AACE/ACE Scientific and Clinical Review – Association of SGLT-2 Inhibitors and DKA

October 24-25; Dallas, TX; Full Report – Draft

Executive Highlights

Our team is back on the West Coast bringing you insights from this weekend’s AACE/ACE meeting on SGLT-2 inhibitors and DKA, which drew 50 attendees, many from industry, to Dallas, TX. The packed 1.5-day meeting was full of discussion on the mechanistic basis of diabetic ketoacidosis (DKA), detailed reviews of specific cases, epidemiologic information on the prevalence of SGLT-2 inhibitor associated DKA, and plenty of healthy debate over the level of risk and how best to manage it. This topic promises to remain at the forefront of discussion, and we applaud AACE/ACE for proactively addressing the issue. A writing group met immediately after the close of the meeting to draft a set of recommendations based on the weekend’s discussion, which they plan to summarize briefly in a press release and outline in more detail in a published paper. Below are our top themes from the meeting, followed by full detailed coverage.

Table of Contents 


  • There seemed to be consensus that SGLT-2 inhibitor use creates a ketogenic milieu that predisposes patients to DKA in the presence of a triggering event. The increased risk was notably characterized as a consequence of the class’s central mechanism of action rather than a random off-target effect.
  • Lowering insulin dose – either intentionally or accidentally – was identified as a major precipitating factor for DKA associated with SGLT-2 inhibitors. Experts in the room generally agreed that insulin dose should either not be lowered or adjusted very carefully with these drugs; presumably this would be the latter in patients who aren’t too far from their glycemic target given hypoglycemia risk.
  • While speakers generally agreed that incidence of SGLT-2 inhibitor-associated DKA in patients with type 2 diabetes is relatively low, there was some debate on how to best define the risk. For one, several speakers noted that the definition of type 2 diabetes itself is not always clear, as some patients who developed DKA in “type 2 diabetes” trials were likely misdiagnosed. Others argued that DKA incidence in clinical trials may have been artificially low since patients were so carefully monitored and those with more severe disease excluded.   
  • There was some debate over whether or not the guidelines from this meeting should address SGLT-2 inhibitor use in patients with type 1 diabetes. On one hand, several speakers and conference organizers noted that use in type 1 diabetes is off-label and thus cannot be included in official AACE/ACE guidelines for liability reasons. On the other hand, other attendees argued that off-label use is occurring regardless and providing guidelines will make the practice safer for both patients and prescribers. We thought the latter was important for HCPs to hear given that many are in the midst of off-label use for a range of compounds.
  • The general consensus was greater education on the topic was needed for all. There were several calls to action to disseminate the recommendations from this meeting as widely as possible once they are drafted. In addition, several speakers emphasized the need to educate providers correctly on the risk of DKA and how to diagnose it, especially in the possible absence of high blood glucose levels. The need to educate patients to recognize symptoms of DKA and seek medical help was also echoed several times throughout the conference.
  • Weighing the risk of DKA against the benefits of the SGLT-2 inhibitor class came up as a consideration. While speakers were asked to stay focused on discussing the DKA risk of the class, some speakers noted that the cardioprotection demonstrated with Lilly/BI’s Jardiance (empagliflozin) – a possible class benefit – could outweigh the risk. In our view, the benefits in type 2 diabetes clearly outweigh the risks. In type 1 diabetes, the situation remains more uncertain given that results are not yet available about the benefits.
  • Several speakers noted that diagnosis of elevated ketone bodies and DKA can be murky. A common test for ketone bodies involves testing the urine, but several speakers pointed out that urine ketone levels may not correlate with blood ketone levels and more research is needed to develop more sensitive tests and more definitive thresholds for dangerous ketosis.

Detailed Discussion and Commentary

Normal Physiology

Aaron Vinik, MD, PhD (Eastern Virginia Medical School, Norfolk, VA), Timothy Garvey, MD (University of Alabama, Birmingham, AL), John Miles, MD (University of Kansas, Kansas City, KS), Roger Unger, MD (University of Texas Southwestern, Dallas, TX)

This opening session on the physiology of energy metabolism likely prompted flashbacks to medical school biochemistry for some attendees. Dr. Aaron Vinik led off by focusing on the brain. He argued that the central dopamine deficiency present in diabetes “unbridles” the sympathetic nervous system and alters the normal balance between sympathetic and parasympathetic activity. This leads to a range of downstream effects: increased glucose and fatty acid production, insulin resistance, and ultimately impaired glucose metabolism and cardiovascular pathology. Dr. Vinik proposed timed release of a dopamine agonist as one intriguing potential solution to overcome those effects. Dr. Timothy Garvey brought the discussion back to somewhat more familiar ground, delivering an impressively comprehensive overview of the pathways of carbohydrate metabolism. He made the point that hyperglycemia in diabetes is a metabolic adaptation to background pathology. Therefore, putting a “spigot” in the glucose supply with an SGLT-2 inhibitor could artificially disrupt that adaptation and potentially induce starvation responses. Dr. John Miles focused more specifically on the physiology of lipolysis and ketone body production. Most relevant to SGLT-2 inhibitors, he cited data demonstrating that glucagon stimulates ketone body production under conditions of insulin deficiency despite not having any effect on lipolysis. Finally, the legendary Dr. Roger Unger offered his signature critique of the insulin-centric view of current diabetes treatment. He presented a substantial body of preclinical evidence suggesting that blocking glucagon action leads to much more stable glycemic profiles than giving exogenous insulin. While much more study would be needed in humans, Dr. Unger closed with the provocative prediction that “the era of insulin centrism may be ending as it approaches the one-century mark.”

Panel Discussion

Dr. Sam Dagogo-Jack: Lately there has been a lot of publicity surrounding the bionic pancreas, in which our colleagues have infused insulin and glucagon into a pump in humans to avoid hypoglycemia. Based on a glucagon-centric argument, would that approach be misguided?

Dr. Unger: They oppose each other. The more insulin you give, the more glucagon will be suppressed, so there’s no protection against hypoglycemia. Glucagon evolved to maintain fuel delivery to the brain during starvation and intense exercise, so creatures on the earth could get through famine without dying. It did not evolve to protect insulin against hypoglycemia – that came only in 1922.

Dr. Ralph DeFronzo (University of Texas Health Science Center, San Antonio, TX): The studies that Ele and I did are consistent with what Bob and Tim showed. When you’re type 2 diabetic and insulin resistant, the cell has a basic need for energy. Sugar rises to a level that by mass action gets sufficient glucose into the cell to meet energy requirements. If you intervene with a drug like an SGLT-2 inhibitor to acutely drop blood glucose, there’s not a way for the cell to adapt quickly, so it switches from glucose to fat metabolism. John has pointed out that once you switch to fat, you don’t need a lot of other changes to drive ketogenesis. That might be a reasonable point to comment on in this scenario because it will come up and play out in future talks.

Dr. John Miles (University of Kansas, Kansas City, KS): Any time you give a drug that exerts its effect through a non-insulin mediated mechanism, you’ll lower glucose concentrations. I’d like to hear discussion about how one class might differ from another. Assuming the mechanism is non-insulin mediated, why would one class be more likely to shift the body to lipid fuel economy than another?

Dr. DeFronzo: For GLP-1 agonists it’s clear. They will kick out insulin and knock down glucagon, so you have a counter mechanism that blocks lipolysis. With SGLT-2 inhibitors there’s some improvement in insulin secretion but there’s such an acute drop, you have to switch to fat.

Dr. Ele Ferrannini: I want to go back to the argument Sam raised, about the bihormonal infusion. The purpose of that is to prevent hypoglycemia and it’s entirely based on the kinetics of insulin action being so much slower than the kinetics of glucagon. If you want to switch off the insulin pump or reduce the infusion, by the kinetics of insulin, it is going to be there for much longer - that’s when you need glucagon to prevent hypoglycemia. I don’t think that contradicts the argument Dr. Unger made. One aspect of glucagon physiology and pathophysiology I would like to ask Dr. Miles and Dr. Unger about is the effect on proteolysis. It’s quite clear that if you use a glucagon receptor agonist in people, you get major weight loss. Its long been known that glucagon is proteolytic. Glucagonemia in type 2 diabetes may be linked with protein degradation. Is that anything we should be thinking about?

Dr. Unger: We think because we eat carbs, the rest of world does as well. We’re one of the few species that eat carbs; other species are carnivorous. Most species on the planet eat grass and each other, so it’s difficult to understand why the glucose system works.

Dr. Miles: Separate from proteolysis, anytime you get weight loss you have to invoke a change in energy balance. We don’t like to think about that because it’s hard to measure. We know that small changes of 50-100 calories per day in intake or expenditure can have a profound impact over time. That’s an additional issue.


SGLT-2 Inhibitors and Ketoacidosis: Learnings from the Scientific Literature

Simeon Taylor, MD (University of Maryland School of Medicine, Baltimore, MD)

Dr. Simeon Taylor delved into the scientific literature in an attempt to illuminate the existing evidence for a relationship between SGLT-2 inhibitors and ketoacidosis and suggest possible mechanisms at play. 

  • Dr. Taylor’s literature review included evidence from several animal model and human studies of SGLT-2 inhibitor molecules. He began by looking at phlorizin, the first SGLT-2 inhibitor to be characterized. Going all the way back to 1956, Dr. Taylor reviewed several studies that suggested phlorizin was associated with an increase in ketone bodies, particularly acetoacetate and betahydroxybutyrate. Phlorizin also decreased glucose reabsorption and increased acetoacetate reabsorption in the renal tubules in dogs. He suggested that these early studies raise the notion that the co-transporters involved in both glucose and acetoacetate reabsorption may be competing for access to sodium in order to reabsorb their respective targets.
  • In terms of marketed SGLT-2 inhibitors, Dr. Taylor showed data demonstrating that dapagliflozin and ipragliflozin (available in Japan only) increased ketone bodies in rats, particularly in fasted rats. In addition, he pointed to human data that shows a clear dose-dependent increase in ketone bodies in patients with type 2 diabetes on tofogliflozin (also approved only in Japan).
  • Dr. Taylor pointed to ketonuria tests as an area for further research. He noted that the tofogliflozin study found that 13.8% of study participants on the highest dose of the drug experienced hyperkentonemia (excess ketones in the blood) but none of the participants had ketonuria (excess ketones in the urine). This raises the question of the relationship between blood ketones and urine ketones and the usefulness of urine ketone tests in the detection of SGLT-2 inhibitor-associated DKA.
  • Dr. Taylor hypothesized that SGLT-2 inhibitors increase glucagon secretion through a direct effect on the alpha cell. There is an increase in renal clearance of ketones, but Dr. Taylor is unsure that this necessarily causes acidosis in and of itself. He noted that decreasing the insulin dose to protect against hypoglycemia will probably also promote ketogenesis and acknowledged the challenge of finding just the right balance to avoid both hypoglycemia and ketosis.

