Public Sensationalism and Clinical Trials: How to Address the Challenges of Science?

Article Outline

• References
• Copyright

The conduct of a clinical trial consists of a time-honored approach of developing a hypothesis and then prospectively going forward and testing it among patients. Successful completion of a trial requires careful attention to important guidelines that govern its conduct. One key element of any randomized trial is oversight by an independent data safety monitoring board that serves to protect patient safety and ensure patients are not exposed to untoward risk. In particular, this requires that patients are not randomized to a clearly inferior or even harmful treatment.¹

In recent years, some trials have encountered rare and unexpected events that modulate future conduct of the trial. Occasionally this has led to unblinding of a trial or even refusal by some institutional review boards to proceed. Rare events that are not part of the primary end point or unexpected safety concerns can be difficult to assess. If an unexpected side effect emerges, which is unlikely if the trial is based on the best existing knowledge of science, it may be unclear whether it simply represents a random observation or a genuinely new finding attributable to the study intervention. In the latter situation, it may be appropriate to unblind a trial or suspend, at least temporarily, further patient enrollment. The challenge may be even more complex in the event that a larger trial is being conducted concurrently. In this instance, greater patient exposure with more events could ultimately clarify the situation.

This issue emerged recently when an unexpected finding of cancer was identified in the ezetimibe–simvastatin investigative program. This led to the unprecedented action of unblinding a trial in progress and the reporting of non-primary outcomes data from 2 ongoing trials.² Needless to say, the investigative community was concerned about the potential for harm, and some sites dropped out of the trial. Furthermore, the public release of the “cancer concern” led many patients to stop participation in clinical trials, and patients, outside trial milieu, also may have stopped other evidence-based, lipid-modifying therapies. When all the data were ultimately combined from 3 large outcome studies, no effect of ezetimibe on cancer occurrence was observed. This unprecedented response to the concern for harm was facilitated by cooperation of the data safety monitoring boards and independent steering committees working together (independently of the sponsor) to provide a report that clarified the issue around cancer. This approach exemplifies how appropriate data safety monitoring board response and independent scientific oversight can address such issues arising during ongoing trials. Ultimately, when all the ezetimibe–simvastatin trials are completed, a clearer and more definitive picture will emerge regarding the risk of cancer. This will allow the scientific community to better define the risk–benefit ratio of this therapy to decrease low-density lipoprotein cholesterol.

In another recent trial evaluating the antiplatelet agent prasugrel versus clopidogrel among patients undergoing percutaneous coronary intervention, a higher rate of colonic neoplasms was observed through adverse event reporting (0.2% vs 0.1%, P=.03).³ This may have been related to excess gastrointestinal bleeding. However, ascertainment bias due to the expected effect of a more powerful antiplatelet agent, such as prasugrel, may have unmasked prior neoplasms. Detailed analysis by the Food and Drug Administration and the sponsors identified that there were 311 cancers in the trial (2.3%) observed through adverse event reporting.⁴ Subsequent analysis and clarification by the site investigators determined that there were only 174 (1.3%) new non-benign neoplasms diagnosed after the start of the study medication. When these were analyzed according to randomization strategy, there was no significant difference in the rate of new cancers with prasugrel versus clopidogrel (risk ratio 1.17; 95% confidence interval, 0.87-1.58; P=.30). The overall final rate of gastrointestinal cancer in the TRITON trial was 0.5% with prasugrel and 0.3% with clopidogrel, with a similar rate among patients without bleeding events (0.1% for both). Also, when additional analysis was performed excluding non-melanomatous skin cancers (although malignant, but generally with a benign course), the findings remained nonsignificant (relative risk: 1.31; 95% confidence interval, 0.95-1.80; P=.09). In addition, preclinical studies with prasugrel have failed to detect any signal for an increased risk of cancer. The use of prasugrel was subsequently approved by the Food and Drug Administration in June of 2009 and is now in use in many countries around the world.

However, this careful analysis by the Food and Drug Administration has not abated the issue about potential cancer and the use of prasugrel. Serebruany⁵ recently published an editorial stating “The significantly high rate of cancer after prasugrel is alarming and can be reasonably explained …,” suggesting that the findings were significant and real. Although we believe that reasonable scientific discourse around cause and effect is of great value and should always be encouraged, this statement fails to report the lack of a significant overall increased risk of cancer. Subsequent posting on various web pages led some sites' institutional review boards to consider withdrawing from an ongoing trial testing prasugrel versus clopidogrel in a more chronic setting, thereby undermining the opportunity to adequately explore this in a larger number of patients.

