Reviews/Commentaries/ADA Statements COMMENTARY

Lessons From the Avandia Controversy A new paradigm for the development of drugs to treat

ROBERT I. MISBIN, MD David Nathan had expressed skepticism about the recent approval of , apparently unimpressed by the extensive ulfonylureas used to be the only oral the desirable pharmacological effects of body of data that was available on the agents available in the U.S. to treat both. was highly effective in FDA Web site (8). What the FDA should S patients with type 2 diabetes. The lowering A1C and had favorable effects take away from Dr. Nathan’s criticism and had been used on serum lipids. But contrary to what one from the Avandia affair is that the time has widely in the 1950s and 1960s but was might expect, muraglitazar appeared to come to reassess what should be expected removed from the market in 1977 be- increase the risk of adverse events related of a new drug to treat diabetes. That a new cause of lactic acidosis (1). Other bigua- to cardiac ischemia (4). This might have drug is more effective than placebo in nides, including , were felt to been considered an isolated event were it lowering glucose levels is no longer be safer and had been used extensively not for the recent brouhaha that rosiglita- enough. New drugs should be tested in elsewhere in the world. But a biguanide zone (Avandia) also increased the risk of comparison with other antidiabetes did not become available in the U.S. until cardiac ischemia. agents that are already in use (9). A plan 1995 (2). The Avandia affair began with a meta- should be in place at the time of approval The year 1995 marked a sea change in analysis showing that in- that will determine what benefit and harm availablilty of drugs to treat type 2 diabe- creased the risk of . can be expected from chronic use (10). In tes. The approval of metformin, a bigua- Its release by the New England Journal of the following sections, I highlight prob- nide, was followed by the approval of Medicine online on 21 May 2007 with an lems with the approval process that are , a galactosidase inhibitor that, accompanying editorial (5,6) generated a illustrated by Avandia and suggest a new like metformin, had been widely used in journalistic blitzkrieg against rosiglita- paradigm for the development of drugs to Europe. A new drug application for acar- zone and the FDA (7). A congressional treat type 2 diabetes. bose had been rejected by the Food and committee hearing took place even before Drug Administration (FDA) 3 years ear- the article appeared in print. But an FDA lier, but in 1995 the regulatory environ- advisory committee concluded by a vote Lessons from the Avandia ment had become friendlier, owing in of 22 to 1 on 30 July 2007 that the evi- controversy large part to the finding of the Diabetes dence against rosiglitazone was insuffi- Control and Complications Trial (DCCT) cient to recommend that it be withdrawn Reducing A1C may not be enough. The that aggressive treatment of hyperglyce- from the market. DCCT and UKPDS (UK Prospective Dia- mia prevented or delayed complications The Avandia affair undermined the betes Study) showed that intensive treat- of diabetes (3). Approval of the thiazo- confidence that patients have in the drugs ment to lower A1C reduced the risk of lidinedione (TZD) and the they take and in the physicians who pre- nephropathy, retinopathy, and neuropa- non- secretagog repa- scribe those drugs. It cast further doubt thy in patients with type 1 (3) and type 2 glinide followed soon after, along with ad- on the FDA’s ability to protect patients diabetes (11,12). Clinical benefit has also ditional members of the newly approved from harm. One hopes that the delibera- been reported for acarbose (13,14). Based classes. tions of the advisory committee will put on the DCCT, the FDA accepted A1C as a In 2005, the uninterrupted succes- the issues into perspective. But the FDA surrogate marker for approval of new sion of new drug approvals came to an should not ignore the perceived short- drugs to treat diabetes. The standards of end. Muraglitazar was a dual peroxisome comings in the regulatory process that al- approval were discussed at an FDA advi- proliferator–activated receptor agonist. It lowed the Avandia affair to get out of sory committee meeting in March 1998. was different chemically from the TZDs hand. Although a guidance to the industry for and fibrates but was believed to combine Even before the Avandia affair, Dr. the development of drugs to treat diabetes ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● was never issued by the FDA, the main points of the draft guidance discussed by From the Division of Endocrinology and Metabolism, Food and Drug Administration, Silver Spring, Mary- the advisory committee were largely in- land. Address correspondence and reprint requests to Robert I. Misbin, MD, 711 Schindler