Questions and Answers

Q: I wanted to know whether it’s a fact that you can have ketonemia without ketonuria?

A: It’s not that you don’t have ketonuria. It’s related to the sensitivity of the test. If the data is correct and blood ketone levels go up and it doesn’t in urine, then that suggest san effect on clearance.

Dr. John Miles: I want to comment on ketosis without ketoacidosis and the question of how we diagnose DKA. If you go to the current consensus statement on the subject (Diabetes Care, 2009), you find that the current criteria are soft. It says the following are usually present: (i) blood glucose of 250 mg/dl or above, but in our study of over 400 DKA cases, over 20% had glucoses <250 mg/dl, and (ii) ketonuria – this may be generally true, but the ADA also has a position paper that says you shouldn’t use ketonuria to diagnose DKA because the test is qualitative. We found that serum bicarbonate will underdiagnose DKA in 15-20% of cases and overdiagnose DKA in another 15-20% because of its non-specificity. We’ve found that betahydroxybutyrate would probably work to diagnose. It’s time for a conversation about requiring more rigor in the diagnosis of DKA.

A: It’s possible that the criterion for betahydroxybutyrate levels would be different in the face of SGLT-2 inhibitors. They could raise the threshold that you need to get a diagnosis even higher. That’s an area for research.

Ketogenesis, Ketosis, and Ketoacidosis

Richard Pratley, MD (Florida Diabetes and Endocrine Center, Orlando, FL)

The great Dr. Richard Pratley provided a masterclass on the major ketone-related processes at play in ketone body metabolism: ketogenesis, ketosis, and ketoacidosis. He began with the point that ketosis doesn’t necessarily lead to ketoacidosis: ketoacidosis only develops when ketone body production is much greater than clearance. After reviewing the sites of ketone body metabolism in the body, Dr. Pratley provided an overview of different scenarios in which ketosis can occur and can even be a positive response. Typically, ketone body production is relatively low during normal feeding with normal physiology. However, a carbohydrate-deficient state can cause the liver to ramp up ketone body production to meet the body’s needs. Ketogenic diets – such as the Atkins diet or other low-carb diets – are promoted for weight loss, management of type 1 and type 2 diabetes, and the treatment of a variety of other surprising medical conditions (everything ranging from epilepsy to cancer to acne). Dr. Pratley pointed out that he often sees patients who may be on ketogenic diets for a variety of reasons, perhaps unrelated to their diabetes (the popular media certainly tells us we’ll lose weight if we stop eating carbs). On top of dietary ketosis, a range of factors can contribute to increased ketone body formation that greatly exceeds clearance, resulting in ketoacidosis. These include classic diabetic ketoacidosis, other endocrine disorders, heart failure, surgery, burns, and multi-organ failure. Starvation states can also contribute. Ketoacidosis can also arise from exposure to toxic substances, including alcohol, and genetic disorders that affect ketone metabolism. Dr. Pratley suggested that it’s important to determine how drugs fit into the ketosis to ketoacidosis process.

DKA: What Is It? Who Gets It? How Does It Happen?

Mary Korytkowski, MD (University of Pittsburgh School of Medicine, PA)

Dr. Mary Korytkowski sought to offer a clearer picture of which patients experience DKA events (independent of SGLT-2 inhibitor use) and why it occurs. Most notably, Dr. Korytkowski tied DKA risk to financial concerns and, in particular, the rising cost of insulin. She pointed to a recent JAMA article that looked at trends in DKA incidence at time of diagnosis in children with type 1 diabetes in Colorado from 1998-2012. Overall DKA prevalence at the time of diagnosis increased 55% in the time period, which correlated with a period of increasing childhood poverty in the state. In addition, the study found that children on government-sponsored insurance (typically an indicator of the family’s income level) were more likely to experience DKA at diagnosis than children with private insurance. However, DKA prevalence among children with private insurance also increased during that time period, which the study authors and Dr. Korytkowski suggested may be related to the increasing proliferation of high-deductible commercial plans that have high out-of-pocket costs. Dr. Korytkowski argued that uninsurance or underinsurance are barriers to families trying to keep up with rising insulin costs and shared anecdotal stories of patients who were trying to make their insulin last as long as possible, resulting in potentially lowered or missed doses and increased risk of DKA. Public and patient outcry over rising insulin costs has been increasing in recent months and we’re glad that Dr. Korytkowski drew attention to this socioeconomic risk factor as a potential contributing cause of DKA. It’s important to keep in mind that the insulin dose reduction that often precipitates SGLT-2 inhibitor-associated DKA could be due to factors unrelated to medically-sanctioned dose adjustments or pump malfunctions.

  • In terms of adults, Dr. Korytkowski cited data from the T1D Exchange indicating similar DKA prevalence in type 1 diabetes across all adult age groups but a higher risk among patients with higher A1cs. Based on a national hospital discharge survey, less than 20% of total discharges for DKA were in children, over 50% were in adults with type 1 diabetes, and about 30% were in adults with type 2 diabetes.
  • In addition to financial concerns, Dr. Korytkowski listed a host of other reasons that patients may miss an insulin dose. These included the notion that insulin should be withheld when illness interferes with eating, weight control attempts, fear of hypoglycemia, emotional and/or diabetes-related stress, rebellion (especially in the teen years), accidental omission of an insulin dose, and pump malfunction.
  • Dr. Korytkowski also helpfully defined “classic” DKA and “euglycemic” DKA in clinical terms. “Classic” DKA involves hyperglycemia with a blood glucose level >250 mg/dl, positive serum or urine ketones, an arterial pH <7.3, and low serum bicarbonate. “Euglycemic” DKA, on the other hand, typically occurs without marked hyperglycemia, with blood sugars <300 mg/dl at presentation.

The Role of the Kidneys in the Clearance or Reabsorption of Ketone Bodies

Matthew Weir, MD (University of Maryland School of Medicine, Balitmore, MD)

Nephrologist Dr. Matthew Weir offered a deep dive into the role the kidneys play in the clearance and reabsorption of ketone bodies. After a brief review of ketone bodies, ketone body metabolism, and ketogenesis, Dr. Weir used a detailed mechanistic understanding of renal ketone absorption and clearance to clarify the processes at play in starvation ketoacidosis and diabetic ketoacidosis. Relating the mechanisms to the topic at hand, Dr. Weir hypothesized how effects of SGLT-2 inhibitors might contribute to increased risk of ketoacidosis. SGLT-2 inhibitors inhibit sodium reabsorption, which in turn increases the electrochemical gradient that drives reabsorption of ketoacids in the kidneys, increasing ketonemia. Dr. Weir also pointed to the effects of volume contraction and reduced glomerular filtration rate (GFR). As a diuretic, SGLT-2 inhibitors decrease arterial blood volume, which in turn reduces filtration and ketoacid clearance. In addition, Dr. Weir referenced animal and human studies that suggest SGLT-2 inhibitors could promote increased glucagon and decreased insulin by interfering with the SGLT-2 cotransporters on pancreatic alpha cells. Dr. Weir concluded that more research is needed in people with type 1 and type 2 diabetes to understand if the mechanisms at play in starvation ketoacidosis are relevant in the case of SGLT-2 inhibitors. In addition, he flagged a possible concern of increased DKA risk in patients with reduced eGFR, such as patients with diabetic nephropathy.

Panel Discussion

Dr. Timothy Garvey (University of Alabama, Birmingham, AL): When I lived in South Carolina, we saw lots of African-American obese children present to the hospital with extreme DKA and no antibodies. The severity is equal to that of garden-variety type 1 diabetes. Within weeks to months they convert to type 2 diabetes and we put them on metformin and low-carb diets. These patients also had a tendency to develop heavy ketosis or DKA more easily than other type 2s. I think, and this has been reported in Hispanics and Native Americans as well, that when we go about gathering more data on ketosis-prone diabetes and the development of DKA in type 2 diabetes with or without SGLT-2 inhibitors, we need to point out what the racial/ethnic identity is. That may not be enough because self-identified race is not a meaningful biological construct. It’s ancestral genetics that are important. I don’t know how much is due to the African ancestral contribution, or Spanish or Amerindian.

Dr. Korytkowski: I think those comments are excellent. The whole idea of ketosis-prone diabetes started with us thinking the obese African-American population with lower socioeconomic status, which was particularly at risk. But ketosis-prone diabetes is found in every group you look at. It’s not limited to one racial or ethnic group, but we’re not really getting good markers for who is at risk though. The VA study listed three risk factors for ketosis-prone diabetes, but those factors are really risk markers for type 2 diabetes as well.

Dr. Richard Pratley (Florida Diabetes and Endocrine Center, Orlando, FL): My colleague has identified genes in the lineage. All of this has been candidate gene work in small populations. Perhaps we’re not focusing on the right candidates. Maybe we should focus on intermediates of metabolism.

Dr. Alan Garber (Baylor College of Medicine, Houston, TX): Do you think we should consider placing more emphasis on the role of dehydration? As a candidate for the oldest potential reference, consider Elliot Joslin’s early work on reversal of DKA with fluid therapy in the pre-insulin era. We know very well it can cure ketoacidosis. Having been the first author of the six-week fasting investigation, I would note that the patients who fasted for six weeks in that study were quite happy and satisfied. They did extremely well on vitamins, fluids, and salt.

Q: We see now that there can be ketonuria and ketonemia, but not necessary ketoacids. With the 176 patients in the FDA database, do you know how many were diagnosed with DKA based on ketonuria?

Dr. Taylor: The short answer is we don’t know. The majority of cases are bona fide cases of ketoacidosis, but we don’t have exact numbers. It’s a good point.

Dr. Aaron Vinik (East Virginia Medical School, Norfolk, VA): I like the data in terms of people not affording insulin. We’re wrestling with that. What does the rising cost of insulin have on this? Is there data on how they’re related?

Dr. Korytkowski: The best data on DKA and insurance is from Colorado. In terms of the affordability of insulin, that’s anecdotal information. I’m not sure we’re seeing more DKA, but we’re seeing cases of people coming in and telling us that they’re trying to make their insulin last longer.

