The timing of the disclosure of these data engendered controversy. Critical issues surrounded the following questions: (a) Can data from retrospective biomarker analyses provide sufficient evidence to change the design of ongoing clinical trials, with additional potential impact on existing standard of care treatments? (b) What level of evidence would be necessary to accept such analyses?
To consider these issues raised by the process, and guide decision making around future analogous circumstances, the authors looked to principles of biomedical ethics while specifically considering the stakeholders involved. Ethical concepts underpinning clinical research include nonmaleficence (not doing harm), beneficence (doing good), autonomy (respect for individual rights), and justice (fair treatment). Failure to warn persons of newly discovered harm could certainly breach the principle of nonmaleficence. KRAS mutations predict a lack of efficacy of anti-EGFR antibodies in patients with metastatic CRC, so individuals whose tumors harbor such mutations are exposed to harm (adverse effects, shorter survival because of time spent taking ineffective therapy, cost) when taking the drugs without any realistic chance of benefit. If the release of early data would have definitively demonstrated the patient subgroups that could not benefit from EGFR inhibitors, then harm occurring during the time that the results were known to the investigators and the ASCO Program Committee prior to the ASCO meeting could have been potentially preventable. Similar considerations apply to the principle of beneficence. It could be argued that patients treated with these drugs, on or off trial, should have known if new evidence predicted “more” benefit for those whose tumors were wild-type KRAS.
On the other hand, considerations regarding not doing harm apply equally to the potentially damaging consequences of prematurely
drawing conclusions or making recommendations based on data that ultimately are not
correct. This possibility remained foremost in the minds of some Colon Cancer Task Force members, at least until the ASCO presentations highlighted above. For example, high-dose chemotherapy with bone marrow transplant was the standard of care for some women with breast cancer, based mostly on data derived from nonrandomized phase II trials. Phase III results, in general, did not replicate the early data [17
]. Another potential example pertains to CRC BRAF
mutational status. Two modestly sized studies identified a total of 20 patients with wild-type KRAS
tumors who had V600E BRAF
mutations; none responded to anti-EGFR monoclonal antibodies [18
]. On the basis of those trials, some clinicians began routinely assaying for BRAF
mutations, withholding anti-EGFR therapies in patients with such mutations on the assumption that efficacy was not realistically possible. More recently, a pooled analysis of the CRYSTAL and OPUS first-line CRC studies contradicted the original finding, suggesting that adding cetuximab to chemotherapy in patients with BRAF
mutations may lead to longer progression-free and overall survival times [20
]. Were those patients with mutated BRAF
, from whom cetuximab was withheld following presentation of the initial two studies, harmed?
Patients entering clinical trials make the decision to enroll expecting that the investigators assume equipoise. The principle of autonomy suggests that, if the investigators' views regarding the potential benefit of all the respective arms of a trial significantly change based on new evidence, then patients should be informed of this and have the ability to reconsider their treatment. Additionally, the principle of justice requires that the decision about whom to share early results with should fairly consider all patients, on or off trial, who are potentially affected by this information. Potential stakeholders in the KRAS decision included patients, physicians, the NCI, industry representatives, the FDA, the Securities and Exchange Commission (SEC), company shareholders, and professional societies (here, ASCO). While individual data and safety monitoring boards (DSMBs) are also charged with oversight for clinical trials, the DSMBs in this case would not have had access to the information from multiple unpublished clinical trials available to the Colon Cancer Task Force.
From an ethical standpoint, the rights and protection of patients are of primary importance, overriding concerns of nonpatient stakeholders. Nonetheless, the others have roles in patient protection and care advancement. Physicians obviously have to care for current and future patients; the FDA is charged with protecting the health of the general public; and the NCI and ASCO are invested in promoting the oncologic research agenda, ultimately leading to better cancer outcomes. Industry, the SEC, and shareholders obviously have a more financially based interest in drugs being proven effective. All groups are potentially affected, albeit for different reasons, by the decision to modify a trial or make a definitive statement about who should or should not be treated with an approved drug. Failure to disclose “convincing” data should invite scrutiny regardless of the group advocating waiting; however, the definition of “convincing” remains ambiguous. Certainly, a disclosure of data later found to be misleading equally invokes criticism.
Should these data have changed eligibility for ongoing clinical trials? The question must first be raised as to how much agreement was required that the data were reliable. Should one single believer and voice have the power to inform the public? Should unanimity among experts be required? Perhaps a simple majority is sufficient? The situation is clouded when the “experts” have signed legally binding confidentiality agreements with industry sponsors restricting public discussion and disclosure.
