks. Dr. Tanizawa concluded that tofogliflozin positively reduced A1c (~0.7-0.9%) and body weight in both mono- and combination therapy. In monotherapy, participants were randomized to either 20 mg (n=63) or 40 mg (n=127) of tofogliflozin, and in the combination therapy, participants were randomized to use of either sulfonylureas (n=171), glinide (n=22), “BG” (n=105; this was not defined), TZD (n=102), a-glucosidase inhibitor (n=99), and DPP-4 inhibitor (n=103), with either a 20 mg (n=178) or 40 mg (n=424) dose of tofogliflozin. In the monotherapy group, A1c levels dropped 0.7% (baseline of 7.8%) and body weight dropped 3.0 and 3.4 kg for 20 mg and 40 mg of tofogliflozin, respectively. A1c dropped an average of 0.8% for all combination therapies for 20 mg tofogliflozin, and dropped 0.9% for 40 mg of tofogliflozin (baseline range from 7.7% to 8.4%). Body weight dropped an average of 2.5 kg for all combination therapies with 20 mg tofogliflozin, and 3 kg with 40 mg of tofogliflozin. Interestingly, Dr. Tanizawa remarked that tofogliflozin in addition to “BG” and an a-glucosidase inhibitor increased weight loss to 3.9 kg on average, but only when used with 40 mg of tofogliflozin. Dr. Tanizawa highlighted that the use of tofogliflozin as a monotherapy may significantly reduce waist circumference, adiponectin, blood pressure, and HDL and LDL, relative to baseline.
Both the monotherapy and combination studies were 52-week randomized, open-label, parallel-group comparison. There were 194 participants in the monotherapy trial and 602 in the combination therapy trial.
The baseline A1c for patients on monotherapy was 7.8%, and patients had an average BMI of 25-26 kg/m2. Additionally, the average age was 58-59 years in the tofogliflozin groups. For patients on combination therapy, the baseline A1c ranged from 7.7-8.4% and BMI ranged from 24-28 kg/m2. The average age of patients ranged from 56 years to 60 years for both doses.
Patients in monotherapy experienced a significant decrease in weight circumference (p <0.001), adiponectin (p <0.001), blood pressure (SBP: p <0.01 and p <0.05, 20 mg and 40 mg, respectively), and HDL (p <0.01 and p <0.001, 20 mg and 40 mg, respectively) for both the 20 mg and 40 mg groups. Additionally, patients taking 40 mg of tofogliflozin had a significant reduction in triglyceride levels (p <0.01). Surprisingly, Dr. Tanizawa noted, patients on 20 mg, but not 40 mg, of tofogliflozin experienced significant reductions in both DBP (p <0.01) and LDL (p<0.05) levels.
There were low levels of cystitis, urinary tract infection, genital infection, and hypoglycemia reported across all groups. There was a particularly low rate of urinary and genital infections – 1-3% in those on tofogliflozin – prompting someone in Q&A to suggest cultural biases. There was no significant difference between groups, and no more than 5% of patients in any group experienced any one adverse events (except for 6% of participants taking a 20 mg dose experiencing mild hypoglycemia). Dr. Tanizawa noted that the three moderate hypoglycemic events all occurred when tofogliflozin was used in conjunction with sulfonylureas.
Nasopharyngitis and upper respiratory tract infection were the most common adverse event. Dr. Tanizawa noted that 70% of patients experienced an adverse event, however, these were likely not due to the medications. There were other changes in safety parameters in the monotherapy group, such as an increase in uric acid levels.
Questions and Answers
Q: This was an interesting presentation, thank you. I am wondering why you are doing this open label? Is it ok for regulations to do this without a placebo control? I am also surprised that there were close to zero genital infections. Could it be that something else is going on? Maybe it is just cultural for women not to complain?
A: We had done a separate, placebo, double-blinded trial.
Q: Do you think it is conceivable that there are cultural differences that cause women not to report their genital infections?
A: I don’t know if this is cultural difference. In fact, genital infections are very rare in Japanese populations.