Dr. Ele Ferrannini (University of Pisa, Italy): One difference between starvation-induced ketosis and DKA is the time course. With starvation it happens quite slowly. You have to go out to six weeks to see anything. With DKA it’s much quicker. With SGLT-2 inhibitors it’s even quicker. The time factor may be important. Going back to the comment Dr. Garvey made about very obese African-Americans with classical type 2 diabetes after being weaned from insulin, in San Antonio this is the phenotype we see. I wonder whether the production of ketone bodies is also a product of the amount of fat on board? Obesity is protective against ketosis so long as you have enough insulin to overcome insulin resistance. When you become diabetic, insulin resistance is as much as it can be and now there’s a bit of insulin deficiency or a defect in beta cell function. If there’s any hormonal background to stimulate ketogenesis and it’s acting upon a big mass of fat, that’s releasing a lot of free fatty acids, which can then be metabolized to ketones. Thoughts?

Dr. Dagogo-Jack: It you look at the etiology of DKA, 40% was due to the involuntary cessation of insulin therapy, driven by using last drop of insulin at home and having no further access to insulin. The rise of DKA could be related to the limitations on insulin access.


SGLT-2 Inhibitors and DKA: The Clinical Stories

George Grunberger, MD (Wayne State University, Bloomfield Hills, MI)

Dr. George Grunberger transitioned the conversation from the basic science-heavy morning sessions to the big picture question of what the guidelines resulting from this meeting should look like. While his presentation included some rough epidemiological estimates of DKA risk, he highlighted several important unanswered questions for the committee to address. First, he noted that the discussion surrounding cases of DKA and overall risk don’t provide a denominator (how many cases of DKA occurred in how many total patients on SGLT-2 inhibitors?) To help roughly elucidate this, Dr. Grunberger pointed to the figure that more than 7 million prescriptions have been written for SGLT-2 inhibitors in the US alone through June 2015. Dr. Grunberger underscored that the denominator is important because it is crucial to present absolute risk to patients and providers. We completely agree and felt at times during the meeting that it was easy to lose sight of this point as speakers delved deep into the weeds of the risk. Dr. Grunberger also pointed out the increasingly blurred line between different types of diabetes and how that can contribute to confusion in the epidemiological data on how often cases occur in patients with type 1 diabetes or type 2 diabetes. Dr. Grunberger also questioned what we know about the mechanism through which the DKA occurs, noting that if it’s through an off-target mechanism that we don’t understand well, then DKA risk is not easily predictable or preventable. Finally, Dr. Grunberger implored the committee to consider how to best spread the recommendations from this meeting quickly and broadly. He emphasized that patients, providers, and the public are waiting for AACE to say something on the topic and provide direction for clinical practice. Again, we completely agree, and we think it is particularly crucial that this message reach non-endocrinologists who may not recognize SGLT-2 inhibitor-induced DKA when they see it.

Questions and Answers

Q: There was the statement that there were only four patients in the canagliflozin experience who had type 2 diabetes. I want to address that in two different ways. In terms of understanding the disease mechanism, is that a fair way to split it out? From a practical view, if you go by the label you’d need to check antibodies before prescribing. But in real world, doctors aren’t always going to check islet antibodies.

A: You reacted exactly the way I hoped you would, because I was trying to provoke. I was looking at it like someone who reads the label – it says in patients with type 2 diabetes. But we don’t even know who has type 2 diabetes. So can we tease out some details that would be clinically useful?

Dr. Dandona (University of Buffalo, NY): The people excluded from these research studies may be treated with these drugs and may be more likely to get it. People are not doing this testing, but if you put it in the label to do so, it’d be very onerous.

Dr. Ele Ferrannini (University of Pisa, Italy): Advanced type 2 diabetes looks a lot like type 1 diabetes. For long-term complications, long-term type 1 diabetes looks like type 2 diabetes.

A: I’m just trying to stroke something here because you guys are going to give the answers. Whether these kinds of things make the label and now clinicians have to do testing – we need to resolve it here. How do we identify the patients most at risk?

Personal Cases

Samuel Dagogo-Jack, MD (University of Tennessee, Memphis, TN)

Dr. Samuel Dagogo-Jack presented five cases of SGLT-2 inhibitor-associated DKA that occurred in patients at his institution. All of the patients had type 2 diabetes, but they also all had marked insulin deficiency. Some of the patients also had high lactate levels, which Dr. Dagogo-Jack suggested could have contributed to the problem. He also noted that some of the cases were not euglycemic (one patient had a blood glucose of 1,113 mg/dl). As other speakers noted during the meeting, this could make it more difficult to determine whether the DKA was truly drug-induced. In all cases, the presentation of symptoms was fairly typical and the patients responded well to standard treatment.

Background Prevalence of DKA in T2DM in Denmark Prior to the SGLT-2 Inhibitors

John Nolan, MD (Steno Diabetes Center, Gentofte, Denmark)

Dr. John Nolan presented data on the background rates of DKA (prior to the introduction of SGLT-2 inhibitors) nationwide in Denmark and at the highly regarded Steno Diabetes Center. He first shared results from a comprehensive Danish national diabetes registry (n=306,624, ~100,000 with type 2 diabetes) showing that there were 621 DKA events in 506 people between 2005-2012, for an overall rate of 1.8 events per 1,000 patient-years. The rate was significantly higher in younger patients and those with a lower BMI, and there was a trend toward higher incidence with worse metabolic control. The most surprising finding for Dr. Nolan was the significantly lower incidence among patients on insulin compared to those not on insulin. Dr. Nolan then presented data from Steno cohorts of patients with type 1 and type 2 diabetes. In type 1 diabetes, his main point was that contrary to criticism the center had received, Steno patients on pumps did not have unusually high rates of DKA. These rates were also lower than the national average. We would also note that they were an order of magnitude higher than in type 2 diabetes, illustrating the clear difference in background risk between the two populations. In the Steno type 2 diabetes cohort (n=6,113), 60 patients got DKA from 1998-2012. The overall incidence was 1.9 events per 1,000 patient-years, with rates declining over time.

  • Steno is now evaluating data from a Danish nationwide cohort on SGLT-2 inhibitor use and DKA rates from 1995-2014. This sort of population data should help the field understand whether the risk in type 2 diabetes is as low in the real world as it is in clinical trials. We sure have to hand it to Denmark for their very comprehensive population registries; we hope the US can make strides in that direction going forward. There is some concern, raised during Q&A, about whether the database will have enough power to detect a subtle increase in risk (i.e., an increase in incidence from 1 in 3,000 to 1 in 1,000), but it should provide valuable information about absolute rates.

Questions and Answers

Dr. Ele Ferrannini (University of Pisa, Italy): The phenotype you identified is younger, leaner, not on insulin, and with a high A1c. Knowing Scandinavia, is this due to binge drinking? [Laughter] But seriously, perhaps there was something in the clinical notes about whether they went to a party the evening before?

A: I don’t think so. We don’t have that data but I doubt it. The relation that surprised us was being on insulin.

Q: How are people referred to Steno?

A: By primary care providers. It’s a system of triage. They’re on multiple medications and have long-term diabetes.

Dr. Simeon Taylor: I asked a statistician at the NIH to give me the power calculation for how large a study you would need to test the hypothesis that SGLT-2 inhibitors cause an increase in DKA from 1 in 3,000 to 1 in 1,000. The estimate was something like 20-25,000 per arm. That’s in a situation where people are randomly assigned. I’m assuming in Denmark the people who receive SGLT-2 inhibitors could be different in one way or another from those who don’t. Do you think you’ll have the power to estimate that level of increased risk?

A: That’s an important question and you already kind of answered it. We don’t know yet what the number of prescriptions are. That will be interesting. I have a feeling it will be in the ballpark to be able to say something useful.

Dr. Zachary Bloomgarden (Mount Sinai, New York, NY): It’s interesting that the rate of 1 per 1,000 is precisely the same as the empagliflozin and canagliflozin rates. In the FDA adverse event database it was about 5 per 1,000 but that’s very subject to bias because of things that occur with adverse event reporting based on news about it. We’re looking at a relatively similar order of magnitude, although there could be a several fold increase. If we’re able to isolate patients taking SGLT-2 inhibitors, we still have the dilemma of not knowing the reason the drug was given to one group vs. another. If it’s because they had worse control, we’ll be biased toward more adverse events.

A: That’s all important. We know in our own clinic we have more problems of power but great phenotyping. Ultimately the statistics and epidemiology are important.

Type 1 Diabetes: Triple therapy With Insulin, Liraglutide, and dapagliflozin – Implications for Ketogenesis

Paresh Dandona, MD, PhD (University at Buffalo, NY)

Dr. Paresh Dandona discussed results from a retrospective analysis (n=16) investigating the addition of AZ’s Farxiga (dapagliflozin) to insulin and Novo Nordisk’s Victoza (liraglutide) in type 1 diabetes. Data on 10 patients was presented as a poster at ADA, and Dr. Dandona stated that the team now has 20 patients enrolled. Results demonstrated significant reductions in A1c (0.7%; baseline = 8%), mean plasma glucose (28 mg/dl), and body weight (2 kg; baseline BMI = 31 kg/m2) and an 11% increase in time in range (70-160 mg/dl) with the addition of dapagliflozin. Importantly, average insulin dose did not decline with the introduction of dapagliflozin, though it had decreased when patients first started taking liraglutide. However, the two patients who experienced DKA did have substantial reductions in insulin dose (from 47 units/day to 28 units/day in one case), which Dr. Dandona described as the “secret” to the whole DKA controversy. While only two patients experienced full-blown DKA, there were significant increases in urine ketones (acetoacetate and beta-hydroxybutyrate) and free fatty acids across the study population. This is consistent with the hypothesis that SGLT-2 inhibitors push patients, particularly those with type 1 diabetes, into a subclinical ketogenic state that can progress to DKA under certain conditions. As Dr. Dandona acknowledged during Q&A, adding an SGLT-2 inhibitor does negate some of the favorable effects of GLP-1 agonist therapy. Fasting glucagon, which had diminished with the addition of liraglutide, rose again when dapagliflozin was added. Carbohydrate intake also increased. However, the overall results certainly suggest that the clinical benefits (A1c reductions, weight loss, reduced variability) are greater with the combination of these two classes compared to either alone.

