Comparing Diabetes Medications – Where Do We Set the Bar?

More than 25 million Americans have type 2 diabetes mellitus and face decisions about which medications to use to lower glucose levels. These decisions are increasingly complex (now involving 12 different classes of glucose-lowering agents) and increasingly costly (resulting in over $18 billion in annual expenditures¹). Yet, despite the enormous health and economic implications of these decisions, there are few comparative effectiveness outcomes studies to guide clinical practice. The major goals of glycemic control in type 2 diabetes are to prevent severe hyperglycemia and to reduce the risk of long-term diabetic complications, with the majority of the effort devoted to the latter. Therefore, clinicians, patients, and health systems need to know how well glucose-lowering agents (or combinations) improve the chances of living longer without cardiovascular events, painful neuropathy, kidney failure, amputation, and blindness, as well as the risks of adverse effects and other burdens of these treatments. Comparative effectiveness studies designed to assess these outcomes may not appear feasible. But what are the alternatives?

Studies designed to compare effectiveness based on glycemic control require fewer patient-years of follow-up and are less expensive but may ultimately mislead us. The recently completed Treatment Options for Type 2 Diabetes in Adolescents and Youth (TODAY) study, funded by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), demonstrates the inherent shortcomings of evaluating diabetes therapies based on surrogate end points.² TODAY tested 3 different strategies for type 2 diabetes in youth: metformin alone, metformin hydrochloride combined with rosiglitazone maleate, or metformin combined with a lifestyle-intervention program focused on weight loss. At the time of trial design, cardiovascular concerns with rosiglitazone³ were not yet apparent. TODAY showed that addition of rosiglitazone, but not lifestyle intervention, was superior to metformin alone in maintaining glycemic control, although this approach was associated with significant weight gain. However, few (if any) clinicians are advocating rosiglitazone for treatment of diabetes in youth, given its known risk profile. If the risks were still unknown, a trial like TODAY would be unlikely to identify rosiglitazone’s cardiovascular safety signal.

The Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness (GRADE) study, a $100 million trial, also funded by the NIDDK, seeks to compare the effects of 4 drugs when combined with metformin in adult patients with recent-onset type 2 diabetes.⁴ The study has recently launched and is recruiting participants. However, like TODAY, the trial is focused on the effects of medications on glycemic durability, not long-term clinical outcomes. The trial will randomize 5000 subjects to combination therapy with metformin and 1 of 4 agents: glimepiride, sitagliptin phosphate, liraglutide, or insulin glargine. The primary end point will be the time to glucose control failure, defined as a hemoglobin A_1c (HbA_1c) level of 7% or higher after having been treated with metformin and started on therapy with the second randomly assigned medication. Although other potentially important aspects of the 4-drug combinations will be assessed (including adverse effects such as weight gain and hypoglycemia, effects on cardiovascular disease risk factors, surrogate measures of microvascular disease, tolerability and quality of life, and cost and cost-effectiveness), the trial is primarily designed to answer the question of which glucose-lowering agents best maintain glycemic control and not which agents are associated with the best outcomes.

Drugs that modify risk factors do not always modify patient risk.⁵ Both low- and high-density lipoprotein cholesterol are strong risk factors for cardiovascular disease, but interventions that decrease low-density lipoprotein cholesterol or raise high-density lipoprotein cholesterol levels do not always reduce the risk of cardiovascular events.⁶ Similarly, glycemic parameters, like HbA_1c, are strongly associated with diabetic complications and thought to be central to their pathophysiology, but interventions that lower glucose levels may not necessarily result in improved outcomes.³ Drugs have multiple effects that may be independent of the intended impact on the disease process. Because these effects are unintended, they are often not anticipated and difficult to recognize.

The Food and Drug Administration (FDA) now requires data to show that the excess relative risk of cardiovascular events is clearly less than 80% for approval of all diabetes agents.⁷ The degree of permitted risk is high, but in their guidance to industry, the FDA advises that lowering of HbA_1c is “reasonably expected to reduce the long-term risk of microvascular complications”⁷(p2) and, therefore “reliance on HbA_1c remains an acceptable primary efficacy endpoint for approval of drugs seeking an indication to treat hyperglycemia secondary to diabetes mellitus.”⁷(p2) From the point of view of a patient, these recommendations may be reasonable as long as there is transparency about what is known and an ongoing effort to expand the evidence. However, patients should want to know whether the new drug will reduce their risk of adverse clinical outcomes, not just whether it will control their glucose levels.

Cardiovascular outcomes trials help expand the evidence about diabetes medications, but they are not sufficient for informed decision making. Two separate, recently completed cardiovascular outcomes trials compared saxagliptin and alogliptin against placebo, and both demonstrated that these 2 dipeptidyl peptidase 4 (DPP-4) inhibitors did not significantly decrease or increase the rates of major cardiovascular events.⁸ In accordance with the FDA guidance, these drugs improve glycemic control without increasing the risk of cardiovascular complications and may be, therefore, accepted as clinically useful. But do these medications provide a benefit with respect to long-term diabetic complications? And how do they compare against other glucose-lowering agents? The standards for drug approval ensure that medications entering the market are safe and effective. However, decisions about management of a chronic disease, like diabetes, require long-term data about outcomes, not just HbA_1c. Clinicians must set higher standards to ensure that these medications are used wisely and result in real health benefits.

To improve decision making in diabetes, we need to invest in long-term outcomes-based comparative effectiveness research. The current approach has led to the approval of 12 different classes of glucose-lowering medications but has generated few data so far with respect to long-term outcomes. With the incidence of type 2 diabetes climbing inexorably and affecting ever-younger patients, we simply cannot afford to continue this short-sighted strategy.

Given the high cost of long-term outcomes trials, complexity of glucose management, and the many years required for the development of diabetic complications, how can this be done? We need to commit at least a fraction of the resources we spend on glucose-lowering drugs to coordinate larger, longer, practical trials focused on end points important to patients. If trials like the GRADE study⁴ were extended over 10 years (instead of 5), they may begin to address important knowledge gaps and provide information about clinical outcomes. Trials focused on durability of glycemic control may be more affordable than what we are proposing—at least in the short run—but they are likely to give us data of questionable value that, in turn, are used to justify therapies for patients that may not improve their health and potentially harm it. That would be costly.