In answer to the question about whether or not ongoing trials should have been amended with the emergence of new KRAS data, the answer is clearly “yes,” if the assumptions underlying the arms were no longer valid. The determination of equipoise rested on a standard of evidence that could be determined only by the principal investigators conducting the trials. If this group believed that uncertainty no longer existed regarding the predictive ability of KRAS mutations, then the trial eligibility criteria should have been changed, as they ultimately were in this case (after the 2008 ASCO Annual Meeting presentations). However, earlier disclosure would have required that the investigators conducting these trials be informed of these results before the ASCO meeting, violating a well-established embargo policy. Currently, the ASCO confidentiality policy states that abstracts submitted to ASCO meetings are considered confidential from the time of submission. Information may not be made available to the public or the news media, nor may it be published or presented elsewhere, nor may it be used for securities trading purposes, prior to public release in conjunction with the meeting. In circumstances in which a high standard of evidence has been achieved and results are sufficiently compelling to warrant early release, then specific exceptions need to be created in order to comply with the policy. Whose responsibility is it to approach ASCO to ask for an exception?
Should these data have been released early to the public? This consideration required a higher level of evidence, in order to avoid inadvertent harm to patients. For reasons discussed earlier, ethical considerations that apply to patients suggest that this information should have been made available earlier if the data were felt to be definitive. However, early data often do not meet this standard of evidence, and misleading, premature conclusions may be drawn after early disclosure. “Formal” vetting by expert commentaries at national meetings like the ASCO meetings and “informal” vetting by discussions among stakeholders at these meetings can help to place these results in proper context for appropriate interpretation and subsequent decision making. However, even these methods of data interpretation are imperfect, and it may be most appropriate to use data to change clinical practice only after they have been presented in manuscript form and confirmed. Thus, it may be that, if there are circumstances in which results are so compelling that they change practice following a national meeting, perhaps even earlier disclosure of these results could and should have had the same effect.
Does the type of predictive classifier affect the level of evidence needed to change trial or treatment behavior? New efficacy markers may be used to include or exclude patients from specific systemic treatments. An example of an inclusionary marker would be the use of trastuzumab in human epidermal growth factor receptor 2–expressing gastric cancer [21
mutations, in regard to anti-EGFR therapies, represent an exclusionary marker. How compelling did the evidence need to be to adopt KRAS
mutational status as a hard stop exclusion from our using these agents in a previously approved setting? More specifically, how definitive do the data need to be that a marker does, in fact, exclude activity to the degree claimed, and to what degree does the marker need to fully exclude activity versus merely decrease the likelihood of activity? Illustrating the difficulties inherent in proposing solutions to these questions, substantial differences of opinion existed even among the national thought leaders in CRC as to when the burden of proof was sufficient to justify routine use of KRAS
testing to make clinical decisions regarding anti-EGFR therapies. Some were comfortable deciding based on modest-sized retrospective analyses of treatment with cetuximab in the setting of refractory CRC. Some became persuaded by the retrospective analysis of the small, randomized, panitumumab versus best supportive care trials. However, others were not comfortable with the use of KRAS
testing until demonstration of its utility in the CRYSTAL trial, the large randomized first-line trial of FOLFIRI with and without cetuximab [10
]. As a result, some experts were routinely testing all metastatic CRC patients for KRAS
mutations as early as mid-2007, whereas sufficient consensus as per the National Comprehensive Cancer Network CRC guidelines committee was not reached until the autumn of 2008. Others chose not to change their practice until the FDA changed the label of cetuximab and panitumumab to exclude use in KRAS
-mutated tumors, which occurred in the autumn of 2009. Clearly, many concerned, thoughtful, individuals, all deeply committed to optimizing care of patients with incurable CRC, had different interpretations of the reliability of the same evidence and reached their individual comfort levels at markedly different times.
The matter is further complicated when considering the other stakeholders involved. One could anticipate that clinicians, patient advocacy groups, and pharmaceutical manufacturers and their shareholders might all have relatively low thresholds for accepting a new inclusionary marker, and might have a higher threshold for embracing an exclusionary one, whereas third-party payers and those charged with containing national health care expenditures could possibly display the reverse, with a higher degree of evidence needed for acceptance of an inclusionary marker and a more rapid acceptance of an exclusionary one. Any solution to further ongoing questions of this nature will likely need to consider the needs, rights, and responsibilities of all these entities.