  • The Q&A session following this talk touched upon bigger-picture questions about the overall benefits and risks of SGLT-2 inhibitors, especially in type 1 diabetes. Dr. Zachary Bloomgarden (Mount Sinai, New York, NY) cautioned attendees “not to throw out the baby with the bathwater.” He emphasized that he has seen impressive improvements in glucose variability with these agents and that they appear to be “a remarkably useful tool” for at least some patients with type 1 diabetes. Dr. Dandona agreed that these drugs could help fill an important unmet need, citing data from last year showing that mortality rates in type 1 diabetes remain significantly higher than those for the general population. By contrast, Dr. Ralph DeFronzo (University of Texas Health Science Center, San Antonio, TX) expressed concern about using SGLT-2 inhibitors in type 1 diabetes. He suggested that most patients simply need better optimization of insulin therapy and stressed that if one is going to prescribe an SGLT-2 inhibitor, it is crucial not to cut back on insulin doses.

Questions and Answers

Dr. Ralph DeFronzo (University of Texas Health Science Center, San Antonio, TX): If you simply decrease clearance, that doesn’t make ketoacidosis. These people, even if they don’t develop DKA, clearly switch from carbohydrate to fat metabolism. You always see a small increase in ketone levels. I agree that’s a precursor for what will happen if something else happens.

A: That’s absolutely spot on. Also in both cases, DKA occurred when we increased the dose of dapagliflozin from 5 mg to 10 mg. That can’t be a coincidence. There’s something else it’s doing. The whole milieu of reducing insulin dose is there, but why did the event occur just after enhancing the dose?

Dr. DeFronzo: One obvious thing is what happened to glucagon. I’m not sure there’s good data on the dose response to changes in glucagon. The role of the sympathetic nervous system has been overlooked. The rebound in glucose production happens within 30 minutes, long before glucagon goes up. That can only be neurologic. I suspect it’s the renal nerves activating the sympathetic nervous system.

Dr. Zachary Bloomgarden (Mount Sinai, New York, NY): Let’s be careful not to throw out the baby with the bathwater. The improvement in variability that I’ve seen clinically is quite remarkable. It appears there is at least a subset of patients with type 1 diabetes for whom use of an SGLT-2 inhibitor adds a dimension to their glycemic control that is not achievable otherwise. It seems quite clear that some people with type 1 diabetes will have DKA. Perhaps most have increased ketone production. At the same time, it is quite clear that this may be a remarkably useful tool. Our challenge for the weekend is to determine how to preserve this remarkably useful set of medications.

Dr. Dandona: Any weapon has two aspects: usefulness and possible toxicity. You have to determine how to balance it. Having said that, my data, although the CGM is impressive, don’t show the effects of SGLT-2 inhibitors alone. We had a GLP-1 agonist beforehand.

Q: Does it matter whether the insulin dose reductions are basal or bolus?

A: When you start with liraglutide from the beginning we’ve shown it’s the bolus dose you can reduce. You can have a gross effect on the whole thing. We’re in the habit of changing the bolus first and then shaving off the peaks by causing glycosuria. You’re peeing off the peak.

Dr. Dan Einhorn (Scripps Health, La Jolla, CA): Are you surprised at the amount of insulin needed to prevent DKA? You can clearly reduce insulin doses but not to 5 or 10 units. That may be one lesson we’re taking.

A: I thought in both cases they needed high doses of IV insulin to treat the DKA. That’s another issue.

Dr. Vivian Fonseca: Do you have data on the time course? Could you have predicted the DKA? What was beta-hydroxybutyrate at one week?

A: In the trial we’re doing with AZ on dapagliflozin in type 1 diabetes the patients have ketone meters, so we will be measuring that.

Dr. Robert Henry (University of California San Diego, CA): Glucagon went up despite the fact they were on liraglutide?

A: Exactly, you’re negating a lot of things. You’re switching on glucagon on top of a glucagon suppressor. There was also increased carbohydrate intake, which is the opposite of what you’d expect. We didn’t show the data on just liraglutide but ketogenesis reduces there.

Dr. DeFronzo: Depending on the assay you use, it clearly cross-reacts with liraglutide. That’s a big, big problem. It needs to be looked at. It’s clear in the paper that in everyone on liraglutide, glucagon goes up without a change in fasting glucose. In part that’s explained by cross-reactivity.

Dr. Ele Ferrannini (University of Pisa, Italy): What you showed is on liraglutide, glucagon levels fell. When you added the SGLT-2 inhibitor they went back up. That also rules out a paracrine effect.

A: The data I showed is fasting glucagon. We haven’t got postprandial glucagon on triple therapy.

Dr. Mary Korytkowski: After hearing your talk, I’m less convinced the reduction in insulin has anything to do with DKA. There was an absolute reduction there but in terms of units/kg, both were the same, on fairly robust doses.

A: One was very impressive weight-wise. But who knows what’s the quality of weight loss and the distribution of insulin. There were two patients who got the thing and only one who had those impressive reductions. You can’t rule out good old insulin that suppresses ketosis dramatically. When you treat DKA and give IV insulin, the first process to switch off is ketogenesis.

Dr. John Miles (University of Kansas, Kansas City, KS): I agree the level of beta-hydroxybutyrate is a cause for concern. I commend you for measuring free fatty acids. What struck me is that those that should shed mechanistic light were relatively modest, not in the DKA range. Can you speculate on that?

A: Free fatty acids were in the same phase as ketone bodies. Those ketones were nowhere near DKA levels. They were three or four times higher than baseline and free fatty acids were 75% above baseline. You’re laying a foundation but not getting there.

Dr. DeFronzo: I have been concerned about the whole story of type 1 diabetes and whether we should be going after them with SGLT-2 inhibitors. It seems to me that the first thing that should be done if you have a patient with type 1 diabetes who’s inadequately controlled is to ask why. Do they need more basal insulin or postprandial? Then make the appropriate dose adjustment. If they’re controlled, then why would you add an SGLT-2 inhibitor? If they’re inadequately controlled and you’re adding it, one thing you don’t want to do is cut back on insulin. To me that didn’t make sense. The idea that you can expect GPs to cut back on insulin to get someone on an SGLT-2 inhibitor and then re-titrate – that won’t happen.

A: You need to read the paper in the NEJM last year that showed mortality in type 1 diabetes with an A1c >9.5% - which is common, including in Steno – was ten times greater than the normal population. If A1c is <6.5%, mortality is twice that of the normal population. We have an epidemic of uncontrolled type 1 diabetes. Try monitoring each one on CGM and putting each one on CSII; think of the cost and effort. We should have alternative therapies. We’re toddlers. We will make mistakes. Insulin got discovered somehow 100 years ago and we haven’t done anything since then.

Dr. DeFronzo: I’m just emphasizing that if you have a poorly controlled patient with type 1 diabetes, it’s insane to cut back the insulin dose.


The European Experience of SGLT-2 Inhibitors and Diabetic Ketoacidosis

Henning Beck-Nielsen, MD (Odense University Hospital, Denmark)

Dr. Henning Beck-Nielsen – a member of the EMA – reviewed the timeline of European regulatory action to address concerns of DKA associated with SGLT-2 inhibitors. By May 2015, the EMA database had identified 101 cases of DKA in patients using SGLT-2 inhibitors. The agency sent out a warning on the issue on June 12, and the United Kingdom followed with a separate notice on June 26. Dr. Beck-Nielsen noted that the overall calculated risk for patients treated with SGLT-2 inhibitors is around 0.1%, but suggested that underreporting is quite likely. Of the European cases, about half occurred during the first two months of treatment, though some occurred much later. Most cases were atypical for DKA since blood glucose was only moderately elevated, which delayed diagnosis and led to inappropriate treatment. Of the cases within the European database, about 50% were treated with insulin. Dr. Beck-Nielsen advocated for the advice advanced by the regulatory agencies with the warning: patients with symptoms of DKA should be tested for elevated ketones, SGLT-2 inhibitor treatment should be stopped if DKA is suspected, confirmed ketonuria requires medical help, the class should not be used in patients with type 1 diabetes, and all suspected cases should be recorded. Dr. Beck-Nielsen argued this advice should be followed until the label is updated. We agree that it is prudent to advise against off-label use, but we also hope that the recommendations from this meeting will include advice for clinicians who do choose to prescribe off-label so that they can do so as safely as possible. Dr. Beck-Nielsen concluded by updating attendees on the expected timeline for EMA action on the subject: a PRAC recommendation to the EU commission is expected February 2016 and a final decision will likely be available in spring 2016. This is somewhat earlier than the original timeline of a final PRAC recommendation by October 2016.

  • Dr. Beck-Nielsen tackled head-on the question some have been raising: how do we weigh this poorly-understood risk of DKA against the substantial potential of cardioprotection? Dr. Beck-Nielsen emphasized the need to weigh the beneficial and harmful effects of SGLT-2 inhibitor therapy in the context of the patient in question. He reviewed the EMPA-REG OUTCOME results that demonstrated a cardioprotective benefit with Jardiance (empagliflozin) and made the provocative statement that the antidiabetic effect of SGLT-2 inhibitors is “not so impressive.” His main supporting argument was that the A1c reduction in the empagliflozin group compared to the placebo group was modest (0.3%) in EMPA-REG OUTCOME. We would note, however, that the study design was meant to produce fairly similar A1c reductions in both arms since the goal was to isolate the independent effects of the drug on CV risk. The difference in A1c after the initial 12-week period when investigators were not allowed to alter glucose-lowering medications was 0.5%-0.6%, closer to the efficacy seen in phase 3 trials. Dr. Beck-Nielsen argued that with so many alternative agents available, there is no need to choose a therapy associated with ketoacidosis in “ordinary” patients with type 2 diabetes without complications. On the other hand, Dr. Beck-Nielsen deemed the effect of empagliflozin on reducing heart failure “dramatic” and stated that “we can accept ketoacidosis” as a potential risk when prescribing SGLT-2 inhibitors to patients with type 2 diabetes and existing heart failure. Dr. Beck-Nielsen’s comments were on the extreme end of speakers at this meeting as few others seemed to agree that SGLT-2 inhibitor use should be limited in those who do not have cardiovascular disease.

Questions and Answers:

Dr. Matthew Weir (University of Maryland Medical Center, Baltimore, MD): Given the EMPA-REG data, to recommend against use of these drugs in people with cardiovascular disease flies in the face of that study. Weighing the risk of DKA more heavily is shortsighted.

A: We have evidence from one study that if you have cardiovascular disease, you have hard evidence that it helps to treat with SGLT-2 inhibitors. When you save lives, you can accept the risk for ketoacidosis. But when you have other drugs that can do the same work, you can’t accept ketoacidosis.

