SodiumGlucose Cotransporter-2 Inhibitors: Lack of a Complete History Delays Diagnosis

Abstract

On 15 May 2015, the U.S. Food and Drug Administration (FDA) warned in a Drug Safety Communication that sodiumglucose cotransporter-2 (SGLT2) inhibitors may lead to ketoacidosis in patients with type 2 diabetes mellitus (1). This discovery should not have been a surprise. For more than 125 years, investigators had described ketosis and ketoacidosis induced by the phytochemical phlorizin, the prototypical SGLT inhibitor, as well as in patients with familial renal glucosuria (FRG), a condition that is considered a natural model of SGLT2 inhibition. The story of SGLT2 inhibitor ketoacidosis offers a cautionary tale about how historical inquiry can provide valuable insights and accelerate recognition of risks associated with new drugs. In Pursuit of History Our exploration of the history of SGLT2 inhibitor ketoacidosis began in early 2014 when one of us (B.R.L.) learned, during employment in the pharmaceutical industry, about the accrual of spontaneous postmarketing reports of ketoacidosis submitted by health care professionals to the manufacturer of canagliflozin. Subsequently, a Freedom of Information Act request for FDA Adverse Event Reporting System (FAERS) data revealed that the FDA s receipt of these same reports began on 8 May 2013 (2). Seeking a precedent for this phenomenon, we did Boolean searches in PubMed and Google Scholar using the key words phlorizin (and its variant spellings), renal glucosuria or glycosuria, ketoacidosis, ketones, ketosis, and acidosis. We iteratively mined the bibliographies of articles and books retrieved by these searches for additional primary sources. Google Translate and native speakers provided translations of nonEnglish-language publications. Two of us (B.R.L. and S.I.T.) had also worked in industry on the development of dapagliflozin, although we had not been aware of the extent of this literature until conducting the searches prompted by these postmarketing reports. Phlorizin: The Basis of a Precedent We focused our search on phlorizin because it is the progenitor of contemporary SGLT2 inhibitors. Initially isolated from the bark of apple trees as a treatment of malarial fever (3), phlorizin increases urinary elimination of glucose through competitive inhibition of SGLT2, the principal transporter for renal glucose reabsorption, and of SGLT1, a lesser glucose transporter in the kidney (4). Reductions in blood glucose after single doses of phlorizin in patients with diabetes (5), as well as favorable effects of phlorizin on glucose metabolism in partially pancreatectomized rats (6), foretold the promise of SGLT2 inhibitors in treatment of type 2 diabetes mellitus. In 1999, pharmaceutical chemists described one of the first orally bioavailable derivatives of phlorizin (7), a prodrug known as T-1095. The authors cited a 1945 review that described acidosis and increased acetone bodies with phlorizin administration (8), although they did not discuss its implications. The manufacturer of T-1095 did not commercially develop this product. Other chemists took a different biochemical approach in developing the SGLT2 inhibitors that are currently marketed. Phlorizin consists of a glucose molecule linked to an aglycone called phloretin. The O-glycosidic bond that joins them is subject to hydrolysis by endogenous enzymes. This results in reduced oral bioavailability and a relatively short half-life. To avert this problem, chemists eliminated the linking oxygen atom to yield a C-glycoside structure (9). They also modified both the glucose moiety and the aglycone to accomplish several objectives: increased selectivity for SGLT2 relative to SGLT1, increased potency for SGLT2, optimization of pharmacologic and pharmaceutical properties, and creation of intellectual property to protect marketing exclusivity. Notwithstanding these modifications, the structures of all marketed SGLT2 inhibitors closely resemble that of phlorizin. Although phlorizin is sometimes called a nonselective SGLT inhibitor, it has 9-fold selectivity for inhibition of SGLT2 relative to SGLT1 (10). This is similar to the 20-fold selectivity of sotagliflozin, albeit less than that of other available SGLT2 inhibitors (11). Moderate doses of phlorizin and dapagliflozin have similar effects on glucose excretion (12, 13). Taken together, these considerations provide a strong scientific basis for concluding that experience with phlorizin could offer insights into the pharmacologic responses to modern SGLT2 inhibitors. From Phlorizin Diabetes to Ketoacidosis By the time pharmaceutical chemists modified phlorizin s basic structure to create contemporary SGLT2 inhibitors, its ketogenic properties had long been known. At the end of the 19th century, the German physician Josef von Mering (18491908) recognized that phlorizin produced glucosuria. This discovery may have occurred while he was testing phlorizin during his studies of antipyretics, which would later lead to the discovery