Sir_Excellence

Copyright 2016 American Medical Association. All rights reserved.

Evaluation of the Cardiovascular Risk of Naltrexone-Bupropion A Study Interrupted Joshua M. Sharfstein, MD; Bruce M. Psaty, MD, PhD

The challenge of assessing the cardiovascular safety of the re- cently approved naltrexone-bupropion combination for treat- ment of obesity originates in the complex issues involving the LIGHT trial, published in this issue of JAMA.1

In 2011, in response to concerns regarding the cardiovas- cular effects of this combination medication, including eleva- tions in blood pressure, the US Food and Drug Administration (FDA) asked the sponsor, Orexigen, to support a definitive safety

study.2 Investigators designed the LIGHT study to randomize approximately 9000 patients

to receive treatment with naltrexone-bupropion vs placebo and follow them up for about 5 years until approximately 400 ma- jor adverse cardiovascular events had occurred. With a nonin- feriority design, the study was powered to exclude a 1.4-fold in- crease in cardiovascular risk associated with the medication.

The LIGHT protocol included a preplanned interim analy- sis after 25% of the major adverse cardiovascular events of in- terest (including cardiovascular death, nonfatal myocardial in- farction, and nonfatal stroke) had been accrued. The goal of this interim analysis was to assess whether naltrexone-bupropion was associated with more than a 2-fold increase in the risk of major cardiovascular events. In most clinical trials, such a re- view is handled in confidence by the data monitoring commit- tee, without disclosure of data to the sponsor or researchers un- less the trial needs to be stopped. In this case, however, for the purpose of initial market approval, the sponsor was also required to share the results of the LIGHT interim analysis with the FDA.

The FDA’s use of interim analyses of ongoing safety stud- ies is relatively new. About 8 years ago, after several medica- tions for diabetes were found to have previously unappreci- ated cardiovascular risks, the agency faced an unappealing choice: either (1) continue to approve drugs based primarily on efficacy studies of glycemic control for diabetes and risk learning about serious cardiovascular risk after approval or (2) delay consideration of new therapies until years of cardio- vascular safety data had been collected.

In 2008, the FDA issued a guidance statement for the de- velopment of drugs for diabetes that reflected a creative com- promise between safety and new drug development.3 For ini- tial approval, the FDA required manufacturers to provide enough data to show that the upper limit of the 95% confidence inter- val for the risk ratio of the risk of cardiovascular events was less than 1.8. Then, as a condition of market approval, sponsors had to conduct a postmarketing trial that, either alone or com- bined with the preapproval data in a meta-analysis, would dem- onstrate that the upper 95% confidence interval for cardiovas-

cular risk was less than 1.3. The new approach accepted the possibility that a medication might meet its initial end point, gain approval, and then fail its final end point—an outcome that might lead to clinical confusion and regulatory uncertainty.

To implement this 2-step safety evaluation more effi- ciently, sponsors of diabetes drug trials asked the FDA for per- mission to handle both parts with a single cardiovascular safety trial. In this model, interim results would be used to secure ap- proval; the trial would continue; and the final results would determine whether the postmarketing requirement had been met. This strategy, however, introduced a new challenge to the process: keeping the interim results confidential.

The 2 primary risks associated with the release of interim results are the potential to compromise the conduct of the trial and the potential to misinterpret the results of the trial.4 Early knowledge of interim results may affect the behavior of spon- sors, investigators, and participants in the trial as well as the actions of physicians and patients in the community. Recruit- ment or retention of trial participants may be jeopardized. If the interim results are perceived to be negative, patients in the study may be inclined to drop out of the trial or stop taking their medications. If interim results are perceived to be positive, and the medication is commercially available, physicians may steer some patients to the therapy rather than to the study. These behaviors are difficult to detect and may jeopardize the integ- rity of the trial. At a public hearing held by the FDA in August 2014, there was broad agreement that, assuming the trial is not being stopped, the clinical community should be able to learn no more about an interim safety analysis than that the pre- specified end point had been met.5 This is essentially the same level of understanding that investigators have when data moni- toring committees permit their trials to continue.

The new approach to cardiovascular safety for diabetes drugs, with its statistically sound design, has increased the number of clinical trials6 and has functioned as planned for several drugs, including saxagliptin7 and alogliptin.8 In both cases, the interim analyses showed a relative risk of less than 1.8 as indicated by the FDA guidance, the companies main- tained the confidentiality of the results, the drugs were ap- proved, and the final studies met their end points.