Dr., Silver Spring, corporated into the standards used by the MD 20903. E-mail: [email protected]. European Agency for the Evaluation of Received and accepted for publication 1 October 2007. Medicinal Products (EMEA), which have The opinions expressed in this article represent those of the author and do not necessarily reflect the been posted on the EMEA Web site since official position of the Food and Drug Administration. Abbreviations: ADOPT, A Diabetes Outcome Progression Trial; DCCT, Diabetes Control and Compli- 2002. cations Trial; FDA, Food and Drug Administration; TZD, . In contrast to older agents, use of A table elsewhere in this issue shows conventional and Syste`me International (SI) units and conversion TZDs has not been shown to decrease the factors for many substances. risk of the complications of diabetes. DOI: 10.2337/dc07-1908 Clinical benefit was presumed to follow as © 2007 by the American Diabetes Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby a consequence of reduction in A1C. In the marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. absence of direct evidence of benefit, the

DIABETES CARE, VOLUME 30, NUMBER 12, DECEMBER 2007 3141 Lessons from Avandia: a new paradigm report alleging increased risk of myocar- trial in which the new agent is added to insulin. Indications for monotherapy and dial ischemia with rosiglitazone was par- background therapy. The problem with use with followed. Troglita- ticularly unsettling. this design is that any benefit observed zone was never labeled to be added to Outcome trials need to be considered. with the new drug can be attributed to metformin monotherapy. By contrast, the The anomaly that muraglitazar and ros- better control of hyperglycemia per se initial approval for rosiglitazone was for iglitazone decreased glucose levels but rather than to a specific action of the new add-on to metformin monotherapy. Thus, appeared to increase the risk of cardiac drug. Thus, the PROactive (PROspective the path chosen for the development of ros- ischemia has led Psaty and Furburg (6) to in macroVas- iglitazone seems largely to have been driven call for large, long-term randomized clin- cular Events) study largely confirmed the by a desire to fulfill a void in the market ical trials to be completed as soon as pos- finding from the UK PDS that lowering (TZD plus metformin) rather than by differ- sible after approval of a new antidiabetes blood pressure and glucose levels re- ences in pharmacology between troglita- agent to identify the health benefits and duced the risk of complications of diabe- zone and rosiglitazone. risk of the new treatment. These authors tes (19). An alternative approach is to That sponsors seek to develop new have previously cautioned against the un- compare two treatment regimens be- drugs to fill a marketing niche can be critical acceptance of surrogate end points lieved to have equivalent effectiveness for partially attributed to the FDA. Accord- (15) and have recently pointed out that control of hyperglycemia. A problem here ing to the standards currently employed, the American Diabetes Association ac- is that there is no gold standard for com- clinical trials are needed for each of the knowledges that lowering A1C to prevent parison. In the RECORD (Rosiglitazone situations in which the drugs will be used: macrovascular disease is based on epide- Evaluated for Cardiac Outcomes and Reg- monotherapy and combinations with miological studies rather than controlled ulation of Glycaemia in Diabetes) trial, metformin, sulfonylureas, insulin, etc. clinical trials (16). rosiglitazone plus metformin was com- This approach needs to be reconsidered. Although I recognize the desirability pared with a sulfonylurea plus met- There are no examples of approved drugs of long-term outcome trials, I have little formin. But there is no body of evidence that were effective as monotherapy or in doubt that hyperglycemia per se is harm- that sulfonylureas plus metformin reduce combination with metformin but were ful. Requiring that an outcomes trial be cardiovascular end points. Furthermore, not effective in combination with other completed before approval would delay good patient care can be expected to re- agents. Thus, requiring efficacy trials for the approval of new drugs. In my judg- duce statistical power because the adverse each situation appears to be unnecessary. ment, approval of drugs to treat type 2 event of interest is less likely to occur than By contrast, safety issues have emerged in diabetes should continue to be based on previously thought. some situations but not others. For exam- change in A1C, but a dedicated safety trial As the medical officer at the FDA who ple, congestive emerged as a should be completed before approval. initially reviewed the Avandia applica- problem in the trial of rosiglitazone with Additional safety