of acetaminophen (14). In a presentation to the German Congresses for Internal Medicine in 1886, von Mering called this phenomenon phloridzin diabetes (15). He ended a second presentation on this topic in 1887 with the observation, In the case of phloridzin diabetes I have repeatedly been able to detect abundant amounts of acetone and oxybutyric [-hydroxybutyric] acid in urine (16). Two years later, von Mering reported that normal dogs given phlorizin during prolonged fasting or with a carbohydrate-poor diet became listless and comatose or died. At necropsy, their livers were depleted of glycogen. In addition to glucose, their urine contained large amounts of ammonia, acetone, and -hydroxybutyric acid, findings that resemble those in patients in diabetic coma (17). These observations were remarkable because dogs are comparatively resistant to starvation ketosis (18). Von Mering also induced ketosis in a man by giving him 10 g of phlorizin by mouth in the early morning, after which his urine contained glucose and abundant amounts of acetone (17). Other investigators confirmed von Mering s findings (19, 20). Von Mering is renowned in the field of diabetes research for his later collaboration with Oskar Minkowski in which they produced diabetes mellitus in a dog by removal of the pancreas (21). Reports of ketosis induced by phlorizin appeared in the English-language literature in 1914. In a short communication, Stanley Benedict (18841936), inventor of Benedict reagent for detection of glucose, noted the ability of protein ingestion to diminish ketosis induced by phlorizin (22). In 1923 at the Rockefeller Institute for Medical Research in New York, Frederick Allen (18791964) published the results of extensive studies of acidosis in phlorizin-treated dogs (23). At the time of these experiments, before the commercial availability of insulin, Allen was the leading proponent of starvation diets in management of type 1 diabetes (24). He described ketonemia, ketonuria, and reductions in plasma bicarbonate concentration (carbon dioxide combining capacity) in phlorizinized dogs that had undergone fasting, diets poor in carbohydrates and high in fat, or partial pancreatectomy. Most dogs became weak and lethargic, and many died. Some exhibited the deep breathing characteristic of ketoacidosis. Acidosis resolved with glucose administration and was difficult to induce when the glycosuric response to phlorizin was diminished, as in a dog with chronic kidney disease. Other investigators confirmed the reduction in blood bicarbonate levels, and they further documented a reduction in blood pH (25, 26). In 1942, I. Arthur Mirsky (19071974), a psychoanalyst with an abiding interest in metabolic research, applied phlorizin to a canine model of type 1 diabetes mellitus. Mirsky noted that although the depancreatized dog developed diabetes, it did not develop levels of ketonemia seen in ketoacidotic diabetic humans, even after starvation and withdrawal of exogenous insulin. He observed, however, that severe diabetic acidosis and coma can be produced in the depancreatized dog by withdrawal of insulin and food plus the administration of a single dose of phlorizin (27). In the absence of insulin in pancreatectomized dogs, ketogenesis was inhibited by administration of glucose in an amount sufficient to restore hepatic glycogen (28). Mirsky concluded that ketogenesis was enhanced by phlorizin when administered to patients with diminished reserves of hepatic glycogen. He presented his observations on the ketogenic effect of phlorizin at a meeting of the Association for the Study of Internal Secretions, the forerunner of the Endocrine Society, in New York in 1940 (29). He did so again at the first annual meeting of the American Diabetes Association in Cleveland on 1 June 1941. In his address to the American Diabetes Association, Mirsky described administering phlorizin to a diabetic patient suffering from a non-specific infection. The glycosuria that ensues causes a decrease in liver glycogen and ketosis follows (30). Thus, the capacity of infection to precipitate ketoacidosis in diabetic patients may be potentiated by concomitant SGLT inhibition. For his lifetime of work in diabetes research, Mirsky received the Banting Medal for Scientific Achievement from the American Diabetes Association in 1965. Numerous publications throughout the past century reported phlorizin-induced hyperketonemia in various animals. Phlorizin caused marked increases in blood ketone levels in starved rabbits and cats (31). Rats given phlorizin while fasting had acetonemia up to 10 times greater than that in fasting controls; however, phlorizin did not produce additional ketonemia in nephrectomized animals, showing the importance of glucosuria (rather than extrarenal actions) in stimulating ketogenesis (32). Arterial concentrations of acetoacetate and 3-hydroxybutyrate (-hydroxybutyrate) in fasting rats treated with phlorizin were similar to those seen in diabetic ketoacidosis induced by streptozotocin (33).