Lessons From the Avandia Controversy

Lessons from the Avandia controversy

Reducing A1C may not be enough.

The DCCT and UKPDS (UK Prospective Diabetes Study) showed that intensive treatment to lower A1C reduced the risk of nephropathy, retinopathy, and neuropathy in patients with type 1 (3) and type 2 diabetes (11,12). Clinical benefit has also been reported for acarbose (13,14). Based on the DCCT, the FDA accepted A1C as a surrogate marker for approval of new drugs to treat diabetes. The standards of approval were discussed at an FDA advisory committee meeting in March 1998. Although a guidance to the industry for the development of drugs to treat diabetes was never issued by the FDA, the main points of the draft guidance discussed by the advisory committee were largely incorporated into the standards used by the European Agency for the Evaluation of Medicinal Products (EMEA), which have been posted on the EMEA Web site since 2002.

In contrast to older agents, use of TZDs has not been shown to decrease the risk of the complications of diabetes. Clinical benefit was presumed to follow as a consequence of reduction in A1C. In the absence of direct evidence of benefit, the report alleging increased risk of myocardial ischemia with rosiglitazone was particularly unsettling.

Outcome trials need to be considered.

The anomaly that muraglitazar and rosiglitazone decreased glucose levels but appeared to increase the risk of cardiac ischemia has led Psaty and Furburg (6) to call for large, long-term randomized clinical trials to be completed as soon as possible after approval of a new antidiabetes agent to identify the health benefits and risk of the new treatment. These authors have previously cautioned against the uncritical acceptance of surrogate end points (15) and have recently pointed out that the American Diabetes Association acknowledges that lowering A1C to prevent macrovascular disease is based on epidemiological studies rather than controlled clinical trials (16).

Although I recognize the desirability of long-term outcome trials, I have little doubt that hyperglycemia per se is harmful. Requiring that an outcomes trial be completed before approval would delay the approval of new drugs. In my judgment, approval of drugs to treat type 2 diabetes should continue to be based on change in A1C, but a dedicated safety trial should be completed before approval. Additional safety data should accrue from extension of pivotal trials for 2 years beyond the date of approval.

It may be appropriate for the FDA to require a commitment to begin an outcomes trial as a condition of approval. But this requirement should be driven by safety concerns related to the drug in question. It is unreasonable to expect that a single drug manufacturer should be expected to bear the burden of answering fundamental questions such as the nature of the relationship between diabetes and heart disease. The ACCORD (Action to Control Cardiovascular Risk in Diabetes) trial, funded by the National Institutes of Health, is attempting to address such issues.

Performance of a successful outcomes trial presupposes the knowledge of what outcomes need to be measured, what the population of interest is, how frequently those outcomes occur in the population to be tested, and what the appropriate comparators are. There are two basic approaches to this type of trial. The manufacturers of pioglitazone and rosiglitazone each employed different approaches to measuring long-term outcomes. Neither has been successful (17,18).

One approach is a placebo-controlled trial in which the new agent is added to background therapy. The problem with this design is that any benefit observed with the new drug can be attributed to better control of hyperglycemia per se rather than to a specific action of the new drug. Thus, the PROactive (PROspective pioglitAzone Clinical Trial in macroVascular Events) study largely confirmed the finding from the UK PDS that lowering blood pressure and glucose levels reduced the risk of complications of diabetes (19). An alternative approach is to compare two treatment regimens believed to have equivalent effectiveness for control of hyperglycemia. A problem here is that there is no gold standard for comparison. In the RECORD (Rosiglitazone Evaluated for Cardiac Outcomes and Regulation of Glycaemia in Diabetes) trial, rosiglitazone plus metformin was compared with a sulfonylurea plus metformin. But there is no body of evidence that sulfonylureas plus metformin reduce cardiovascular end points. Furthermore, good patient care can be expected to reduce statistical power because the adverse event of interest is less likely to occur than previously thought.

As the medical officer at the FDA who initially reviewed the Avandia application, I recommended that a postmarketing safety trial be a condition of approval. This recommendation was based on an imbalance of cardiac ischemic events, weight gain, and lipid alterations in controlled trials of 6–12 months’ duration (20). That a safety trial was not performed was noted by Congressman Waxman to be a failure of FDA management (21). On the other hand, the postmarketing trial that was performed, A Diabetes Outcome Progression Trial (ADOPT), has provided useful safety data even though its primary objective was to assess durability of efficacy. Troglitazone had been removed from the market because of an unacceptably high risk of liver failure (22). But ADOPT showed that chronic use of rosiglitazone was safe to the liver. Of particular interest is the finding from ADOPT that rosiglitazone increases the risk of fracture in postmenopausal women (23). The same was found for pioglitazone in PROactive, a long-term study designed to examine cardiac effects (24). The increased risk of fracture was unexpected and illustrates that a postmarketing safety study should cast a broad net.

Combination therapy trials should be reevaluated.

Troglitazone was initially approved to be used in combination with insulin. Indications for monotherapy and use with sulfonylureas followed. Troglitazone was never labeled to be added to metformin monotherapy. By contrast, the initial approval for rosiglitazone was for add-on to metformin monotherapy. Thus, the path chosen for the development of rosiglitazone seems largely to have been driven by a desire to fulfill a void in the market (TZD plus metformin) rather than by differences in pharmacology between troglitazone and rosiglitazone.