For the review of naltrexone-bupropion, the FDA agreed to a similar approach for an obesity drug for the first time. Problems first became evident in fall 2013, when the interim analysis was presented to the data monitoring committee. The committee ob- jected to the sponsor’s data access plan on the grounds that “busi- ness interests do not provide a sufficient justification for gaining access to confidential information on efficacy or safety.”9 The com-

Related article page 990

Opinion

EDITORIAL Editorials represent the opinions of the authors and JAMA

and not those of the American Medical Association.

984 JAMA March 8, 2016 Volume 315, Number 10 (Reprinted) jama.com

Copyright 2016 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ by University of New England Library, CARISSA LEE on 09/03/2016

Copyright 2016 American Medical Association. All rights reserved.

mittee asked the sponsor to revise its data access plan, with the result that the company’s February 2014 plan, according to the FDA, stated “that the unblinded team is limited to those individu- als needed to facilitate or manage global regulatory submissions.”9

Nonetheless, by April 2014, the company acknowledged to the FDA that “more than 100” people, including the board of direc- tors, had had access to the unblinded interim results.9 These re- sults included a hazard ratio of 0.59 (95% CI, 0.39-0.90) for ma- jor cardiovascular events. In other words, the interim analysis excluded the 2-fold increase in risk that was the purpose of the review and did so with a point estimate favoring the treatment against placebo with respect to adverse cardiovascular events.

The FDA delayed the decision date for the approval of naltrexone-bupropion to assess the situation. In terms of the data, the agency accepted that the interim analysis excluded a 2-fold increase in risk but did not make any further conclusions. It is well recognized that early results are unstable and may change when the study is completed.10 In terms of the breach of confidenti- ality, the FDA determined that the sponsor’s actions had “the po- tential to compromise the integrity of the LIGHT trial.”9 Because the interim analysis had met the prespecified end point, the FDA approved the medication in September 2014. At the same time, however, the agency required that an entirely new study be con- ducted after approval to exclude an increase in cardiovascular risk of 1.4. The FDA also required, among other stipulations, a study to evaluate cardiac conduction; a clinical trial to evaluate potential interactions between naltrexone-bupropion and other drugs; and clinical trials to evaluate dosing in patients with re- nal or hepatic impairment. The agency “strongly encouraged the applicant to continue the LIGHT trial since it may provide sup- portive evidence when the effects of … [naltrexone-bupropion] on [cardiovascular] outcomes are ultimately reviewed.”9

It was not to be. Based on interim results, the sponsor had filed a patent claiming a cardiovascular benefit for naltrexone- bupropion. In March 2015, the sponsor filed a report with the Securities and Exchange Commission on the patent claiming a “positive effect…on [c ardiovascular] outcomes” that “appears to be unrelated to weight change.”11 The company’s public disclosure of the interim results prompted the LIGHT investigators to terminate the study and issue their own press release. This announcement included unblinded results that covered the second 25% of the outcome data, which were less advantageous to the medication.12

The sponsor’s public statements were highly misleading. The interim analysis did not support the conclusion that the drug had a “positive effect” on cardiovascular outcomes. If the LIGHT study had specified the possibility of stopping at the first interim analysis (and it did not), the appropriate stan- dard for declaring superiority against placebo would have been the O’Brien-Fleming boundary, which is a far stricter thresh- old than an upper 95% confidential interval of less than 1.13 The results did not meet this conventional standard for superior- ity. Moreover, there were multiple signs that the interim LIGHT findings were not reliable; for instance, nearly all of the car- diovascular mortality in the interim analysis occurred after par- ticipants had stopped taking their medications.

In this issue of JAMA, Nissen et al1 present findings from the LIGHT study, including the original prespecified interim

analysis based on 25% of the expected end points, the addi- tional analysis based on 50% of the end points, and a final analy- sis using all available data before the study was stopped, with 64% of originally planned end points. The study was meticu- lously conducted and the findings are carefully reported. The 3 analyses, using prespecified intention-to-treat analysis, found the hazard ratio point estimates for major cardiovascular events to be 0.59, 0.88, and 0.95, respectively, and the upper bound of the 99.7% confidence interval for the most complete data is 1.38. As more data on major adverse cardiovascular events ac- cumulated after the first 25% of the expected outcome, the treat- ment group experienced more strokes, more myocardial infarc- tions, and greater overall mortality than the placebo group. The change in point estimates from 0.59 toward the null and the in- crease in the upper bound of the confidence interval as more data were analyzed clearly illustrate why caution is required in interpreting interim results.

