Toward a Rational Regulatory Policy for Development of Drugs to Treat Diabetes Mellitus

1. Toward a Rational Regulatory Policy for Development of Drugs to Treat Diabetes Mellitus Steven E. Nissen MD Chairman, Department of Cardiovascular Medicine Cleveland Clinic Disclosure Consulting: Many pharmaceutical companies Clinical Trials:AstraZeneca, Eli Lilly, Takeda, Sankyo, Sanofi-Aventis, Roche, and Pfizer. Companies are directed to pay any honoraria, speaking orconsulting fees directly to charity so that neither income nor tax deduction is received.

2. Why are We Here? • The ACCORD trial demonstrated that a drug regimen designed to lower blood glucose is capable of increasing mortality in diabetic patients. • Multiple rosiglitazone meta-analyses of CV outcomes showed improved glycemic control, but an increase in myocardial ischemic events. • Many agents to treat diabetes have failed duringdevelopment, some to cardiotoxicity. • No robust cardiovascular outcomes data existfor any current diabetes therapies.

3. The Problem: Insufficient Clinical Trial Data • In the absence of randomized CV outcomes trials, we are left with unsatisfactory methods to assess benefit and risk. • These include meta-analyses or post hoc data “dredging” of randomized trials not designed to determine the benefits or risks of specific therapies. • This was particularly evident at the recent ADAmeeting with attempts to determine the sourceof the excess mortality in ACCORD. • No amount of “torturing” of the data will enable riskassessment when specific drugs usage notrandomized.

4. ACCORD Bottom line: A regimen in which more patients received repaglinide, rosiglitazone, insulin and/or an a glucosidase inhibitor showed increased mortality risk.

5. More than 50 years after the initial introduction of oral anti-diabetic drugs…. Although cardiovascular disease is the cause of death in 75% of diabetics, there exist no well-designed, adequately-powered comparative effectiveness trials evaluating macrovascular outcomes for diabetes drugs

7. Level of Evidence: Comparative Effectiveness Trials of Diabetes Drugs

8. The Knowledge Gap The absence of information on macrovascular effects of diabetes therapies is the unfortunate consequence of current regulatory policy that emphasizes the importance of glucose lowering, not health outcomes, as a therapeutic goal.

9. This Approach Has Created a New Disorder Glucocentricity glu-co-cen-tricity |ˈgloōkō senˈtrisitē noun The irrational belief that lowering blood sugar using virtually any pharmacological means will produce a reliable reduction in adverse outcomes

10. Consequences of a “Glucocentric” Regulatory Approach to Diabetes Drugs • Pre-approval studies focus on demonstratingmaximal glucose lowering effects. • Patients are selected with relatively high HbA1clevels to enhance apparent “efficacy.” • Studies seek “bragging” rights- “my drug lowersblood sugar more than your drug.” • Patients at high CV risk are deliberately avoided.Why take a chance of an adverse safety signal?

11. Consequences of a “Glucocentric” Regulatory Approach to Diabetes Drugs • When safety signals arise, physicians stampede to the newest diabetes therapies (for which we know the least about safety). • Example: After the rosiglitazone concernsemerged, the fastest growing diabetes class isa new class, a DPP-IV inhibitor (sitagliptin). • However, this agent has limited glucose lowering efficacy (about half the effect of establishedtherapies) and virtually no long-term safety data.

12. The Principal Dilemma How do we balance the need to bring new diabetes agents to patients in a timely fashion. vs. The need for more robust outcome datathat inform physicians of howto use these drugs safely and effectively

13. Proposal: A Rational Approach to Approval of New Diabetes Drugs Requiring a large CV outcomes trial prior to approval is undesirable, because this approach would delay new diabetes therapies by 5-7 years. A Compromise with Two Components • A pre-approval set of clinical trials designedto rule out a high level of CV risk. • A large randomized outcomes trial that must be underway at the time of approval.

14. Pre-Approval Development Program • Pre-approval trials of sufficient size andduration to rule out a HR of 2.0 for MACE(upper 95% confidence interval). • Required: Pre-specified pooling of CVoutcomes in all trials with adjudication by anindependent clinical endpoints committee. • Useful: At least one study in patients at highCV risk, perhaps 1000 patients for 1-2 years.

15. Pre-Approval Cardiovascular Studies MACE = major adverse CV events including death, MI, stroke, and hospitalization for revascularization

16. 87 Events to Exclude Upper 95% CI of 2.0 0.98 1.23 0.67 1.56 0.4 0.5 0.67 1.0 1.5 2.0 2.5 Active better Control better

17. Implications of This Regulatory Approach • Positives: • Encourages sponsors to include patientswith higher levels of cardiovascular risk. • Provides more reliable pre-approval databy adjudication of CV events for pooled trials • Negatives: • Modestly slows development programs,thereby delaying introduction of newdiabetes medications by 6-12 months

18. Step II in Approval: An AdequatelyPowered Cardiovascular Outcomes Trial • Assuming no evidence for excess CV risk (upper95% CI of HR <2.0), a new diabetes drug would beapprovable based on glucose-lowering efficacy… • If an adequately-powered ongoing CV outcomes trialis underway (already enrolling patients). • This outcomes study should also address any otherongoing safety issues (renal, fractures, skin) • This policy is a compromise designed to balancespeedy approval with the need to promptly obtainevidence of benefit (or risk).