data should accrue from tion, I recommended that a postmarket- insulin but not in trials of rosiglitazone extension of pivotal trials for 2 years be- ing safety trial be a condition of approval. monotherapy. It should also be noted that yond the date of approval. This recommendation was based on an it is problematic to evaluate the efficacy of It may be appropriate for the FDA to imbalance of cardiac ischemic events, a new drug in insulin-treated patients be- require a commitment to begin an out- weight gain, and lipid alterations in con- cause of the need to adjust the dose of comes trial as a condition of approval. But trolled trials of 6–12 months’ duration insulin based on changes in glycemia. For this requirement should be driven by (20). That a safety trial was not performed these reasons, combination trials with in- safety concerns related to the drug in was noted by Congressman Waxman to sulin should be structured to evaluate question. It is unreasonable to expect that be a failure of FDA management (21). On safety. a single drug manufacturer should be ex- the other hand, the postmarketing trial pected to bear the burden of answering that was performed, A Diabetes Outcome New paradigm for the development fundamental questions such as the nature Progression Trial (ADOPT), has provided of drugs to treat type 2 diabetes of the relationship between diabetes and useful safety data even though its primary heart disease. The ACCORD (Action to objective was to assess durability of effi- Trial design. Approval of new drugs to Control Cardiovascular Risk in Diabetes) cacy. Troglitazone had been removed treat type 2 diabetes should recognize the trial, funded by the National Institutes of from the market because of an unaccept- recommendations of the American Diabe- Health, is attempting to address such ably high risk of liver failure (22). But tes Association that patients with type 2 issues. ADOPT showed that chronic use of ros- diabetes should generally be treated with Performance of a successful outcomes iglitazone was safe to the liver. Of partic- metformin at the time of diagnosis (25). trial presupposes the knowledge of what ular interest is the finding from ADOPT Implementation of this recommendation outcomes need to be measured, what the that rosiglitazone increases the risk of will make it difficult to conduct placebo- population of interest is, how frequently fracture in postmenopausal women (23). controlled monotherapy trials. Drug- those outcomes occur in the population The same was found for pioglitazone in naı¨ve patients will become increasingly to be tested, and what the appropriate PROactive, a long-term study designed to scarce and ethical considerations prevent comparators are. There are two basic ap- examine cardiac effects (24). The in- them from remaining untreated. For these proaches to this type of trial. The manu- creased risk of fracture was unexpected reasons, the period of placebo-only treat- facturers of pioglitazone and rosiglitazone and illustrates that a postmarketing safety ment should be 8–12 weeks (Table 1, each employed different approaches to study should cast a broad net. trial A). The pivotal monotherapy trial measuring long-term outcomes. Neither Combination therapy trials should be should be a 6- to 12-month comparison has been successful (17,18). reevaluated. Troglitazone was initially of the new drug, drug X, to metformin One approach is a placebo-controlled approved to be used in combination with (trial B). A second pivotal trial (trial C)

3142 DIABETES CARE, VOLUME 30, NUMBER 12, DECEMBER 2007 Misbin

Table 1—Paradigm for development of new drugs to treat type 2 diabetes end point should be reports of serious ad- Suggested pivotal trials verse events. A secondary end point, Trial A 8- to 12-week monotherapy trial in treatment-naïve change in insulin dose, should assess ef- patients to study effects of three of more doses of ficacy. Subset analyses should be based drug X versus placebo. on background oral antidiabetes treat- Primary end point can be A1C, fasting, ment. In some circumstances, enrich- postprandial, or mean daily glucose depending ment with patients using TZDs or other on the circumstances. oral agents may be appropriate. Trial B 6- to 12-month active comparator trial in patients A safety trial has not previously been to study effects of two or more doses of drug X required before approval of oral antidia- versus metformin. betes agents currently marketed in the Primary end point is A1C. Events related to cardiac U.S. However, concerns about vascular ischemia should be adjudicated. events in postpartum women taking bro- Trial C 6- to 12-month add-on trial in patients on mocriptine led the FDA to require that a metformin to study effects of two or more doses safety trial be completed before approval of drug X versus placebo. of Cycloset. Details of this trial