How can the scientific community best manage reasonable differences of opinion, avoid undue alarm within society, and ensure that the scientific rigor required for good science is allowed to proceed in a safe and ethical way? First, we in the scientific community should use the word significant carefully in scientific writing. We presume that, in general, a P value of less than .05 denotes a statistically significant finding after statistical testing. Thus, describing an association as significant when the P value is greater than .05 is likely to confuse the public. Furthermore, when describing rare events (<2%), we should be even more cognizant that a random effect can appear as a nominally statistically significant finding, yet with reallocation of 3 to 4 patients within a trial of 10,000 patients it could become nonsignificant.

Second, investigators and institutional review boards should look more to the data safety monitoring board of any trial to provide guidance when thorny and complicated issues arise in ongoing clinical trials. As was described in the ezetimibe experience, the data safety monitoring board of multiple trials worked together to provide valuable insight into the safety of patients already enrolled, as well as potential patients. A similar approach has been advocated by others.⁶ Timely communications of the data safety monitoring board to investigators and institutional review boards would minimize any confusion and ensure that investigators and institutional review boards are kept abreast of the science as guided by the steering committee and data safety monitoring board. For large-scale trials or those with a large risk of adverse events, it seems prudent that the data safety monitoring board should meet at least every 6 months, and more frequently if a signal of an adverse effect appears. This should be followed by regular communications on safety and unexpected events to the steering committee, institutional review boards, and investigators.

Third, investigators and leaders of clinical trials should be encouraged to report any rare unexpected cluster of events to the Chair of the data safety monitoring board. This communication would help the data safety monitoring boards to review rare events and put them into a perspective for the trial as a whole. For example, in a trial in which a survival benefit seems probable, a cluster of abnormal liver function test results among patients receiving an investigative agent may be interpreted differently from a trial with similar liver test result abnormalities and survival curves favoring standard therapy.

Finally, because the risk factors for coronary artery disease and malignancy are similar (ie, smoking, physical inactivity, obesity, and inappropriate diet), we should not be surprised at the detection of cancers among patients in cardiovascular trials who are followed long term over several years. The current methodology for collecting and characterizing adverse event reporting is too susceptible to investigator bias, as was noted by an approximately 45% reduction when adverse event reports were clarified in the TRITON trial, which makes subsequent analysis difficult and often ambiguous. These observations become especially relevant to the ongoing TRILOGY-ACS (NCT00699998) trial, comparing clopidogrel with prasugrel among medically treated patients with acute coronary syndrome. In this trial, we are studying older patients who may be at higher risk of cancer over an expected median follow-up of at least 15 months. Accordingly, we are prospectively collecting information at baseline and occurrence of new cancers with a detailed case report form to minimize any biases. We expect this approach to set a new standard for cancer surveillance in trials for heart disease. We believe the scientific community should collaborate on the development of a standardized method for prospectively collecting information on cancer (either new or ongoing) in all cardiovascular trials in which the average follow-up period is long enough to expose the patient to therapy that may induce cancer or allow detection of preexisting cancer. A reasonable time may be 9 months.

Allowing important trials to rationally proceed and avoiding unnecessary sensationalism will help to better define the risk and benefits of any therapy. Central to this approach is the need for trust and transparency between the scientific community and industry. A wise and diligent data safety monitoring board is the cornerstone of an approach to minimize biases and allow clinical trials to reach their full potential, thereby best serving the needs of our patients.

Back to Article Outline

References 

1. DAMOCLES Study Group, NHS Health Technology Assessment Programme. A proposed charter for clinical trial data monitoring committees: helping them to do their job well. Lancet. 2005;365:711–722
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (456 KB)
   • CrossRef
2. Califf RM, Harrington RA, Blazing MA. Premature release of data from clinical trials of ezetimibe. N Engl J Med. 2009;361:712–717
   • View In Article
   • CrossRef
3. Wiviott SD, Braunwald E, McCabe CH, et al. TRITON-TIMI 38 Investigators Prasugrel versus clopidogrel in patients with acute coronary syndromes. N Engl J Med. 2007;357:2001–2015
   • View In Article
   • CrossRef
4. Available at: www.fda.gov/downloads/AdvisoryCommittees/CommitteesMeetingMaterials/Drugs/CardiovascularandRenalDrugsAdvisoryCommittee/ucm129219.pdf. Accessed March 2010.
   • View In Article
5. Serebruany VL. Platelet inhibition with prasugrel and increased cancer risks: potential causes and implications. Am J Med. 2009;122:407–408
   • View In Article
   • Full Text
   • Full-Text PDF (206 KB)
   • CrossRef
6. Peppercorn J, Buss WG, Fost N, Godley PA. The dilemma of data-safety monitoring: provision of significant new data to research participants. Lancet. 2008;371:527–529
   • View In Article
   • Full Text
   • Full-Text PDF (63 KB)
   • CrossRef