Diabetes Care Volume 40, May 2017 629 COMMENTARY Modern Sulfonylureas: Dangerous Matthew C. Riddle or Wrongly Accused? Diabetes Care 2017;40:629–631 | DOI: 10.2337/dci17-0003 Except for insulin, sulfonylureas and bi- presumably because they are very inex- sulfonylurea (relative risk 1.53, 95% CI guanides are the best studied and most pensive, allow once-daily oral dosing, re- 1.43–1.65) in studies with an identified widely used glucose-lowering agents. liably reduce glucose, and rarely cause potential bias, metformin as compara- However, neither class of drugs has symptomatic side effects other than hy- tor, and mortality as the outcome. Rela- had an easy life because of concern poglycemia. More than 40 years after tive risk was not increased (1.06, 95% CI about safety. Phenformin, a biguanide, the UGDP, their risks versus benefits 0.92–1.23) in studies with no major was associated with lactic acidosis and are still debated (9–11). Statistical as- bias, a comparator other than metfor- withdrawn from use (1) after causing in- sessments of data pooled from random- min, and all cardiovascular events as creased mortality in the University ized studies and clinical databases the outcome. Presumably these differ- Group Diabetes Program (UGDP) (2). continue to be published, with con- ences contribute to the inconsistency The UGDP also found a sulfonylurea, tol- flicting conclusions (12,13). of the literature. butamide, to be associated with in- This issue of Diabetes Care includes a The authors further commented on creased mortality (3). Since then, a thoughtful contribution to this discus- difficulties posed by properties of the newer biguanide, metformin, has risen sion by Azoulay and Suissa (14). These treatment to which sulfonylureas are to its current place as the leading oral experienced epidemiologists describe compared. All the studies judged free of therapy for diabetes based on its relative the potential pitfalls in designing and bias compared use of a sulfonylurea with lack of hazard from lactic acidosis and interpreting analyses of observational metformin, except one that compared evidence, especially from a subgroup (real world) data on treatment with sul- sulfonylurea plus metformin with metfor- of participants in the UK Prospective Di- fonylureas or other agents. They identify min alone. The bias-free studies directly abetes Study (UKPDS), that it can reduce three difficulties. The first is “exposure comparing sulfonylureas with metformin cardiovascular risk and mortality (4,5). misclassification,” a failure to identify showed more frequent deaths or cardio- Even though the main randomized com- the time each patient is actually taking vascular events during treatment with a parison in the UKPDS (sulfonylurea or in- the drug in question. A second is “time- sulfonylurea (relative risk ranged from sulin vs. lifestyle therapy) showed that lag bias,” in which the analysis does not 1.16 to 1.55, with lower boundaries of cholorpropamide, glyburide, or glipizide account for the effect of studying patients the 95% CI above 1.00). A possible inter- also can reduce medical risks (5,6), the at earlier versus later stages of diabetes. pretation of this finding is that sulfonyl- reputation of all sulfonylureas has re- The third is “selection bias,” resulting ureas increase cardiovascular risk. An mained tarnished. A warning of “in- from exclusion of certain patients be- alternative is that metformin is beneficial, creased risk of cardiovascular mortality” cause of changes of regimen or clinical while sulfonylureas have a neutral effect. remains in their labeling information. events during the period of observation. The cardiovascular benefitofmetformin Treatment guidelines and publications After assessing 20 observational studies in the UKPDS supports the second inter- reporting effects of new drugs in other of patients with type 2 diabetes who pretation. The bias-free study that in- classes often emphasize the risk of hypo- were using sulfonylureas, they judged cluded metformin in both arms showed glycemia and weight gain from sulfonyl- that only 6 were free of these kinds of no difference in risk, also suggesting ureas. And yet, at least 25% of patients with bias. They found cardiovascular risk to a neutral effect of the sulfonylurea. Be- type 2 diabetes are using sulfonylureas (7,8), be increased during treatment with a cause metformin, with its favorable Division of Endocrinology, Diabetes & Clinical Nutrition, Oregon Health & Science University, Portland, OR Corresponding author: Matthew C. Riddle, [email protected]. ©2017 by the American Diabetes Association.Readersmay use this article as long as thework is properly cited, the useis educational and not for profit, and the work is not altered. More information is available at http://www.diabetesjournals.org/content/license. See accompanying article, p. 706. 