That sponsors seek to develop new drugs to fill a marketing niche can be partially attributed to the FDA. According to the standards currently employed, clinical trials are needed for each of the situations in which the drugs will be used: monotherapy and combinations with metformin, sulfonylureas, insulin, etc. This approach needs to be reconsidered. There are no examples of approved drugs that were effective as monotherapy or in combination with metformin but were not effective in combination with other agents. Thus, requiring efficacy trials for each situation appears to be unnecessary. By contrast, safety issues have emerged in some situations but not others. For example, congestive heart failure emerged as a problem in the trial of rosiglitazone with insulin but not in trials of rosiglitazone monotherapy. It should also be noted that it is problematic to evaluate the efficacy of a new drug in insulin-treated patients because of the need to adjust the dose of insulin based on changes in glycemia. For these reasons, combination trials with insulin should be structured to evaluate safety.

New paradigm for the development of drugs to treat type 2 diabetes

Trial design.

Approval of new drugs to treat type 2 diabetes should recognize the recommendations of the American Diabetes Association that patients with type 2 diabetes should generally be treated with metformin at the time of diagnosis (25). Implementation of this recommendation will make it difficult to conduct placebo-controlled monotherapy trials. Drug-naïve patients will become increasingly scarce and ethical considerations prevent them from remaining untreated. For these reasons, the period of placebo-only treatment should be 8–12 weeks (Table 1, trial A). The pivotal monotherapy trial should be a 6- to 12-month comparison of the new drug, drug X, to metformin (trial B). A second pivotal trial (trial C) should compare drug X to placebo in patients who are receiving metformin but whose hyperglycemia is not adequately controlled.

A safety trial that does not exclude patients who may be at risk of adverse events should be completed before approval. As illustrated by the trials with rosiglitazone, cardiac adverse events are more likely to be found in insulin-treated patients. Trial D in the proposed paradigm compares one or more doses of drug X (including the maximal dose to be marketed) with placebo in patients who are receiving insulin with or without oral agents. The insulin dose should be adjusted to account for the activity of drug X such that there be little difference between the arms with respect to change in glucose levels. Under these circumstances, safety issues that emerge can be attributed to drug X and not to differences in glycemic control. The primary end point should be reports of serious adverse events. A secondary end point, change in insulin dose, should assess efficacy. Subset analyses should be based on background oral antidiabetes treatment. In some circumstances, enrichment with patients using TZDs or other oral agents may be appropriate.

A safety trial has not previously been required before approval of oral antidiabetes agents currently marketed in the U.S. However, concerns about vascular events in postpartum women taking bromocriptine led the FDA to require that a safety trial be completed before approval of Cycloset. Details of this trial have been published (26).

Trials A–D should provide the data on efficacy and safety to form the basis for a new drug application. Extensions of these trials should continue during the time it takes the FDA to review the application (10 months for a standard review and 6 months for a priority review). As already required, a safety update on ongoing trials should be submitted to the FDA before regulatory action. Extensions of these trials for an additional 2 years should be required for approval.

The extension trials will provide long-term comparisons of drug X to metformin (trials A and B), metformin plus drug X versus metformin plus standard therapy (trial C), and insulin plus drug X versus insulin plus standard therapy (trial D). New trials may be required to begin after approval to examine safety issues that emerged during the FDA review.

Approval.

Drug X should be approved if its efficacy is superior to that of placebo (trials A and C), noninferior to metformin (trial B), noninferior to standard therapy with respect to serious adverse events (trial D), and does not have other serious safety problems. It might still be appropriate to approve drug X even if it is less effective than metformin, particularly if it represents a novel mechanism of action and raises no safety concerns.

The FDA does not have the authority to require that a new drug be superior to existing drugs. On the other hand, the FDA is not required to have proof that a new drug is unsafe to deny approval and should set a high standard for drugs that offer no advantage over existing therapy. Even for a novel agent, a few cases of a rare but life-threatening event, such as agranulocytosis, may be sufficient to prevent approval.

Labeling.

New drugs should be labeled for treatment of hyperglycemia in patients with type 2 diabetes, but no claim for reducing the risk of complications should be allowed unless directly supported by data. There should be no distinction between first- and second-line therapy or between monotherapy and combination therapy. Data from pivotal trials should be shown in the label to provide insight into what clinical effects can be expected under what conditions. Gaps, uncertainties, or inconsistencies should be acknowledged, such as lack of information in certain populations. In some circumstances, sponsors should be required to perform postmarketing trials to address remaining issues. Failure to complete these trials should be posted on the FDA Web site. From the clinical review and action letter posted by the FDA on its Web site, professional organizations can develop recommendations about how a new drug should be used in relation to other approved agents.

Looking ahead

Table 1 is a summary of a suggested paradigm for clinical trials to develop new drugs for type 2 diabetes. With better knowledge and experience, this paradigm may need to change. But the goal should remain the same. New drugs should continue to be developed to reduce the global burden of diabetes and its complications. Regulators should set high standards but should also be pragmatic. Fear of uncertainty should not be allowed to stifle innovation.