have been Primary end point is A1C. published (26). Trial D 6- to 12-month add-on trial in patients on insulin Trials A–D should provide the data to study effects of one or more doses of new drug on efficacy and safety to form the basis for versus placebo. Dose of insulin is titrated a new drug application. Extensions of according to the standard of practice. these trials should continue during the Primary end point is reporting of serious adverse time it takes the FDA to review the appli- events. Events related to cardiac ischemia should cation (10 months for a standard review be adjudicated. and 6 months for a priority review). As Suggested extensions to pivotal trials already required, a safety update on on- (at least 2 years beyond approval) going trials should be submitted to the Trial A Patients on drug X are continued on drug X. FDA before regulatory action. Extensions Patients on placebo are started on metformin. of these trials for an additional 2 years Rescue due to hyperglycemia with insulin should be required for approval. secretagogue, insulin sensitizer, or insulin as The extension trials will provide long- appropriate depending on the nature of drug X. term comparisons of drug X to metformin Events related to cardiac ischemia should be (trials A and B), metformin plus drug X adjudicated. versus metformin plus standard therapy Trial B Patients on drug X are continued on drug X. (trial C), and insulin plus drug X versus Patients on metformin are continued on insulin plus standard therapy (trial D). metformin. New trials may be required to begin after Rescue with insulin secretagogue, insulin sensitizer, approval to examine safety issues that or insulin as appropriate depending on the emerged during the FDA review. nature of drug X. Approval. Drug X should be approved if Trial C Patients on drug X are continued on drug X. its efficacy is superior to that of placebo Patients on placebo are started on insulin (trials A and C), noninferior to metformin secretagog or insulin sensitizer as appropriate (trial B), noninferior to standard therapy depending on the nature of drug X. with respect to serious adverse events Rescue with insulin. (trial D), and does not have other serious Trial D Patients on drug X are continued on drug X. safety problems. It might still be appro- Patients on placebo are continued on placebo. priate to approve drug X even if it is less Insulin is given as needed for glycemic control. effective than metformin, particularly if it Patients who cannot achieve adequate control are represents a novel mechanism of action withdrawn. and raises no safety concerns. The FDA does not have the authority to require that a new drug be superior to should compare drug X to placebo in pa- more doses of drug X (including the maxi- existing drugs. On the other hand, the tients who are receiving metformin but mal dose to be marketed) with placebo in FDA is not required to have proof that a whose hyperglycemia is not adequately patients who are receiving insulin with or new drug is unsafe to deny approval and controlled. without oral agents. The insulin dose should set a high standard for drugs that A safety trial that does not exclude pa- should be adjusted to account for the activ- offer no advantage over existing therapy. tients who may be at risk of adverse events ity of drug X such that there be little differ- Even for a novel agent, a few cases of a rare should be completed before approval. As ence between the arms with respect to but life-threatening event, such as agran- illustrated by the trials with rosiglitazone, change in glucose levels. Under these cir- ulocytosis, may be sufficient to prevent cardiac adverse events are more likely to be cumstances, safety issues that emerge can approval. found in insulin-treated patients. Trial D in be attributed to drug X and not to differ- Labeling. New drugs should be labeled the proposed paradigm compares one or ences in glycemic control. The primary for treatment of hyperglycemia in patients

DIABETES CARE, VOLUME 30, NUMBER 12, DECEMBER 2007 3143 Lessons from Avandia: a new paradigm with type 2 diabetes, but no claim for re- and progression of long-term complica- D, Fleming TR, Wagner EH: Surrogate ducing the risk of complications should tions of insulin-dependent diabetes mel- end points, health outcomes, and the be allowed unless directly supported by litus. N Engl J Med 329:977–986, 1993 drug-approval process for the treatment data. There should be no distinction be- 4. Golden J: Muraglitazar safety analysis. of risk factors for cardiovascular disease. tween first- and second-line therapy or Presented at the Food and Drug Adminis- JAMA 282:786–790, 1999 tration Metabolic and Endocrine Advisory between monotherapy and combination 16. Psaty BM, Furberg CD: Rosiglitazone and Committee Meeting, 9 September 2005, the FDA. N Engl J Med. 29 August 2007 therapy. Data from pivotal trials should Silver Spring, Maryland [Epub] (DOI: 10.1056/NEJMc076347) be shown in the label to provide insight 5. Nissen SE, Wolski K: Effect of rosiglita- 17. Nathan DM: Rosiglitazone and cardiotox- into what clinical effects can be expected zone on the risk of myocardial infarction icity: weighing the evidence. N Engl J Med under what conditions. Gaps, uncertain- and death from cardiovascular causes. 