Comment: If that is the case, then maybe we will start to have country-specific drugs.
Q: Are you examining triple therapy?
A: No, not at this moment.
Pablo Lapuerta, MD (Chief Medical Officer, Lexicon Pharmaceuticals, The Woodlands, TX)
LX4211 is Lexicon Pharmaceutical’s dual SGLT-1/SGLT-2 inhibitor. SGLT-2 is the major glucose transporter in the kidney, and SGLT-1 primarily influences glucose and galactose absorption in the gut, although it has a lesser role in the kidney. In this phase 2b 12-week study of 299 patients on top of metformin, LX4211 didn’t show an excess of GI side effects (nausea, diarrhea) or urinary tract infections over placebo. Vaginal infections were only seen with LX4211, but their incidence was reasonably low and no patients discontinued as a result.
LX4211 is an interesting dual SGLT-1/SGLT-2 inhibitor under development by Lexicon Pharmaceuticals and is currently undergoing phase 2b studies. SGLT-1 is the primary glucose and galactose transporter in the gastrointestinal tract, although it has a minor role in the kidney, where SGLT-2 is the primary transporter. SGLT-2 inhibition has been associated with an increase in genitourinary infections, and it is thought that the incidence may relate to the degree of glycosuria. Lexicon believes that the SGLT-1 inhibition of LX4211 might lead to lower glycosuria and therefore a better safety profile. On the other hand, SGLT-1 inhibition is often linked with GI side effects such as diarrhea, although studies of people with natural SGLT-1 mutations do not support this.
Accordingly, this study seeks to investigate the safety profile of LX4211 added to metformin, over a 12-week period. At baseline, 299 subjects inadequately controlled on metformin were randomized to four doses (from 75 mg to 400 mg) and placebo. BMI was 33 kg/m2 and baseline A1c was 8.0%.
After 12 weeks, the highest dose of LX4211 (400 mg) showed a 0.9% A1c reduction from baseline, and the lowest dose (75 mg) reached 0.5% reduction. There was a nice dose response effect. However, higher doses of LX4211 didn’t increase glucose excretion in the urine above that of the 200 mg dose
Genital tract infections were observed in 7%-10% of women at the three higher doses, with none occurring in men or in the placebo group. There wasn’t a strong trend with dose, and there were no discontinuations as a result. Urinary tract infections were the same as placebo. Diarrhea ranged from 4%-10%, compared to 7% in the placebo group. Nausea, vomiting and constipation were also comparable with placebo. LX4211 also exhibited no changes in hypoglycemia, orthostatic blood pressure, LDL or urine electrolytes.
Dr. Lapuerta concluded that the favorable GI profile of LX4211 suggests that there is a therapeutic window for SGLT-1 inhibition, and that the dual inhibition approach is promising. Longer duration studies will be forthcoming.
Questions and Answers
Q: Is there an increase in GLP-1?
A: We didn’t measure it in this study, but we have documented sustained GLP-1 elevation for several hours in other studies.
Q: What was the impact on body weight?
A: There was an approximately 2% reduction in body weight. So we see similar weight loss to other members of the class despite less glycosuria.
Q: What was the change in blood pressure?
A: There was a 6mm Hg decline in systolic blood pressure.
Q: Isn’t SGLT-2 blocking better, because of SGLT-1’s effect on the kidney?
A: Looking at the pharmacokinetic profile, we don’t think systemic SGLT-1 inhibition is likely. If it was, there would be more glycosuria. We believe that the inhibition is in the GI tract.
Q: How much was water weight loss? Did you check serum electrolytes?
A: There was no change in serum sodium and potassium. We didn’t check weight loss via DEXA or waist circumference. But the weight change was gradual and did not plateau. A diuretic effect would have been more dramatic, and we didn’t observe this.
--by Adam Brown, John Close, Poonam Daryani, Hannah Deming, Jessica Dong, Hannah Martin, Manu Venkat, Vincent Wu, and Kelly Close