Dr. Weir: This is only study I’ve ever seen that showed a drug can lower blood glucose and prevent CV deaths.

A: I don’t think it has anything to do with the blood glucose. That was also the conclusion of the presenters in Stockholm.

Dr. Sam Dagogo-Jack (University of Tennessee, Memphis, TN): There is a suggestion that we attempt to profile who would benefit from SGLT-2 inhibitors on the background of higher morbidity. Before we consider that seriously, we need to understand whether DKA rates are fundamentally higher in type 2 diabetes patients treated with SGLT-2 inhibitors. Because reduction of insulin on any background can trigger DKA, it’s not rare in type 2 diabetes.

A: I agree. When the FDA comes out with regulations, they can help us answer that question. If the risk is not higher than background, we shouldn’t do anything.

Dr. Daniel Einhorn (Scripps Health, La Jolla, CA): Since you’re getting granular about the selection of drugs – other factors can come into play, like adherence. In general these are well tolerated. One of the things the writing group will have to include is not the frequency of adverse events but relative importance of them.

Dr. Aaron Vinik (Eastern Virginia Medical School, Norfolk, VA): The writing committee won’t do that. We’re not referring to the class, we’re just trying to understand the DKA. This is not a general discussion on the role of SGLT-2 inhibitors on the management of type 2 diabetes.

Dr. Simeon Taylor (University of Maryland School of Medicine, Baltimore, MD): One thing that’s important to realize is the data are not perfect and I’m not sure we’ll ever have perfect data. You have a choice – you can either try to do best you can with the data available or be purist and say you just don’t know and it’s not knowable and throw up your hands. I think we should do the best we can with the data available.

Dr. Vinik: That’s correct, that’s the issue this group will have to face. We hope to balance it and not pick one side over another. This is just the beginning.

Town Hall Meeting

Sustained Response of SGLT-2 Inhibitors on the Kidney

David Polidori, PhD (Janssen Research & Development, San Diego, CA)

Dr. David Polidori presented pharmacokinetic and pharmacodynamic data on J&J’s Invokana (canagliflozin) that could potentially shed some light on the timing of DKA risk. The data showed that the induced increase in urinary glucose excretion persists for several days after stopping the drug. There is still a ~90% effect after the first day, at least with the 300 mg dose, and some noticeable effect three or four days later. Dr. Polidori noted that this has also been observed with other agents in the class. He also suggested that it is purely a function of plasma concentration of the drug and that there is no desensitization or other regulation taking place. From our perspective, the main potential implication is that patients may need to stop treatment several days in advance of surgery or another anticipated stressor in order to avoid DKA. By contrast, canagliflozin’s effects on volume are largely attenuated with sustained treatment. In a study in patients with type 2 diabetes, the initial ~10% volume depletion had almost completely corrected itself after 12 weeks. This is consistent with the observation that adverse events associated with volume depletion occur primarily in the first six weeks of treatment. Dr. Polidori did note that small changes in hemoglobin and hematocrit do persist, suggesting that some changes in fluid status are more sustained.

  • While less pertinent to this meeting, it is interesting to think about how the volume effects of SGLT-2 inhibitors relate to the cardiovascular benefits seen with Lilly/BI’s Jardiance (empagliflozin) in EMPA-REG OUTCOME. While the mechanism of benefit in that trial remains a matter of much debate, a reduction in heart failure due to empagliflozin’s diuretic effect appears to be the closest thing to a prevailing hypothesis at the moment. However, if the volume depletion with empagliflozin is similarly transient, such a sustained and dramatic effect on heart failure and mortality would be surprising. We assume this is one reason why most of the pre-EMPA-REG OUTCOME speculation about cardioprotection with SGLT-2 inhibitors focused on other factors like blood pressure and weight reductions.

Dapagliflozin and DKA Cases: Clinical Development & Post-Marketing REports

Nayyar Iqbal, MD (AstraZeneca, Princeton, NJ)

Dr. Nayyar Iqbal presented data on rates of DKA with AZ’s Farxiga (dapagliflozin) in clinical trials and post-marketing reports and provided an update on steps the company has taken to address the issue. In the entire type 2 diabetes clinical trial program for dapagliflozin, there was one case of serious DKA, two cases of non-serious ketonuria, and one case of non-serious metabolic acidosis in over 6,000 patient years of exposure. Dr. Iqbal estimated the incidence at 0.01% and noted that event rates in the ongoing DECLARE CVOT have been <0.1% thus far. In type 1 diabetes, there were no cases of DKA in the company’s 14-day phase 2b study (n=70) or in a 24-week investigator-initiated study (n=12). Excitingly, we learned that AZ is conducting a phase 3 trial (DEPICT 1) in type 1 diabetes in which DKA is being prospectively adjudicated. As far as post-marketing data, which Dr. Iqbal stressed is likely incomplete, 46 cases of DKA with dapagliflozin had been reported to the EMA as of its warning on the issue in June. Of those, 12 cases occurred in people with type 1 diabetes taking the drug off-label. Dr. Iqbal also noted that AZ has distributed a letter to 700,000 healthcare providers in the US about the risk in collaboration with other SGLT-2 inhibitor manufacturers. The company has also distributed information in the EU and globally, and to the investigators in ongoing clinical trials. Right now, the type 1 diabetes trial is the only one in which DKA is being prospectively adjudicated, but AZ is implementing that process in all other ongoing studies involving dapagliflozin. The company has also developed a detailed questionnaire for those reporting DKA post-marketing and is initiating mechanistic studies to learn more about the pathogenesis.

Questions and Answers

Dr. George Grunberger: You mentioned that there are 7,000 patients in DECLARE. Is there any indication of the reports from that trial?

A: I can’t give you the exact number, but it’s a very small number, less than 0.1%.

Dr. Simeon Taylor: In the previous presentation, the speaker made the point that we’re really at the top of the dose-response curve with canagliflozin at the high dose. The data with dapagliflozin suggests that it didn’t have as large of a pharmacodynamic effect on glycosuria, so perhaps we’re not at the top of the dose-response curve. That could mean it has a slightly lesser effect in terms of glycemic efficacy. On the flip side, it could have a benefit in terms of fewer mechanism-based side effects. Is it possible the selected dose of dapagliflozin could have less of an issue?

A: Obviously we’ll monitor it closely. It’s possible there are minor differences between different drugs in the same class but we can’t be certain.

Dr. Ele Ferrannini (University of Pisa, Italy): Are you also collecting data on the fixed-dose combination? The DPP-4 inhibitor could balance out the tendency by potentiating glucose-induced insulin release.

A: In three studies we didn’t see any cases. Up to this point DKA was an event we were monitoring for. In any trial where dapagliflozin is present, we’re now putting adjudication in. It’s not implemented now but that’s the plan.

Dr. Beck-Nielsen: What was your definition of type 2 diabetes? Did you include measurement of C-peptide?

A: I can’t speak to all of the trials, but the majority did exclude people with <0.7 C-peptide. A few might not have but the majority used a cutoff.

Sandra Williams, MD (Cleveland Clinic, Cleveland, OH)

Dr. Sandra Williams used a case presentation of SGLT-2 inhibitor-related DKA to argue that the phenomenon is more similar to starvation ketosis than classical diabetic ketoacidosis. The DKA occurred in a patient on metformin, liraglutide, and canagliflozin who had been eating significantly less than usual following cosmetic surgery. Dr. Williams suggested that SGLT-2 inhibitors likely amplify the process seen in a normal fasting state: glucagon increases even more than usual due to glycosuria and secondary effects on the alpha cells, and this decreases glucose reuptake, depletes glycogen stores, and inhibits gluconeogenesis. This can rapidly trigger ketogenesis and possibly ketoacidosis. Her main point was that the mechanism of ketoacidosis, especially euglycemic ketoacidosis, with SGLT-2 inhibitors may be more similar to that of starvation ketoacidosis than classical DKA. The main clinical implication, echoed throughout the meeting, is that patients should not take these drugs if they are fasting or on a very low-carb diet. During Q&A, other speakers suggested that the metformin could also have contributed to the risk in the context of low food intake.

Questions and Answers

Dr. John Miles: I noticed this patient continued to get metformin and liraglutide even though they weren’t eating. If I wanted to make someone sick with metformin, I’d give it on an empty stomach. If someone’s not eating and I wanted to make them worse, I’d give metformin and liraglutide.

A: She had the procedure at an outside hospital, got ill at home, and was admitted to our hospital. We discontinued all drugs.

Dr. Miles: My mistake, but the point should be made that a lot of these patients are also taking drugs that can cause nausea. One piece of advice we don’t always give people on metformin is if you’re not eating, don’t take the drug.

A: I totally agree. When she was admitted, the case the interns put together was that the nausea and vomiting was narcotic-induced.

Dr. Ralph DeFronzo: The theory is ok but it has some problems. Metformin is a powerful inhibitor of gluconeogenesis. She was almost assuredly intravascularly volume depleted. She probably had a very high metformin level, which has its own independent effect on ketogenesis. Some things may have been related to metformin.

Q: Was she desperately trying to lose weight?

A: She had some personal issues going on. She wanted her age to be reported as only in her 40s, she probably had an event coming up and seemed to have a target weight but she didn’t disclose that.

The Boehringer Ingelheim Available Experience

Thomas Seck, MD (Boehringer Ingelheim, Ridgefield, CT)

Dr. Thomas Seck offered a look at BI’s experience with ketone-related adverse events in Lilly/BI’s clinical development program for Jardiance (empagliflozin). He began with a brief review of ketone body measurement data from phase 2 and 3 trials for the drug, which found that empagliflozin increased ketone body production. In a phase 2 study, empagliflozin was associated with a reduction in the insulin/glucagon ratio, though the effect was attenuated after four weeks of chronic use. Dr. Seck then moved on to a discussion of specific cases of DKA within the clinical trial program, of which there were five in the placebo group, two in the empagliflozin 10 mg group, and one in the empagliflozin 25 mg group. He pointed out that none of the cases were deemed by the study investigator to be related to empagliflozin use. Next, he gave a detailed review of the cases of DKA from the EMPA-REG OUTCOME study (three in the empagliflozin 10 mg group and, one in the empagliflozin 25 mg group; he did not review the one case in the placebo group), noting that the EMPA-REG OUTCOME was an especially relevant population to look at since its participants had advanced diabetes and 50% were on insulin. Finally, Dr. Seck offered a look at both of the cases of DKA that occurred among the 117 patients enrolled in the two phase 2 trials for empagliflozin in type 1 diabetes. Notably, insulin delivery was limited in both cases (in one, insulin dose was reduced by 70% when empagliflozin was started and in the other the insulin dose was reduced by 50% and there was an insulin pump catheter malfunction). Overall, Dr. Seck concluded that since seven of the cases within the type 2 diabetes program occurred on empagliflozin and five occurred on placebo, the incidence of DKA was overall low and balanced between the two groups. He did acknowledge that the studies were not powered to exclude a clinically meaningful increase in DKA. The situation in type 2 diabetes is shaping up to be somewhat analogous to the incretin/pancreatitis controversy: there are hints of an increase in risk but the absolute numbers are too small to be detected in most trials.