357:64–66, 2007 ties, or inconsistencies should be ac- N Engl J Med 356:2467–2471, 2007 18. Skyler JS: PROactive: a sad tale of inap- knowledged, such as lack of information 6. Psaty BM, Furberg CD: Rosiglitazone and propriate analysis and unjustified inter- in certain populations. In some circum- cardiovascular risk. N Engl J Med 356: pretation. Clinical Diabetes 24:63–64, stances, sponsors should be required to 2522–2524, 2007 2006 perform postmarketing trials to address 7. Bloomgarden ZT: The Avandia debate. 19. Fonseca V, Jawa A, Asnani S: Commen- Diabetes Care 30:2401–2410, 2007 remaining issues. Failure to complete tary: the PROactive Study: the glass is half 8. Nathan DM: Finding new treatments of full. J Clin Endocrinol Metab 91:25–27, these trials should be posted on the FDA diabetes: how many, how fast. . . how Web site. From the clinical review and 2006 good? N Engl J Med 356:437–440, 2007 20. Food and Drug Administration: Medical action letter posted by the FDA on its Web 9. Misbin RI: A possible drug fix? Washing- Officers Review of Avandia. Rockville, MD, site, professional organizations can de- ton Post, 24 Aug 1998: Sect. A, p. 18 Food and Drug Administration, 1999 (ap- velop recommendations about how a new 10. Misbin RI: Evaluating the safety of diabe- plication no. 021071) drug should be used in relation to other tes drugs: perspective of a Food and Drug 21. Committee on Oversight and Govern- Administration insider. Diabetes Care 28: approved agents. ment Reform: Hearing on FDA’s Role in 2573–2576, 2005 Evaluating Safety of Avandia [article 11. UK Prospective Diabetes Study (UKPDS) online], 2007. Available from http:// Looking ahead Group: Intensive blood glucose control oversight.house.gov/story.asp?IDϭ1325. Table 1 is a summary of a suggested par- with sulfonylureas or insulin compared Accessed 6 June 2007 adigm for clinical trials to develop new with conventional treatment and risks of drugs for type 2 diabetes. With better complications in patients with type 2 di- 22. Misbin RI: Troglitazone: associated he- knowledge and experience, this paradigm abetes (UKPDS 33). Lancet 352:837–853, patic failure. Ann Intern Med 130:330, 1998 1999 may need to change. But the goal should 23. Kahn SE, Haffner SM, Heise MA Herman remain the same. New drugs should con- 12. UK Prospective Diabetes Study (UKPDS) Group: Intensive blood glucose control WH, Holman RR, Jones NP, Kravatz BG, tinue to be developed to reduce the global Lachman JM, O’Neill MC, Zinman B, Vib- burden of diabetes and its complications. with metformin complications in over- weight patients with type 2 diabetes (UK- erti G: Glycemic durability of rosiglita- Regulators should set high standards but PDS 34). Lancet 352:854–865, 1998 zone, metformin or glyburide mono- should also be pragmatic. Fear of uncer- 13. Chiasson JL, Josse RG, Gomis R, Hanefeld therapy. N Engl J Med 355:2427–2443, tainty should not be allowed to stifle M, Karasik A, Laakso M; STOP-NIDDM 2006 innovation. Trial Research Group: Acarbose treatment 24. Food and Drug Administration: Medical and the risk of cardiovascular disease and Officers Review of Actos. Rockville, MD, hypertension in patients with impaired Food and Drug Administration, 2007 (ap- References glucose tolerance. JAMA 290:486–494, plication no. 021073) 1. Misbin RI: Phenformin associated lactic 2003 25. American Diabetes Association: Stan- acidosis: pathogenisis and treatment. Ann 14. Hanefield M, Cagatay M, Petrowitsch T, dards of medical care in diabetes—2007 Intern Med 87:591–595, 1977 Neuser D, Petzinna D, Rupp M: Acarbose (Position Statement). Diabetes Care 30 2. Misbin RI: The phantom of lactic acidosis reduces the risk of myocardial infarction (Suppl. 1):S4–S41, 2007 due to metformin in patients with type 2 in type 2 diabetic patients: meta-analysis 26. Scranton RE, Gaziano JM, Rutty D, Cin- diabetes. Diabetes Care 27:1791–1793, of seven long-term studies. Eur Heart J 25: cotta MA: A randomized, double-blind, 2004 10–16, 2004 placebo-controlled trial of safety and tol- 3. Diabetes Control and Complications Trial 15. Psaty BM, Weiss NS, Furburg CSD, Ko- erability during treatment of type 2 diabe- Research Group: The effect of intensive epsell TD, Siscovick DS, Rosendaal FR, tes with either Cycloset or placebo. BMC treatment of diabetes on the development Smith NL, Heckbert SR, Kaplan RC, Lin Endocr Disord 7:3, 2007

3144 DIABETES CARE, VOLUME 30, NUMBER 12, DECEMBER 2007