630 Commentary Diabetes Care Volume 40, May 2017 Table 1—Arguments for and against the modern sulfonylureas Evidence yet to be For the prosecution For the defense presented Ischemic preconditioning Tolbutamide and glyburide interfere Gliclazide, glipizide, and glimepiride with ischemic preconditioning do not alter ischemic preconditioning Hypoglycemia Hypoglycemia and weight gain Gliclazide, glipizide, and glimepiride occur with all sulfonylureas cause less hypoglycemia than glyburide Observational studies Cardiovascular risk is higher with Cardiovascular risk is mainly higher vs. Studies with better design to sulfonylureas vs. comparators in metformin, which decreases risk limit bias some studies Randomized studies Short-term mortality was increased Long-term mortality was decreased CAROLINA (glimepiride vs. with tolbutamide in UGDP with glyburide, chlorpropamide, and linagliptin) glipizide in UKPDS GRADE (glimepiride vs. other second-line agents) The modern sulfonylureasdgliclazide, glipizide, and glimepiridedare accused of “increased risk of cardiovascular mortality.” The table summarizes the main arguments for prosecution and defense, including evidence from studies of older sulfonylureas. cardiovascular effects, is the preferred taking gliclazide or glimepiride than for this question (19). Also, in the Glycemia first-line oral agent, a sulfonylurea would those taking glyburide (odds ratio 0.15, Reduction Approaches in Diabetes: A Com- most helpfully be compared not with 95% CI 0.04–0.56, P , 0.005). There is parative Effectiveness Study (GRADE), metformin but with alternative second- also evidence that glyburide causes treatment with glimepiride or alternative line therapies. more hypoglycemia than other cur- agents is randomly allocated (20). To summarize, the main findings of rently used agents. A dramatic example Meanwhile, metaphorically, the jury Azoulay and Suissa (14) suggest that is an analysis of emergency department is still deliberating as to whether all some of the harm attributed to sulfonyl- admissions for hypoglycemia in Ger- sulfonylureas are unsafe based on wor- ureas may be related to unintended many that showed .80% fewer events risome evidence from studies of tolbuta- bias in the design or interpretation of with glimepiride than glyburide (0.86 vs. mide and glyburide (Table 1). Gliclazide, studies rather than an effect of this class 5.6 events per 1,000 patient-years) (17). glipizide, and glimepiride are reliably ef- of agents. Their description of several Another issue not emphasized by fective in lowering glucose, but are they categories of bias is illuminating and could Azoulay and Suissa (14) is selection too dangerous to use? As suggested by improve the design of future analyses bias related to the clinician’s judgement Azoulay and Suissa (14), more skillful anal- of observational data. However, some re- in choosing a treatment well suited to an ysis of observational data are possible, lated questions deserve further comment. individual patient. Treatment allocation and some randomized trial experience is One problem lies in the assumption bias creates an imbalance that is difficult soon to be reported. If new evidence that all sulfonylureas are alike. Sulfonyl- to neutralize by statistical methods, in- supports a not guilty verdict, the modern ureas differ in at least two ways that are cluding calculation of a propensity score, sulfonylureas should regain respect and relevant to cardiovascular risk. One con- especially in databases lacking detailed continue to be an important option for cerns an effect on vascular KATP channels information on concurrent illnesses. It controlling glucose. that interferes with ischemic precondi- is a persistent limitation of observational tioning and may increase the risk of car- studies and can be entirely avoided only diac events. This undesired effect occurs by random allocation of treatment. Funding. SupportforthisworkwasfromtheRose with tolbutamide and glyburide but not For these reasons, both well-designed Hastings and Russell Standley Memorial Trusts. with gliclazide, glipizide, or glimepiride observational studies focused on the Duality of Interest. M.C.R. has received re- (15). Whether this difference alters car- newer sulfonylureas and randomization search grant support through Oregon Health & diovascular outcomes is not well estab- comparisons are needed. Notably, it Science University from AstraZeneca, Eli Lilly, and Novo Nordisk and honoraria for consulting lished, but some evidence suggests it would be good to know whether cardio- or speaking from AstraZeneca, Biodel, Elcelyx, does. A well-conducted, prospective ob- vascular risk differs when a dipeptidyl