The LIGHT trial and the surrounding events also highlight why strong regulatory oversight is essential in the development and marketing of prescription drugs. The sponsor disregarded the multiple harms caused by serial breaches of confidentiality, treated the academic investigators unfairly, ignored the data monitoring committee, and defied the FDA. Yet instead of fac- ing sanction, the sponsor won approval for its medication, was relieved of the obligation of paying for the rest of the LIGHT study, and gained the ability to market naltrexone-bupropion in the absence of final results of a cardiovascular safety study until the new study is completed, which is expected to be 2022.14

Orexigen is supporting the expense of the new cardiovascular safety trial, which is estimated to be more than $200 million.17

There remains the basic question of whether naltrexone- bupropion is safe for clinical practice in light of data on efficacy. Nissen et al found modest weight loss; at trial completion, there was a mean decrease in body weight of 3.9 kg in the naltrexone- bupropion group and 1.2 kg in the placebo group. According to the FDA’s review, across several studies, about 2 to 3 times more patients achieve 5% loss of body mass index while taking naltrexone-bupropion than patients taking placebo.9 Balanced against this benefit is an unknown cardiovascular risk. Based on the evidence known at the time of approval, LIGHT study par- ticipants experienced less than a 2-fold increase in cardiovas- cular risk. Other than this, the available data are difficult to in- terpret and therefore largely uninformative. The product label now states, “The effect of [naltrexone-bupropion] on cardiovas- cular morbidity and mortality has not been established.”

In addition to the complexity of interpreting the LIGHT study, clinicians should be aware of a number of issues. Car- diovascular risk factors, including hypertension, were largely under control in the study population. As a result, even greater caution in the use of naltrexone-bupropion is necessary for pa- tients at elevated cardiovascular risk. In addition, medication adherence was poor in the trial; at 1 year after randomization, only 37.5% of those in the naltrexone-bupropion group and 26.3% of those in the placebo group were still taking study drug. Given the uncertainty about safety and the long delay before ad- ditional data will become available, the FDA should impose additional restrictions on use—such as a Risk Evaluation and Mitigation Strategy that mandates certain disclosures about the

Editorial Opinion

jama.com (Reprinted) JAMA March 8, 2016 Volume 315, Number 10 985

Copyright 2016 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ by University of New England Library, CARISSA LEE on 09/03/2016

Copyright 2016 American Medical Association. All rights reserved.

uncertainty on cardiovascular risk to counter misinformation disseminated by the sponsor—until more is known.

The conventional reasons for stopping a trial early are convincing demonstration of efficacy, safety, or futility.15

These criteria do not include the justification of Nissen et al that “the ongoing trial was no longer viable after the original sponsor…publicly released interim data.” To the best of our knowledge, disclosure of interim data results has never been used previously as the reason to stop a large clinical trial early. Although the investigators found themselves in an enor- mously difficult position, the decision to stop the trial, which diminished the contributions of the thousands of participants and investigators, was not the only option.

An alternative approach would have been for the data moni- toring committee, with the support of investigators, to release the second 25% of the accrued data in the trial. Dissemination of these data, which were not favorable toward the medication, might have helped undo the public misinformation distributed by the sponsor and help reestablish equipoise. While unconven- tional, the strategy of responding to one disclosure of interim results with another, in this case, might have worked to restore integrity to the LIGHT study and would have avoided setting an unfortunate new precedent for stopping a clinical trial.

In the shadow of LIGHT, the FDA should review its policy of permitting approval based on interim analyses of ongoing safety studies. At a minimum, when a company violates its commitment to confidentiality and the FDA requires a new trial, the agency should delay approval at least until a viable replacement study is being conducted. If the drug has al- ready been approved, the FDA should use its authority to re- quire Risk Evaluation and Mitigation Strategies to counter mis- information or restrict use of the medication as appropriate. Such an approach would reestablish the balance between safety considerations and new drug approval. The FDA should pur- sue additional safeguards to prevent breakdowns in sponsor- investigator relationships and avoid the dissolution of future trials. For example, the agency should consider having data monitoring committees report interim results directly to the FDA, not to the sponsor.

The LIGHT study should serve as an important message to sponsors of clinical trials. As articulated by Fleming,16

repeated breaches of confidentiality may leave the FDA no alternative other than to require that full safety studies be con- ducted prior to product approval. The demise of the LIGHT trial is a reminder that basing approval on interim safety data is a carefully drawn compromise, not an entitlement.

ARTICLE INFORMATION

Author Affiliations: Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland (Sharfstein); Cardiovascular Health Research Unit, Departments of Medicine, Epidemiology, and Health Services, University of Washington, Seattle (Psaty); Group Health Research Institute, Group Health Cooperative, Seattle, Washington (Psaty).

Corresponding Author: Joshua M. Sharfstein, MD, 615 N Wolfe St, W1033D, Baltimore, MD 21205 ([email protected]).

