SGLT Inhibitors for Type 1 Diabetes: An Obvious Choice or Too Good to Be True?

In the recent past we have been fortunate to find new treatments based on entirely new mechanisms . . . or is that really so? The sodium–glucose cotransporter inhibitors (SGLTi) offer an instructive case in point. These apparently “new” agents were preceded by phlorizin, a naturally occurring compound that has been used for physiologic studies for over 100 years (1). Given orally it interferes with intestinal absorption of carbohydrates and causes glucosuria. Studies in animal models of diabetes have provided useful insights, such as evidence that reduction of hyperglycemia can restore insulin sensitivity toward normal (2). However, phlorizin was considered unsuited for use by humans due to undesirable side effects. Recently its mechanism of action has been traced to effects on sodium–glucose cotransporters (SGLT) in the intestinal mucosa and the renal proximal tubule (3), and drugs with more nuanced effects on these transporters have been developed (4,5). These newer agents have greater affinity for SGLT2, which is expressed mainly in the kidney, and less for SGLT1, which facilitates intestinal absorption of glucose. Clinical effects of the newer SGLTi drugs in type 2 diabetes (T2D) include modest reductions of plasma glucose and A1C, weight, and blood pressure and are mediated in large part by glucosuria and sodium diuresis. These agents are taken once daily by mouth and do not require dose titration. Symptomatic side effects—mostly urinary or genital irritation or infections—are common but tolerated by most patients.

Importantly, the new SGLTi drugs have highly desirable—and somewhat unexpected—cardiac and renal protective effects, as least in selected cohorts. Large trials of empagliflozin and canagliflozin, aiming to demonstrate safety in patients with T2D and high cardiovascular risk, have shown favorable effects on heart failure, cardiovascular death, and progression of albuminuria (6–9). Unwanted effects have been reported as well, including dehydration, lower-extremity amputations, and diabetic ketoacidosis (DKA), but they are relatively uncommon.

Although their underlying mechanisms are not fully understood, the cardiovascular and renal benefits have led to great enthusiasm for the SGLTi drugs and increasing clinical use. Consideration of their use immediately after metformin is advised for patients with T2D with established cardiovascular disease (10), and questions have arisen about other groups of patients. We need to know whether these agents can be as effective for patients with T2D not accompanied by overt cardiovascular or renal disease, or for cardiac patients without diabetes, and studies are exploring these questions. But what about prescribing these agents along with basal-bolus insulin for type 1 diabetes (T1D) (11)?

Supplementing earlier small studies of combining an SGLTi with basal-bolus insulin for T1D, six reports of larger studies have now been published (12–17), two of them in this issue (16,18) and three in the September issue of Diabetes Care (14,15,17). These studies were designed to determine whether this form of combination therapy can improve glycemic control without unacceptable risk of hypoglycemia, DKA, or other immediately apparent adverse effects in this population. Their main features and results are summarized in Table 1. Canagliflozin was the drug used in one study, sotagliflozin in three, dapagliflozin in two, and empagliflozin in one. The designs of these studies were similar. All were two- or three-arm, randomized, placebo-controlled studies with duration between 18 and 52 weeks. The mean age of participants was between 41 and 46 years, mean BMI between 27 and 30 kg/m², and mean baseline A1C between 7.6 and 8.5%. Outcomes reported at the end of the studies were also similar. The mean placebo-adjusted improvement of A1C associated with use of the SGLTi ranged between 0.25 and 0.52%, with little difference between lower and higher doses in the studies that included this comparison. Changes of weight relative to placebo averaged between −2.2 and −4.4 kg, with a tendency toward greater reductions at the higher doses. Of note was the incidence of DKA, which ranged between 1.5 and 6.0% greater with active treatment than with placebo. Other than DKA, no notable adverse effects beyond the urogenital symptoms previously observed in T2D appeared in these short-term studies.

How can we interpret these results? On the positive side, the reduction of A1C by about 0.3–0.4% (1.5–2.0 mmol/mol) is at a level generally considered clinically significant. If sustained over time, this improvement would likely reduce the risk of microvascular complications. Better overall control was obtained without increasing the risk of hypoglycemia. The modest but consistent reduction of weight might also be important for some patients and could limit later cardiovascular risk. These considerations are naturally appealing to patients and providers who are frustrated when glycemic targets cannot be attained and concerned about weight and other cardiovascular risk factors.