Questions and Answers

Dr. Zachary Bloomgarden (Mount Sinai School of Medicine, New York, NY): One of the concerns is that because of increased reabsorption, ketonuria is not a good measure of ketoacidosis. It might be worth it to go back to see if we can get a sense of the false negative rate of ketonuria. At present time, ketonuria has to be revised. If it turns out it’s good enough for a screening, that would make a big difference.

Dr. Matthew Weir (University of Maryland Medical Center, Baltimore, MD): With regard to ketone stick testing, is it not concentration dependent? If it’s concentration dependent, these people are sick as hell so what’s the directionality? Is it “DKA” causing the gastroenteritis or the other way around and we’re getting fooled by volume depleted people who have a concentrated ketone stick?

Dr. John Miles (University of Kansas, Kansas City, KS): This test of urine ketones not quite worthless but it’s pretty close. It’s like trying to put lipstick on a pig.

The Public Perspective

Emily Regier, BS (Close Concerns, San Francisco, CA)

Our very own Ms. Emily Regier offered a change of pace with a big-picture perspective on the DKA controversy (see her slides here). She began with a review of the timeline in which awareness of DKA has grown, from an alert by Dr. Anne Peters in January to the first public presentations on the topic in March to its current status as a significant focus of discussion at every major diabetes conference. Next, she evaluated the risk of DKA associated with SGLT-2 inhibitor use in type 2 diabetes (low) and type 1 diabetes (significant but possibly manageable), and reminded attendees of the high patient engagement involved in actually managing DKA risk through active monitoring of ketone levels. While many speakers at the meeting wanted to focus the conversation and guidelines only on type 2 diabetes, Ms. Regier argued that practical recommendations for their safe use in type 1 diabetes are needed since off-label use is clearly happening (as evidenced by data from diabetes market research company dQ&A). Ms. Regier also called on the industry representatives in the room to conduct more randomized controlled trials of the class in type 1 diabetes, saying that “the bottom line is, we need more data.”

  • Ms. Regier acknowledged that the phenomenon occurs in type 2 diabetes, but argued that the risk appears low (0.1% or less in clinical trials to date) and linked to specific precipitating factors. She reminded the writing committee that while its important to outline those factors, it’s equally important to place the risk in perspective.
  • Moving onto the more uncertain area of type 1 diabetes, Ms. Regier noted that while the risk appears to be higher, off-label SGLT-2 inhibitor use also appears to be associated with meaningful clinical benefits, including A1c reductions, weight loss, and reduced glycemic variability. She suggested that the consensus in the scientific community based on commentary at conferences is that the risk of in type 1 diabetes seems to be “significant but manageable.”
  • Ms. Regier reminded attendees that managing this risk in type 1 diabetes will likely require active engagement from patients. Specifically, it will likely require active monitoring of ketones, which assumes that patients know what normal ketone levels should be, own a ketone meter or urine test strips, and know under what circumstances they need to check ketones. She cited a study (n=120 patients with type 1 diabetes) that found that 17% of patients regularly monitor their urine for ketones, 16% regularly monitor their blood, and 58% do neither. When asked if they would monitor their ketones under specific circumstances like nausea, vomiting, or feeling unwell, only 9%-33% said yes. We don’t think there is nearly enough education on ketones in type 1 or type 2 patients – we also believe to some extent this is an addressable problem.
  • Off-label use of SGLT-2 inhibitors in type 1 diabetes is clearly happening. Ms. Regier cited data from diabetes market research company dQ&A that shows 3% of patients with type 1 diabetes surveyed (36 out of 1,204) had taken an SGLT-2 inhibitor this past quarter and another 1% had taken one previously. The majority of these patients reported positive changes in their diabetes management after adding the SGLT-2 inhibitor (A1c reductions, reductions in fasting and postprandial glucose, and improvements in confidence), and the majority lowered their insulin dose while on the drugs. While citing Dr. Anne Peters’ personal protocol as an example of safe off-label use, Ms. Regier acknowledged that most patients are not being treated by a physician as well-versed in the topic as Dr. Peters (we wish!).
  • Ms. Regier concluded her presentation with specific calls to action for the various stakeholders in the room. To the writing committee, she suggested that it would be prudent to advise against off-label use in type 1 diabetes, but that guidelines are needed for providers who do choose to prescribe off-label. To the many industry representatives in the room, Ms. Regier encouraged them to continue studying these drugs in type 1 diabetes, noting that too many patients with type 1 diabetes are not at goal and more options are urgently needed. SGLT-2 inhibitors may represent a promising option, but we need to figure out how to maximize their benefits while reducing the risk of DKA – a feat that can only be accomplished with more large-scale randomized controlled trials. Finally, Ms. Regier thanked AACE/ACE for proactively addressing this issue and urged everyone to transmit the recommendations that result from the meeting as widely as possible – we’ll certainly do our part to do so!

Questions and Answers

Dr. Simeon Taylor (University of Maryland School of Medicine, Baltimore, MD): For off-label use in type 1 diabetes, it’s worth remembering that DKA is not the only unknown safety issue. If the bone loss issue turns out to be important, that also needs to be taken into consideration. As you have enthused for some amount of off-label use, it’s important to remember this is not the only unknown. I’m a great enthusiast for phase 3 randomized controlled trials.

Dr. Ele Ferrannini (University of Pisa, Italy): Your presentation was so articulate and full of good sense. My recommendation is don’t go to medical school, you’ll lose it [laughter].

Dr. Tim Garvey (University of Alabama, Birmingham, AL): What the writing team should do is write down what Emily said so we can all go home early.

Q: Do you have any information on patients enthused by the ability to lose weight by adopting a low carb diet? I’ve been thinking, how do clinical trials differ from the real world? In trials, patients are given scrupulous dietary info. In the real world, patients see ads about losing weight. To use clinical trials as incidence of ketoacidosis in the real world may be overlooking the messiness of real world.

Ms. Regier: That’s a good point. The CEO of market research company dQ&A is here in the room. Richard, do you have any data on that?

Mr. Richard Wood (dQ&A, San Francisco, CA): We pulled some data this afternoon on folks taking SGLT-2 inhibitors. Among type 2’s, 87% have a BMI over 27 – they’re all interested in losing weight. About a third have a BMI over 40. About a quarter are on GLP-1 agonist and about a third have tried GLP-1 agonists previously. We haven’t specifically asked if they’ve adopted a low-carb diet, but we have another survey going out in December, so we’ll ask them then.

The Lexicon Experience: Type 1 Diabetes

Paul Strumph, MD (Lexicon, The Woodlands, TX)

Dr. Paul Strumph provided a detailed look at cases of DKA within Lexicon’s phase 2 program for sotagliflozin. Sotagliflozin is particularly interesting because it is a dual SGLT-1/SGLT-2 inhibitor and is being developed exclusively for a type 1 diabetes indication at the moment (we are not sure if partnership discussions in type 2 diabetes are still ongoing). Though the product is still several years away from the market (it recently advanced into phase 3), Lexicon already fielded questions from investors concerned about the possibility of DKA risk during its 2Q15 update. Both cases of DKA that occurred in sotagliflozin’s phase 2 study occurred in patients on an insulin pump. Dr. Strumph made the point that if someone is more prone to starvation ketosis, there might be something about insulin pump malfunction specifically that can exacerbate it more than other forms of insulin. In both cases, the patients presented with “classic” DKA with high blood glucose levels and there was no clinically meaningful decrease in the basal insulin dosage leading up the event. Dr. Strumph noted that the phase 3 program is progressing as planned and is being monitored carefully by a data monitoring committee. Lexicon has suggested in the past that sotagliflozin could pose a lower risk of DKA than SGLT-2 selective inhibitors, primarily since the SGLT-1 inhibitor component promotes a potentially protective increase in GLP-1.

Type 2 Diabetes Patient Profile: DKA after SGLT-2 Inhibitors and Ways to Prevent DKA

Lance Sloan, MD (Woodland Heights Medical Center, Lufkin, TX)

Dr. Lance Sloan synthesized evidence from case studies to provide his hypothesis for the driving factors behind SGLT-2 inhibitor-related DKA in patients with type 2 diabetes. He began by noting that the patients who experienced DKA without a specific precipitating factor at play typically had an A1c >12%, meaning they have very low insulin secretion and are prone to dehydration. He argued that SGLT-2 inhibitors exacerbate that dehydration and cause a rise in stress hormones including glucagon. These stress hormones will further increase blood glucose, triggering a need for increased insulin production to prevent ketosis. Since beta cell function is low, this increased insulin production is impossible and DKA occurs. Dr. Sloan noted that according to AACE and ADA guidelines, insulin therapy should be initiated for all patients with A1c >9%-10%. He also advocated making sure patients initiating SGLT-2 inhibitor therapy have normal intravascular volume and recommended starting long-acting basal insulin simultaneously at a dose of at least 0.5 units/kg. Dr. Sloan also advised educating patients to stop their SGLT-2 inhibitor usage and seek medical help if they experience nausea, vomiting, and abdominal pain. In “classic” type 1 diabetes (lean individuals on low insulin doses), Dr. Sloan suggested that SGLT-2 inhibitor usage should be avoided entirely to prevent DKA. He signed off on SGLT-2 inhibitor use in “non-classic” type 1 diabetes, namely obese patients, as long as the insulin dose does not drop below 0.5 units/kg/day and patients know how to monitor for symptoms. Dr. Sloan concluded with the assertion that DKA is generally rare in both type 1 and type 2 diabetes (though more common in type 1), SGLT-2 inhibitors do not cause DKA but may make patients more susceptible, insulin deficiency is the key culprit, and the condition is preventable with proper insulin therapy. Dr. Sloan also argued that providers need to be educated to recognize and diagnosis DKA independent of blood glucose.