Conflict of Interest Disclosures: The authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Sharfstein served as principal deputy commissioner of the FDA from 2009 to 2011. Dr Psaty reports serving on the data monitoring committee of a clinical trial funded by the manufacturer (Zoll LifeCor), on the steering committee of the Yale Open Data Access Project funded by Johnson & Johnson, and on the FDA Science Board.

Funding/Support: This work was supported in part by grants HL078888, HL103612, and HL105756 from the National Heart, Lung, and Blood Institute.

Role of the Funder/Sponsor: The National Heart, Lung, and Blood Institute had no role in the preparation, review, or approval of the manuscript or decision to submit the manuscript for publication.

Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Heart, Lung, and Blood Institute, the National Institutes of Health, or the FDA.

REFERENCES

1. Nissen SE, Wolski KE, Prcela L, et al. Effect of naltrexone-bupropion on major adverse cardiovascular events in overweight and obese

patients with cardiovascular risk factors: a randomized clinical trial. JAMA. doi:10.1001/jama .2016.1558.

2. Pollack A. FDA declines to approve diet drug. New York Times. February 1, 2011. http://www.nytimes.com/2011/02/02/business /02drug.html?_r=0. Accessed February 11, 2016.

3. Guidance for Industry: Diabetes Mellitus—Evaluating Cardiovascular Risk in New Antidiabetic Therapies to Treat Type 2 Diabetes. FDA website. http://www.fda.gov/downloads/drugs /guidancecomplianceregulatoryinformation /guidances/ucm071627.pdf. December 2008. Accessed February 11, 2016.

4. Fleming TR. Protecting the confidentiality of interim data: addressing current challenges. Clin Trials. 2015;12(1):5-11.

5. Confidentiality of interim results in cardiovascular (CV) outcomes safety trials; part 15—public hearing before the commissioner. FDA website. http://www.fda.gov/Drugs /NewsEvents/ucm405023.htm. August 11, 2014. Accessed January 23, 2016.

6. Bethel MA, Sourij H. Impact of FDA guidance for developing diabetes drugs on trial design: from policy to practice. Curr Cardiol Rep. 2012;14(1):59-69.

7. Scirica BM, Bhatt DL, Braunwald E, et al; SAVOR-TIMI 53 Steering Committee and Investigators. Saxagliptin and cardiovascular outcomes in patients with type 2 diabetes mellitus. N Engl J Med. 2013;369(14):1317-1326.

8. Zannad F, Cannon CP, Cushman WC, et al; EXAMINE Investigators. Heart failure and mortality outcomes in patients with type 2 diabetes taking alogliptin vs placebo in EXAMINE. Lancet. 2015;385 (9982):2067-2076.

9. Center for Drug Evaluation and Research, FDA. Summary review of application number 2000063Orig1s000.

http://www.accessdata.fda.gov/drugsatfda_docs /nda/2014/200063Orig1s000SumR.pdf. Accessed January 23, 2016.

10. Zhang JJ, Blumenthal GM, He K, et al. Overestimation of the effect size in group sequential trials. Clin Cancer Res. 2012;18(18):4872- 4876.

11. US Securities and Exchange Commission. Form 8K [filed by Orexigen Therapeutics Inc]. http://www.sec.gov/Archives/edgar/data/1382911 /000119312515074251/d882841d8k.htm. March 3, 2015. Accessed January 23, 2016.

12. Cleveland Clinic. Clinical trial testing safety of obesity drug Contrave halted; 50 percent interim data released by the study’s executive committee. http://my.clevelandclinic.org/about-cleveland -clinic/newsroom/releases-videos-newsletters /2015-5-12-clinical-trial-testing-safety-of-obesity -drug-contrave-halted. May 12, 2015. Accessed January 23, 2016.

13. O’Brien PC, Fleming TR. A multiple testing procedure for clinical trials. Biometrics. 1979;35(3): 549-556.

14. O’Riordan M. LIGHT stopped: Contrave CVD safety study halted following premature release of data. http://www.medscape.com/viewarticle /844575. May 12, 2015. Accessed January 23, 2016.

15. Psaty BM, Rennie D. Stopping medical research to save money: a broken pact with researchers and patients. JAMA. 2003;289(16):2128-2131.

16. Fleming TR. Response. Clin Trials. 2015;12(1):21-23.

17. Helfand C. Orexigen picks up new trial tab to settle Contrave dispute with marketing partner Takeda. http://www.fiercepharma.com/story /orexigen-picks-new-trial-tab-settle-contrave -dispute-partner-takeda/2015-08-07. August 7, 2015. Accessed February 14, 2016.

Opinion Editorial

986 JAMA March 8, 2016 Volume 315, Number 10 (Reprinted) jama.com

Copyright 2016 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ by University of New England Library, CARISSA LEE on 09/03/2016