Unfortunately, there are also limitations. The increased frequency of DKA is the main cause for concern. The incidence of confirmed DKA averaging close to 3% during these studies lasting no longer than a year is not a trivial problem. This risk might be even greater under conditions in routine clinical settings where monitoring of patients may be less close than in clinical trials. Experience in populations of patients considered to have T2D has shown that treatment with an SGLTi can be associated with DKA accompanied by no more than moderate hyperglycemia and nonspecific symptoms (19–21). In many cases the event is preceded by an illness or procedure or by an inappropriate reduction of insulin dosage. The correct diagnosis may be delayed by lack of marked hyperglycemia. Some patients affected clearly have T2D, but many appear to be individuals with unrecognized T1D. Thus, the consistently increased frequency of DKA in recent studies of SGLTi in T1D is not very surprising. If SGLTi drugs are to be approved and prescribed for T1D, heightened awareness about this complication must be emphasized. This is also true for patients thought to have T2D who actually have T1D and are already being prescribed SGLTi, especially given recent reports of increasingly frequent hospital admissions for DKA generally (22,23). Education and support of both patients and medical providers to prevent euglycemic DKA are key clinical strategies and must be improved. They should include advice on ketone testing as well as recognition of atypical symptoms (nausea, vomiting, malaise) in the absence of significant hyperglycemia (24).

Moreover, even if we are able to limit the risk of DKA, there are other questions to address. Is the excess of fractures found in the canagliflozin development studies in T2D (25), posing the possibility of a class effect of SGLTi on bone integrity, a real concern? No such effect has yet been seen with empagliflozin and dapagliflozin (26,27), but physiologic studies of bone metabolism provide some support for this hypothesis (28–30). In addition to altered handling of calcium and phosphorus in the kidney, ongoing mobilization of gluconeogenic substrates provoked by fasting glucosuria and accompanying hormonal changes might lead to a long-term effect on bone matrix. Weight loss due to caloric restriction is known to be associated with loss of bone mineral and fractures (31–33), and women with anorexia nervosa have greatly reduced bone mineral content at an early age (34,35). Whether long-term use of SGLTi drugs by younger T1D patients could have effects on bone like those induced by voluntary dietary restriction cannot be determined by short-term studies. We need to weigh known short-term clinical benefits against potential long-term risks such as this one.

These observations on the efficacy and safety of SGLTi have important clinical implications. Balancing benefits versus risks of SGLTi will require thoughtful decisions by regulatory and professional groups, care providers, and also people with diabetes. The most obvious responsibility falls upon regulators and groups creating guidance for clinical care. Does the evidence provided by these short-term studies justify approval of these agents for use in T1D? If so, what tactics for mitigation of risks are needed? Are we at a stage where we can stratify individual patients according to benefits versus risks so as to allow appropriate use of these new drugs by selected cohorts? We can expect a vigorous discussion of these questions. To support such decisions, we need more information. Of particular interest is whether a direct renal effect limiting progression of nephropathy—beyond that provided by improved control of A1C and blood pressure—can be demonstrated in T1D, and the present studies do not address this question. Also, confirmation of the short-term risk of DKA in T1D reminds us that long-term risks also need further evaluation in both T1D and T2D. As suggested in earlier commentaries (36–38), drugs in this class should be prescribed cautiously until longer-term, prospectively collected experience with them is available.

The story that began with phlorizin is still in progress. The newer SGLTi drugs offer a greatly improved balance of benefits versus risks, and their use in T1D is clearly an exciting possibility. It would offer the first adjunctive oral therapy for this important cohort of patients. But we do need more information to personalize the use of these powerful new agents in T1D. So, is use of SGLTi agents in T1D “an obvious choice,” or are the promises of this approach “too good to be true”? The key lies in precise identification of which patients will obtain the greatest benefit with the least risk.

• © 2018 by the American Diabetes Association.