Cases, DKA, & Concepts I

Dan Einhorn, MD (Scripps Health, La Jolla, CA)

Dr. Dan Einhorn shared data on the incidence of DKA in the clinical development program for J&J’s Invokana (canagliflozin) in type 2 diabetes. The database included 17,596 patients and nearly 24,000 patient-years of exposure. Patients with a history of type 1 diabetes or DKA were excluded, and reporting was not perfect, as ketoacidosis had not been identified as a significant concern during most of this period. A total of 12 patients in the canagliflozin trials experienced DKA. Of these, two were treated with placebo, four with 100 mg canagliflozin, and six with 300 mg canagliflozin. Importantly, Dr. Einhorn noted that six of the 12 patients actually had autoimmune diabetes – they got into the trials because “things happen.” In addition, C-peptide levels were generally quite low, which Dr. Einhorn suggested could be a helpful indicator of risk. Most of the patients presented with some precipitating factor, and ten of the 12 had a blood glucose level >300 mg/dl. Of the ten patients who got DKA while on canagliflozin, eight were on insulin. All of those patients had either questionable adherence to insulin or concomitant illness. Of the two patients not on insulin, one had type 1 diabetes and the other had serious concomitant illness. It is clearly possible to slice the data in any number of ways, but Dr. Einhorn’s main conclusion was that these cases were predominantly classical DKA with a precipitating factor, and that the incidence rate was consistent with background rates of DKA in type 2 diabetes. He raised the question of whether this issue could be “metformin/lactic acidosis redux,” where there was enormous public concern about a side effect that ultimately proved to be very rare. We appreciated this big-picture perspective and agree that while it is crucial to better understand the causes and solutions to this risk, it is also important to consider it in its proper context.

Questions and Answers

Dr. John Miles (University of Kansas, Kansas City, KS): We heard mention of Flatbush diabetes, which reminded me how plastic the beta cell can be. When you get low C-peptide like this, do you have a sense of whether it’s acute or chronic?

A: For the purposes of this, it kind of doesn’t matter. In my opinion you would use C-peptide more in an acute setting. To get into the study it wasn’t measured, so we don’t know baseline C-peptide. For the purposes of the DKA story, it’s acute.

Dr. Yehuda Handelsman (Metabolic Institute of America, Tarzana, CA): Typically when people get DKA their insulin is low, so C-peptide is low during the acute event and then comes back up.

Dr. Zachary Bloomgarden: (Mount Sinai, New York, NY): In most studies we’ve had incidence expressed per patient-year. This is a lower incidence but many of the studies were shorter than a year. Can you re-calculate it to make it more comparable to everything else?

Dr. Gary Meininger (Janssen, Raritan, NJ): In the article, the mean duration for the entire population was 85 weeks. Most patients who contributed to that come from CANVAS, where the median duration was 2.5 years.

Dr. Bloomgarden: You can’t just take the overall population duration. It needs to be done right.

Dr. Sam Dagogo-Jack (University of Tennessee, Memphis, TN): Is there solid evidence that these same people, if they were not on an SGLT-2 inhibitor but were in the same metabolic state and exposed to the same precipitant, would have escaped DKA?

A: We certainly don’t have that. That’s why the background rate concept is important. Long before the SGLT-2 inhibitors, people with type 2 diabetes went through this. Why we haven’t talked about it is something else. The key is that the drug itself may not be the cause. The drug may be along for the ride when a person with type 2 diabetes of a certain type gets sick.

Efficacy and Safety of Canagliflozin, a Sodium Glucose Co-transporter 2 Inhibitor, as Add-on to Insulin In Patients Wiith Type 1 Diabetes Mellitus

Robert Henry, MD (University of California San Diego, CA)

Dr. Robert Henry shared data, first presented at EASD, from a phase 2 trial (n=351) of Invokana in type 1 diabetes – the largest study to date of an SGLT-2 inhibitor in this population. A higher percentage of patients achieved the composite primary endpoint of A1c reduction ≤0.4% and no weight gain with both doses of canagliflozin vs. placebo (41% and 37% with the higher and lower doses, respectively vs. 15% with placebo; p<0.001). Dr. Vivian Fonseca (Tulane University, New Orleans, LA) critiqued this primary endpoint as overly modest during Q&A, but as Dr. Henry noted, the importance of weight loss vs. weight gain for many patients on insulin should not be underestimated. Canagliflozin also led to greater mean A1c reductions (~0.3% placebo-adjusted) and weight loss (~3-5% placebo-adjusted). Insulin doses were also reduced, mostly from bolus insulin. Patients were also advised to consider down-titrating basal insulin prior to beginning the study to avoid hypoglycemia. Rates of total documented hypoglycemia were consistently high (>90%) in all groups. Rates of severe hypoglycemia were low, but there was a numerical increase in the 300 mg canagliflozin group (7% vs. 3% with 100 mg canagliflozin and 2% with placebo). Most relevant to this meeting, there was a significant increase in ketone-related adverse events with canagliflozin: 11 events with the 300 mg dose and 6 events with the 100 mg dose vs. no events with placebo. All events were associated with precipitating factors, though as Dr. Anne Peters (USC, Los Angeles, CA) noted in her subsequent presentation, the same type of events occurred in the placebo group and did not cause DKA. As at EASD, Dr. Henry argued that future trials can help illuminate what needs to be modified to mitigate the risk of DKA. We agree, which is why we feel so strongly that companies should continue studying these agents in type 1 diabetes. We are also very eager to see CGM data from this trial, as reduced glycemic variability seems to be one of the most meaningful benefits of SGLT-2 inhibitors in type 1 diabetes based on anecdotal evidence. Dr. Henry said that this data should be presented at ADA next year.

Questions and Answers

Dr. Vivian Fonseca (Tulane University, New Orleans, LA): Your criteria for success were relatively modest: a 0.4% reduction in A1c and very little weight loss. Maybe people want it, but the tradeoff doesn’t seem worthwhile to me.

A: You might be right, but it’s not how patients think. (Editor’s note: From a patient perspective, we add that we continue to hope that “success” with therapies and diabetes management more broadly isn’t always viewed solely in terms of A1c changes. We point readers toward Dr. Anne Peters’ comment on CGM traces after using SGLT-2s – “A: I have many, many CGM traces. Everyone’s gets better.”)

Dr. Fonseca: They feel they’ll do better than the mean?

A: It’s reasonable. Some patients really do well.

Q: Is there a difference between doses?

A: The only difference was the change in body weight. The 300 mg group had more weight loss. And this is fairly substantial weight loss for people with type 1 diabetes with a BMI of 28 kg/m2.

Dr. Dan Einhorn (Scripps Health, La Jolla, CA): Would you suggest that cutting back the insulin dose is part of why DKA developed?

A: I don’t think so. People got into trouble not because they backed off. It looks to me like these folks might have gone into DKA even if they were not on canagliflozin. It seems to me that what happens is the bar is a little lower, so maybe they would’ve gotten ketosis or moderate ketones. That’s my take on the data. You’ll have to make a judgment

Dr. Einhorn: Did the placebo group have the same precipitants that didn’t result in DKA?

Dr. Anne Peters (USC, Los Angeles, CA): Yes.

Dr. John Miles (University of Kansas, Kansas City, KS): Did you say the incidence of DKA was 5% per year?

A: In 18 weeks it was 5% and 7%.

Dr. Miles: Yesterday Mary reviewed the type 1 diabetes registry and the number was 5% per year, just to remind us. We can’t make a formal comparison.

A: There were no cases with placebo. That’s the real issue.

Dr. Ele Ferrannini (University of Pisa, Italy): Why did you leave the option of reducing the insulin dose by 20% or 10% before the study when you were titrating to target anyway?

A: It was an optional thing that was done because previous studies had shown that if you don’t reduce the dose you’ll flatten out. People had an A1c between 7% and 9%, and if you’re 7.2% and put on 300 mg canagliflozin you’ll flatten out. If you’re at 8.8% it’s unlikely. That was the thinking at the time. I think it’s something for us to discuss, whether there should be a reduction.

Comment: Clearly this drug has benefited some patients with type 1 diabetes. I would ask the writing committee to write their paper in a way so payers won’t be able to come back and say you can’t use this in your patients, just look at this AACE report. I think you should make the statement that doctors should use their best medical judgment and when the benefit outweighs the risk, it’s reasonable to use it in most cases. It’s legal to do so.

A: Remember that right now it’s off label in type 1 diabetes.

Dr. Simeon Taylor (University of Maryland, Baltimore, MD): If you saw 5% incidence of DKA in 18 weeks in the context of a clinical study with among the most skilled physicians in the world, do you want to extrapolate about off-label use by people who are less aware of all the concerns and who may not have the same level of supervision?

A: I have lots of suggestions. That will come out when the writing committee sits down. The DKA paper has gotten favorable views so far, and that’ll come out within a few months. We don’t have CGM data. We’re hoping to present it at ADA next year. The CGM data was pretty impressive. That’s something that stuck with me.

Dr. Sam Dagogo-Jack (University of Tennessee, Memphis, TN): You had some non-obese type 1 diabetes patients on insulin doses of 30-40 units, and the dose reduction ranged from 4-7 units per day. Could that have played into the DKA events?

A: Each of the patients were seen by individual investigators, who have lots of leeway in type 1 diabetes. Some didn’t lower doses initially and some did. It’s hard unless you tell them they can’t. I don’t think anyone will say that or the investigators will listen.

Update on Ketoacidosis with SGLT-2 Inhibitors

Anne Peters, MD (USC, Los Angeles, CA)

Dr. Anne Peters opened by half-jokingly accepting responsibility for the entire SGLT-2 inhibitor/DKA controversy, noting that she was the one who first brought the issue to the FDA’s attention, at the urging of Dr. John Buse (UNC, Chapel Hill, NC). She stressed that her primary concern was the fact that cases in type 1 diabetes were being missed. Particularly since the companies couldn’t comment on off-label use, she felt (and still feels) that physicians had a public health duty to raise awareness about the risk. She also noted that she has now developed a protocol, which she outlined in detail at EASD, for off-label use in type 1 diabetes that has so far been successful at avoiding DKA. Dr. Peters emphasized that many of her patients with type 1 diabetes have experienced very meaningful benefits with SGLT-2 inhibitors, sharing a compelling video of a patient discussing the dramatic improvements in her quality of life due to lower glycemic variability with J&J’s Invokana (canagliflozin). However, she also stressed that the increased risk seen in the phase 2 canagliflozin trial should be taken seriously. Even though all events occurred along with precipitating factors, those same factors occurred in the placebo group and did not lead to DKA. Dr. Peters noted that it was the FDA that expanded the conversation to include type 2 diabetes, and she stressed that while she, Dr. Buse, and others have seen some cases in that population, it appears to be very rare. All of the cases she has seen personally have been related to surgery, and she argued that the main concern in that situation is that the cases are misdiagnosed. For example, she shared a harrowing story of a patient who almost died from DKA after bariatric surgery because he was misdiagnosed by hospital staff. Dr. Peters closed with what we thought was a very astute summary of the situation: (i) physicians who prescribe SGLT-2 inhibitors off-label in type 1 diabetes need to have a clear protocol for managing the risk; (ii) frontline providers need to be aware that euglycemic DKA is possible; and (iii) “the lawyers should stay out of it … I’m not afraid but we should be aware.”