19. Ongoing CV Study at Time of Approval MACE Endpoint: Death, MI, stroke or ACS hospitalizationalpha=0.05 (two tailed) and 80% power

20. Promises Not Kept: Phase IV Commitments † Public Citizen data * FDA report to Congress

21. Quality of Cardiovascular Outcomes Trial: A Critical Issue

22. Those who cannot remember the past are condemned to repeat it. George Santayana, a Spanish-born American author in 1905

23. Dual PPARs: A Promising Approach • Both hyperlipidemia and insulin resistance appear to promote atherosclerosis in diabetics. • Accordingly, pharmaceutical companies have sought to develop dual  and  PPAR agonists combining: • Fibrate-like effects, raising HDL and lowering triglycerides • TZD-like effects, improving insulin sensitivity thereby lowering blood sugar.

24. Muraglitazar Advisory Panel(Sept 8, 2005) The first dual PPAR to reachthis stage of the approval process

25. Muraglitazar Development Program

26. HbA1c: Change from Baseline (%) Placebo Mura 2.5mg Mura 5mg Open 5mg Mean baseline 8.0 8.0 7.9 10.7 -0.32 -1.05 -1.23 -2.62

27. Lipid Parameters: Change from Baseline (%) Triglycerides HDL-cholesterol LDL-cholesterol 16 10 2 -2 -18 -27

28. The Panel Meeting September 5, 2005 • Although there appears to be an higher incidenceof major adverse cardiovascular events, the sponsor argues there there is: • “Lack of biologic plausibility for CV risk with muraglitazar based on”: • Beneficial effects on markers of CV risk • Broad diversity among reported CV events • No increase in CV events with increasing dose • Absence of off-target CV toxicity in non-clinicalor clinical studies

29. The panel votes 8:1 to approve muraglitazar as monotherapy and 7:2 to approve use with metformin. They vote 7:2 against approval in combination with sulfonylureas, requesting more data.

30. Six Weeks after the Advisory Panel

31. Individual Components of Composite

32. Commonly Used CV Composite Endpoints

33. Muraglitazar: Epilogue • The FDA issues an “approvable” letter requesting additional CV safety data. • After ongoing extension trials confirm the CV hazard, all development of the drug halted. • However, a risky agent came close to approval. • A clear standard requiring an upper CI <2.0for the HR for CV events would have precluded the necessity for an advisory panel.

34. The slippery slope of surrogate endpoints in diabetes drug development • Ezetimibe was approved to treathyperlipidemia on the basis of a reductionin LDL-C averaging 16-18% • What should we do with a diabetes drugthat lowers blood sugar (HbA1c by 1.1%),but increases LDL-C by 16-18%? If a 16-18% reduction in LDL-C is sufficient to demonstrate benefit, what inference should we draw when a drug increases LDL-C by a comparable amount?

35. Rosiglitazone

36. Rosiglitazone Approval Package (1999) “Weight/lipids – Patients treated with RSG manifest undesirablechanges in weight and lipids…”† *Statistical Review Joy D. Mele M.S. Mathematical Startistician † Medical Review Robert I Misbin MD.

37. Rosiglitazone Advisory Panel: CV Events FDA Reviewer: “A post-marketing study to evaluate long-term safety of RSG should be required for approval.”

38. What Happened to Mandated SafetyStudy of Rosiglitazone? The Adopt Trial • Not a safety study, a “marketing” study, designedto show greater durability of glucose loweringwith rosiglitazone. • Cardiovascular events collected in a haphazardfashion (not adjudicated!) • Because of the LDL-raising effect of rosiglitazone,more patients received statins (p<0.01). • HR 1.33 for myocardial infacrtion (0.80-2.21).The pre-approval signal never goes away!

39. Rosiglitazone Post-marketing Studies • Many small, short term “marketing” studies to show glycemia reduction in various populations. • No well-designed randomized trials to measurehealth outcomes (RECORD underpowered). • By 2007, 42 trials were completed with 14,237patients. • FDA analysis in July 2007: Odds ratio formyocardial ischemia 1.4 (95% CI 1.1-1.8) • The pre-approval signal never goes away!

40. Rosiglitazone in Retrospect • If an upper CI <2.0 for the HR for CV events had been required for approval in 1999, rosigltazone would not have been approved without more safety data. • If a large, well-powered outcomes trial had been mandated in 199, we would not have to wait until 2014 (15 years after approval) to determine if this drug is safe or not.

41. PPARs: a Special Case • At least 50 IND’s filed following last approvalof a TZD, nearly all terminated due to toxicity. • Toxicities observed in animals are also evident clinically - cardiac, skeletal muscle, renal, bone marrow. (El Hage) • Most development programs terminated withoutany publication of the toxicities encountered. • PPARs all activate different genes and must beconsidered individually- this is NOT a “drugclass”

42. Gene Expression with PPAR g Agents Pioglitazone 12 2 0 Troglitazone Rosiglitazone 13 5 10 23 2 Matrixmetalloproteinase 12 5 26 Pooled TZDs Bioinformatics. 2004 Nov 22;20(17):3108-27. Epub 2004 Jun 24.

43. Diabetes, Glycemia and CV Disease • The goal of merely lowering blood glucose levelsin diabetes is too simplistic. We must reduce the complications of diabetes, including CV disease. • With respect to CV disease it appears importanthow you lower blood sugar as well as how much. • Diabetes drugs, even within the same “class” may yield dramatically different CV outcomes. • Clinical outcomes trials comparing alternative diabetes therapies are essential to determine the optimal approach to prevent CV morbidity-mortality.

44. Two Components: A new paradigm for Diabetes Drug development • Sufficient pre-approval data to exclude a excess hazard of CV events (an upper CI for the hazard ratio not to exceed 2.0. • A robust post-approval outcome programto provide data in a timely fashion (anongoing outcomes trial at the time of approval)