Questions and Answers

Dr. John Miles (University of Kansas, Kansas City, KS): I was particularly entertained by the cases in which doctors diagnosed lactic acidosis with normal lactate. Were admission lactates obtained in all cases save outpatients?

A: Everyone who went into acidosis had lactate measured, and they all had some elevation. Those who come in in DKA have lactate levels that are a little high but not lactic acidosis. It goes away afterwards. Ketones were much higher.

Q: Can you speculate whether patients with type 1 diabetes with C-peptide may be protected against DKA compared to garden-variety patients?

A: It depends on how you define C-peptide. People have done studies where they actually measure it and some people have levels that wouldn’t be detected with a different assay. But people don’t have enough insulin, so I don’t think it protects them. The average duration of diabetes was 22 years, and most people were C-peptide-negative. To me measuring it in type 1 diabetes is complicated, all of these people aren’t making enough insulin, and I don’t think it protects them.

Dr. Jaime Davidson (University of Texas Southwestern Medical Center, Dallas, TX): Did those who developed DKA have a longer duration of diabetes?

A: No, it was all over the place.

Dr. Dan Einhorn (Scripps Health, La Jolla, CA): It’s difficult to classify what type of diabetes a person has, which makes a big difference. Is there a way of looking back and looking at those that were called DKA and how they presented? Has this been going on longer than we realized?

A: In the study there were zero cases in the placebo group. In a placebo-controlled trial, people living through the flu season didn’t get DKA, and it excluded people who had recent DKA. In my practice people don’t go into DKA. People can contact me and I can keep them out of DKA. I started seeing patients who’ve never gotten DKA get it on an SGLT-2 inhibitor because I didn’t understand what to do. Now I’m doing it and not seeing cases. I didn’t historically see much DKA, there was a big uptick, and now it has come back down.

Dr. Einhorn: In type 1 diabetes it’s a clearer story. In type 2 diabetes, should we look back and characterize the background incidence better so we understand whether it’s brand new?

A: I think type 2s get DKA because they’re in between. In my practice type 2s never get DKA. On the other side of town with no access to medical care they’re always getting DKA. DKA occurs more in type 2 diabetes in another setting than in the patients I see. I hadn’t seen type 2s with DKA ever. I’ve seen it now in post-surgical cases and I see it in the county hospital all the time, so it depends. SGLT-2 inhibitors might tip someone who wouldn’t have gotten it into it under rare circumstances. I’m not worried in type 2s. We just need to know it can happen.

Dr. Yehuda Handelsman: Were any of those patients also on a GLP-1 agonist?

A: Only a few, most weren’t. In John’s case, he doesn’t think it protects against it.

Dr. Sam Dagogo-Jack: You showed a video of an eloquent and fluent lady who’s a poster child for SGLT-2 inhibitors in type 1 diabetes, who alluded to sensing highs and lows that are now flattened. You also shared CGM data on the tendency to flatten, but it seems not to be universal. Can you comment on your experience?

A: I have many, many CGM traces. Everyone’s gets better.

Cases, DKA, and Concepts II

SGLT-2 Inhibitors and Ketosis

Ele Ferrannini, MD (University of Pisa, Italy)

Dr. Ele Ferrannini shared clinical data on how Lilly/BI’s Jardiance (empagliflozin) might predispose individuals to. He demonstrated that empagliflozin was associated with an increase in plasma glucagon levels that was detectable only half an hour after dosing. However, he noted that this effect was attenuated after four weeks of chronic administration. Empagliflozin was also associated with excess endogenous glucose production (EGP) and decreased plasma insulin. Dr. Ferrannini shared data showing similar increases in glucagon and EGP in individuals with normal glucose tolerance.  In addition, both individuals with type 2 diabetes and those with normal glucose tolerance experience increased glycerol turnover while on empagliflozin, a measure of increased lipolysis. Overall, Dr. Ferrannini emphasized that the effects of empagliflozin in type 2 diabetes applied in individuals without diabetes as well, just scaled down. Looking specifically at lipid oxidation and ketogenesis, Dr. Ferrnannini showed that lipid oxidation is already elevated in individuals with type 2 diabetes, as evidenced by higher plasma levels of the ketone body beta-hydroxybutyrate. There is also an association between less well-controlled diabetes and increased lipid oxidation. With the addition of an SGLT-2 inhibitor, lipid oxidation increased in both groups but increased more in the individuals with type 2 diabetes. Dr. Ferrannini also found that ketonemia in patients with type 2 diabetes was associated with lower fasting insulin, lower mealtime insulin, and decreased beta cell sensitivity to glucose. His main conclusion was that the DKA associated with SGLT-2 inhibitors occurs through its central mechanism of action, not through unpredictable off-target effects.

The Pathophysiology of DKA in Patients with Diabetes on SGLT-2 Inhibitors

Ralph DeFronzo, MD (University of Texas Health Science Center, San Antonio, TX)

Dr. Ralph DeFronzo provided a mechanism-focused summary of how SGLT-2 inhibitors predispose patients to DKA. Dr. DeFronzo argued that, at its heart, DKA is caused by insulin deficiency, which causes the body to switch to ketone metabolism for fuel. As Dr. Ele Ferrannini (University of Pisa, Italy) demonstrated in his previous talk, SGLT-2 inhibition increases endogenous glucose and glucagon production in addition to promoting glycosuria. Dr. DeFronzo argued that typically, in very insulin-resistant individuals, a rise in blood glucose past a certain level will push glucose into the cell through non-insulin-mediated mechanisms so that glucose can still be used as a fuel source. However, with massive glycosuria, blood glucose levels are never allowed to rise to that level and the body must initiate ketosis in order to supply sufficient energy to cells. In addition, SGLT-2 inhibitors create the potential – but not the necessity – of euglycemic DKA, because the renal glucose excretion may or may not match the excess blood glucose depending on how much insulin is on board. However, Dr. DeFronzo made it clear that the “euglycemic DKA” label is a false distinction since the same mechanism is at play in both cases. The situation can be exacerbated by precipitating factors like insulin deficiency, hyperglucagonemia, stress, and hypovolemia. Dr. DeFronzo concluded that the “metabolic milieu of DKA is set by this drug.” He also asserted that while incidence isn’t surprising based on the drugs’ mechanism of action, it will be difficult to predict exactly who will develop DKA (since unpredictable events can occur such as an emergency surgery or insulin pump malfunction). In his view, the most important thing is to educate patients to recognize the symptoms and seek medical help.

Questions and Answers

Dr. Sam Dagogo-Jack (University of Tennessee, Memphis, TN): Given the very logical precursors to a ketogenic environment these drugs create, can you speculate on relative rarity of DKA events?

A: I would agree with you. Despite the fact that we switched people over to lipid metabolism, my take is this is an unusual and rare event. I would feel remiss if the FDA started to put limitations on the use of these drugs. We and pharmaceutical companies need to alert people that this is potential problem. I think Anne’s idea of drawing up some key points so you can tell the patient when they might be in trouble needs to be done. However, the benefits of SGLT-2 inhibitors far outweigh any potential side effects in type 2 diabetes. In type 1 diabetes, it needs to be evaluated. I’m not against its usage, but the way it’s being done is all wrong. I think we need lower doses and I’m against cutting back on insulin.

Q: Do you think the negative caloric balance matters?

A: In Japan, they believe in putting you on a high-fat, low-carb diet, which of course is going to exacerbate situation. It’s giving them the substrate to drive ketogenesis. Not enough attention is paid to issue of dietary intake.

Q: These two talks imply that ketosis is far more common than ketoacidosis. Can you envision long-term morbidity as a consequence of chronic ketosis on SGLT-2 inhibitors?

A: If you look at diabetics who are poorly controlled, beta-hydroxybutylate levels will be up and lactate levels are up for same metabolic reasons. Diabetics have the propensity to form ketones, they’re already in this metabolic milieu. Will SGLT-2 inhibitors make it worse? I don’t really have an answer to that.

Dr. Paresh Dandona (University of Buffalo, NY): What has emerged for me is that we have to be really conservative with the reduction of insulin doses. Another thing from this discussion this morning is we ought not to be frittering around with A1cs around 7%. The other thing in my mind is, until now, our teaching of diabetic patients has been carb-centric. Now we’re into a contradictory situation as to how we’re going to play this game.

Common Features and Contributing Characteristics of DKA in Patients on SGLT2 Inhibitors

Jaime Davidson, MD (University of Texas Southwestern Medical Center, Dallas, TX)

Dr. Jaime Davidson offered a closing take on what the writing committee should consider as they craft their recommendations. First, Dr. Davidson reminded attendees that the guidelines are distributed widely and AACE guidelines historically do not recommend off-label use of drugs, as that opens the organization to liability. Next, Dr. Davidson summarized the “common sense” takeaways for clinical care with SGLT-2 inhibitors. Such advice included hydrating patients and especially using insulin instead of additional oral agents if A1c is >9%. Dr. Davidson argued that in all cases of SGLT-2 inhibitor-associated DKA in which there were no precipitating factors, the patients had A1c ~12% and should have been on insulin instead of an SGLT-2 inhibitor. Dr. Davidson recommended implementing physician education programs since most of these patients were not appropriate candidates for SGLT-2 inhibitor therapy in his mind. Also, given the low incidence of DKA in clinical trials for the class, Dr. Davidson called the issue largely “much ado about nothing” and reminded the committee of the class’s many benefits, including its improvement of fasting and postprandial glucose, A1c, diastolic and systolic blood pressure, weight reduction, and – last but certainly not least – its mortality reduction of 38% (at least for empagliflozin).

-- by Helen Gao, Emily Regier, and Kelly Close